Thorold Theunissen, Ph.D., of the Washington University School of Medicine in St. Louis shares his work using naive stem cells to model trophoblast development.
[MUSIC] My lab studies human pluripotent stem cells. We’re especially interested in understanding different stem cell states and their applications. Recently, we’ve become really excited about using naive stem cells to model trophoblast development. I know many of you here are interested in that as well, including a monos lab so I’m going to focus on that side of the lab today. Just as a very brief basic intro, I think most of you are well aware of the amazing capacities of pluripotent stem cells.
They can be isolated either from the blastocyst or from somatic cells through directory programming and multiple species. Of course, in the mouse system we can take these embryonic stem cells and use them to generate transgenic mice through gene targeting, blastocyst injection, and germline transmission. In the human context, we can take these embryonic stem cells and induce pluripotent stem cells and differentiate them towards a variety of clinically relevant cell types and organoids, either for disease modeling and hopefully, also for cellular replacement therapy in conjunction with gene correction by CRISPR editing. My work has really focused on the concept that pluripotency is not a unitary states. We now appreciate that there are various flavors of pluripotent stem cells that can be isolated from early embryos.
This was first worked out in the mouse system, and now increasingly, we understand how to capture these different states from human embryos and through human reprogramming as well. Work over the past 10,15 years has clarified that the canonical human embryonic stem cells that refers to derived by Jamie Thomson in the late 1990s, even though they’re derived from the pre-implantation blastocyst, actually acquire a developmental state that is more akin to the late post-implantation embryo.
They seem to progress from their blastocyst identity towards this more advanced developmental state. We refer to this as the prime state pluripotency. This is a terminology introduced by Austin Smith and Jenny Nicols.
These cells, if you look at their transcriptome and you match that transcriptome to various stages of monkey development or human embryos cultured in vitro, it’s quite evident that they resemble the post-implantation epiblast around e13, e14 of developments. That’s right around when the cells begin to gastrulate, when they start to specialize into different lineages.
So based on transcriptional data, they’ve clearly moved on from their blastocyst origin. If you look at epigenetics, these prime pluripotent stem cells have high levels of DNA methylation. If you quantify global levels of CpG methylation, in the prime cells, we see around 70-80 percent DNA methylation, which is far higher than what is seen in the pre-implantation blastocyst, where you have around 20-30 percent of CG dinucleotides that carry a methyl mark on the cytosine.
Another important epigenetic aspect of prime pluripotency is the fact that most female prime stem cell lines, both ES and iPS show signs of X chromosome inactivation, where one of the two X chromosomes is already inactivated.
This is again different from the situation in the blastocyst where we know that in human embryos, you actually have two actively transcribed X chromosomes before the embryo implants. Then finally, in terms of a more functional property that we should all be concerned about is we use these cells. It’s been shown both in the human and the mouse contexts where we have a similar cell type called epiblast stem cells, that these prime pluripotent cells often show lineage bias if you try to differentiate them. So you can have lines that are terrific and making ectoderm, but cannot at the same time efficiently differentiate towards, let’s say mesoderm or vice versa.
That type of variability in differentiation outcome complicates a lot of applications related to these cells.
We and others have been interested in trying to push these cells back toward some more primitive stem cell state that we refer to as the naive state of pluripotency. As Amanda mentioned, I developed this 5i/L/A cocktail for accomplishing this prime to naive conversion. There were many other methods out there. Another one that gives a very similar transcriptional state is the t2i/L/Go method developed by Austin Smith in Cambridge.
They use an HDAC inhibitor to get this process started. What’s important is that with either of these methods, you end up with a naive stem cell state that clearly has a pre-implantation transcriptional identity. If you look at, for instance, transcription factors expressed in these naive cells, they correspond quite nicely to the same transcription factors that you see expressed in the human epiblast before implantation. These cells have globally reduced levels of DNA methylation, again, consistent with a blastocyst identity and they show X chromosome reactivation in female cells.
Now, more recently, we and a number of other labs have shown that these cells also have an enhanced potential for extra-embryonic differentiation, which opens the door to modeling extra embryonic development diseases associated with the placenta.
So I’m going to talk more about that today. Just briefly, what are the potential applications of these naive stem cells? Why are we and others excited about it? The first is this concept of looking at their lineage potential in relation to prime cells. Perhaps these naive cells, by virtue of being in this less committed ground state, have less bias in differentiation outcomes.
This has not really been tested yet with respect to germ layer differentiation, hasn’t been compared side-by-side, but we do find they have this enhanced extra embryonic potential. They also have potential importance for modeling X-linked disorders. I mentioned that the prime cells often show X-inactivation. They’re also known to undergo erosion of that inactive X chromosome with increased passaging where the inactive X becomes partially re-expressed and then when you try to differentiate those cells, they fail to properly inactivate their X chromosomes. This can complicate modeling of X-linked diseases.
Many autism spectrum disorders carry mutations on the X chromosome. We’ve shown in collaboration with Catherine Clarke at UCLA that these cells converted under our naive media actually undergo X chromosome reactivation and when you differentiate them, they can also initiate de novo X-inactivation from the naive state. Another broad point to make is that the naive cells may offer windows into aspects of early development that are difficult to model from the prime state. A good example of that are so-called transposons, jumping genes that are uniquely expressed in the naive cells as well as in early human embryos, families like SVA and HERV-K, but are lowly expressed in the prime states. You can actually figure out what those transposons are regulating, how they are in turn regulated using the naive stem cells.
