A Cell Atlas of the Developing Human Testis


(mellow music) (soft electronic music) – Science really is a journey, and I think a lot of us
are influenced by mentors, and influenced by environment, and we follow the data where it takes us. And I think that is
something that’s wonderful about our field. We never stop being trainees,
we never stop being students, we never stop learning and
changing to try to attack and solve the most interesting problems that we run into on our journeys. We still continue to agree seriously on Chromatin remodeling complexes. We actually published
in a collaborative work in Science a month ago or two months ago on the structure of risk, a
switching and flight complex and its ejection and sliding mechanism. We also work, as Miles mentioned,
on chromatin development, especially in the early zebra fish and in early mouse and
human embryos as well to understand how you sculpt chromatin before, zygotic genome activation, to properly activate a
repressed developmental in housekeeping genes, but today I’m gonna focus
solely on one story. It’s split into three papers, our work to understand
germ line stem cells. And the way I got into this work is I was fascinated, as Miles mentioned, about relationships between
chromatin and development and how you might use
chromatin to poise genes for subsequent activation and development. And I was thinking, what
would be the most interesting place you could look for poising of genes? And I thought, the germ line. If you could actually test
whether there was evidence for packaging and poising
genes in a particular way back in sperm and eggs, that would then have an influence
on subsequent development, that that would be an
interesting area to look at. So we actually conducted
the first examination of chromatin structure
in nucleosomes in sperm. And this was published
maybe 10, 11 years ago. And the basic result was
that you do package genes both for development and
for housekeeping genes, particular ways all the way back in sperm. In fact, the developmental
genes are packaged very much the same way they’re
packaged in the ES cells. So that was our initial, I’d
say, interest and observation that’s really motivated a lot of the work. And since then I’ve been
very interested in germ line, germ line stem cells, and
developmental potential. So today I’m gonna tell
you about our progress in that area. We’ve been working in applying genomics to this area for quite some time. And this is a group of people who know a lot about development, so I won’t spend too long on this. But this cartoon is not to scale, but it’s meant to demonstrate, I think, this fascinating stem cell
system in the germline. We’re interested in the test,
in the testis as a whole, the niche and the germline stem cells. The seminiferous tubule is shown here, and basically it’s a developmental pathway where you have the undifferentiated or quiescent center slowly
self renewing stem cells here at this basal membrane. And then development goes
from lamina to lumen. And if this blown up here, basically, and this is exaggerated
a little bit for size, but it illustrates the point, that you essentially
have a laminar component that houses the stem cells
themselves, the self renewing, and then they start differentiating. And once they commit to meiosis, they pass through this junction and then undergo meiosis,
become spermatocytes and spermatids, elongating spermatids, and then finally, mature sperm. They’re supported by a variety of cells. Sertoli cells, which
are inside the tubule, and then these myoid, tubular myoid cells form the outer lamina. And then on the outside
they’re also Leydig cells, which make hormones like testosterone. And so we’re very interested
in many of the aspects of this. What determines how do
you build this structure? What determines the self renewal versus differentiation
of these stem cells? How do you maintain this
compartment for a lifetime? So how do you balance self
renewal and differentiation for a lifetime, right? And also, how is the niche
influence this process and makes sure that it works
again well for a lifetime? And I don’t have time, but I want to make sure
I give proper credit to the decades of work
that has been done in mice. For time, I can’t compare. This is such a complex process that I can’t in every result
tell you what’s been known in the mouse, and the rat, etc. That comparative stuff
is very interesting. But there has been beautiful
work done in rodents, including mice, on physiology, genetics, molecular work, genetics especially, which you can do in the mouse, genomics and single cell
analysis in the mouse. In humans, however, when we
started this in about 2016, there wasn’t any genomics or single cell work that had been done. But it was already known that
there was notable differences between the human and the mouse system. A lot of similarities for sure, and we could use those similarities to kind of pin the human
work to see where we are, so that we had confidence
then in the differences that we were observing as well. For example, there’s a major difference in the percentage of undifferentiated spermatogonial stem cells. So it’s maybe .1% or less in a mouse, and it’s several percent in human cells. They don’t undergo the same
level of expansion in humans. They don’t have a long syncytia in humans like they do in mice. They’re gonna be 16 aligned,
16 cells at some point in the spermatogonial differentiation. An also, there’s puberty in humans. We go through this long time. I’ll be talking a lot more about this, of 10 or 11 years where the
stem cells are undifferentiated and basically quiescent
and then transition, as I’ll be telling you more about. It doesn’t happen in the mouse. The mice essentially have
what’s called a first wave where there’s tons of
stem cells at that time, and they either have one of two fates. They either take up shop in the niche, and they become long term, or they can contribute to the long term stem cell population, or they go on to differentiate
into sperm at that time. So it’s that synchronized first wave, which has been fantastic for studying the process of gametogenesis
by many people in the mouse, but that just doesn’t
happen in the human cells. So I think, or in the human system. So I think there are similarities
and lots of differences, and of course, we wanna
understand those differences in detail and study the human. And we thought with
the type of single cell and all the types of genomics’ approaches that we have in our lab,
we had an opportunity with the sample pipeline,
which I’ll tell you about, to really make some inroads
into the human system, we and others. Okay, the physiological, I’m gonna tell you three parts today. I’m gonna tell you about
