Scientist Stories: Donald Ingber, Co-Founder of Emulate
In this hour‑long seminar, Donald Ingber, founding director of the Wyss Institute and co‑founder of Emulate, presents the evolution, technology, and applications of human organs‑on‑chips. He begins by framing the drug development crisis—high cost, long timelines, and poor predictive power of traditional cell culture and animal models—and introduces microengineered chips as a solution that combines living human cells with precise microfluidic and mechanical control. Ingber explains the foundational technologies: soft lithography using PDMS, photolithographic etching, and the creation of laminar micro‑flows that enable tissue‑tissue interfaces and dynamic mechanical cues. He details the first lung alveolus chip (2010), showing how cyclic stretch mimics breathing, enabling realistic inflammatory responses, pulmonary edema modeling, and the discovery of a TRPV4 inhibitor that prevented edema in both chip and animal models, leading to a phase‑2 trial. He then surveys a portfolio of organ chips—airway, intestine, kidney glomerulus, blood‑brain barrier, liver, bone marrow, lymphoid follicle, and orthotopic cancer models—highlighting how each recapitulates organ‑level function (e.g., villus formation, glomerular filtration, BBB electrical resistance) and enables mechanistic insights that are inaccessible in static cultures or animal models. He emphasizes the role of mechanobiology, showing that mechanical forces (stretch, flow) can dominate drug responses, tumor dormancy, and fibrosis. Ingber describes integration of multiple chips into a “human body‑on‑chips” platform, using a common blood‑substitute medium and robotic sampling to link gut, liver, kidney, heart, lung, brain, skin, bone marrow, and pancreas. This system can predict human pharmacokinetic parameters (Cmax, half‑life) for nicotine and cisplatin, matching clinical data. A major portion of the talk focuses on rapid response to viral pandemics. Using airway and alveolus chips, his team modeled influenza infection, identified host‑targeted duplex RNAs that boost interferon‑beta, and screened repurposed drugs for SARS‑CoV‑2. They discovered that hydroxychloroquine and chloroquine lack efficacy in the physiologically relevant chip, while amodiaquine shows potent antiviral activity, now in clinical trials in Africa. The workflow combines chip screening, transcriptomic‑based drug‑repositioning, molecular dynamics, and validation in BSL‑3 labs and hamster models. Ingber concludes with reflections on translation: the Wyss Institute’s funding model (large, time‑limited gifts), the creation of Emulate to commercialize the technology, and the broader vision of biologically‑inspired engineering beyond medicine. He discusses future challenges—adding stromal, immune, and neuronal components, modeling sex, age, and rare diseases, and convincing reviewers to accept chip data without animal validation. He also offers career advice on navigating academia versus industry, emphasizing the fluidity between the two. The talk ends with a Q&A covering cell sources, chip failures, matrix stiffness, and the role of public‑private partnerships.