Finally, there’s this exciting concept of interspecies chimerism of trying to get the human pluripotent cells to incorporate into an animal host embryo in order to generate human stem cell-derived tissues within the developing animal.
This has been mainly attempted using mice. It doesn’t work very well. Now, groups are trying this with large animal host embryos. I think that’s probably a more sensible approach, but the efficiency of this is still very low.
I think it’s going to require optimization both of the naive stem cell media and potentially engineering suitable host niches within these animal host embryos. Definitely though an exciting space to watch. My lab is focused on four interrelated aspects of these different stem cell states. The one that’s really taken off in the past few years is modeling placental development using naive cells. I’m going to tell you about that work in particular today.
We are also interested in mapping the transcription factor networks that control these different stem cell states. We’ve used biochemical approaches with a collaborator at Columbia, Jianlong Wang, to identify proteins associated with Oct4, which is a master transcription factor of pluripotency. We’ve identified different epigenetic subunits of the birth complex that are associated with Oct4 in either naive or primed conditions.
We also have this strong interests in further improving the signaling environment of these naive stem cells. Even though these are very useful tool for modeling the early embryo, they’re known to undergo imprint eraser.
You have parents specific imprinting marks that are present in the prime state, but become erased due to the excessive demethylation in the naive state. There’s also issues with long-term genetic stability under naive conditions. So we’re interested in refining these culture conditions. Finally, we, like many others, are exploring the ability of these naive cells to self-organize into 3D models of the human embryo in order to study human development and also potentially implantation failure. But today, I’ll strictly focus on this first arm of our research program.
I want to tell you how we became interested in looking at placental differentiation.
We performed this ataxic analysis in collaboration with DJ Trona’s lab in Switzerland, comparing open chromatin in even primed stem cells. We found that the naive cells are enriched and open chromatin sites associated with the morula and blastocyst stages of human development as you’d expect. But surprisingly, they also showed significant open chromatin found in the placenta. That was a bit of a curiosity.
We did not expect that, but it reminded me of a prior finding where based on RNA seek data, we found that these naive cells show increased expression of various transcription factors associated with the trophoblast lineage, which is the main epithelial cell type found in the placenta. This raised the question, do these naive cells perhaps have an enhanced potential for extra embryonic differentiation specifically towards trophoblast fage? You can imagine that because we’ve pushed these cells back in terms of their developmental state more towards the point when the epiblast and the trophy ectoderm initially segregated in the human embryo, they may have increased plasticity to turn into trophoblast cells. Just to remind you, a trophoblast is an extra embryonic lineage derived from this outer trophy ectoderm layer of the blastocyst, which gives rise to the placenta.
Critical functions of trophoblast cells or to mediate implantation of the embryo into the uterus and to promote maternal fetal exchange of various gasses, nutrients, and waste products.
If you look at the hierarchy of trophoblast cells in vivo from that preimplantation trophy ectoderm, you generate so-called Villa cytotrophoblast. These are the in vivo trophoblast stem cells and these in turn can differentiate either into extra villas trophoblast. These are the invasive cells or the multinucleated and hormone secreting sincere show trophoblast. The first question we looked at is whether these naive cells are able to respond to conditions for human trophoblast stem cell isolation.
This was really inspired by a paper published in 2018 by the ARIMA lab in Japan.
I consider this to be a landmark paper for the trophoblast field. To contextualize that back in the 1990s, Janet Rossen in Canada succeeded in deriving mouse trophoblast stem cells and they’ve become a really widely used tool to look at mouse trophoblast development. Since then, it’s been very difficult to isolate human trophoblast stem cells. But this paper was really the first to succeed, so they worked out culture conditions for deriving human trophoblast stem cells that are self renewing and by potent that can differentiate either into a DVT or STB fate, both from blastocysts and from first trimester placental tissues so very exciting paper. The one limitation however, is really the source of these TSEs at the time.
Neither blastocyst nor first trimester placental tissues are widely accessible for many researchers. In fact, where we are in Missouri, the use of first trimester placental tissues in research is very difficult, very controversial, and I fear this will not improve in the current political climate. We ask this question, can we derive similar trophoblast stem cells from human pluripotent sources, and we tested both are primed and naive lines.
We found that the naive ES cells and iPS cells when treated with those same media, those TSC media shown here on the left, readily adopted a morphology closely resembling laying these primary trophoblast stem cells. The prime cells, on the other hand, shown on the left here, both the ES cells and the IPS cells did not acquire the same morphology.
They actually acquired a very differentiated, normal looking morphology and barely proliferate it. There’s a clear difference in morphology depending on whether you start from a naive or a primed stem cell. But what about molecular markers? Here we perform flow cytometry analysis for two trophoblast associated cell surface markers, EGFR, an integran alpha six. And we find that the naive cells treated with these TSE media show strong induction of these surface markers while the prime cells treated with those same culture conditions actually show downregulation of these surface markers.
We’ve also looked at a number of transcripts by qPCR, again, noticing high activation of various trophoblast associated transcripts in the naive cells switch to TSC media, but not the prime cells. You’ll notice on this graph that to some extent, these genes are already deeper pressed in the naive state compared to the prime state that’s consistent with what I mentioned earlier, but this is further increased upon addition of these trophoblast media. We then ask the question, can these naive TS cells actually differentiate into specialized trophoblast fate? We applied the two methods for lineage directed trophoblast specialization developed in the o chi protocol, either using direct Yulen and a TGF beta inhibitor to push them towards extra villas trophoblast fate or the use of force Glenn, which is the cyclic AMP agonists degenerate since the show trophoblast, I’ll show you the EVT data first. We’re able to generate this typical mesenchymal morphology of EVTs using the naive derived TS cells.
