Scientist Stories: Hans Clevers, Organoids and Drug Discovery

Welcome to the Wednesday afternoon lecture series. I'm Cathy Kelly, Chief of the NCI Laboratory of GU Cancer Pathogenesis and a CCR Deputy Director, and I'm very excited to introduce today's speaker, Dr. Hans Clevers, who has pioneered the field of organoids. Our speaker today surely is well known to most of you. Hans Clevers has long held the position of Professor of Molecular Genetics at Utrecht University, as well as principal investigator at the Hubrecht Institute in the Netherlands. In March 2022, he accepted the position of head of Roche Pharma Research and Early Development. Indeed, he's giving his lecture today from Basel, where it is now 9 PM. So Dr. Clevers obtained his MD degree in 1984 and his PhD degree in 1985 from Utrecht University. His postdoctoral work from 1986 to 1989 was done with Cox TerHurst at the Dana-Farber Cancer Institute and Harvard University. Throughout his career, he has worked on the role of Wnt signaling in stem cells and cancer. His discoveries include TCF as the nuclear Wnt effector, the role of Wnt in adult stem cell biology, and of Wnt pathway mutations in colon cancer, Lgr5 is a marker of multiple novel types of adult stem cells and as receptor for the Wnt-amplifying ARS bondins, and new methods to grow ever-expanding organoids derived from a range of healthy and diseased human tissue. This work has led to 750 publications, at least, and 90,000 citations. Dr. Clevers is a member of the Royal Netherlands Academy of Arts and Sciences, the American Academy of Arts and Sciences, the US National National Academy of Sciences, and the French Academy of Sciences. His honors have been many. He is a recipient of a prestigious European Order of Merit of Science and Art. He is a knight in the Order of the Netherlands Lion, and in 2013, he was among the first recipients of the Breakthrough Prize in Life Sciences. The title of Dr. Cleaver's talk is Organoids to Model Human Disease. Hans, welcome to the NIH. Thank you so much for being here, and we look forward to your talk.

Thanks very much, Cathy, for, for a kind introduction and for the invitation to speak. What I'll try to do is give to all of you a little bit of background as to how we more or less by chance discovered this technology to grow human organoids from normal and diseased tissues. And for that, I have to take you back 25 years or so. Go when my lab had just discovered that the TCF transcription factors are the effectors of the Wnt pathway. And at the same time, we noted that when we knock out one of these, TCF4, that the process you're now looking at comes to an abrupt halt. And so this is the gut which you see here, stem cells in action. And we also then with Bert Vogelstein showed that in colon cancer in particular, activating Wnt mutations activate the stem cells leading to adenoma formation, and eventually to carcinoma formation. So, wind in the gut is a driver of normal stem cells and of cancers. Here you see a villus in the small intestine of a mouse or human being. At the base of the villus are about 6 to 10 so-called crypts of Lieberkühn. And this is where it was long known that stem cells should live near the base of the crypts. And when we started looking at this in mice, the TCF4 knockout mice, where this process that you see here doesn't work anymore. This process of cell renewal is probably the fastest in a mammalian body. So cells are born near the base of the crypt from these elusive stem cells. The daughter cells only take 2 days while they proliferate to exit the crypts. They then take up rapidly one of about 10 different cell types, cell fates. Paneth cells will actually then settle at the base of the crypt. You see these large cells here, all other cell types keep on moving up. And when they're about 5 to 6 days old in the mouse, they reach the tips of the villi, and they undergo apoptosis. So in a mouse, this tissue cell renews every week or so. In humans, it might be closer to 2 weeks. But still, it's probably the fastest, as I said, self-renewing tissue. So we had found that in cancer and in physiology of crypts, Wnt is the driving force. It took us a while then to find out what the target gene program is that's activated by Wnt. Actually, we had to wait till microarraying was invented in the early 2000s. Now you would probably do that by RNA-seq. But this allowed us in colon cancer cell lines to see about 200 genes that were under the control of this activated Wnt pathway, we found that the same list of 200 genes is active in crypts. And then we thought possibly— and we here is Nick Barker, a British postdoc— we thought that hidden in this list of about 200 target genes, most of them actually expressed in the entire crypts as they were expressed in the entire adenomas and carcinomas of the gut. But that hidden in there, there should be maybe one or more genes that would mark the stem cells at the base of the crypt. And here you see a gene that Nick picked after several futile exercises earlier on. And this, this is a knock-in in the Lgr5 gene GFP, but also a recombinase CreER, I'll show in the next slide. And this lights up these tiny cells. So this is one cell here on the right, the larger cells are the panicked cells. So these are post-mitotic cells, fully differentiated, active in keeping bacteria out of the crypts. And these cells, we then thought they might be good candidates to be the stem cells. They're at the base of the crypt, they are proliferative, And indeed, when we now perform a so-called lineage tracing experiment, so here we cross this mouse with a Cre reporter at will. I won't give you too much details here, but we can now turn one of these candidate stem cells permanently blue. And when that cell divides, its daughter cells will also be blue. And that's what you see here summarizing the experiments of Nick. The first few days, he indeed saw blue daughter cells arising from this marked stem cell in an adult mouse. After 2 or 3 days, they exit the crypt, they differentiate out into to one or more of the cell types, goblets, enterocytes. Day 3, 4, 5 of the experiment, as hoped for, we saw that this blue ribbon slowly creeps up the flanks of the villus. And then by day 5 or 6 of the experiment, the first blue cells arrive at the tips of the villi, and they undergo apoptosis. Now, if you don't terminate the experiment here, but we leave the mice for another 2 years, that blue ribbon is still there. So the cell that we marked at the start of the experiment is long-lived. Every day, it keeps cranking out these blue daughter cells, 2 years long, 3 years long. And when we do the experiment well, and we only mark here and there a single cell, and we produce these ribbons, we can then read out in the ribbons what kind of cell types can be produced by this cell. And what we found, what Nick found, was actually that all of the other 10 or so cell types that you see in the gut epithelium were present in one of these clonal ribbons, implying first that the cell that we mark is long-lived, unlike all other cell types, and second, that it actually is multipotent because can make all cells