our initial attempts to understand spermatogonial
stem cell development using the initial fluidigm system, which involves cell sorting. Then we progress to the 10X system where we could look at the whole testis and do the adult atlas. And then the third part, which was published this
morning at 9:00 (laughing), the atlas of puberty. So it’s an opportune
time for me to be here. So we have a set of short term goals, many of which I’ll go through
here today in the work, but we also have some long term goals ’cause I wanna give you
some of the perspective, what the opportunities are. So most of what I’m gonna
tell you about and we’re doing is this and understanding the stem cells, and the niche, and the problems
I was telling you about. Self renewal, development,
potential, etc., and I’ll tell you about
our progress there, but what we’d really like
to be able to use that work to take the field forward
as well in defining human spermatogonial stem cell cultures. And this has not been done with human. It’s been done with a
mouse very effectively, and this is another major difference between humans and mice. It’s actually not that hard to grow spermatogonial stem cells from the mouse, and but nobody’s been
able to do it from human. And if you’re on your
computer you can check it, and you’ll say, Brad,
there are four papers that’s say it’s been done. Well, it’s never been done
twice (chuckles), okay. (laughing)
So that’s the problem. It’s fair, fair statement? Okay, so, and they almost immediately lose their germline identity. They essentially become, they first become ES like cells, and then they differentiate very quickly. So that is a huge goal for the field. We also wanna understand germ cell tumors, how they form and how
fertility might be treated. So a better understanding
of these stem cells which are the origin of
these gene cell tumors, I think is critically,
we need to know that, especially the drivers are known. As you can sequence you can find what some of the mutations are and they often upregulate KRAS or this transcription factor DMRT one, which you’ll see more of. So, and understanding these stem cells, I think is important
for many, many reasons. From a fertility perspective, infertility in men is
actually reasonably common, and if you had an in vitro system, you might have in the future, be able to transiently
express a wild-type gene for which they’re defective, which you would know by sequencing, and you could then enable the
in vitro production of sperm to restore fertility. So I think we have a
medical, fertility reasons to understand this system in great detail. We have cancer reasons, and we have massive, I
think, basic science reasons to study this system. So we’d like to contribute to this field. So the work began leveraging
what other people had told us before about human
spermatogonial stem cells, and that is that they were
basically two flavors. I’m not gonna get into the A dark A pale. I can ask questions, answer
questions about that later. But in terms of being able
to separate cells out, it was known from the work of others there was a lot of physiology
immunohistochemistry that the quiescent stem
cells, the earlier cells, many of them would have
SSEA four as a cell surface glycolytic mark-like lipid. And the more differentiating
cells were KIT positive, okay? So this is true in
humans, not true in mice. This is shared between humans and mice. So the first paper on
this, we did MACS Sorting and enrichment for SSEA four and KIT, and we would stain them, and yes, there were many
double positive cells and that was fine because you’re gonna have cells
going through a transition from this state to this state, and they’ll be double positive, and then you can do their transcriptomes and see if they’re discrete stages that these stem cells are going through to help understand how they
make this journey, okay? So we did bulk RNA sequencing. I’m not gonna tell you anything about that because I think all the major information came from the single cell and totally backed up the bulk sequencing but really deconvoluted
it in an interesting way. We did ATAC sequencing, which is a way of
looking at open chromatin in a genome using a transposase, which hops in where
there are no nucleosomes. So you can mark where the
open chromatin is in cells. This is a powerful technique
because it gives you strong hints about what
transcription factors might be active at that time because you can dovetail
the open chromatin with binding sites for
transcription factors. We also did DNA methylation
profiling to see whether DNA methylation was
changed during this process. These were the early days. I’ll tell you in the step three,
it has 10,000 cells in it. Step one had 93 cells that
passed all of our QC measures. This is the fluidigm system. The sequencing is very deep, but you don’t get as many cells. But still, I think as we
were able to go a long way I think in providing classes,
states of stem cells, which are then validated
in the later experiments I’ll tell you about. Then we did analysis of that
single cell data using tSNE, clustering, Monocle, pseudotime, and a set of validation experiments, which I won’t show you. Since this work was published
a couple of years ago, and I wanna tell you two more
stories that relate to it. I’m just gonna tell you
the take home message, but I need to tell you
because it’s the foundation for the next two stories, okay? So, the take home is that
we were able to identify four separate states that
these cells go through. There are cells, which you can see have, are transitioning between those states, but we think these states
really are reasonably discrete based on all of the data
that I’m not showing you. So, the first state we call the undifferentiated stem cell state. It has some markers that are high in it, that which are familiar from the mouse, like this marker ID4. I’m not gonna, I’m gonna try not to get into too many markers. I’m gonna try to keep it at concepts, but I have to show some markers to provide sort of sign post for people
who familiar with the area. So it was known from working
the mouse, for example, that the undifferentiated
stem cells were ID4 positive. There were, like they were FGFR3 positive. So these are markers known in the mouse. So we were confident that
a lot of the data aligned. There are other factors which
had not been characterized in the mouse before that
we were able to show were fairly unique to
the human system as well. I think what we, when you move
from state one to state two something very striking happens. You move from a state of
quiescence to one of proliferation. So what turns on is Ki-67,
