Scientist Stories: Jeremy Nathans, Molecular Genetic Studies of CNS Vascular Development and Disease
In this lecture Jeremy Nathans reviews his laboratory’s multi‑decade investigation of how canonical Wnt/β‑catenin signaling controls vascular development and barrier function in the central nervous system (CNS). He begins with the discovery that the secreted protein Norrin, acting through Frizzled‑4 and LRP5/6, is the ligand for retinal angiogenesis and the cause of Norrie disease and familial exudative vitreoretinopathy (FEVR). Genetic knockout of Norrin or Frizzled‑4 in mice produces hypertrophic, abortive retinal vessels, a phenotype that is rescued by endothelial‑specific expression of a stabilized β‑catenin, confirming that loss of canonical signaling underlies the disease. The talk then shifts to two CNS‑specific co‑activators, GPR124 and RECK (Rec), which act as ligand‑specific enhancers for Wnt7a and Wnt7b. Luciferase reporter assays, live‑cell binding studies, and mouse knockouts demonstrate that GPR124/Rec are required for Wnt7a/b‑driven angiogenesis in brain and spinal cord, and that Norrin can compensate for their loss when β‑catenin is stabilized. Further CRISPR knock‑in of point mutations in Rec abolishes Wnt7a/b signaling, producing region‑specific vascular defects that reveal partial redundancy with Norrin. The lecture then connects canonical Wnt signaling to the blood‑brain barrier (BBB): single knockouts of Norrin or GPR124 have modest effects, but combined loss leads to massive tracer leakage, loss of tight‑junction protein Claudin‑5, and ectopic expression of the fenestration marker PLVAP. Circumventricular organs (area postrema, subfornical organ, choroid plexus) naturally lack BBB properties; stabilizing β‑catenin converts their vasculature to a barrier‑competent state, while loss of the Wnt inhibitor WIF1 partially restores barrier features. Finally, Nathans outlines genome‑wide profiling (RNA‑seq, ATAC‑seq, whole‑genome methylation) of CNS versus peripheral endothelial cells, identifying a BBB‑specific transcriptional program and candidate transcription factors that may be therapeutically exploitable. The lecture emphasizes mechanistic specificity, redundancy, and the therapeutic relevance of modulating Wnt signaling for vascular and neurological disease.
Chapters
It's a pleasure to join you and give you an update on our work in the last several years, and in particular, what we've been doing to look at the mechanistic basis of CNS vascular development and disease. Let me begin somewhat unconventionally with the names of the people who did the work. I don't want this to be an afterthought. And in my own lab, there are a number of students, postdoctoral fellows, technicians and research associates who've worked on this project. I want to particularly point out Chris Cho and Amir Ratner, Mark Sabah, Phil Smallwood, Yanshu Wang, Xin Yi, and Yulian Zhu, who were main movers here. And we've had many wonderful collaborators over the years. They're listed down below. And just to give us the picture from 30,000 feet, here's our object of interest. It's the human circulatory system, the vasculature. Of course, virtually every tissue in the body is supplied by the vasculature, but the particular tissues that are of interest to us are the brain and retina, the central nervous system. And in particular, we're interested in the way in which canonical Wnt signaling affects both the development and the function of the CNS vasculature. And here's a very much simplified view of canonical Wnt signaling. Just by way of review, probably for most of you, the signal is initiated when an extracellular ligand, a Wnt protein—these are roughly 40,000 molecular weight proteins—binds to its receptor called Frizzled, a 7-transmembrane protein, and that receptor-ligand complex interacts with a co-receptor called LRP5 or LRP6—they're nearly identical. Single-span transmembrane proteins. And this, this little cartoon, this is the lipid bilayer at the plasma membrane, and this combination of Wnt, Frizzled, and LRP leads to a stabilization of beta-catenin, a rather unusual mechanism. Beta-catenin is a cytoplasmic protein that is normally turned over at a very high rate, and its stabilization leads to its accumulation in the cytoplasm and then eventually its migration from the cytoplasm to the nucleus. Where it and its partners control transcription. Now the story for us begins with a ligand that is actually not Wnt, it's called Norrin, but it also— I'll show you in a minute— it behaves very much like a Wnt, and it uses one Frizzled, Frizzled 4, and it signals in retinal vascular development. And this story really began in the early 1960s when Meta Warburg who was a Danish ophthalmologist, described a disease that she named Norrie disease after the original ophthalmologist who saw the very first patients in the 1920s. And here is one pedigree from