Chapters
Okay, so that's it. So, so I would like to thank Don Ingber for having accepted to talk to us today. It's a great pleasure. So he's in the line with our previous various prestigious speakers that we had, like Rakesh Jain, Bob Weinberg, and Hans Clevers. And so, so I will briefly introduce him here. So Don Ingber is the founding director of the Wyss Institute of Biomedical Inspired Engineering at Harvard. He's the Judah Falkman Professor of Vascular Biology at Harvard Medical School, Boston, and he is also the professor of bioengineering at the Harvard John Paulson School of Engineering and Applied Science. He is a member of a number of, of organizations like the American Institute of Medical and Biological Engineering, National Academy of Medicine, the National Academy of Inventors, and so forth. And more recently, I have seen this on your Twitter account, because I'm following you on Twitter also, that you is— you are a member now at the National Academy of Engineering. This is pretty recent. So Don is really a founder of this emerging new field of biological inspired engineering. So he is He has done many contributions in all kinds of disciplines on mechanobiology, on cytoskeletal biology, matrix, and so forth, integrin. He has done also work on angiogenesis and blood vessels in the past, systems biology and translational studies. He has around 500 publications, extremely highly cited. And Of course, he has also done translation into the industry of his inventions. So he has found a number of companies, and a recent one, Emulate, who is transferring his organ-on-chip technology to the general public and the science community. So, um, Don Ingber is from the state of New York, and he, he graduated, however, in Yale and got his PhD there. And then he moved, in fact, to Boston, also did a long postdoc with Judith Folkman and was his close, his, his close associate. So now, in terms of science In fact, Don has really pioneered the field of discovery of the mechanical forces in the development control of cells in general. He has done also work on cancer. He has promoted a number of things with new concepts in that area and also new technologies. So what is important is that in fact, in his idea, the mechanical forces are extremely important that control the biology and gene regulation. And it's like the philosophers would call it a downwards causation. Philosophers talk about, about downward causation, which is in fact another level of control of biology that is in fact more outside of the genetic realm, but closely linked to it. So he studied in this respect also the transducers of the mechanosensors like integrins and others like this. Besides that, he is, done work on angiogenesis, of course. He has discovered, uh, and, uh, the first angiogenesis inhibitor TMP470. And, uh, now of course his— all his work is centered on the organ-on-chip, organ-on-chip technology. And he, for instance, designed the first human organ chip, which is the lung chip.— which has been created already in 2010. And now, of course, he is with this organ technology all around. He is now promoting a full human organ-on-chip, with a number of organs on this chip. And it's really extremely impressive work. Somehow, you can ask how he can do all this. But in fact the publications are there and his output is stunning. He is, of course, he was what I already said, he the funder, the funding director of the Wyss Institute. The Wyss Institute has emerged from a Swiss philanthropist Hansjörg Wyss who has financed it and maybe he could explain to us how he got this funding through, through Harvard, or was it, was it direct funding to him and, and to his institute. But it's really extremely impressive, I mean, all what he has done so far. And so he will talk about us, about his, this research on the organ and chip technology. And it's extremely impressive and why, and let's say eye-opening to all of us, what can be done with this approach. So besides the, besides the other thing also, Don is interesting also in the art because he has done art and design exhibitions with a number of artists. He has been working, and also, for instance, his organ chip technology has been displayed in museums like in New York at the MoMA and so forth. So it's a great pleasure to have Don here today. So it's really a great pleasure. So just for the format, So the format is the following: he will talk about 1 hour and then you may have questions for half an hour, so you should write down the questions on the chat please. I will ask them afterwards to Don and of course afterwards the students, when, if the students want to, if the students want also to ask him directly. They can ask questions to him afterwards. So, Don, thank you very much for having accepted it. Of course, you will get, as I promised you, wine bottles sent to you. I mean, all our prestigious speakers will get them, and the floor is yours.
Thank you. I'm going to share my screen. Let's make sure that this is working again.
Yes, this is perfectly working.