lots of cell cycle genes, lots of genes involved in
DNA replication and repair. So this, as I’ll show you now, and you’ll see very clearly in the data slides I show you later, is a major transition from
quiescence to proliferation. The other thing that happens at this time is many of these factors turn down. So your stem factors, some of them really fall very strongly. The next thing we notice, which
is very clear in the data, was an effect on metabolism, okay? And we noticed that in
these undifferentiated cells had very high levels of
this guy called TXNIP. This is the master regulator, or negative regulator of glucose uptake. So these cells actually, if
you look at the transcriptomes and infer the proteome from
them, they are glycolytic, but they have very little glucose, okay, because they have high
levels of this protein. It then drops like a stone
when you go from one to two, and proliferation happens. But what also happens now is
that metabolism then kicks in. So NADPH dehydrogenase factors turn up. ATP synthase factors turn up. So you are now utilizing glucose
to make ATP, for example. So these guys are becoming
quite metabolically active. You also see differentiation
transcription factors, some of which were known in the mouse to turn on at that time. So we can see differentiation phasing in. We see changes in splicing. And these metabolic and
differentiation factors phase in as the stem cell factors fall. Right at the end, when you transition from state three to state four, basically the proliferation
takes a nose dive. Ki-67 turns off as these guys
prepare the enter miosis. So taken together, we
think we have a logic, a basic logic for how
stem cells transition from a fairly quiescent and
metabolically inactive state to one that’s proliferative,
metabolically active, and then finally make a
commitment to differentiation by going post mitotic. We were also able to
identify that yes indeed, these stem cells have
the signaling pathways known in the mouse to be
important for growth in vitro. So they the receptor for GDNF
and GFRA1, FGF2 receptors, FGF receptors one, two, and
three, LIF, for example. So those are all present. This is sufficient for
growing mouse stem cells. It’s not for humans,
but they have receptors. But they also have a lot else
going on in the Wnt pathway, PBGF, BNP receptors, T-span receptors, which I’ll be talking
about endocrine, etc. So we think that it’s a more complex probably ligand relationship and maybe attachment relationship that you need in the human stem cells than has been needed in the mouse system, and we’re interested in
using this information about ligands-receptor interactions
to help us to culturing. We have not been successful yet, but we are working hard on
it, and we have made progress, and I can take some questions on it later. So, after that first paper, I think the strengths were the staging, and some of the insights,
and mouse/human differences, and some potential ligands for culturing, but it was not complete in
that we did it by max sorting. And you sort of only get to
sort out what you know, right? So if there are cells that
aren’t SSCA-4 positive or aren’t KIT positive, you don’t see ’em. So what are you missing, right? So you wanna be able to get other states. We knew nothing about the
development of the niche cells, the presence, or properties, or development to those niche cells, and we don’t know anything
about the interesting conversation that might be happening between the niche and and germline. So 10x Genomics and their approaches was really helpful for that when it could be coupled with
a system of testis isolation that we could exploit in many,
many, many different ways. So the first samples we had were basically from patients with idiopathic pain who wanted a testicle removed, okay. So that doesn’t happen very often, but we were able to get through
the first work to do that. During that work, we decided
to set up a collaboration with Intermountain Donor Services,
now called Donor Connect, who are basically the rapid autopsy group that provides hearts, and
livers, lungs to donor recipients all across the Intermountain West. They were not harvesting
the testis at that time. We’ve set up a collaboration with them where we can get samples. Samples are available from people all the way from infants to very old men. And we get them basically
within a couple hours of death, and then we process them immediately. And we were very regimented
about how we do the collections and the processing so that we can minimize technical variation in
all the work we’re doing. So we are now, this is
enabled by Jim Hotaling, a surgeon at Utah, is a
close collaborator of mine. We’re getting one to three pairs a month, again, with a wide age range. The first study was just men
at peak reproductive age. So late teens and 20s. But we’re doing a lot
more with this collection, which I’ll tell you
more about in am moment. So, actually I have the
statistics slide on here, and I guess it went away. So what do we do? We do three biological replicates, okay. We also do two technical
replicates for each of the samples. So we have six samples total. The slide I took out would of shown you that the R values are extremely good, both for the technical replicates and comparisons of the
biological replicates. We really can’t tell much
of a difference at all between any of these three
biological or technical. So, we derive six very large datasets, which could be combined
as a reference dataset that we hope will be used by us and others as the reference dataset going forward for all the other studies
we’re gonna do in puberty, fetus, infancy, adult
aging, and fertility, etc. So, I’m gonna show you, some of you may be familiar
with this, some of you may not. This is a common way of
showing single cell data called a tSNE projection. This is a program called t-Distributed Stochastic Neighbor Embedding, which is a non-linear
dimensionality reduction algorithm for exploring high-dimensional data. So basically, each of these
dots is a single cell. And you take the entire
transcriptome, right, all the features of it from
the multi-dimensional data, and you basically crush
it down to two dimensions. And you let each of these cells, if these cells are like another
cell they’re placed close on this dimensionality production program, and if they’re very different
then they’re far away, okay? It’s non-linear though. Okay, so it’s just meant for a projection. You have to use other types
of analyses like clustering, and etc. to analyze really
how far apart things are, but it’s a great way just to
show the data, and discuss it, and know what to think
about what to do next. So, just from this projection though, as soon as I saw it I was