Warburg's paper. Every square is a male, every circle is a female. I think you can appreciate that this is classic X-linked inheritance. Here's a carrier female, she has several affected sons. She has a daughter who also has an affected son, and so on. And Norrie disease turns out to be due to mutations in a small secreted protein. It's a far-flung member of the TGF-beta superfamily, and the phenotype is congenital blindness. In fact, the retinas are very much altered, and we now believe that the result— that is the result of a severe hypovascularization, a failure of the retinal vasculature. To grow and a subsequent scarring and degeneration of the retina. Some years later, a milder retinal hypovascularization disorder, or really family of disorders, was described. These are called familial exudative vitreoretinopathy, FEVR, or FEVR for short. And this is what FEVR looks like through the ophthalmoscope on the right compared to a normal retina on the left. You can appreciate that the The blood vessels come out of the optic disc where the optic nerve goes in towards the brain, and they cover the surface of the retina. And in the fever retina, many of the blood vessels are present, but there's a central region which is hypovascular, and in this instance, has retracted from the retina. There's a scar reaction, a scarring reaction, which involves retraction, and there's a very substantial loss of vision for this individual in this zone of retracted retina. Now, uh, where this got, um, these two worlds of fever and Neurie disease collided and got very interesting was when we and, uh, others made knockout mice. In particular, Burger's group in Switzerland made the first Norrin knockout mouse. And the protein name is named Norrin, but the gene is named NDP, Neurie disease protein. When that gene was knocked out in mice and when we knocked out Frizzled4, we saw nearly identical phenotypes. And this is just our repeat of the Berger lab's knockout for Norrin. And the phenotype is quite striking. So these are cross-sections of the retina. Let's just look at the wild type first. In blue is DAPI, so we see the three nuclear layers: the photoreceptor layer, the inner nuclear layer, the ganglion cell layer. And in red are the blood vessels. And there are a series of layers also. There's an outer tier, a middle tier, and then an innermost vitreal tier of blood vessels. I think you can appreciate that in both the Fissured-4 knockout and the Neurii disease knockout retinas, the vasculature is present. In fact, it's hypertrophied on the vitreal face of the retina, and it has begun to invade the retina, but it has abortively invaded. That is, it has failed to invade at a sufficient distance and to link up in capillaries. These are just dead-end balls of cells that we're seeing here in the mutant retinas. Well, this virtual identity in the retinal vascular phenotypes of a putative ligand, a member of the TGF-beta superfamily, for which at this time no receptor was known, and a known receptor, Frizzled-4, for which the ligand was already known, suggested to us that maybe this receptor actually had a different ligand, an additional ligand. That is, it had Norrin as a ligand, and that turned out to be the case. This is work of Chung Zhu, who was a postdoc in my lab at the time, and what Chung did was he simply transfected the receptor, the ligand, and one of the co-receptors into a reporter 293 cell line, and he saw very impressive canonical Wnt signaling. This is a luciferase readout. When all three are present. But if you leave out any one of the three, for example, in this case the ligand is omitted, there's virtually no signaling. And the same is true whether one uses the LRP5 or the LRP6 receptors, or co-receptors, I should say. Now, in addition to mediating cell signaling, Norrin also binds with high affinity to Frizzled-4, and this was demonstrated by with, uh, Genshi Shea in my lab, uh, in multiple ways. For example, if we produce the Frizzled-4 ligand binding domain, which we call CRD, cysteine-rich domain— it's a small domain of about 100 amino acids— as an IgG fusion protein, and as a control, we produce the ligand binding domain of a different Frizzled, also as an IgG fusion, we see very specific binding of Norrin to the Frizzled-4 ligand binding domain. And the same if we do binding to whole cells expressing the full-length receptor, and the affinities are in the low nanomolar range. So FRZL4 seems to be acting as a receptor for Norepinephrine, at least in these cell culture and biochemical experiments. But you might be wondering, well, does this really hold up in vivo? And so let me show you one piece of evidence. It's an experiment from Yuli and Zhu that argues strongly in favor of The answer being yes. And let's just focus on the bottom 3 panels here. These are enphosphos views of the retinal vasculature, and on the left is a wild-type mouse