Thank you. Great. Well, thank you so much for the opportunity and the kind introduction. As you heard, I've been at Harvard for 37 years. I did start out as a postdoc with Judah Folkman, worked with him for 25 years, and then for the last 12, I've been the founding director of the Wyss Institute for Biologically Inspired Engineering. Let's see if I can get this to turn here. This is our logo up top. And yes, we were kickstarted with what was then the largest single gift to Harvard University in its history of $125 million, which we've tripled since then. We can talk in the question period about it. But what we were tasked to do is to take on big problems that would be extremely hard to solve, that you wouldn't solve with a conventional government grant that gives you sort of incremental funding to do incremental research, but that if solved could lead to transformative impact. And what I'm going to talk to you today is what we call human organs on chips, and I'm going to show you that these have really gone from the realm of experimental models to clinical mimicry. And so when we started, the biggest problem that I could see at the time was that the drug development model was broken. It costs over $3 billion a year to go from a discovery at the lab bench to get a drug approved. The many reasons for this, some of them is that we work with cells cultured in dishes that don't function like our bodies. We have to do animal studies, uh, before you go to the clinic, but it takes years to complete them. Uh, very big ethical issues, but the real problem is that more often than not, and sometimes depending on the area, 75 to 95% of the time, they do not predict clinical responses, and that's why drugs fail so often. And, and so as a result, there's been a search for new models that can better mimic whole organ function. For those of you interested in cancer research, for example, where biologics are becoming— coming more to the fore, like therapeutic monoclonals, and people are exploring CRISPR-based therapeutics, there's a real need for new models because animal models can't be used. 40% of drugs in the pipeline are biologics. They now often require non-human primate models. If you've read the newspapers with COVID-19, monkeys are in short supply, and they're not even enough of them for just vaccine development alone. Major ethical concerns. But the real problem is that new drugs such as biologics are so specific for human, they don't even cross-react with antigens or RNA targets in, in, uh, non-human primates. So there are real needs for preclinical models to, to basically get over that huge abyss between basic research, preclinical research, and, and clinical drug approvals. So we have developed what we call human organs on chips. These are engineered microchips that contain living human cells that reconstitute organ-level functions, not cell or tissue like an organ organoid, but organ-level functions to accelerate drug development, replace animal testing, and advance personalized medicine. Now, why microchips? Well, computer microchip manufacturing offers control of features at the same nanometer to micrometer scale that living cells and tissues live at. And this is done using what's known as photolithographic etching, where you often have a silicon chip, you make a mask that will obscure UV light, in a— you can make little holes in it using a computer to create a pattern, and where it— the light goes through it, it'll cause a chemical polymer layer to dissolve, and then you could make patterns and etch for— in, in electronic chips. Now, 25 years ago, George— over 20 years ago, George Whitesides and I collaborated together and developed what is now known as soft lithography. You can now buy these the materials for cell culture from companies, but we developed this technique which George had created, an inexpensive way to make computer chips for the— microchips for the computer industry, whereas you'd use photolithographic etching to make a pattern, but now the trick was you pour a liquid polymer, it's called polydimethylsiloxane, PDMS, it's silicone rubber, but it's a liquid, you pour it on, polymerize it, and you retain surface topography down to 60 to 90 nanometer resolution. And then you can stamp chemical inks such as extracellular matrix molecules for living cells. And we use this in a series of Science and Nature papers early in the '90s to show that we can control cell adhesion, cell shape, and cell function, including growth, differentiation, apoptosis, all by how far cells stretch and change their shape, so that mechanics control cell fates— also directed motility, sorry. George also started to use this method to create what are called microfluidics, where you have little inlets, like tributaries to a river, that would come together and fluids would join. Interestingly, they're so small, less than a millimeter wide, you don't get turbulence. You only have laminar flow. So if you two different liquids of different colors, they don't mix. We actually adapted this to develop a sepsis therapeutic device, which interestingly has just entered clinical trials for COVID-19. I don't have time to go into that. However, as someone who's a vascular biologist, to me, these were engineered microvascular networks. So we combined