They also show activation of HLA-G key surface antigen associated with EVT fate, as well as up-regulation of MMP2. And we then performed a transwell matrix gel invasion assay to see if these EVs can actually invade, which is their key functional attribute in vivo, and indeed the EV T-cells show strong invasive potential through this matrigel trans well or as the parental cells could not invade as well. Similarly, we performed directed sincere show trophoblast differentiation and were able to obtain these multi-nucleated cells that show induction of CGB, a key component of hCG, which is a major placental pregnancy hormone, as well as up-regulation of syndication and down-regulation of TI-84, which is more associated with the trophoblast stem cells. And we also confirmed that many of these markers are expressed at the protein level in these STBs. We then did RNA seek analysis, this is a bulk RNA seek, and I want to take my time to explain what’s on the PCA here, the principal component analysis.
On the right are the starting prime stem cell lines that we used. When we treat them with these phi via media, we generate naive cell shown at the bottom.
When you treat those with trophoblast stem cell media, we obtain these naive derive TS cell shown in red, which have a very similar gene expression profile as the primary trophoblast stem cells isolated by the group in Japan shown in green. We actually obtain those same cell lines from the Japanese lab, grew them side-by-side with our own cells in the lab, and then perform the RNA seek to really make sure that we’re comparing apples and apples. What happens to the primed cells that were treated directly with these hTSC media, as you can see on the right, they actually acquired a very different states, nowhere near the primary hTSCs.
If you look at the volcano plot, you can see that the naive cells treated with hTSC media really strongly upregulated many trophoblast associated transcription factors. While the primed cells treated with those same conditions actually activated the number of ectodermal genes, genes like DCX, PAX3, SOX11 which really have nothing to do with trophoblast fate. Just again, to make that point, the response to these culture mediums seems very much determined by the starting state of your stem cells.
Putting that together, we know from prior work in the field, much of which was actually developed here by Mana’s lab that primed cells treated with BMP4 can acquire cytotrophoblast fate. But we find that the naive cells treated with those same media, BMP4 do not actually turn into trophoblast.
It seems that the naive cells are not responsive to BMP4. There’s increasing reports coming out that naive cells may not respond to the same growth factors, the same differentiation cues as primed cells. That’s not in and of itself surprising. We did find that the naive cells when treated directly with these trophoblast stem cell media can turn into trophoblast stem cells and we don’t see the same response starting from the prime state. Similar observations were reported by a number of other labs over the past two years.
Now, last year there were two other papers that came out that modified this protocol. They showed that if you treat the naive cells with a MEK inhibitor and a TGF Beta inhibitor, you can actually transition the cells through a pre-implantation trophectoderm like state, which they refer to as naive trophectoderm before pushing them further towards trophoblast stem cell fate.
I think that is a nice modification of the protocol because one thing I haven’t mentioned yet, the trophoblast stem cells that we derive clearly have a post implantation identity. When you look at their gene expression, they do not match the gene expression program of the trophectoderm in the blastocyst. They’re far more similar to cytotrophoblast and the post implantation environment.
With this modification, we can now also perhaps not stably capture, but at least study this pre-implantation trophectoderm step. To make matters more complicated, there are also reports coming out that it is possible to generate trophoblast stem cells from primed cells if you first treat them with BMP4 or variations thereof and capture this cytotrophoblast like state and then apply the TSC media. There is a big debate and I look forward to discussing that with many people here today. What really is the nature of the trophoblast stem cells that you generate starting either from primed or naive stem cells.
I think that’s an important question for the field.
I’ll point out that from our perspective with our interest in studying really how early trophectoderm is specified. The naive state does offer a unique perspective there because you’re able to model how this lineage is established from a pre-implantation state, which is of course what happens in vivo in the embryo.
Recapping this why are these studies important. We’ve proposed that the ability to directly convert naive cells into extra embryonic stem cells may present a functional test for naive human pluripotency. I deliberately talk about extra embryonic stem cells more broadly here.
This is not restricted to trophoblast. There’s work from Josh Brickman’s Lab in Denmark showing that you can also turn these naive cells into primitive endoderm, which is the other major extra embryonic lineage found in the blastocyst. The conversion of naive cells into trophoblast stem cells gives us a tool with which to study early mechanisms of how trophoblast is specified. It also offers potentially renewable, more accessible source of patient-specific human trophoblast stem cells to study diseases associated with the placenta. Now, this field moves very rapidly.
There are now also methods for generating trophoblast stem cells from somatic cells through directory programming. Our methodology is not the only way of generating patient-specific TSC, so I want to make that clear as well.
Finally, there’s a lot of excitement in the field about building 3D models of the human embryo. Just in the past year, there were three actually more groups reporting the generation of so-called blastoids that contain all the lineages of the blastocyst in a proper 3D architecture. Again, they exploited the extraembryonic potential of the naive cells to do so, the fact that they can turn both into epiblast, trophectoderm, and primitive endoderm.