retina, and the vasculature has been color-coded based on its depth within the retina. So blue is most superficial, green is in the middle, and red is deepest. And in the second panel, the middle panel, we see the same retinal flat mount now for a Norrie disease knockout retina. And you can see this hypertrophy of vessels on the surface. You see these balls of vasculature that really have not accomplished anything. They penetrated a little bit into the retina, but they haven't formed the vasculature. And now on the right is a flat mount of a mouse retina. It's a Norrie disease knockout retina. But we have genetically engineered these vascular endothelial cells, the cells that line the blood vessels, to express a stabilized form of beta-catenin. That's the CTNNB1 gene. Here's the trick. This is from Mark Takedo's lab in Japan. Turns out that the phosphorylation site on beta-catenin that controls its stability, that leads to its degradation, resides— is encoded on a single very tiny exon. That exon happens to be a multiple of 3 nucleotides, so if you delete it, you get an in-frame deletion and the protein appears to be perfectly fine. It doesn't need this little segment of polypeptide, but it's stabilized. It can't be degraded nearly as efficiently as the wild-type protein, and so it bypasses any defects at the receptor and ligand level and simply stabilizes beta-catenin within the cells, and the result is a very nice rescue. So we take this as strong evidence that what's wrong in vivo in a Norrin knockout mouse, and also presumably a Norrin mutant human, is a failure of canonical Wnt signaling in vascular endothelial cells. Now let's look at a second chapter in this story because it gets more interesting and more complex, and that is GPR124 and REC. These are two additional players, two additional proteins, and they are ligand-specific coactivators of beta-catenin signaling, canonical Wnt signaling, and they are important in the brain and spinal cord, but they do essentially the same sort of signaling. They contribute to the same sort of signaling. And by way of background, let me just note that our lab showed a number of years ago, Max Tischfield, when he was a postdoc in the lab, that canonical Wnt signaling is required in the embryo in the vasculature only in the central nervous system. So what Max did was he knocked out beta-catenin with this conditional knockout using a Cre that drives expression in all vascular endothelial cells, and what he saw was a little embryo that's shown on the right— the right is the mutant embryo, the wild type is on the left— what he saw is a little embryo that was normal in size and shape and morphology outside of the CNS seemed perfectly normal. It had somites, it had a heart, it had an aorta, and so on. But within the CNS, the vascularization, which you see on the left here in the normal embryo in green— this is the neural tube going down the middle— all these beautiful blood vessels growing into the neural tube were growing abortively, essentially the neural tube equivalent of what we saw in the retina with the Norrin knockout mouse. And that would argue that there's something special generally about CNS tissue that requires canonical Wnt signaling for angiogenesis, for vascular invasion. To this day, we don't know what that is. We don't know what's special about CNS tissue. Now, this initial experiment got more interesting thanks to the work of my colleagues here at Johns Hopkins, Ken Kinsler and Bert Vogelstein. In the Oncology Department, shown on the right, and their postdoctoral fellow Brad St. Croix, who when he was a postdoc here— that was 20 years ago— did a very remarkable piece of work. He looked with one of the earliest transcriptome technologies, actually invented by the Vogelstein-Kinser lab, called SAGE, which allowed him to ask in a systematic way whether the vascular endothelial cells in tumors were different in their transcriptomes from the corresponding normal vasculature of non-transformed tissues. And in particular, they looked in the colon, and what they saw was something remarkable, which was there is in fact a change in gene expression. It doesn't involve that many genes, it's in the tens to 100 range, and among them were 8 genes, 8 membrane proteins, encoding 8 membrane proteins, which showed dramatic increases in tumor vasculature. And they decided to call these tumor endothelial markers, and they numbered them 1 through 8. And TEM5, tumor endothelial marker 5, is the one that we're going to talk about here. It was a known protein, not a protein of known function, but a protein that was that was in the database as a sequence. It was called GPR124, and it had 7 putative transmembrane segments, so it was considered potentially a G protein-coupled receptor, hence the name GPR. And it was 124 of these orphan putative receptors in the collection. Well, here's just an example of the