this all together and created this idea of organs-on-chips, and the first model is we called a human breathing lung-on-a-chip, published in Science in 2010. And what's important is we're not trying to build a whole organ. I like to think of these as living 3-dimensional cross-sections through a major functional unit of a human organ. And on this one, we started with the alveolus, or the air sac. And as you probably know, this is a structure where we have gas exchange, aerosol-based drug delivery, pneumonia, cancer metastasis, et cetera. So very, very important. Important structure, but it's relatively simple in that it has air, it's lined by a single alveolar epithelial cell, it sits on a porous basement membrane extracellular matrix, and then there's a capillary blood vessel cell on the other side, and then there's blood. What it doesn't show you, it's incredibly mechanically active in that every time you breathe in and out, the tissue-tissue interface and the cells stretch and then they relax back. And it's known in respiratory physiology that these breathing motions are absolutely critical for development of the lung as well as for continued functionality of the lung. So if you're trying to distill down to like, what are the minimal design principles that make an organ an organ? Well, the first thing is that you have to have a tissue-tissue interface. This is two or more tissues come together, new functions emerge, often one being a vascular tissue to provide oxygen and nutrients and remove breakdown product. The other feature that we felt was critical were the mechanical motions, in this case, dynamic flow of blood, flow of air, and the cyclic breathing motions. So the next video animation shows you how this works. The top right is the organ-on-a-chip. It's the size of a computer memory stick made out of that optically clear silicone rubber. You cut it in cross-section, it has 3 parallel chambers less than a millimeter wide. The middle one has top and bottom cut by a thin membrane with pores. We coat this with extracellular matrix. We then put human lung alveolar cells on the top, human lung capillary cells at the bottom. We just recreated the alveolar-capillary interface. The trick is we put cyclic suction through the side chambers. This is a flexible polymer, and so this will stretch and relax at the same rate and degree as when we breathe. We then put air over the top to create an air-liquid interface. And then we could flow medium with or without immune cells, or even whole blood if we put endothelial cells on all four sides, as we do nowadays. Now, if this were to work, it should mimic organ-level responses. So imagine you have an infection like pneumonia. So bacteria will induce the epithelium to signal with cytokines the tissue next to it, the endothelium. So there's a tissue-tissue signaling. The endothelium then are activated to express adhesion receptors like ICAM, which then recruit white blood cells that were just flowing by, they now stick, roll, diapodes, migrate across, and engulf. Now I'm gonna show you imaging through the device. These are fresh white blood cells. I know they're fresh because we took them out of my postdoc. They're labeled fluorescently white. You can't see the endothelium because they're not labeled, and the epithelium is behind the screen. So to begin with, quiescent vessel, they just flow by. Now we put bacteria. Now there's that signaling response, ICAMs expressed, and you're watching the white blood cells being pulled out under flow, And it's well known that shear stress is actually critical for that initial adhesion. Now I show you a higher mag. That's one white blood cell right about here. It goes between two endothelium that are unlabeled. Then it migrates through the matrix-filled pore. It goes out of focus to the other side. And now you're going to see it come through by phase contrast. So now I'm going to show you the white blood cells in red and the bacteria in green because they're labeled with GFP. And you watch them being engulfed. So you just watch the entire human inflammatory response in high resolution in this little rubber chip. Okay, so we started to talk to pharmaceutical companies, and they thought this was really neat, but they, they were more interested in disease models. That's where animal models really fail, and also drug toxicity models. So we hit two birds with one stone by using the cancer drug, approved cancer drug, interleukin-2, which is normally given intravenously. So we injected it through the vascular channel at time 0. Oh, I should say that its dose-limiting toxicity is pulmonary vascular leakage that leads to pulmonary edema. So we injected at the same doses in patients intravenously. Time 0, looking from above, this is the airspace, alveolar epithelium with air above. Over 2 to 4 days, you see a meniscus of fluid filling the airspace, and it's completely filled by 4. This is the same time course at the same dose that pulmonary edema is seen in patients. We could quantify this by using a trick from kidney physiology, a fluorescent dye called inulin, um, that travels with the water. And what you see here is this increase in fluid shift into the airspace over time. But what is very amazing was that if you did this with No breathing motions, that's