I want to tell you about a different 3D model next, not of the entire embryo, but specifically of the placenta. These are so-called trophoblast organoids. These were derived by two groups in Europe few years ago, starting from first-trimester placental tissues. They isolated cytotrophoblast progenitors, cultured them in matrigel, and then under appropriate organoid, supporting media were able to create the self-renewing trophoblast organoids. Really offering a 3D model of the early placenta that encompasses a variety of different trophoblast cell types.
Now, there’s one important caveat about these organoids, which is the fact that their architecture does not actually match the precise architecture of the placental villas. In vivo, the placental villus has this outer layer of syncytiotrophoblast and a more interior CDH1 positive progenitor compartment where the cytotrophoblast are.
This is shown both here in the micrograph and in the cartoon at the bottom. Whereas the trophoblast organoids actually have a more interior syncytio compartment and an outer layer of these trophoblast progenitor cells to fill a cytotrophoblast. With that architectural caveat in mind, still these organoids offer a useful 3D tool to study placental development.
But again, they were derived from first-trimester placental tissues which are not readily available in many places.
So, we ask the question, this is work from my postdoc, Rowan, which was just recently published. Is it possible to derive similar trophoblast organoids starting from human pluripotent cells using the methodology that we develop where we convert the naive cells into trophoblast stem cells. We actually compared the ability of trophoblast stem cells from various sources. Now, if cells blastocyst and first-trimester placental tissues to self-organize into trophoblast organoids using matrigel droplets and the organoid media that were previously developed by Margarita Turco’s lab.
We find that in all cases, the TS cells that we test have the ability to self-organize into 3D structures that can be maintained for up to 10 passages, at least. Just a note about the culture media. They’re quite similar with some modifications. When we switch from the 2D hTSC culture to the 3D organoid culture, we actually remove this HDAC inhibitor. We add the combinant FGF as well as HGF.
There are some modifications, but the cells actually adapt quite easily to the 3D culture environment. We did immunostaining to look at the architecture of the stem-cell-derived trophoblast organoids as we call them. Again, they have this inside out architecture with a more exterior layer of cytotrophoblast progenitors marked by ECAD and more interior syncytio compartment, we do find that a subset of these trophoblast organoids are almost entirely syncytio as show here by CGA.
When we do an over-the-counter pregnancy test to detect hCG secretion, they express abundant levels of hCG. That’s a quick test to see if you’re actually working with trophoblast cells.
But importantly, we see that same inside-out architecture that was reported with the primary trophoblast organoids. To get a better understanding of the cellular composition of these organoids, we performed single-cell transcriptome analysis using the 10X genomics platform. For this we use both a primary TSC derived organoid line and a naive derived organoid line that I refer to as H9. In both cases, we generate quite similar cellular composition of five major subpopulations to cytotrophoblast shown on the right, a smaller primitive extra villas cluster in the middle, and two syncytiotrophoblast clusters on the left. Very reassuring to me when I first saw this data is the fact that the overall cellular composition is so similar regardless of whether you start from primary trophoblast stem cells on the left or naive hTSCs on the right.
I say that because these naive hTSCs have a very different history.
They came from primed cells that were converted to the naive state from there to trophoblast fate, and they do have some epigenetic differences with primary hTSCs. Yet at the overall level of this, you map analysis, they look very similar to each other. We propose that this organoid culture is perhaps a stable attractor state. If you look at specific markers, the CTB clusters show appropriate progenitor marker expression on the right.
We see the presence of some EVT markers in that more primitive EVT cluster as well as syncytiotrophoblast markers towards the left. There is some heterogeneity, so shown here, for instance, is CGB3. This is a more mature STB gene. This is specifically expressed in that upper STB subpopulation.
There is some variability between these clusters.
Now we then map these data to the human embryo. I think that’s a critical experiment. To what extent are we actually modeling trophoblast identities that are seen in human embryo. We used a dataset from a group in China, who cultured human embryos in vitro up to day 14, also in the presence of matrigel. At the top left here you can see in color code the subpopulations in our organoids and superimposed on the UMAP in gray are the trophoblast identities in the human embryo.
If you then look to here towards the bottom left, you can see that the embryonic cytotrophoblast really cluster nicely with the two CTP clusters that we have on the right.
As do the EVT cells from the human embryo, as do the STB cells from the human embryo. There’s a nice one-to-one correspondence of trophoblast identities in vivo and in these organoids models. On the right here we’ve also added temporal data showing that the STBs that were isolated from a later stage of human development day 14 actually cluster more with that more mature STB clustered towards the top compared to earlier STB. There is developmental progression that this model captures as well.
Now, one issue that we noticed early on is that the extra villas cluster is quite small and perhaps not fully mature, so we also switched the media to try to capture more specialized EVT organoids. We did this by reducing the level of wind signaling. This was previously worked out for primary trophoblast organoids, and we applied the same protocol to the stem-cell-derived organoids.
We find that these EVT organoids indeed show more migratory morphology. You see these invasive projections that emerge.
They upregulate MMP2 as well as a number of other EVT marker shown at the bottom right. We then wanted to do a more functional test. You can look at all these molecular markers, but the real function of EVT cells in vivo is to invade into the endometrium. We performed a co-culture assay with human endometrial cells in the presence of Matrigel. We add it either the standards organoids or the low wind EVT organoids to this co-culture environment where we use two types of human endometrial cells together with Matrigel.
The degree of invasive projection, the length of these projections was measured in this graph at the bottom and you can see that the EVT organoids only in the presence of endometrial cells show really extensive invasive projections. The moment you remove the endometrial cells, the degree of invasion significantly drops.