very excited about the data. Because what you can do is we can show is that basically these clusters represent the different somatic cell types, and this swoosh from here through here is basically all of germline development. Okay, end of swoosh. Starting with the
spermatogonial stem cells and ending with mature sperm, okay? So how do you establish that? The computer has no idea about biology. It just knows numbers. And it’s just putting
cells as crunched numbers either together or far apart. Well you can do something called casting. We know about the biology. We know the difference
between a Leydig cell and a spermatogonia from
lots of work in the mouse. So what you can do is you can take what you’ve learned in the mouse, and you can say, okay, what
lights up in my clusters? So casting means what you
do is take all the cells and you turn ’em gray, and then you let them be red according to how much gene expression is in that cell for that gene
you’re interested in, okay? And you now start casting
everything you know about the mouse onto it, okay? So, for example, VIM is
a gene that’s basically in the somatic cells
but not in the germline. So immediately, the somatic
cells from the germline swoosh. And you don’t just do one, right? I’m showing you a gallery of 20 here. We do like 300, just to make sure we know that we’re assigning things correctly. But I won’t go through all of these, but you can take markers, which are the key markers in
the mouse from myoid cells, and it only lights this
cluster, Leydig cells. DLK1 only lights these. So you can, and then you
can basically go around gametogenesis from markers
known in the mouse to be in, and spermatogonia stem cells just kinda walk your way around, around
germline development, okay? And what you find is a
lot of what you thought was true in the mouse
is true in the human, and then some things are different. But the extent to which
things are the same gives you confidence
about the differences. You can also use a program called Monocle, and there are various versions of it, but I really like that there
are others out there too, which is called the pseudotime. What does pseudotime mean? Pseudotime means that the program is trying to find sets of
cells that are on a continuum. They’re not that
different from each other, but you can see the changes moving in kind of a linear fashion. It will also find branch points. It’s really effective at
finding cells, progenitor cells, and then branch into two different types. So this is an arrow that’s
essentially pseudo-drawn by the program that says, I
see a developmental trajectory through these guys. They are actually similar to each other and following a course. The other way you can look
at this is you can take all the guys and crunch them
down basically to the line, the pseudotime line, which I’ve done here. I’ve maintained their color so you can see them going from here. Purple one down to mature sperm here. And you keep them on the line, but then you say, okay,
let’s look at the inventory of gene changes, right? This is not a subtle number of genes or subtle changes, right? And aggregate. These are thousands
and thousands of genes. And each one of these stages is actually characterized by very
large numbers of genes changing as you go from one
step to the other, okay, through spermatogonial changes, through the gametogenesis process. So, it provides then an
inventory of over 8,000 genes in an order that can be used as a dataset to compare anything you want to now. Elderly, puberty, infertility, etc. You actually have this going
now in biopsy samples too. We and others. So the next thing you can do is that t-SNE is great for looking at everybody, but you can also start saying,
okay, this is a blob here, but is this set of spermatogonia here, are there more subtle
differences that you could see if you didn’t have to
compare them to myoid cells? And so what you can do is
you can take the cells out and recluster ’em. So we take out this guy and this guy, which are the early and
later spermatogonia, as I’ll show you, and you can recluster them
and look at them in isolation. So if you ask tSNE to
recluster, this is what you get. You get basically one,
two, three, four, five along a continuum, and then pseudotime puts an
arrow through them like this. And then the first thing we asked was, okay, we’ve got these
cells from this procedure, but this procedure is
like no sorting, right? You look at all the cells. There’s a completely
different library prep, etc., etc., etc., etc. So very different than the first paper. But if we compare the markers
that we got in the first paper for states one, two, three, and four, how do they compare with what we now have when we did lots and lots of cells? So it turns out that there
were definitely four states, and they put them in the right order, and all these cells are state
one, two, three, and four. Okay, so that aligns with the first paper, but we had a whole group here which did not look like these guys, okay? Computer put them at the
start of the pseudotime, but that’s again, the computer
doesn’t know development. But if you crunch ’em down
then to pseudo timeline and you look at the genes, again, we’re looking at
a large number of genes that are changing along this timeline, what you find is the guys
that are called state zero have hundreds of genes, right, that are up in state
zero, down in state one, conversely higher in state. Sorry, down in state zero
but higher in state one. Okay, so, it’s not just three genes. It’s hundreds of genes that are difference between those two states. There’s probably some
heterogeneity between them. That’s fine. But they are quite
different from one another. And so we were interested
in whether these guys might be a distinct state. So here is the node I told
you a lot about before. That is the node between
state one and state two. It is a virtual flip of the
transcriptome at that time. Okay? What was on turns off, and many things that were off turn on, and you get basically
cell cycle in mitosis, and then you go through
the stages of up-regulation of the mitochondria, etc., and then finally commitment
to miosis and gametogenesis. So everything the first paper
validated by the new work, and a new insight on a state zero type. So, we now have, we
thought we had five states, and if you can compare them
to the mouse work, etc., again, I don’t wanna get
too much into markers here, but there are a few things
that caught our eye. For example, I hadn’t shown Ki-67 yet, but look, it’s basically
off in state zero and one, and state two, bang, goes up, and then in state four it turns off ’cause you’re gonna go