in situ hybridization data from the Kinsler-Vogelstein lab. Here, for example, is TEM-5. The InSightU signal is in red, and this is a colon carcinoma. And you can see it's very nicely expressed in the vasculature that has infiltrated this tissue, but it's essentially not at all expressed in the normal human colon. The same is true, shown on the left, for TEM-1. Now, this got even more interesting about 10 years after its discovery. This is now —now we're up to about 2010—when 3 groups essentially simultaneously showed that knocking out GPR124 in mice led to a defect in CNS angiogenesis and also in blood-brain barrier integrity. And let me just show you a little bit of that data. This is now reproduced in our lab. This is the developmental data. If we look at an embryo, this is about day 13, And we section through the head. We see lots of blood vessels in the CNS. Let's just look at the cortex on the right. You can see the blood vessels in the cortex at this point, and the second from right is the medial ganglionic eminence. And in a GPR124 knockout, first there's a lot of bleeding in the CNS. This is secondary to vascular breakdown, but most impressively, there's very little vasculature. So if you look at the cortex, it's essentially avascular. And likewise for the median ganglionic eminence. And there's also a fair bit of cell death occurring. Okay, quite a dramatic phenotype. Now, the, uh, the view in the field at that time was that this was a novel G protein-coupled receptor, and therefore there was a novel G protein-coupled receptor pathway that must play an integral role in CNS vascularization. And that all seemed very exciting. When we saw this data, we had a somewhat different reaction. We thought this is amazingly similar to the defects one gets with canonical Wnt signaling knockouts, and maybe GPR124 has something to do with canonical Wnt signaling. So at this point, Julian Zhu, who was a grad student in the lab, did a very simple experiment. She asked— and this is the left panel, I'll take you through this step by step— she asked in the luciferase reporter assay if you plot on the horizontal axis the relative luciferase activity that you get when you transfect in any of each of the 19 Wnts— and there are 19 Wnts in the mammalian genome, and we've tested each one of them— if you test a Wnt plus GPR124 plus LRP5, and in some cases you get more or less signaling. And I should add here that these cells, these are modified 293 cells, they actually express at low level essentially every one of the 10 Frizzleds encoded in the mammalian genome. So we didn't have to add an additional Frizzled for this experiment, and in any case we weren't sure which one we would add, so it's just as well that they express all 10 at low level, saved us the trouble. But we see that many of the Wnts do in fact induce signaling in this reporter assay. And then on the vertical axis is the relative luciferase activity ratio of this experiment in which Wnt, GPR, and LRP are present, divided by the level that you get if you omit the GPR124. You just put in the original empty vector PRK5. So we're looking for the level of stimulation afforded by GPR124 and we're asking this for each of the 19 WINCs plus Norrin. There are a whole bunch of the data points down here at the left which are not labeled, but the ones on the right, many are labeled— all are labeled, I guess. And the striking observation was that, uh, for almost all of the WINCs and for Norrin, there's essentially no effect of adding GPR124. That is, the ratio on the vertical axis is very close to 1. Shown by this horizontal line. But for two of the Wnts, Wnt7a and Wnt7b, there was clearly a stimulation by GPR124. And the reason this was interesting is because at this point we knew from Andy McMahon's work and Ben Barris's work that Wnt7a and Wnt7b are critical for CNS vascularization. They're doing in the brain and at least part of the spinal cord very much what neuron is doing in the retina. And here on the right, we just have a more complete assay. Now we're transfecting in Frizzled4, the receptor, as well as LRP and a Wnt and GPR124, and we get a huge stimulation compared to the control of either the empty vector or related orphan GPCR, GPR125. So this would argue that GPR124 is somehow and specifically stimulating canonical Wnt signaling through Wnt7a and Wnt7b, which made perfect sense given its phenotype of CNS vascular deficiency. Now, without showing you all the gory details, I'll just say this has gotten more interesting and complex in the years since. Turns out there's yet another protein called Rec. It's a GPI-anchored protein with many domains which is also involved in promoting Wnt7a and Wnt7b signaling. And so the full complex at the surface of the cell, we now believe, is at least these proteins: the co-receptor LRP, the receptor Frizzled, the ligand Wnt7a or Wnt7b, and the what we call