of this particular tissue. Now, we tried to publish that these cells are the stem cells of the gut. There was a lot of resistance because it was generally assumed that stem cells divide very rarely. And actually, one characteristic or defining characteristic of an adult stem cell would be the fact that if you labeled its DNA, that cell would then retain the DNA label for long periods of time because it would not divide for a long period. That was quite a reasonable hypothesis, because when you divide, you have to copy your DNA, you can make mistakes. So stem cells might not want to do that. However, the gut stem cell, and we now know the stomach stem cell, the hair follicle stem cell, go through many, many cell cycles. In a mouse, we think that these cells go about through about 1,000 cell cycles in a lifetime of mouse, consecutive cell cycles. In humans, maybe 20,000. Logistics of this must be enormous, but somehow these cells manage. Now, based on this, another postdoc in the lab, Toshi You see him here, and me sat down and we asked, can we possibly create an environment in a plastic dish, in a petri dish, that mimics the environment in a crypt? And the intention was really not what you see here. The intention was to take a single stem cell, see it here top left, add these growth factors. We know that Wnt was key. We knew also that our spondin is an amplifier of Wnt signals. We didn't know it was the ligand of our Lgr5 receptor, but actually that we found out much later. But actually explains what we were seeing. So we had our spondin in our cocktail, and epithelial growth factor, so an activator of tyrosine kinase receptor. And then from an old transgenic mouse experiment, we knew we had to block BMP signals. If you express Noggin in the gut of a mouse, you'll see many, many more crypts. So BMP inhibition somehow creates a crypt-permissive environment. Now, the combination of these three recombinant growth factors no serum, but we do this in Matrigel, as I guess pioneered by Mina Bissell over the past 30 or 40 years. Matrigel is a mix of laminin, collagen, fibronectin, mimics the extracellular matrix of epithelia. When we stick in a single stem cell, the idea was really, you know, create many stem cells from a single stem cell, like people were growing ES and iPS cells at the time. But we got something very different. We got these epithelial structures. And when Toshi started looking carefully at what was inside these things, he realized that these buds are crypt equivalents. So there would be multiple stem cells at the base. Also paneth cells, this other cell type that sits there. The other cells in this bud would be the rapidly dividing daughters. And the central lumen was lined by all of the other 9 or 10 cell types that we knew from, from the primary epithelium. And actually, a number of years later, by single-cell sequencing, we discovered two novel cell types that we had not seen seen earlier in organoids. We then went back to a real gut and we found that they really existed. So it's really remarkable that a single stem cell, 3 growth factors, Matrigel, nothing other is special in this culture medium, apparently remembers how to build a complete version of the tissue that it emerges from. And Toshi called these structures mini guts. I guess they would now be called intestinal organoids. So they grow for very long periods of time, we've probably had them in culture for 10 years, and mouse lives only 3 years, as you probably know. Telomeres stay long, we've never seen senescence by sequencing, we never see oncogenic mutations. Also phenotypically, they stay very stable over time. The ultimate experiment to see if they really stay normal would be a transplantation. And this was done by Toshi in my lab and Mamoru Watanabe's people in Tokyo. So Toshi sorted a single stem cell from one of our mice where where the stem cells were green, but also the entire mouse was RFP positive, so it was red fluorescent. This allowed us to keep track of the tissue growing. So from a Dutch mouse in Utrecht, Toshi isolated a single cell, grew it up to about 100 million cells in the form of these miniguts, sent them to Tokyo, and there Mamoru's people had treated several dozen mice with DSS, which is a chemical that induces an inflammatory bowel-like syndrome in these mice. They infused these red fluorescent Dutch organoids through the anuses of these several dozen mice. And over the next few hours— this movie summarizes this— they float around. Always the basal side of the epithelium is on the outside, so the integrins are on the outside. The moment they see these open lesions of collagens, fibronectins, they will attach, they will open up, and like a living Band-Aid, these organoids— and I stress they were grown from a single stem cell a few months earlier— these organoids will, will seal these lesions. And this is the real experiment. Colon of one of these mice. What is black is the Japanese tissue. Here you see the Dutch transplanted tissue. I stress again, grown from a single stem cell used to transplant maybe 40 mice. And you see these large patches of red fluorescent tissue. So the offspring of the stem cell that was isolated and grown fully integrated in the tissue. And the only way you can find it is with a confocal microscope. And any marker analysis we do is this is normal tissue. Now, one prediction would be that this stem cell should have been exhausted by now because it's been expanded so dramatically over time in the lab. So we should lose these patches. That is not what's happening. Other prediction would have been from people who believe the stem cells should really be quiescent is that these cells probably now are somehow transformed. So we might see neoplasms, polyps, adenomas. Also, we never saw these things happening. And based on this, Mamoru has gone on to do this with human organoids to immunodeficient mice. They also integrate nicely, and he currently is performing a first-in-man trial where he treats inflammatory bowel disease patients with an unstable therapy-resistant form of inflammatory bowel disease, and he treats them with organoids grown from healthy parts of their own colon. So this would be an autologous transplant, and we're of course very curious to see if this, if this will work out. Well, these are probably all experiments that are now 10, 12 years old. First, we thought this was unique to the gut epithelium because these gut stem cells are incredibly active. But then Toshi showed that this could be done for pancreas, where the pancreas is not at all a proliferative tissue under normal conditions. And as you see, there's now a long list of protocols available, human and mouse, where typically you can start with a tiny bit of tissue from a biopsy, for instance. You can send them around the world as long as you keep it on ice at 0 degrees in complete medium. You can basically send them around for 2-3 days before you put them in culture. They will need an activator of the Wnt pathway. They will need an activator of a tyrosine kinase receptor. And they typically need inhibitors of TGF-beta and BMP signaling. In addition, like memory gland, for