postmitotic to become miotic. This is a common marker used,
in fact, a classic marker, one probably two or three
that’s used in the mouse to say, I am a
undifferentiated spermatogonia. These are actually a very high. GFR alpha one is the
receptor for GDNFN state one, but it’s really very few state zero cells have a GFR alpha one. So it’s one of the distinguishing features between zero and one and the difference
between mice and humans, which we don’t thing, or we don’t have evidence that they have a
equivalent state zero cell. So, what is unique about state zero? They have PIWIL4, which is
a marker we’re using a lot. They also have PIWIL2. So they have a think
a small RNA repertoire that we think might be
very, very interesting for suppressing retrotransposons
and other targets, and we are very interested in looking at this work more closely. They also have specific
transcription factors like EGR4, chromatin factors like MSL3, and cell surface markers that we can and have used to sort in the paper, TSPAN33 here, which high
in state zero cells. So we think this is a
state that is distinct and can be studied based on the fact that we can now fax ’em out. And I’m not gonna show the data for time, but we did sequential FISH
with Long Cai at CalTech to validate several of these key markers and immunohistochemistry
and other standing methods with Ann Goriely, a very
good collaborator at Oxford. I think for time, I won’t go into the single
molecule validation. It’s published and it’s in the paper, but it’s really a nice technique, the single molecule FISH, and Long Cai is now, I
think, ramping this up to multiplex at thousands of (mumbling) They’ve had a couple nice
papers on this recently. So, I said before that the computer says state zero comes before state one, and meaning it’s the
most, possibly the most undifferentiated stem cell in the adult. And we hypothesize, if that’s
true, how could we test it? We can’t, they’re not like mice. You gotta do it with humans. What we thought would be
a good test to the model is if you actually looked
at what infants have. So infants don’t make sperm. They only have
undifferentiated spermatogonia. So what do their spermatogonia look like? Do they look like state two? Do they look like state one? Do they look like state zero? So, we, from our donor connect collection, we had two infants, both of which were
between 12 and 13 months, and they had very few germ cells, but all those germ
cells, infant germ cells, when you combine them
with the adult germ cells, overlapped very well in
the state zero range. So they were all positive for the markers that I showed you in the previous slides. So we think this provides evidence. And Miles Lab has evidence from
this from neonates as well. That there is an early stem
cell we call state zero that is present at birth and
with you for your whole life and from which all these other
states are likely derived. And you need a balancing system
to make sure that you can keep them at the right state
and rate for your life. So, we did, one of the things that, one of the other ways we
analyzed that the data was using a program from the Kharchenko Lab called Velocity or Velocito. And actually this came out in 2018, but we saw the beta version
of this in our archive early and applied it early, ’cause I thought this was
a very interesting program. ‘Cause what it can do is it
can take your single cell data and say, okay, let’s look
at the introns and say, where is this, by just
looking at the introns, what is the cell trying to become? Is it trying to go somewhere? So is the intron repertoire similar to the next state or not? And that’s a basic way of looking. So it’s basically inferring
nascent transcription by looKing and doing intron analysis, and comparing it to the total
RNA pool, the splice pool. So, what I thought was of
interest was two things, is that there’s a population, a sub-population of state zero cells that seem to be wanting to move into the state one direction, and population of state two cells that seem to be wanting to move back into the state one direction, okay? And it made us think a lot about the potential of plasticity, that maybe one of the ways that you wanna balance the stem cell populations is that actually they can
go back and forth, okay? And remember, it’s not just a
few genes that are changing. It’s hundreds to thousands of genes that are changing in these state changes. So how, normally when you
go through development, you open and close chromatin,
you DNA demethylate, you methylate, etc., you
make yourself go forward so you can’t go backwards. So how much is the chromatin changing? So what we did is we
did ATAC-seq profiling of these early state cells with SSEA4 and these late state cells with KIT, and we ask the extent to which their ATAC-seq maps are similar. And we also compared them to ES cells, for example, as a comparator. And what we found was
that the SSEA4 purified and the KIT purifieds are actually very, very much like one another. There are a few dozen
changes in the genome, and we’re interested in
knowing whether those are potential drivers in helping state move forward and back, but we’ve done a lot of
ATAC-seq profile development in other contexts. And usually, when you
see 500 genes change, you see 500 low side. I mean, 200 or 1,000 either
promoters or enhancers open up as part of that process. Okay, to see thousands of genes change, and to see 23 ATAC-seq changes is a real disconnect in my view. The second thing is there’s almost no change in DNA methylation. There is nothing above a
false discovery rate of 1% that changes in DNA methylation between this state and this state. So what I’m speculating and will propose is that this is the chromatin
basis for plasticity. That if you wanna move
forwards and backwards, and you want to be able
to do it fairly easily, you don’t commit yourself to large changes in your open chromatin, and you don’t DNA methylate your way through the developmental process. It allows you to go
backwards without putting in the kind of enzymatic
and organizational power like you might have to with TET proteins, which Anjana was a big discoverer of, to do that type of process. So, if the chromatin is open
right from the beginning, where is it open and does it make sense? Okay, so what I’m showing you here is the difference between
a spermatogonial stem cell and an embryonic stem cell
and in the chromatin map. And what we decided to
do is to look at our maps and say, okay, what is open? Open means you get the transposase
to hop in in a stem cell but not open in the embryonic stem cell. Then this is the unfiltered
list here of what’s open. So, unfiltered. So this is the top 12 guys. And it’s basically CTCF, DMRT1, which is famous both in sex