co-activators GPR124 and Rec. And just to give you an idea of how specific this stimulation is, I'll show you data from some— this is not all the Wnts, we've done them all, but just so they all fit on the slide, I'm showing you just some of them. If we look at, for example, empty vector transfection, GPR124 transfection, Rec transfection, or Rec plus GPR in blue, I think you can appreciate that for Wnt7a and Wnt7b, and this is the work of Chris Cho, a grad student in the lab, there's a very impressive stimulation of signaling when both Rec and GPR are present, a much less impressive stimulation when just one or the other is present. But the other Wnts, and this is true of the ones I'm not showing you as well, are essentially insensitive to Rec and GPR124. And the same specificity is seen in a direct binding assay. So let me just show you, this is an assay that Chris developed. It's a live cell binding assay with an alkaline phosphatase fusion probe, and what we're doing is we're probing live cells with either the N-terminal region of Rec, and we've shown that that's the business end of Rec as far as Wnt signaling stimulation is concerned, or we're probing with the N-terminal region of GPR124. And again, we've shown this is the business end with GPR124. Now, if we transfect the cells— this is 293 cells with GPR124 and Frizzled4, and then each of the indicated Wnts, Wnt1 through Wnt16, or Norrin, and then probe those cells with Rec, which was not transfected into them, so it's the missing piece, we see that the Rec probe binds only to cells that were co-transfected with Wnt7a or Wnt7b, not with any of the other Wnts or with Norrin. And likewise, the GPR124 N-terminal region probe binds only to cells transfected with Rec and Frizzled4 and the two appropriate Wnts, 7a and 7b, but none of the other ligands. So this complex, we think, is assembling in a very specific manner and inducing signaling very specifically by these two Wnt ligands. Now let me just show you a little bit of in vivo data that this this canonical signaling story really is happening in vivo. The top two rows you've seen before. This, the top one's a wild-type mouse, the second one is the GPR124 knockout mouse. And what we've done here is we've asked, what if we just flood the central nervous system with norepinephrine, the norepinephrine ligand? And we have a CRE-dependent mouse transgene that will do that. It's called Znorepinephrine.. And so here we're showing in a wild-type background, there isn't much effect in a wild-type mouse of having some more extra Norrin on board. And I should mention here, Norrin is expressed in the embryo, in the brain and spinal cord, so it's not just in the retina. And we had noticed this and were supposing that Norrin was also acting in the brain and spinal cord, but maybe not giving much of a phenotype because of redundancy with this Wnt Wnt7a and 7b system. But now if we take a GPR124 knockout, which we presume is defective in the Wnt7a, Wnt7b signaling arm, and flood the CNS of that little embryo with Norrin, and you can compare it to the second row here, which is the knockout alone, you can see there's a very nice rescue of the CNS vascularization defect. That is, it doesn't really matter how you activate canonical Wnt signaling as long as you do. In, of course, the right place at the right time, in this case, vascular endothelial cells in the embryo. But they will work just fine if they're activated by Norrin signaling as opposed to Wnt7a or Wnt7b signaling. Now, we've refined this even further. I won't go into all the details, but it turns out we've been able to narrow it down to just a couple of amino acids, critical residues in Wreck in this N-terminal domain that are essential for Wnt7a signaling in cell culture. And when we use CRISPR to knock in a double mutant— it's a proline and tryptophan which each get mutated to alanine, and this is just the genomic PCR and sequencing to prove that the mutations are there— we see this very impressive bleeding in the CNS, and we also see a a spectacular hypovascularization. I should just— let me just go back a second and just say, I won't go into all the details, but by Western blotting, the mutant protein is present. It's— so this does not destabilize the protein. The protein is there, and whatever else Rec does— and has many other domains, and it probably does other things, we presume it's able to do all those other things— but it just is unable to activate Wnt signaling, which with Wnt7a and Wnt7b. And we do that and compare the wild type at the top here and the mutant down below, we see this very dramatic hypovascularization, all this green and hypoplastic, that is, it failed to proliferate as much as it should. No surprise since it has no blood vessels, CNS tissue. And here, same thing, and it's very region-specific. So this is cortex, essentially completely hypovascular. But the