instance, likes estrogens. The prostate, here it is, likes testosterone. So there are additional components. Many of these culture conditions have now been published. And then what you do is you basically get your primary tissue, grind it up in small pieces, stick it in, put it in Matrigel, and then typically what you would see is that everything dies in your, in your biopsy. But for the epithelial cells, they very rapidly close up, they form a cystic structure, and then they start slowly growing. And depending on the conditions, you'll start seeing stem cells, but also all of the differentiated cell types of that particular tissue that you're trying to culture. This can be used for all sorts of basic science questions where you would use knockout mice, or where you would use maybe cell lines. We think that this tissue is much closer to the real tissue that you'd like to study. But it also can be grown directly from diseased tissues like cancers. We've done a lot on cystic fibrosis, for instance, where you can really measure very well in vitro the drug response, and then show that actually in vivo patients that in vitro responded to a drug in vivo also respond to drugs. And actually in Holland now, this is the way that cystic fibrosis doctors will treat their patients. By making organoids and then decide whether the Vertex drugs that are available will work, and then they get reimbursed. Also, I'll give some examples. We can, we can model infectious diseases quite well. One other tissue that has always been very hard to grow are the hepatocytes from human liver. And as you probably realize, these are very important cells in drug development process, but also understand the control of metabolism, the synthesis of many serum proteins, etc., etc. The liver essentially only has two real liver cell types, hepatocytes and cholangiocytes that build the bile ducts. And all the other cells you see here, Kupffer cells, stellate cells, typically are migrants that move into the liver when it's forming. But it's the cholangiocytes and hepatocytes that are the real cells that build the liver. Liver is also probably the most regenerative organ of the human body. And there's two ways by which the liver can repair itself. When all the hepatocytes are sick, for instance, after a viral infection or an intoxication, the bile duct cells can dedifferentiate into oval cells proliferate and then recreate bile duct cells as well as hepatocytes. And the other way by which the liver can repair is, for instance, after surgery for metastasis from colon cancer. If you leave a third of the liver in place as a surgeon, the liver will actually very rapidly grow back. And then this is done by hepatocytes because the remaining hepatocytes are healthy. They can divide as hepatocytes, so you would never call them real stem cells, yet they create a lot of tissue. And then in the, in the restored liver, you'll see hepatocytes, you'll see the bile ducts formed as well from these hepatocytes, and you'll see the other cells migrating in. So here you can see that cholangiocytes sorted from a rejected human donor liver grow exceedingly well, almost 1 to 1. So a single cholangiocyte, a fully differentiated cholangiocyte, so one of these bile duct cells, then put in the right cocktail of growth factors, as shown by Mary Hoof from mouse a little bit earlier, but this is human here, which you see almost every cholangiocyte in a matter of days dedifferentiates and create a cholangiocyte organoid. These will have early cholangiocyte and early hepatocyte markers. In vitro, we can differentiate them towards hepatocytes, but they never get very adult-like. They really are immature, and we have to transplant to see that they really can make hepatocytes. And here are the hepatocytes— large cells, beautiful cells. You can buy them, primary hepatocytes. They're often used in the drug development industry. But it's been almost impossible to keep these hepatocytes in their healthy stay for more than a few days or maybe a week. So it's really a salt-alkaline cell type. The cancer cell lines that many labs use are cancer, are carcinomas of liver cancer, and they express liver genes to some extent, but for instance albumin will typically be 100 or 1,000-fold lower than a real hepatocyte, and cytochromes would be the same. So this was actually, we worked very hard for multiple years until Huili Hu, a Chinese postdoc who was in our own lab back in China, China, came up with a complicated cocktail with about 10 growth factors and small molecules that indeed allowed her to take a single hepatocyte, convince these hepatocytes to start dividing. So they're slower than the gut organoids, but you can see they grow pretty well. This is still mouse marked for albumin expression. We then could also do something similar for human liver, where the younger the donor, the better this works. And we use now routinely fetal human liver, if you probably a lot harder to work with in the US. But they look— what we grow looks very similar to adult livers. The only thing is that they actually seem to be able to grow forever, whereas adult livers stop after 4, 6, 7, 8 months. But you might appreciate, you know, these large cells, 40, 50 micrometers in diameter, large nuclei, big nucleolus. The HNF4-alpha transcription factor marking mature hepatocytes is clearly expressed. You even see a dividing hepatocyte here. And the protein that we stain for could have been albumin, but this is alpha- antitrypsin, very abundant expression. And when you do single-cell sequencing, these cells are indistinguishable from real hepatocytes. With the one difference is that you see a lot of cell cycle markers expressed in many of the cells because they're proliferative. Even better, they create supracellular structures. And maybe the next slide is better. Yeah, here you see again these large hepatocytes. This is now a stain for a bile duct bile acid transporter, MRP2. And MRP2 is expressed in so-called bile canaliculi, small channels that are formed by fusing the apical domains of two neighboring hepatocytes together and this creating a little channel. And the hepatocytes will then secrete bile acids that they make in these cultures very abundantly into these bile canaliculi, and they feed into a central space, which normally we think would be the space where another bile duct will connect created by the other self-replenishing sites to take the bile acid out of the liver. And actually what you see under the microscope, that these spaces swell up and eventually burst, and that's how they release their, their bile acids. So you might also see that probably every hepatocyte has managed to create a bile canal with its neighbor. We now have very nice movies where we see how this process occurs. You can transplant these again like what I showed earlier for For the gut here, the experiment was done by Yipe Di Yong in New York, and in my lab, Helmut Geijhardt made the organoids. So this is human albumin staining a large islet of transplanted human hepatocyte organoids. And in yellow, you see nuclei of dividing cells stained by Ki-67. This is about 2-3 months after transplantation, you