We also don’t see any invasion really using the regular trophoblast organoids in the high wind media. This suggests that there is some interplay between the endometrial cells and the EVTs. That the endometrial cells maybe encouraging the EVTs to invade.
I think this provides a really interesting assay, to begin to figure out what are the factors on the endometrial side that are encouraging these EVT cells to actually invade in this co-culture assay? I think this sets up the question a lot of labs are now interested in, how can we model interactions between these trophoblast models and human endometrial tissues or stem cell-derived endometrial models.
Now, we submitted this paper. A reviewer asked a very interesting question, which is, can we use this system to also look at this question of X chromosome inactivation? Which is not well understood in the human placenta.
In mice, it’s always the paternal X, that’s inactivated both in the placenta and the yolk sac. There’s evidence that in the human placenta X inactivation is more patchy, where you’ve got entire areas of the placenta that show maternal versus paternal X chromosome silencing. Can we use this system to figure out at what point is the X chromosome shutdown and is it random or perhaps imprinted? For this purpose, we used a biallelic reporter line that we have in the lab. In which both alleles of an X-linked gene MECP2 are labeled with different fluorescent reporters.
Because of X-inactivation in the prime state, we find that only one of these alleles is actively transcribed. Here it’s the green one. When we then apply Naive media, we can induce reactivation of that second allele. You see cells that are both red and green.
Very nice here on the facts.
What happens when we turn these naive cells into trophoblast cells? Interestingly, in the hTSC media already, we find that that double positive population gets resolved into two discrete single green and single red’s positive cells. Suggesting that there is X inactivation. At least at this locus, bear that in mind. We’re looking at one locus here as a proxy for X inactivation.
But it happens already during that transition from naive cells into trophoblast stem cells. It’s not quite entirely random, there’s some bias here. If you look at that Q3 proportion for the tomato allele. Then what happens when you make organoids? Interestingly, we find that organoids seem to maintain at a clonal level, whatever XCI pattern was established in the hTSC.
We get organoids that are either entirely red or entirely green. There’s a couple that are biallelic, suggesting that there is a clonal expansion of the inactivation status and the trophoblast stem cells and also pointing out that XCI may occur quite early in the human trophoblast lineage.
There was a paper just recently in Science from the [inaudible] lab in Japan who looked at X inactivation in a non-human primate model. They actually reported that the trophoblast is perhaps the first lineage to shut down its X chromosome compared to both other embryonic and extra embryonic lineages. We propose that also in humans this may be the case.
Now, what is the status of the X and the human placenta? I already mentioned, there’s been some debate about whether it is homogeneous, patchy or mosaic. Most of the evidence at the moment suggests that it’s quite patchy, where you have entire parts of the placenta that show either maternal or paternal silencing with some bias towards the paternal X.
The model is that, this reflects the generation of entire villus trees from trophoblast progenitors at a fairly early stage of gestation. You have an early choice as to which X chromosome to inactivate, followed by clonal expansion, which is similar to what we see in our in vitro model.
We then ask the more topical question, which is, can we use this model to investigate susceptibility to emerging pathogens? This is a topic that’s widely debated. There’s lots of evidence that COVID-19 can lead to pregnancy complications. There’s much less evidence that COVID-19 infected women can pass on that virus to the fetus, so-called vertical transmission.
We looked at our single-cell data.
We find that the entry factors, ACE 2 and TMPRSS2 are quite restricted really just to the STB population within our single-cell data especially that more mature STB population towards the top. We then used a chimeric virus that was generated by Sean Wayland at WashU, which contains the VSP backbone in eGFP reporter and either the VSV glycoprotein or the SARS-CoV-2 spike proteins. This allows you to study SARS-CoV-2 infection under reduced biosafety containment levels. We find that the VSV G version of this readily infects the organoids, everything turns green while VSV S barely infected the organoids. We saw very low level of infection with the SARS-CoV-2 spike protein.
Of course, you can say this is a very artificial scenario with this chimeric pseudo virus. What about the live virus? We find that the live virus likewise did not readily infect the trophoblast organoids. This is infected at the top, uninfected at the bottom. In contrast, ZIKA readily infected these organoids and this is consistent with prior data on ZIKA.
Of course, being readily transmitted to the fetus via the placenta and also we find that ZIKA entry factors are more widely expressed in the organoids data. We see a different response to these two emerging pathogens, which we believe is correlated to the level of expression of their entry factors. Now, that said, you might argue this is not the right model as reviewers did to study trophoblast infection because you have that inside-out architecture. Remember we have that outer layer of cytotrophoblast, which are ACE 2 negative and then a more interior ACE 2 positive STB compartment.
Is it possible that that inside-out architecture is hindering, preventing the virus from accessing the STBs?
To address that, we actually generated STBs using the lineage directed protocol from the [inaudible] group starting from TS cells without the organoids context and then re-infected them with the SARS-CoV-2 virus, the chimeric virus and shown in G here versus H, we find that VSV S again with the SARS-CoV-2 spike protein present does infect a subset of STB.
We do see some that are getting infected, but the level of infection is far lower than what we see with VSV gene shown on the right. Our interpretation is that there seems to be heterogeneity within the STB. Some may be able to get infected both into 3D context, perhaps more some of the 2D context. But overall we see far less infection than with other viruses like VSV G or ZIKA virus.