midbrain and hindbrain are actually pretty well vascularized, and that's because Norrin is there. So Norrin is partially redundant with the Wnt7a and Wnt7b systems, but it is not redundant in the cortex or the medial and lateral ganglionic eminences. Okay, now this actually goes some way towards addressing what I would say is a long-standing challenge in biology, Wnt biochemistry, and that is how Wnts interact with their Frizzled receptors in a specific manner. This is a crystal structure that Chris Garcia's group at Stanford did now about 8 years ago, and in pink is the Wnt and in blue is the ligand binding domain of a Frizzled. And you can see this most remarkable co-crystal structure in which the Wnt, the pink part, is holding the ligand-binding domain like a, like a hand holding a hot potato with, with one thumb and one finger, but nothing else is touching except a little bit of protein on each side and a lipid here which is sitting, sitting in a lipid crevice. But most, nearly all of the contact residues in fact are highly conserved across different wnts and across different frizzles. So the question of ligand-receptor specificity, which very clearly exists from in vivo data, was left substantially unanswered by this crystal structure. And we think that accessory proteins like GPR124 and Rec are a large part of the story. They are somehow, maybe for other Wnts, there are other accessory proteins, but at least for Wnt7a and Wnt7b, they are providing much of the specificity. And so just in sum, these are the players now as we know them for canonical Wnt signaling in vascular endothelial cells in the CNS. And these two accessory proteins, GPR124 and Rec, are the players that make this 7A and 7B signaling possible and specific. Now let me move on to a related topic, and that's the blood-brain barrier. I mentioned it very briefly in connection with the original GPR124 knockout mice. And, uh, there's a fascinating, uh, story here. Recall that the blood-brain barrier is really a whole cell biological and molecular genetic specialization of vascular endothelial cells in the CNS. Here's a cartoon of a capillary in the CNS, and what these capillaries have that makes them special is they have tight junctions, so there's no intercellular diffusion of small molecules. They lack the transcellular diffusional pathways like fenestry and bulk transcytosis. Everything that moves from the blood space into the parenchyma of the brain has to do so with specific receptors as shown here for receptor-mediated endocytosis or for a transporter here transporting, for example, glucose or an amino acid. A very expensive system. But almost certainly the reason the system is designed this way is to keep toxins out of the brain. I should mention that the retina being part of the CNS has exactly the same system. This is a whole pattern of gene expression which has altered the properties of the CNS vasculature. It's a big deal because blood-brain barrier integrity is compromised in many neurologic disorders. So for example, in stroke, multiple sclerosis, CNS infections, traumatic brain injury, tumors, epilepsy, Alzheimer's, the list goes on and on. There is a decrement in blood-brain barrier integrity and a leakage of serum contents into the brain. It's also important because the BBB excludes drugs that act in the periphery and that would have intolerable side effects if they entered the CNS. For example, blood pressure medication. And conversely, drugs that have to act in the CNS, like antidepressants, antipsychotics, anxiolytics, analgesics, and so on— they need to have the right combination of properties to essentially fool the blood-brain barrier and penetrate it, and that's not so easy— easily done. So this is a big deal in the pharmaceutical industry. Now let me just show you the one really sort of critical result And that is that canonical Wnt signaling is not just used for CNS angiogenesis, it is also repurposed later in life, in fact throughout life, to activate the blood-brain barrier program. So here's just one example showing how this works, and I should add that like the developmental case, both the Wnt7a and Wnt7b systems, that is the system that also uses GPR120, 24 and REC, and the Norrin system are redundant, substantially redundant, in maintaining the blood-brain barrier. Both those systems, the components, are expressed throughout life. And if we look at a wild type— and now we're using a little tracer, sulfo-NHS biotin, which we have introduced into the vascular system peripherally— and that little tracer is kept out of the brain. You can see the brain parenchyma is blue. This is DAPI staining. It enters the cerebrospinal fluid in the lateral ventricles as shown here, left and right, but otherwise it's kept out of the brain. And that's true whether the Norrie disease gene is knocked out— in this case, a heterozygote for 124, a conditional liver