reversal and in cancer, tumors, and then you see pioneer factors, and you see hormone receptors. So, we’re interpreting this as, it’s an open landscape upon which pioneers work together with hormone
receptors and other factors to maybe guide you, help guide you through this developmental
transition process, okay? So, what this work, which
is published last year, I think told us was we have five states. (stuttering) I think bigger contribution here is this identification of the state zero cell, which we think is right there at birth. I won’t tell you the data, but we’ve been able to trace
this back now into the fetus to identify the time at which this cell is actually identified. We think it goes through
this regimented steps that I told you about before. And we think might be
the reserve stem cell that you always have, and I’ve told you about
our speculation about a chromatin basis for plasticity. I will say that there
is complementary work, which I don’t have time to
go into all the comparisons from several other
labs, the Tang/Qiao Lab, the Hermann/McCarrey work,
and also nice work from Miles that was published that I
think compliments the work. And I think between the labs,
there has been major progress in using these genomics
technologies to understand all of these questions
that I’ve been discussing. So the last thing I’m
gonna do is talk about our, or it was unpublished until this morning, atlas of male puberty. And we were able to, these
are incredibly difficult samples to obtain. We were able to obtain a spectrum of samples through puberty. It’s a small number, but basically these samples
are almost impossible to get. So we felt fortunate to get them, and I thank the families who
were able to donate these for these efforts. And we did the same
replicate structure too, technical replicates for each sample, and then we can compare
them to our infant data and our adult data to understand
the process of puberty. So I’ll just say, puberty, the male testis and the female breast are the two organs that
change dramatically after birth, right? So if you look at the testis in humans, and at a seven year old, for example, it’s basically these little chords. There’s no seminiferous tubule. There’s basically no lamina. It’s basically a chord of sertoli cells together with spermatogonial stem cells. The stem cells are disorganized among the small sertoli cells, and then an amazing
physiological change happens to form the thick lamina, and the lumen, and set up a developmental progression. So, we about 10,000
cells to look at puberty. Again, this goes from one,
seven, 11, 13, 14, 25. We did the casting, so we
know what all the cells are. I won’t go through that again because I went through it previously. We were able to recluster then
just he germ cells lineage and explore that and do the casting to get confidence in the order, and the order was what
we hade seen before, but there’s a bias in terms
of where the samples fall in terms of the order with the young, the infants and the youth basically having all their spermatogonia here, as I’ll show you more in a moment. So this is the depiction
of just the spermatogonia and germline laid out either in four steps of whether they’re undifferentiated, differentiated spermatogonia,
spermatocytes, or spermatids. This is percentage, okay? Whereas this is total numbers of cells. So this is both germline
and the niche cells, okay? So I’ll toggle back and forth here. It’s a little bit hard
to see so I’ll help you. So this is spermatogonia. So for example, only
about 3% of the cells, total cells in the
testis are spermatogonia in a one or a seven year old, and 100% of those cells
are undifferentiated, okay, by our criteria. And then things change at 11. At 11 you get this huge expansion. About 15% of the cells are,
15-20 are spermatogonia. So you have had a spermatogonia expansion, and, but you don’t yet have
any sort of mature spermatids coming out at that point, okay? You have differentiated. You have immature spermatids, but you don’t make any mature sperm yet. And then between the 13 year old sample and the 14 year old sample
a huge change occurs, and our 14-year old looks
a lot like our 25-year old. So I wanna say, it’s not meant
to mean that these samples pin every person of that age, right? This is one person who
happened to be that age, and we would like to have
dozens and dozens of samples where we could do test weight
and other types of criteria that endocrinologists use for staging, but that is just not possible, right? It’s extremely hard to get these samples. You can do certain validation tests, which we have done as
well with a collaborator, but I just wanna leave you with that. So, you can even see the
expansion between the seven and the 11-year old of these UTF1 positive undifferentiated spermatogonia. Low levels in the seven year old, and then lots of them in the 11-year olds. You can see this expansion
in the validation setting. One of the things we were
really interested in seeing is their myoid cells and Leydig cells are two of the cells in the testis that are the supporting niche. Very different cells, but it turns out that they
come from a common precursor. So, the pseudotime what
told us that actually are one and seven year old, you could see these early cells, and that they actually went
down two different lineages. So you could see in these early cells actually markers that you
would normally only see in later cells on both sides. So they would be both IGF1 positive and ACTA2 positive, for example. So these cells would be, would express at low levels
markers normally used to discriminate what the
two cell types actually are. So they think that they
derive from a common precursor in the juvenile. Whereas this process
would have been defined much earlier in the
field case in the mouse. You can also see this
in a pseudotime format where you can now project the cell. So these are the only
the one year old cells, only the seven, only the 11. You can sort of see them
progress in pseudotime down into the two different lineages. We also were able to
demonstrate computationally that we think there
are two sertoli states, an immature we call one and immature two that are present early in development. You can see these, especially
in these early samples. And then the mature sertoli seem to be much more homogeneous. We do not know if these are
two developmental states. I actually don’t think they are. I think they’re probably
physiological states, but without lineage tracing
there’s no way to know for sure. But computationally, they
are clearly different. You can sort of see the populations converging by pseudotime. They have certain properties,