To summarize this part, I’ve shown you that hTSCs of various origins can self-organize into these 3D trophoblast organoids, they contain a number of different trophoblast identities that closely correlate with trophoblast cell types in the embryo. These organoids display clonal X inactivation patterns that are consistent with the patchiness of X inactivation in the human placenta. We can use this model system to investigate placental vulnerability to a variety of emerging pathogens. I want to briefly tell you in the remainder of my talk about another story that we’ve been working on.
Having established these 2D and 3D models of the trophoblast, we were wondering, can we use this system now to investigate the genetic regulation of human placental development?
A very popular way of doing this is CRISPR screening. You can do these genome-wide CRISPR screens to identify both essential and growth restricting genes in various cell lines. We use the Brunello library, which has genome wide coverage. Transduced our trophoblast stem cells apply puro and then passage these cells up to day 18 and collected DNA at various time points in this experiment for deep sequencing. This allows you to measure the abundance of the guide RNAs over time.
The principle is that essential genes, are the genes for which the guide RNAs actually dropout over time. If you look at essential versus non-essential, you can see that the essential genes are the ones where the guide RNAs drop out towards the end of this CRISPR screen.
We identify around 2000 essential genes in human trophoblast stem cells. Encouragingly, these essential genes tend to be quite highly expressed in a variety of different hTSC lines. We’ve also compared this with so called core essential genes, which were identified by other groups and a variety of different cell lines, including many cancer cell lines.
Those genes are also quite highly expressed in the trophoblast stem cells compared to the non-essential genes.
Here’s an illustration of what these data look like. You can see that the guide RNA for the essential genes drops out over time, shown in red. Here are two examples, TFACP2 and TEAD4. These are known trophoblast regulators, so they’re shown to be essential and on the left and right of each of these graphs you can see the neighboring genes.
The neighboring genes do not show that same drop in guide RNA abundance. These essential genes tend to be enriched in processes such as in utero development, as well as a number of signaling pathways that are directly controlled by the trophoblast stem cell growth factors and inhibitor. That makes a lot of sense. You would expect these pathways to be essential given that they are directly regulated by the factors that we give to support the trophoblast stem cells in the culture media. Now we then wanted to identify not just essential genes and trophoblast stem cells, but once that are specific to trophoblast stem cells.
In the Venn diagram, we compare the essential genes in the TSC cells to core EGs, as well as EGs in prime stem cells that were previously identified by other labs.
We identified around 872 essential genes that are unique to trophoblast stem cells that are not unique in these other datasets. We then asked another question which is how many of these are actually enriched in the cytotrophoblast in vivo. Ideally, we would identify factors that are not simply essential for the cells in vitro, but also play a role in trophoblast development in vivo. Doing this comparison, we were able to identify a number of genes that are not only essential specifically in trophoblast stem cells, but also upregulated in cytotrophoblast in the human embryo compared to epiblast or primitive endoderm.
Some examples are shown here. We’ve got a number of known trophoblast regulators, genes like GATA2, SKP2, this is known to regulate the cell cycle as well as ARID32A.
These are more highly expressed in cytotrophoblast than other lineages in the human embryo. We also pulled out some previously unreported trophoblast regulators such as TEAD1, TCAF1 and ARID5B. TEAD1 really caught our eye.
This is a transcription factor which is dispensable for mouse defected from specification. You can knock it out in mouse embryo and the defected arm still develops just fine. It’s paralog though TEAD4 is known to be a key determinant of trophectoderm specification in the mouse. Yet TEAD1 had a higher rank in our CRISPR screen compared to TEAD4 if you look at its essentiality based on guide RNA abundance. Now we looked at expression of TEAD1, we find that it is upregulated in the trophoblast stem cells compared to either print or naive pluripotent cells.
It’s also highly expressed in the invasive extra villus cells, much lower in the STBs, and we see a nice nuclear localization for TEAD1 in our TS cells. We ask the question, what is the role of this transcription factor in human trophoblast biology? To do that we make TEAD1 knockout cells. We knocked out TEAD1 using CRISPR first and prime cells. Then made naive and then made TS cells.
The reason for doing the experiment this way is that we can actually study the requirement for TEAD1 during the transition from prime to now naive as well as from naive to trophoblast stem cells. In the end, we were able to make TEAD1 knockout naive cells quite easily. There was no real phenotype going from prime to naive. When we make TS cells, however, there was a delay in differentiation ability. You can see that the TEAD1 knockouts initially struggled to reach the trophoblast state, but with continuous culture, we were able to make TEAD1 knockout cells eventually.
They were able to catch up with the wild type cells. We then perform transcriptome profiling comparing the TEAD1 knockouts with their wild type counterparts. We found that these TEAD1 knockout trophoblast stem cells showed substantial upregulation of a number of syncytiotrophoblast genes. Genes like CGB2, CGB7 and PSG11, while they actually show down regulation of markers associated with EVT fates such as HLAG, FN1 and integrin Alpha 5. This is also reflected in the goal analysis on the right, you can see that STB associated terms like hormone response were upregulated in the TEAD1 knockout trophoblast stem cells, while many EVT associated terms were going in the other direction.
That prompted the question based on this dysregulated gene expression, do these TEAD1 knockout trophoblast cells have a defect when you try to make specialized trophoblast cells. We push them either towards extra villus trophoblast fate or towards sensitial trophoblast fake both wild type and knockout in parallel, and we find that the TEAD1 knockout TSCs did a really poor job of making extra villus trophoblast.