minus— or in this next to the right panel We have a GPR124 conditional knockout, which is now eliminated in adulthood with a CreER driver. We have to do it in adulthood, of course, because we need the embryo to develop its vasculature normally. But again, the blood-brain barrier is normal. But if we knock them both out simultaneously, we get massive leakage of this little tracer everywhere within the brain. And the same thing can be seen by immunostaining. So Without going into all the details, Claudin-5 is a tight junction protein. PLVAP is a marker of leaky vasculature. It's a fenestration protein. And in the wild-type, or any of the brains that have wild-type phenotypes with respect to the blood-brain barrier, there's lots of Claudin-5 and there's no PLVAP, so it's green and not purple. But in the double knockout, the vasculature anatomy hasn't changed, but the gene expression has changed. PLVAP is now expressed, CLAUDIN5 is repressed, and that has led, as well as changes in other genes that normally would distinguish blood-brain barrier competent from incompetent, that has led to the loss of barrier function. Now finally, the last topic, circumventricular organs. And these are regions within the CNS that have highly permeable vasculature. So these are the exceptions that prove the rule.. And where are they? Well, they're just in a few, uh, very distinct places. They're in places where the contents of the serum need to be accessed by neurons. So for example, the area postrema, which is right under the cerebellum, is a small region of CNS tissue which is sensing toxins within serum, and it makes you vomit. That makes perfect sense because if you've ingested something that's toxic, as that material is starting to get absorbed into your bloodstream, you have a little time window to vomit up the rest of it and get it out of your stomach. And that's what the area post-trema does. In fact, it's also— this is of medical significance— it's the region that's responsible primarily for the nausea that accompanies chemotherapy in cancer patients, because that is essentially a toxin and acts on the area post-trema. The subfornical organ again needs access to the serum because this is an organ that assesses the osmolarity of serum, the sodium content in particular, and determines whether the organism should be thirsty or not. Now, each of these areas has a vasculature which is leaky. So here's, uh, the area postrema. This is a sagittal section, posterior to the left, anterior to the right. Let's just focus on this region and the surrounding green. This is the same set of markers we saw before— sorry, it's the same colors we saw before, but the green marker is different. This is now the glutamate transporter, GLUT1, which is yet another marker of the blood-brain barrier. It's required because glucose can't freely diffuse across a barrier-competent endothelial cell, so you need a specific transporter there, and this is the one the one that does it. So barrier-competent cells are green, leaky cells are purple, and this is the wild type, and here's the area post-trema, has leaky vasculature. And here we've just injected that red tracer, and you can see it lighting up the perivascular region. It's just— it's leaked into the immediate surroundings of the blood vessels. I should just say, this very leaky region on the right is the choroid plexus, the the place where cerebrospinal fluid is produced. So no surprise that that's also leaky. Now, if we play the same trick by stabilizing beta-catenin in all vasculature everywhere, we see an amazing conversion. This is now done in the adult. The area postrema has flipped from leaky to tight. It's now become barrier competent, and there's very little leakage. There's an occasional cell which has failed to undergo combination. But except for those cells, the rest of the cells are barrier competent. Here's the same thing with the subfornical organ. I just want to draw your attention to this protein Lef1, which is one of the co-activators with beta-catenin. It's a partner of beta-catenin, but it is also part of a positive feedback loop. The Lef1 protein itself is a target of canonical Wnt signaling or beta-catenin signaling and is upregulated. To give even more signaling. And in a wild-type mouse, if we look at the subfornical organ— this is a coronal section— we see again the purple staining shows that it's leaky vasculature. The surrounding area has tight vasculature. I say this green at the top is just an artifact of nonspecific sticking to myelin. But here we see that Lef1 is in the nuclei of the normal CNS vasculature. It's absent from the nuclei of the subfornical organ's vasculature. But now if we play the same trick and stabilize beta-catenin in all vasculature, we see that not only has the purple marker gone away, the PLVAP, to be replaced by Claudin-5 in the subfornical organ's vasculature, but now Lef1 is