very striking properties like the percentage of
mitochondrial RNA expression. And ribosomal protein expression is very different between
these two populations, and they resolve to a
single one as they mature. I think I’ll just go through
these quickly for time. For example, androgen
receptor is one clear example where that develops later. You see many more androgen
receptor positive cells. This is a little bit bright in there. And the the other thing
we were able to show is changes as sertoli cells mature, and transcription factors,
and in signaling pathways like notch pathway, for example. One of them was striking though, which I had no idea even existed, was this class of defensin peptides and other types of small
anti-microbial peptides. A whole suite of them, which
are highly up-regulated in sertoli cells. So you’re across the blood-brain barrier. You can’t get immune cells to help protect your seminiferous tubule. So it looks like the strategy is that sertoli cells secrete
once they’re mature a whole lots and lots of these small anti-microbial peptides. So it turns out, yeah, you can see it. We can even stain for some of them. Now the important is something
I tell you in just a minute. So, this is kind of a summary. For time, I’m not gonna go through it. I’ll just sort of go
through it at the end. But I will tell you that we notice that as the Leydig cells make testosterone, and they split from the myoid
cells early in this process and start to make testosterone. So they’re almost, making the testosterone is almost part of the developmental
pathway of the process. So we wondered whether, what does testosterone
do in the human system to help either establish or
maintain testis development. We can’t do establishment, but we could do maintenance
because there are, you can get people who are
on testosterone suppression. So, trans females who are undergoing gender confirmation surgery go on long-term testosterone suppression, and we were able to get
two samples to study and do all this genomics with. And the results I think
are quite interesting. Here is an example with one of them. So a ton of spermatogonia,
and almost no spermatocytes, and very, very, very few
spermatoids or mature sperm. It’s known that this greatly reduces or eliminates sperm production
in people over time. I won’t go through all the casting, but it was quite clear. We could show that in one patient, the early spermatogonia
are highly up-regulating in terms of percentage
relative to untreated. And in some cases there’s actually 100% of the germline is actually is undifferentiated spermatogonia. One of the things we wanted to know is, what is the state of that spermatogonia? So are they very much
like untreated males, or are these spermatogonia
very, very different now than the untreated? So we did tSNE and other
types of analysis that basic, and we were able to look
at the top 100 genes, basically let’s say, of state zero genes or state one genes in the untreated case, and say, how much is
that like the treated? And the untreated and treated are almost indistinguishable, okay? So they’re very much like one another. I think that makes sense because you have
undifferentiated type zero and type one spermatogonia
as a juvenile or an infant, and there’s very, very
little testosterone. So having no testosterone
is actually a normal state before you hit puberty. So the cells simply either stay in or revert to that state
before puberty, okay. That wasn’t the only thing that reverted. We could actually see an
effect on the sertoli cells. So this is the transgender
does not look exactly like the 14-year old or the 25-year old in terms of where these cells position. It looks regressed. So we tried to figure out
what is the difference at the gene level between them. And what we found is,
I told you about these defensin peptides here, this whole suite. Basically they are extremely high. This is a log scale. Extremely high in the untreated males, but in the treated males
they are extremely low or undetectable. So we think these are androgen
testosterone dependent, testosterone driven enate peptides, and I think the obvious
possibility is that you don’t active them until
you’re sexually mature because you don’t have any risk. You have much less risk of infection before you become sexually active. So I think I’ll, for time,
I see the time is gone. So remember, there are other things I don’t have time to show today, but I think it’s quite interesting that we’re able to look at all
sorts of signaling pathways between the germline and the niche, and ideas about going forward, what types of things we might look at. The retinoic acid pathway is interesting. Retinoic acid being created and destroyed, and where that’s occurring. We’re very interested
in the activin pathway. We couldn’t help but notice that basically the activin receptors
are in the spermatogonia, but spermatogonia are
making actually inhibitors to the activin path, or to that pathway, and neighboring cells are making
activators of that pathway. And some of those greatly
change at particular times. Like here, the inhibitor,
inhibin A is actually very high when you’re not going through
spermatogonial expansion and then low when you do. So we think we might have
some clues about human, or primate specific, or
applicable ligan systems that we can explore that
haven’t been explored before in the mouse for these types of issues. So taken together, the
puberty part of the story, which was published this morning, shows three phases I think
for the spermatogonia, all undifferentiated, then
activated, or expanded. You start making spermatogonia, but you don’t let ’em make sperm, and then you progress
into sperm production. We think two sertoli states
can form down to one. We would like to understand what these two states are different, but how they’re really different. We have a progenitor cell that bifurcates into making the lamina
and making testosterone. So very, very different cell types. And we’ve been able to work
on testosterone suppression by having these precious
samples from the trans females and been able to show
that those spermatogonia basically look like the
undifferentiated state you would see in a juvenile. And we hope next to use this information to help culture spermatogonial stem cells and progress from undifferentiated
to differentiated. And we’re working on this establishment of state zero in the fetus. And I thank you for you time, and I’m happy to take questions. (applauding) – [Student] One thing I
was kinda curious about, you have that one pseudotime
where you had conversions of the spermatogonia to maturity. – Sertoli cells. – [Student] Or sertoli cells.