can see that the human islet is still growing. And this is done in the background in FAH mutant mice, which on normal Ciao will actually slowly lose its own epithelium. So that creates space and allows these human islets to grow and slowly fill up the entire mouse liver. Again, showing that what we're growing here from very few, with maybe probably only one cell, remains entirely normal despite the fact that we can expand them very dramatically over time. We are currently— this is unpublished, but we're trying to now see if we can use these organoids for various disease modeling. Here we mimic chronic HBV, hepatitis B, infection, which is a huge challenge, particularly in Africa and in Asia and China. And because hepatocytes cannot be maintained long-term in culture, as I said, days, maybe a week, you could never study chronic infections. So here, essentially, we actually plate them out in 2D after we grow them, and then we keep them in 2D. Daizhong Wang, Chinese postdoc currently in the lab, he's now actually managed to to maintain HPV infection for 2 months in these, in these cultures. And for the inside, I see the cccDNA, which is the sort of the latent form of the hepatitis B virus genome, is abundantly present in these cells and keeps on cranking out the HBs antigen. So you can measure it outside the cells. But as you can see, also viral particles, infectious viral particles are produced constantly in these cultures. So we think this might be a very nice platform to study chronic HBV infections and possibly develop therapies for these chronic HBV infections. And another model that we actually currently— we just submitted a paper by Delilah Hendricks, see her name here, who's been trying to model steatosis, the first stage of NASH, of non-alcoholic fatty liver disease. See the names here. And one way of creating these models is by knocking out genes that are at the end of the lipid synthesis, lipid secretion pathway. So ApoB is a central element of LDL particles, and the hepatocytes needed to secrete de novo synthesized lipids. And if you knock out ApoB, there's also hereditary syndrome that has the same phenotype. You can see these large lipid droplets of lipids synthesized by the hepatocytes, but the hepatocytes are not able to, to secrete this. Surprisingly, even despite they're so heavily lipid loaded, they will continue to grow. But I don't show here, we can do the same thing with wild type organoids if you just grow them in in high concentrations of fatty acids, they also show these same lipid droplets. There's no registered drugs at the moment for steatosis, even though it's estimated about maybe 20% of all adult Americans have a fatty liver, which might develop into cirrhosis and fibrosis and eventually into liver cancer. There's a large number of compounds in development. We've tested a whole series of them, and we see a few that actually in a matter of 1 or 2 days will totally lead to the removal of these lipid particles. So basically restoring the normal situation. So we think that this would be a very nice model to study how hepatocytes accumulate lipids and what you can do about this. Other applications, this is infectious disease. I just showed this because the animation is that Jeroen Haaijboom makes in Holland. The animation is very nice. But here we work with Rob O'Connor, then in Boston. We tried to see if we could grow grow microbes that could not be grown in the lab on normal cell lines. And one of them is Cryptosporidium parvum. It's a malaria-related eukaryotic parasite. It has a very complicated, like malaria, asexual and sexual part to its life cycle. And here you see the experiment. So what we wanted to do is assuming that because it's a gut infection that they cause chronic, chronic gut infections, particularly in third world countries. So we were going to inject so-called oocysts, which is the infectious stage, just like an egg with 4 sporozoites inside, and hoping that in our organoids we would have the cell type that would be the preferred site for these parasites to proliferate. Here you see the oocyst landing inside a human small intestinal organoid. The most abundant cell type in these organoids are enterocytes. These are the cells on the villi that do all the absorptive activities of the gut. You see the sporozoites coming out of the oocyst. They now actually— what we find in the real experiments— that they prefer the enterocytes. So this abundant cell type, you see the microvilli here. And once they do this, they form a type I merozoite. Inside the merozoite, they multiply to about 8 of these type I merozoites. So this is essentially an amplification step. After about a day or so, this stage comes out of this structure and now infects a secondary enterocyte. Similarly, again, it forms a meront, but now this is a type 2 meront. So now actually it is ready to go into its sexual stages. We're now about 2 days into the infection cycle. Now we get 2 types of structures. In front we see a male structure, in the back we see a female structure. And the male structure has the microgametes, so it's called the microgametes. They're like the sperm equivalents. When they leave the microgametes, they can now fertilize. This is the lucky guy, first fertilized. The macrogametes. And now we're actually back at stage 1, where an oocyst is formed, and this oocyst is then released and now can leave the body of the infected animal— can be cattle, but also a human being— and then it's ready to infect the next parasite. Now, this works very well, as you can see here by EM. I won't spend too much time on this, but indeed, the reason probably that this parasite could never be grown in the lab was because cancer cell lines, essentially all cell lines, at least that we know, are transformed cells. They do not have these fully differentiated enterocytes, and therefore they cannot propagate this, this particular microbe. Marie Estes earlier showed that the same thing holds for norovirus, which also was a virus that could never be cultured in the lab. It also prefers enterocytes. And actually, we tried this, it never worked for us. And Mary reasoned that possibly bile acids would be important, and that was the key to get this working. So she could show that you can actually propagate norovirus in the lab using organoids with fully differentiated cells. Well, I think what's very productive is to use this technology for cancer research. We and many other labs have published extensively on generating organoids from all sorts of carcinomas—gut, lung, prostate, breast. Typically, when you use the basic protocol, something will grow. You can then improve. Also, side by side, you can grow healthy tissue and cancer tissue from the same with CRISPR, you can take healthy tissue and turn it into cancer by introducing 4 or 5 mutations. From cancers with CRISPR, you can remove mutations and see what that particular mutation contributed to the cancer phenotype. So that's all described, you can use it for personalized medicine. There's a large number of studies now, sort of validating organoids as an avatar of patients. So test drugs on organoids might predict what