also expressed and accumulates in those vascular endothelial cells. So a toggle switch here in which canonical Wnt signaling toggles the cell back and forth between leaky, if beta-catenin signaling is suppressed, and barrier-competent, if beta-catenin signaling is turned on. Just to give you an idea how incredibly specific this is, this is a flat mount looking down on the dorsal brainstem. Just, we've peeled away the cerebellum. We're looking down at the area postrema,, and we're looking at its vasculature and staining for PLVAP in green. Uh, now I'm sorry, I switched the colors on you, so this was potentially confusing, but the green is the leaky vasculature and the red with GLUT-1 is the normal CNS barrier-competent vasculature. And you can see that the vasculature that supplies the area postrema in green is uniformly leaky. What's amazing is the sharpness of the boundary between the two. Going from one vascular endothelial cell to its immediate neighbor, we go from green to red. There's no gradation in between. That is, the state is— the states are quantized. The cell is either completely leaky or completely tight, and it seems, probably thanks to various positive feedback loops, that there's no in-between. And here I'll just note we've been wondering what the source of this difference in gene expression and of course cell biological properties is, and so we've looked at known proteins, extracellular proteins, that inhibit Wnt signaling. You might expect that some of them would be in these regions, these circumventricular organs, and in this case the area postrema, now shown in a coronal section. And in fact, here's one of them. It's called Wnt- Wnt inhibitory factor 1. We actually discovered this about 25 years ago. It's a secreted protein that binds tightly to essentially all Wnts and serves as a competitive inhibitor. And you can see by in situ hybridization, it's precisely expressed in the area postrema, not in the adjacent brainstem or the cerebellum. And when we knock that out, what we see is a partial conversion. I emphasize partial because there are clearly other players here, and I don't want to overstate this, but if we look at an area post-trema in a wild-type WIF1++, we see here in purple the cells that are in the area post-trema and that are leaky, and then we can ask how many of those cells have GLUT1. There are very few, but if we do the same analysis in a WIF1 knockout, we see quite a few more cells, and we've quantified this and it's highly statistically significant. But I think what this is doing is it's pushing the toggle switch a little bit, but there are other players for sure, which must be critical for pushing it the rest of the way, and we don't know at this point what those players are. Let me just conclude by saying that our current approach is really to take a deep dive at the whole genome analysis of barrier competence and the gene expression program that's underlying the CNS vascular endothelial cell phenotype. And so we've done, for example, RNA-seq, and here's just a comparison of mouse brain, liver, lung, and kidney. These are FACS-purified endothelial cells. R1 and R2 are the two replicates in each case. So for this particular instance, this is the glutamate transport, GLUT1, we can see that there are lots of transcripts in the brain endothelial cells. If we look at accessible chromatin with ATAC, we see multiple regions upstream that are accessible. The promoter region is largely accessible in all cell types, even though it's not expressed in the non-CNS cell types. But these upstream accessible regions are presumably enhanced enhancers. In fact, when we look at methylcytosine— and this is done in collaboration with Joe Ecker's lab at the Salk— whole genome methylome analysis, we see that in the brain there are clear regions of hypomethylation that correspond exactly to the regions of chromatin accessibility. You can see them here in each of these places, as well as the gene body being less methylated because it's expressed. So with this in hand, Mark Sabat has done a very thorough computational analysis. We have a pretty good idea, based on the putative transcription factor binding sites within the promoters and enhancers, which transcription factors are relevant to the CNS blood-brain barrier phenotype and distinguish it from the non-CNS cells. And so we hope at some point this will, just to circle back, give some clinical dividend. I think the big question from the clinical point of view is what's different about these cells and the disease condition and what can we do about it? So thank you very much. I'll stop there. And I look forward to questions and to meeting many of you one-on-one. Thank you.
More from Joshua
Scientist Stories: Hans Waterham, The value of phenotypic assays in peroxisomal disorders
Scientist Stories: Joel Zlotogora, Genetic carrier screening programs for reproductive purposes