– Yeah. – I guess the thing that
I’m wondering about is, did you try, do you think that they
could be actually making, they could be pro-generators
of subtypes of the mature cells and then trying to recluster
out those mature cells, and then see if they
are actually making more specific cells types, I guess is my– – We did try to sub-cluster them, and we didn’t see anything that would give us confidence in that. I think it’s really hard. I mean, I think we have to be careful when you interpret ’em. Without lineage tracing,
etc., we really don’t know, but what we are interested in pursuing it. My gut tells me that they’re
probably just different states, and you don’t have the
same type of cycling that you have in mature cells. – [Female Student] So I’d loved
your theory about plasticity as being non-chromatin,
non-DNA methylation changes. So is that mostly signaling
of different kinds? – It would have to be. So thank you for asking. It leaves me a little bit
more time to expand on what I mean by that, okay. So, I love, ’cause you
know I love chromatin. I love chromatin changes, and I love chromatin modifications, and I think they’re all
important and interesting. That said, I think we have to know when, what is cause and what’s
consequences, etc. So I’m not saying that they’re are no chromatin modifications
that are happening, right? We’ve checked. We and others have checked for, during gametogenesis, for example. Things get acetylated when
they get active, and K4, and all that stuff happens. Okay, so chromatin
modifications are still, I think regulating these pathways. But I think the logic is
that you’re starting off at the beginning fixing two things. You’re fixing DNA methylation. I don’t mean that anything’s broken. You’re establishing a
state of DNA methylation and a state of open chromatin. One, you wanna keep completely unchanged. And that is, actually this is true also if you go through gametogenesis. So if you don’t just stop
at the end of spermatogonia, you go through gametogenesis,
there is no change. We spent a lot of money
on mouse and the human showing there is no change. I wanted there to be a change, okay. So, that I think is a
fundamental principle. I think that’s something from an epigenetic inheritance standpoint. You wanna lock that down really strongly. It seems that that’s also, you are also opening
that platform initially for spermatogonial differentiation. So I think that you’re gonna
have signaling control. That’s absolutely right
with transcription factors, and that interplay will determine which transcription factors
are maybe getting translated, are moving into the nucleus, and are then executing
those programs, yep. – [Student] Were you able to
look at chromatin confirmation in the transgender individuals and compare them to the
non-transgender and infant? – No, we were not. We did not. We didn’t look at ATAC-seq or
a Hi-C or anything like that. Yeah. – [Student] Brad, it’s an
interesting possibility that the chromatin is
maintained in an open state. – Yeah. – [Student] Which would be like embryonic and susceptible to many changes. So I’m curious about whether
you’ve actually looked at testicular cancer to
see whether it actually evokes that kind of organization, what you do see in a
number of other cancers, which reprogram to an embryonic state. – Yeah, so that’s a good question. I’ll give a little of nuance
to the first question, or first comment. So, when we say embryonic, one
of the contrasts I made there was between an ES cell and
a spermatogonial stem cell. They are actually incredibly
different in what’s open. So they are, I take very, very different. And but– – [Student] I’m talking
about the embryonic form of the spermatic– – The early spermatogonia. – [Student] Not an embryonic cells. – Right, so and those two the undifferentiated
spermatogonia are actually very, very different from
the embryonic stem cells. What we haven’t done, and you ask a very good question there, compared either the open
chromatin state to tumors. And I think if you look
at the RNA seq though, I mean, it’s incredibly different. So many germ cell tumors will have, for example, an activation
of the pluripotency network. They’ll either be Nanog positive, or they can be Oct4
positive, not all of them but more than you see in
some other cancer types. And what the spermatogonia can do, actually in a very quick way in both, in the mouse system at least, is they convert from
these germline stem cells to being just like embryonic stem cells, and then they differentiate, right? So I think this is a
fundamental principle. If you look at what is distinctly off in a germline stem cell,
it’s Nanog, Oct4, and SOX-2. I mean, they are, they are completely off. But if you look at the
accessory transcription factors that people associate with pluripotency, they’re actually on. So I think this is a
fundamental part of the strategy of germline stem cells. They have to demethylate. All the Hox genes, FOX,
have to demethylate all the transcription factors that are gonna drive
development in differentiation. They leave, the pluripotency, everything about the pluripotency network is open and on, and basically
poised by valent chromatins. So it’s all just like an ES cell. But you hollow out the middle. You DNA methylate Oct4, Nanog, SOX-2, so that you can’t enter
into the implementation of development. That I think is a
fundamental principle, okay? And what goes wrong in
caner is you don’t do that. You don’t keep that system completely off, and that’s why the, so the germ cell tumors
can be of the three, all three germ layers. Okay, so they have tremendous
developmental potential. And if you look on the female
side, you have teratomas, and those have ears, and
teeth, and hair, right? They have unlimited development potential. That actually was one of
the things that drove me into working on this issue
is the cancer biology. I mean, it’s amazing. You’ve got a stem cell
that can become anything, anything, right? Before fertilization, right? It’s on one side or the other, right? So it has the potential
but hold it in check. So I think that’s a key concept. – [Student] Hi, Brad.
– Hi. – [Student] I was just
curious about the defensins and the peptides that you. So do you have any insight into how testosterone is regulating those genes? Do you find sort of regulatory elements that are associated with those genes, or? – So yes, as we have dug into this, it turns out that one of these classes that’s in the literature already has an androgen receptor binding site. And they have already
done experiments in vivo, and in culture systems
to show that they are androgen dependent to turn them on. So I think that may extend into this whole suite that we’re seeing. – [Student] Do you sense
that the trans female will be a good opportunity to help you with the culture system? Like it seems like they are much more homogeneous potentially in terms of the undifferentiated state, or I don’t know, I’m just guessing. – I think we’re gonna use information everywhere we can get it. We have, we’ve made progress with it, but I think there is an, I think it’s much more
complicated in the human system. I think there are ligands
and attachment needs that aren’t being satisfied
in either a cell culture or an organoid system. So we’re making the
most progress right now in keeping the tubule going in the dish. So yeah. – [Announcer] Well, with
that, Brad, thanks for coming. – Okay, thank you. (applauding) (soft electronic music)

Add a Comment

Your email address will not be published. Required fields are marked *