would be a good drug combination to treat patients. This is a recent study, very different from what I just described, published last year. And actually, these 3 PhD students, Cayetano, Jens in my lab, and Axel in Ruben's lab, came up with this experiment. And actually, they showed the results before Ruben or I knew that they were doing this project. They started reading about a version of E. coli, so very common microbiome component. And these particular E. coli have an extra bit of DNA, about 60 kilobases, called polyketide synthesis, PKS positive, containing about 10 genes, mostly enzymes that together constitute a synthesis pathway for a polyketide. And this particular polyketide I show in the next slide is called colibactin. And E. coli that can produce colibactin are known, or have been known for about 30 years to be so-called genotoxic. So when you culture PKS positive E. coli, so E. coli with this extra piece of DNA in their chromosome, when you culture them on top of human cell lines, you rapidly AC double strand breaks. That's why they're called genotoxic. And the question that the three PhD students asked is essentially, okay, if PKS can secrete this substance that is genotoxic, can exposure to this bacterium actually lead to mutations? And in particular, can it lead to cancer mutations? So not only causing double strand breaks, but also in the resolution of these double strand breaks, can this result in mutations that would then ultimately lead to cancer? Now, this is colibactin. While we were doing these studies, this paper was published in Science, or actually several other papers have been published on proposed structure. We believe that this is the correct structure of colibactin. Importantly, there are two warheads here that in this paper are proposed to covalently bind to very specifically to A residues in DNA. And when we model that here, here you see one warhead covalently binds to an A in one strand of the DNA, the other warhead is covalently bind to the A in another strand. Now the cell sustains an interstrand crosslink. This is not compatible with life of the cell. So it has to remove this somehow. And apparently, in this process, the cell can undergo double strand breaks. And that would most likely lead to the death of the cell. So that's why colibactin has been called genotoxic. Now what Ruben Caetano and Axel asked, maybe in the resolution of getting rid of this adduct, maybe the cell introduces mutations in its own DNA, and possibly it would be— that would be an oncogenic event. And the experiment they did, so we got from our collaborators in France, we got some very potent E. coli pks strain. We also got a mutant derived from this strain on the left, so you see it on the right, which lacks one of the essential genes in this island. And so this is the bad guy, on the right, you have the control for that bad guy. And the idea would be is that we now we grow them on agar, on Petrifilm plates. We're not microbiologists, but that's easy. We would then pick them up in a pipette. Actually, they have glass needles. We use zebrafish injectors for this. And then we would inject individual organoids with these bacteria, leave them in for 1 or 2 days, and then ask, you know, if we see what was published previously. Indeed, do we see double strand breaks? And of course, on the left we would see them, on the right that would be negative control. And the outcome you see here on the left, about maybe 40 or 50 organoids, one by one injected by Enzo Cayetano with these PKS bacteria. And indeed, on the blow-up of one of these organoids, you see many of these red dots, which essentially are gamma H2AX foci, which reveal the presence of a double-strand break. So indeed, in the control, we don't see it. So yes, PKS also in organoids when introduced where they normally are, they normally sit in the lumen of the gut. And we think this is The lumen of a mini-gut is the equivalent of the lumen of a real gut because the basal side of the epithelium is here and the apical side is here. So they really are where they are in a real gut. They cause those breaks. And then the real experiment came up. So now they were going to expose these organoids chronically to PKS E. coli on the left and the control on the right. They do so by injecting on Mondays and on Fridays they treat with antibiotics to be able to passage the organoids. Because as soon as the E. coli escapes from the organoids, it'll overgrow the culture, and that would be the end of the experiment. So it is for about 3 months. So really mimicking chronic exposure to PKS equiline. So if you now would sequence these, and assuming that they have sustained many more somatic mutations than the controls on the right, you would probably still have a very hard time to see the mutations, because every cell in this organoid might have a different complement of somatic mutations, point mutations, small deletions, and insertions. So what they then did is is they subcloned after this 3-month exposure, they subcloned single cells, grew them up essentially to amplify the genomes of these single cells to an amount where we could sequence these organoids. And then in essence, read out the genome of the cell that led to the subcloned organoid. And this now allows us to count mutations. And indeed, on the left, the exposed organoids, we saw many more mutations than on the right. But more importantly, and this was using stratums, a way of identifying DNA signatures. So looking for what kinds of mutations do we introduce, because there are many different types of mutations you can induce. And I'll show this better in the next slide. But we found out that indeed, so this would be the normal mutations you see in cell culture. So cells in culture sustain random mutations at some level, you see the same things here. But then particularly when we score for T to C changes, or T to G, we now see peaks that we would not— that we don't see in the controls. And when we very carefully map the single base changes or the single base deletions in the exposed organoids, we make a striking discovery. So apparently, colibactin goes for a very unique motif, almost like a restriction enzyme or a transcription factor. So you see a T-A pair is either deleted or is mutated. But it's done in a context of an AT-rich short domain, where there is an A on the minus 3 position. And if you model this again in the way I showed you earlier, so we believe what's happening is that colibactin binds covalently to an A on the positive strand. So that would be this A. It also binds covalently to an A on the negative strand at position 0. So that would be the A opposite the T here. Now the cell has a problem, it has to get rid of this adduct. And while it gets rid of it, it apparently changes the TA pair here into any other base, or it deletes the TA pair. Now, if this happens in a reading frame, changing it at any other base could cause a missense mutation. And deleting this single base pair would actually lead to a frameshift. And both cases, of course, can contribute to cancer. Now, these mutations occur randomly in the genome. So currently, we're not— in this