They don’t acquire that nice migratory morphology that you expect from EVTs. They also showed reduced expression of HLAG, as well as reduced chromatin accessibility at this EVT locus by ATAC-Seq. Whereas they actually showed increased expression of CGB, which is again more associated with STB fate, which you really don’t expect from EVT cells.
On the other hand, when you push them towards syncytiotrophoblast they were actually more efficient degenerating syncytiotrophoblast and the wild type cells. They also showed increased expression of some genes associated with STB fate and reduced expression of keratin 17, which is more associated with EVT and trophoblast progenitor fate. This suggests that if TEAD1 may be a critical regulator for this balanced lineage potential of the trophoblast stem cells, it seems that in the absence of TEAD1 the cells cannot complete EVT differentiation faithfully and yet they adopt STB fate more readily.
Now, just as a final point, we not only identify essential genes in this analysis, but these CRISPR screens also allow you to do the inverse analysis and look at genes whose guide RNA abundance actually increases over time is shown. Here, these are called growth restricting genes, so they likely play an opposite role in the cell type of interest, namely to restrict the growth of in this case trophoblast stem cells.
We identified 619 growth restricting genes in this assay. One of these is GCM1. This is a factor associated with syncytiotrophoblast fates suggesting that that factor actually rains in the growth of the trophoblast stem cells.
We validated a number of these hits including tattoo. Interestingly this seems to be growth restricting in the human contexts it’s been reported to do the opposite and the mouse context and PTPN14 and other growth restricting gene is a phosphatase that is known to interfere with yap and yap is a key trophoblast determinants.
That also makes intuitive sense. When you knocked down these genes, the proliferation of the TS cells actually increases. Now these GRGs seem to fall into two broad categories. There are ones that are enriched in cytotrophoblast in vivo, like the example shown here. These factors are probably preventing the excessive proliferation of trophoblast cells.
There again more abundant in trophoblast in vivo than in other lineages. We think these are great candidates for thinking about what’s responsible for chorionic carcinoma, which is a tumor of trophoblastic origin. Potentially genes such as these are dysregulated in that cancer and then cause excessive trophoblast proliferation. There’s also the opposite category. Growth restricting genes that are actually more highly expressed in the neighboring cell types like epiblast and primitive endoderm compared to cytotrophoblast.
Some examples are shown here.
These include genes like Nodal and TBX3, which are actually pluripotency associated. We propose that these factors may be preventing trophoblast programs from becoming activated in these other cell types. Now, finally, we looked at conservation with mouse data. We wanted to know to what extent are the essential and growth restricting genes in the human context shared with mouse trophoblast.
There was very important paper from Miriam Hamburgers Lab a few years ago who did a large scale mouse phenotyping analysis and identify genes that cause placental defects. It’s probably not an exhaustive resource, but still it gives you a pretty good idea of what genes are necessary for placental development in the mouse. When you intersect these data with our trophoblast stem essential and growth restricting genes, there’s a little degree of overlap. It’s not overwhelming, but there are 20 candidates that are actually conserved between these datasets shown here. These 20 genes are more highly expressed in human villa cytotrophoblast compared to other cell types found at the human eternal fetal interface.
Again, that’s somewhat reassuring. You would expect these conserved regulators to be more highly expressed in the villa cytotrophoblast compartment.
What was interesting though is when you look at some prominent examples within this list of 20, there are several mitochondrial genes included here. Again, these are genes that are essential for human trophoblast stem cell growth have also been implicated in placental dysfunction in mouse. Again, it’s not a very long list, but this does suggest that metabolism, oxidative phosphorylation is a really conserved regulator both of mouse and human trophoblast biology and you can see that these candidates tend to be indeed more specifically expressed in the trophoblast cells at the human fetal maternal interface.
I think that opens up an avenue for further work into how does mitochondrial regulation effect trophoblast biology. To summarize this part, I showed you our data reporting genome-wide CRISPR screen where we identified both essential and growth restricting regulators. We did a lot of follow up on one of these candidates, the essential transcription factor TEAD1 and proposed that it plays a human specific role, particularly in EVT differentiation from trophoblast stem cells, although it also has some phenotype in both specification and maintenance. We also report the number of conserved trophoblast regulators which are enriched in mitochondrial genes.
Overall we think this is really a starting point for now following up on these candidates both in the 2D model, potentially the trophoblast organoids model.
One can also thinking about looking at the role of these regulators in the trophectoderm within the blast steroids that are now being created by a number of groups to begin to understand what are the factors that control trophoblast specification and potentially placental dysfunction. With that said, I want to give a big shout out to the people who actually did the work, including my student Chen Dong.
He was actually the one who pioneered our naive to trophoblast stem cell efforts. He also let the CRISPR screen effort very productive, graduate student who just defended his thesis, the first one from my lab. The organoids work was entirely performed by Rohan Carfus, who’s a post-doc in the lab with help from all the others as well as our generous collaborators at WashU.
I also want to give a big shout out to our generous funding sources both at NIH as well as a number of private foundation grants. With that, I’ll be glad to take your questions. [APPLAUSE] Thanks. Great talk. I have two questions actually.
Can you comment on the ability of the naive versus the primed to induce trophoblast stem cells? There seems to be a lack of a repression complex to me in the naive. I’ve never seen anyone look at that, and it’s probably a complex of Polycomb related groups and a number of things.