version of the experiment, we're not assigning them to individual genes. We just sequence the entire genome, and we see where Colibactin hits the genome. And we don't think that Colibactin selected has a particular preference for genes. It just goes for AT-rich sequences, and then that look like this, and then it'll change or remove this TA pair. So very artificial. We then asked, now, does this happen in real life? Do we ever see these mutations? They were very unusual. They've never been seen before. We're lucky to have access to the Hartwig Medical Foundation collection in Amsterdam. So they've been sequencing whole genome metastases of a large variety of different human cancers. We asked which of these cancers do we see signs of this particular mutational process. And lo and behold, we do see that about maybe we now think about 15% of all colon cancers show clear signs of these types of mutations. So this essentially scores how many mutations fit what I just showed you, this signature. By comparison, there are quite a few breast cancer cases in this collection, there was not a single breast cancer case where we saw any evidence of this type of mutations. You get the occasional head and neck cancer, we see a few more now, the occasional bladder cancers here. But those are also sites where E. coli can live. So possibly these are also caused by PKS E. coli. And then in a second cohort, this is colorectal cancers, Genomics England, again, the same prevalence was found of about maybe 5 to 10% of the colon cancer patients had this particular signature. You now have an improved algorithm. And Ruben van Boxtel, whose lab is pursuing the bioinformatics of this, now believes that probably about 15% of of colon cancers have large numbers or fewer numbers with clear evidence of this particular mutational process. We then asked, so far we've only looked at mutations anywhere in the genome. Now we ask, do they actually hit cancer genes? Now, the KRAS mutations which are quite common in colon cancer, KRAS, G12D, or G13, for instance, cannot be created by this mutational process because the mutational footprint is not there, the ATRH sequence. APC, which is the most common tumor suppressor in colon cancer to be mutated. Up to 90% of colon cancers are believed to carry APC mutations. Actually, when you lose APC in a stem cell in the gut, you'll immediately become an adenoma. So you grow out right away. And indeed, we see that in many of these cancer cases, we find stop codons or frameshifts right at the hotspot of where typically the truncations occur in APC. So yes, we believe that in many colon cancer cases, it is the APC gene is a very good target for this mutational process. To summarize this bit, so PKS E. coli is not only genotoxic, it actually induces mutations. And we see these mutations specifically in colon cancers in the subset. We already know that from literature that about 10% of the human population carries these bacteria at any given time. Of course, the bacteria don't have to be actually present in a colon cancer patient when colon cancer emerges. Mutations might have happened much earlier in life. But based on this, if this is all confirmed, Colibacter-positive E. coli would be a mutagenic carcinogenic bacterium. So far, I guess Helicobacter is the only proven bacterium that causes cancer indirectly in humans. And this might eventually lead to maybe find out who carries these bacteria that are quite easily removed with antibiotics. And that could lead to a situation where you prevent colon cancer, particularly if we think about high-risk patients like hereditary colon cancer patients or patients with chronic inflammatory bowel disease where you probably don't want another mutagenic event. Yeah, and then a final story, I think I have time for this. So when COVID broke out 2 years ago, we, like many countries, went into lockdown and we were only allowed to do the most essential experiments, but we are encouraged to do COVID experiments. And so we teamed up with a corona lab in Rotterdam, led by Marianne Koopmans, and particularly Mark Lammers and Bart Haagmans were collaborators there. And the question was quite simple that we asked. We asked, is the gut— and this is March 2020— so is the gut also a target organ for this virus? And then we knew a few things then. ACE2, we knew to be extremely highly expressed in the gut, much higher than in the airways. And ACE2 was then already recognized as a receptor for SARS-CoV-2, was rapidly found because it's also the receptor for the original SARS virus 20 years ago. We also noted that quite a significant percentage of COVID-19 patients could show up with the GI symptoms, so have diarrhea, nausea, stomachaches. And there were some— the first PCR tests were becoming available based on the genome of the virus, and people had shown that there was some viral RNA detectable in stool, so that either came out of the lungs and ended up in stool, or possibly was produced by infected cells in the gut. So we asked very simply, can we infect, um, airway organoids, and can we infect gut organoids? Now, the airway organoids worked, but actually the gut organoids are much easier to infect, and they produce much larger viral numbers of viral particles. Here we score for live virus. So in only 2.5 days, we get a 4-log increase in viral particles. Quite surprisingly, these cells keep on looking very healthy. They don't— they keep on growing. Viral RNA, the same thing. And actually, they only after maybe a week and a half, we start slowly seeing problems in these cultures, possibly explaining why. And also, we did a lot of single-cell sequencing here, and there's a very weak interferon response, for instance, possibly explaining why infected individuals could produce so much virus without being really sick in the beginning of their disease, and in that way being fantastic spreaders of the viral infections. So in order to By optical you see a single infected cell, by EM we actually see that the virus will attach to the apex of the cell. So this is where ACE2 sits. So the infection comes from the lumen, much like they enter the patient, either the intestinal tract or in the lungs. ACE2 sits here, doesn't sit on the basal side. We stain for nuclear protein here of the virus in white. We also by EM see that the virus is secreted again to the luminal side of the cell. And indeed, after 60 hours, you can see that in an infected organoid, many get infected, but neighboring organoids that were not primarily infected don't appear to be sensitive to the virus. So the virus really has to enter through the apex and then leaves the cell again through the apical membrane. And that becomes a little bit important in the last bit of the talk. So essentially, it was published very rapidly, reviewers reviewed overnight, the journals published overnight, and there are actually several other papers showing something very similar. And then one I found very nice in Nature Medicine from Jane Zhao in Hong Kong, She showed that you can quite easily grow bat organoids from the guts of bats. And there, because there are many more coronaviruses than that can infect humans, and so they have been isolated as sequences, but they've never been grown. They come from bats, and we would propose that actually these bat organoids would be great platforms to grow these coronaviruses. But so far, I guess regulatory authorities are very much afraid of growing additional species of coronaviruses in the lab. So I don't I don't think this has been followed up by Jane. This is a study that is now actually published. So here we ask, what are the host genes that are used by these viruses? And there were then already a few papers, mostly in Cell, where genome-wide CRISPR screens have been done in the workhorse cell lines like Vero E6 cells. And about maybe 50, 60 genes have been proposed to be key for the viral infectious cycle. ACE2, I already mentioned it, the receptor, this was a control here. So what we do here is we don't do barcoded whole genome library transfections, we basically go clone by clone and we take candidate genes. So with CRISPR, we isolate a number of mutants, we sequence them, confirm that they are nulls, and then we use them. And here, for instance, you see two ACE2 mutant clones. And as you can see here, compared to wild type, they cannot be infected by SARS-CoV-2. Also, they cannot be infected by the original SARS SARS virus, but they are happily infected by a third species here, related species MERS. The MERS virus is an endemic coronavirus, has been around in Arab countries and appears to be jumping from camels to humans. And there's very few reports of this virus jumping from humans to humans. One big difference, particularly with SARS-CoV-2, is that people get very sick right away and have a very high lethality. I believe it's in the order of 50% or so. But we also see that the cultures essentially are dead in 2 or 3 days. So the, the moment the virus starts propagating itself, it kills the cells. And that's maybe why MERS hasn't really spread so far as compared to SARS-CoV-2 here. So that works, our positive control. Another control here, DPP4, is known to be the receptor of MERS, has also been proposed to be a receptor for SARS-CoV-2. However, when we knock out DPP4, SARS-CoV-2 happily infects in red compared to green, the control. But now the virus, the MERS virus, is no longer infected. So this is the wild-type control. And these are two DPP4 mutants. So this confirms that you can essentially read out in these mutant human organoids, you know, what the virus needs to propagate. And I would stress this is primary epithelium, this is not, you know, some fibroblast-like cell line. We then went through a long list of these published genes. And much to our surprise, none of these gave us a phenotypes. So we could knock out quite a few without any effect. There was a single gene already well studied, TMPRSS2. TMPRSS2 is a transmembrane protease with the protease domain on the outside of the apical side of the cell. Actually, TMPRSS2 is known to cleave the spike protein of the virus, and this then allows fusion of the viral membrane, the viral envelope, with the host cell. So we confirm when you knock out TMPRSS2, the virus cannot infect. This is the blue lines here. We knocked out a number of other TMPRSSs that are expressed in epithelial tissues. They had had no effect. So we would conclude that TMPRSS2 would be probably a great drug target. TMPRSS2 knockout mice are fine as published, so it should also be a safe target. And actually, just last week, I think this paper appeared in Bio Archives with an inhibitor of the TMPRSS2 protease. So it's a host protease, but will not mutate when the virus tries to get around the drug. And it was shown to be a very effective blocker of viral infection. And this will not only work for SARS-CoV-2, but probably will work for many other coronaviruses. We think it's really a fantastic drug target. None of these others really worked. I draw your attention to cathepsin L in purple here. It's a protease that's involved in the endocytic pathway, came up in most of the screens. In our case, it just enhances infectivity by the virus, so it doesn't block it at all. Quite surprising. And then we repeated some experiments from the literature, chloroquine, hydroxychloroquine, chloroquine when used on Vero E6 cells. So these are also the cells that were used in these genome-wide screens, at least in most of these. Chloroquine is a fantastic inhibitor of viral replication. At 1 micromolar, we reach the IC50. At 10 micromolar, there is essentially no viral propagation in these Vero E6 cells. Let me now take 10 micromolar chloroquine in a primary epithelial structure such as an organoid in here in blue it does nothing. Now, how do we put all of this together into a single story? I should stress again, we are not the only ones that have proposed this. There were some other papers that sort of at the same time as we proposed this. So what happens in, in these workhorse cell lines that are typically more fibroblast-like the way they grow, they're definitely not like primary epithelia. Often they don't express ACE2, so you need to, you need to transfect ACE2 to make them sensitive to the virus. The virus can then bind, is then endocytosed. This endocytosis step is readily blockable by hydroxychloroquine. That's how that works also in malaria. And then to release the virus from the endosome, this is where the cathepsins come in. So it's no wonder that endocytotic genes like cathepsins came up in these genome-wide screens. What happens in primary epithelia, so this would be the apical side of the epithelium, ACE2 is naturally expressed, TMPRSS2 is also naturally expressed on the apical membrane. The virus binds binds to ACE2, TMPRSS2 moves in, activates the, by proteolytic cleavage, the spikes that sit on the outside of the coronavirus. And now the virus directly fuses its envelope with the host cell membrane and can inject the RNA molecule straight into the cytoplasm of the cell. This does not involve endocytosis, so there is no sensitivity to hydroxychloroquine. And there's also no reason why CATEPSINS, for instance, should be involved in this step. And that's what we they are not enveloped. So from this, we would deduce it's probably a good idea that if one does, you know, screens, high-throughput screens in easy cell lines, easy to work with, the hits, maybe it would be a good idea to also confirm them in tissues that are more close to primary epithelium, in this case, such as airway or gut organoids. And again, viruses that enter the airways or the intestinal tract will always come in through the apical side. So there we believe the organoids are particularly suited because you can inject your whatever agent into the lumens of the organoids would be here. You can also grow organoids when they're expanded in 2D, you can actually keep them for a few weeks, and then just add whatever you want to test on the, on the top side of the 2D culture on the apical side. And in transwells, you can then play around with the basal side of the epithelium. And with that, I reached the end of my talk. I believe I mentioned all of our collaborators during the talk, so I'll stop share, and I hope I left some time for Q&A.