Can you comment on that? Yeah, so the initial thought we had, it’s a great point, what is really the molecular basis for this extra embryonic potential?
My initial thought was it’s all about DNA methylation. Something really striking about both naive human stem cells and trophoblast stem cells is that they’re both very hypomethylated. They have only 20-30 percent CpG methylation. That was my initial thinking. I know, through the grapevine, that there might be other molecular causes as well.
To give a clue there, is when we looked in our organoid data, we find that a particular microRNA cluster that’s implicated in trophoblast biology in chromosome 19 is significantly induced in the organoids and the TS cells, but also in the naive cells compared to the prime cells. One speculation is that potentially a key non-coding element like these placenta specific microRNAs could play a functional role in allowing the naive cells to convert, but not the prime cells.
But again, we can talk in black and white terms about this. There is lots of evidence including work being done here, that you can push these prime cells towards a trophoblast fate with some modifications of the method where you first pre-treat them with BMP4. I think there’s a lot of work to be done in resolving what is the molecular cause.
But I think potentially the combination of hypomethylation and activation of some of the specific microRNAs may give part of the answer. The second question is on your organoids, your 3D culture, the X-inactivation. Have you looked at exist exact, is that expressed, and have you found any escaping genes? Yeah, so there are escapees, we see that in our single-cell data. We had really two lines of evidence here.
We have to reporter line and we looked in the single-cell data, looked at mono versus biallelic expression of X-linked genes.
It’s a little tricky to do that with single-cell data, but we were able to do it. We haven’t done the detailed analysis exist exact by RNA fish, and that really needs to be done in order to properly investigate X-inactivation. We weren’t even doing this work, then reviewer came in and said, can you look at this aspect? But I think it’s a great question.
How is H3K27 trimethylation regulated, for instance, during the process of trophoblast stem cell induction. I can’t comment on it yet, but I think it’s a great model system to study X-inactivation from human naive cells. Can I ask a follow-up question to that actually? I think Jose Paulo and his group had looked at the reprogramming fate map going from a somatic cell to trophoblast stem cells, and their conclusion was that it skips over, that the reprogramming is done with the same factors as reprogramming to pluripotency.
It doesn’t seem to have to go all the way to naive in order to make it to trophectoderm.
Can you comment on that and any insights into that? Yeah. This is a great point. Do you have to go through pluripotency to reach trophectoderm fate? Is it possible to directly turn a somatic cell into trophoblast cell without actually proceeding through pluripotency?
I suppose it’s possible. I do believe there’s data within the Polo paper suggesting that there is some activation of an intermediate like totipotent state. One possibility is that the cells go through an eight cell-like stage. You may have seen these papers come out recently where labs are trying to create even earlier cells that are not quite naive, but perhaps an early totipotent precursor. I can’t exclude that possibility.
I guess developmentally it might make sense if trophectoderm is specified very early on after that eight-cell point. So, yeah, I can’t rule it out. Then a question from the virtual audience, from Dr.
Alan Goran. He says, thanks for the interesting talk.
How do you decide which factors inhibitors to add when optimizing the media conditions? I suppose he’s talking about naive. Yeah. Yeah, this is a big effort in the lab. So we did a chemical screen in collaboration with Novartis where we identify small molecules that can maintain naive stem cells using specific reporter, linked to the distal enhancer of Oct4, this is what we previously used to identify the [inaudible] cocktail.
We were especially interested in factors that can replace MEK inhibitors, because MEK inhibitors have been identified as a major cause of instability both in mouse and human stem cells. Somewhat disappointingly, most of our hits hit that same pathway. They’re FGF pathway hits. Interestingly though, we still pull out a number of alternative inhibitors of the pathway both upstream and downstream, that give us different phenotypes in culture including more stability, potentially faster reprogramming kinetics as well. There’s not a lot of room to work with.
I think you have to hit that pathway at the level of the receptor Raf-MEK-ERK. I’m hopeful that by reaching an optimal combination of these enzyme kinase inhibitors, we can further improve the naive phenotype. But this is very much an active area, and many laps are active in this space, but hopefully we’ll get there.
[LAUGHTER] You made a comment that the naive PAC derived TSCs are post-implantation like whereas naive PAC are pre-implantation like. I was just wondering if you had any thoughts about why that is, if you expected that result?
Yeah, I think that the answer is that the media developed by Hokkaido simply captured the cells in a post implantation identity. From that point of view, it’s very similar to the original Jamie Thomson media, the FGF active and base media for capturing human prime stem cells. They’re just not consistent with the pre-implantation identity. When you treat naive cells with the Hokkaido media, these initial human trophoblast stem cell media, you push them towards that later state. Now, is there a way to trap them at an earlier stage?
Again, based on these papers last year using dual MEK and TGF Beta inhibition, it is possible to pass the cells for a couple of days through a pre-implantation trophectoderm state, but no one has yet been able to successfully capture that state in vitro.
Another way to do it though, is to build these blastoid models because it’s pretty clear that the blastoids have an actual pre-implantation trophectoderm that is not post implantation trophoblast. I think there are two ways of getting there. Perhaps both start from naive. You can either directly make this pre-implantation TE for a couple of days or make a blastoid.
But there is still this unanswered question, can we stably arrest human trophectoderm stem cells in vitro? I think that’s maybe the next objective for the field to do so.
*** All content on NationalStemCellTherapy.com is for informational purposes only. All medical questions and concerns should always be consulted with your licensed healthcare provider.