Great, thanks so much, Hans. That was really a beautiful presentation that's really given us an appreciation of the huge potential of organoids to accurately mimic human tissues to study disease. Very, very interesting. There are a few questions, and maybe I'll take the first. If I understood correctly, I think that your cultures of fetal versus adult liver are really mimicking the aging process. So, you have an in vitro model of aging there that hopefully is accurate. And I'm wondering if you've analyzed it, is it telomere erosion, epigenetic like differences? Do you know what's going on?

Yeah, so actually, we did a lot of all sorts of genomics. But the principal observation is that the telomeres shorten, and we don't see that in most other cultures. So for instance, TERT is highly expressed in our gut organoids, but in the liver, it's sort of a limiting thing. Also, in liver cancers, TERT activating promoter mutations are quite common. So it looks like in the liver, if you want to proliferate, you need to do something about your TERT. You can actually overcome it, not entirely, but largely by TERT expression, by transfection of TERT. And also, you see the typical senescent P21, P16 come up. And then you basically, and then this looks like classic senescence, which we never see in other— still, we don't know whether it's lack of one growth factor that would solve this, or whether it's by whether it's sort of baked in the liver. But the fetal liver, indeed, not all samples, but many samples we can grow for years. They look, if we do RNA-seq, they look very much, if not, they're almost identical to the adult livers. We lack, now, some cytochromes might not be at the adult level. You can play around with growth factor conditions, but there is this one big difference where they will grow forever and the adults don't.

Yeah, that's really interesting. Prostate, actually, synesis as organoids as well. Let me see, so there are a few questions here. Antonella Pleggi had a question about the bowel repair experiments, and she's really wondering about the plasticity of the stem cells in the gut. So, if you were to use stem cells from esophagus, gastric, small, large bowel, would What would you expect?

That's a very good question, actually. So there, my former postdoc Toshi Sato, who essentially developed organoid technology in my lab, has in his own lab has done a lot of transplants. And what he finds in essence is that, so yeah, once the tissues are established, so say if you take samples, you know, late fetal or adult, they know what they are, and you cannot change that fate. So they can actually— stem cells can become goblet cells, enterocytes, and they can move back and forth. So there's plasticity in their lineage. But if you transplant a small— so I should say, yeah, a small intestinal organoid to colon, and colon, for instance, doesn't have villi, it doesn't have paneth cells, you'll actually get a colon with villi and with paneth cells. He's gone on, he had a beautiful Nature paper recently where he shows with short bowel disease, which is a terrible syndrome in infants where they lose a large part of the small small intestine. The colon is still fine, but doesn't really help because the small intestine is where you absorb all your nutrients, and the colon only absorbs water. So he actually converted colon into small intestine. So he basically— and this was done on animals, we hope to do this in these babies— so you can actually take part of the colon, strip the epithelium off, then seed it with small intestinal organoids. Could be even grown from the patient you're transplanting in vitro. And then in a second— so First he creates a loop with two ends on the outside through the skin, so then he actually can manipulate the colon. So when it's then seeded with small intestinal epithelium, he then takes that loop surgically and places it in the small intestinal tract, and then it works. And this is a very old developmental biology experiment where the subepithelial structures instruct the epithelium what to be. But apparently later in life, so once they have their fate, you cannot move them. You can actually not turn stomach into small intestine, and even in the length of the small intestine, there are regions that are quite different in what they take up. They know what they are, and even after a year in culture, they are still ileum or jejunum or duodenum. If you transplant them, that's what they will make. Yeah.

Well, in some ways, that makes the system, you know, gives you confidence that you're looking at the right the right thing when you develop these organoids to mimic the tissue.

Yeah, you essentially make them after they fully fade. It's very different from iPS cells where you can essentially push them in any direction, and the iPS gut organoids tend to be a mix of the different areas.

And yeah, absolutely. Okay, there is a question from Jim Fleet. He says, others have now shown that traditional culturing of intestinal organoids leads to many more mutations from oxidative stress than what one sees in vivo. Have you cultured organoids in a relative hypoxic environment to mimic normal physiology?

Yeah, this is true. Actually, it was shown by our collaborators, and we're a little bit involved, as well as other labs showing this. Yeah, the mutational rate is 4 or 5 times higher than it is in vivo. In vivo, actually, we also could use this technology to show that a gut stem cell in humans sustains about 40 single base per year of life, so 1 per week, and it's linear throughout life. Liver, the same thing, although the cells don't divide much, also about 1 per week. It's higher, it's significantly higher in culture. Yeah, we've worked with oxygen scavengers and also low oxygen. So low oxygen, you get fewer mutations, but also they grow slower. So there's people that believe that stem cells like hypoxic conditions. It is true that they sustain fewer mutations, but also they undergo fewer— so if you calculate, you don't win because they grow slower and you have to grow them longer and then they will still have all those mutations. But it's clear that our, the cultural environment is more mutagenic than what they sustain in real life.

Yeah. And then there are multiple messages about what a wonderful and informative talk it was and significant and thank you, thank you, thank you. So you'll probably be able to see those. So I know it's 10:00 PM there. I feel like we should let you go. But I'd just like to thank you again. It was really a wonderful presentation. OK. We all learned a lot.

Well, thanks to all for listening.