shown in red. So breathing motions appear to be required to see this toxicity and this disease manifestation of pulmonary edema. Now, this is not mimicking physiology. No one's ever seen this before. So this was a prediction. So we developed an ex vivo ventilation-perfusion model in the mouse where we could control whether or not they breathe while we're perfusing and instilling through the air and measuring fluid in the airspace. And as you can see here, that the whole lung in the mouse exhibited this exact same behavior. Now, I've been working in the field of mechanobiology for 40 years, or since I was an undergrad, and this idea that mechanical forces may be as important as chemicals and genes for development and regulation of cell, tissue, and organ level functionality. And I had found that a drug that would inhibit mechanical signaling within 5 milliseconds after cells being stretched through their matrix— and this was forces going through integrins, through the focal adhesion to an ion channel called TRPV4 that lets calcium into the cell— and I heard that GlaxoSmithKline was working on an inhibitor of this and that the program really, it didn't seem to be progressing. And I called them and I was able to get the drug to test in our model. And we found that this TRPV4 inhibitor completely prevented pulmonary edema in this model. They in parallel tested this in dogs and rabbits in a cardiogenic pulmonary edema caused by back pressure from heart failure. And they got the same result. And we had back-to-back papers in Science Translation Medicine. And this drug is now in phase 2 clinical trials. And it's an interesting story because we— the reviewers of our paper said, one of them said, this should not be published, it's too simple a model, there's no immune cells. The other said, this is amazing, this is like synthetic biology at the cell, tissue, and organ level. They just showed you don't need immune cells for pulmonary edema induced by interleukin-2, and it was accepted. This is something we've seen again and again and again, is that we get insight into mechanisms because we could control every parameter individually—flow, mechanical motions, cell types, etc.—that you can't get in animal models, organoids, or in humans. So this one model, this human lung alveolus chip, has demonstrated proof of principle for human disease model drug toxicity. We have another paper where we've shown we could mimic the pulmonary thrombosis induced by a drug that failed in clinical trials because of deaths. Even though there was no animal toxicity, and you could see that in these chips with whole blood. Drug efficacy, therapeutic target discovery, and a new drug discovery, as I showed you. As you heard, this has won many awards. I'm very proud we won the International Design Award in 2015. We beat out the Google car and a Frank Gehry building. My family is very proud that this is now in the Museum of Modern Art's permanent design collection. And it was also— and they're proud because I'm now a permanent member of MoMA, and I get a discount at the gift shop and the restaurant for life, which is wonderful. And it's one honored by the World Economic Forum as well. Now, we have integrated immune cells, as you saw. We now integrate stromal cells for a look at pulmonary fibrosis. And we can integrate any cell you want. It really is synthetic biology at the tissue organ level. Next chip that we did was a lung small airway chip, because these companies said, well, we're interested in asthma and COPD, chronic obstructive pulmonary disease. So we basically made the chip taller, a millimeter high, which is the radius of a bronchiole. We use primary human bronchiolar epithelial cells that normally are transitional epithelium, have cilia and mucus. We grow them at an air-liquid interface on top. For 3 weeks, only feeding them from the blood channel. And at the end of 3 weeks, this is what they look like on chip. This is ciliary motion. They move in a directional manner. This is an EM colorized on the chip. And if you put small fluorescent particles to look at mucociliary clearance and look at it in real-time video, this is what it looks like. This is exactly the same rate of mucus being cleared as in our lungs as you listen and I speak. We then made chips with cells from patients with COPD, and again, 3 weeks at an air-liquid interface on chip, and we find that they retain the COPD phenotype. No immune cells here, just the cells. It must be epigenetic in that they have downregulation of TollR3 and 4 receptors, as in patients. Now, what brings these patients to the ER is usually exacerbation, either by bacterial infection viral infection, or cigarette smoke. Here you can see that if we use a mimic of viral infection, poly(I:C), or a bacterial infection, LPS endotoxin, you can see that only the COPD chips respond by either IL-8 secretion or MCSF. These are cytokines, whereas the healthy do not. So you can see exacerbation. Now, at the Wyss Institute, we have amazing engineering capabilities. And so my postdoc at the time, Camus Benham, I reached out and they developed a cigarette smoking robot. So this is a Gatling gun of 10 real cigarettes. That is a cigarette lighter from a car. And this actually breathes in and recreates the breathing puffs, pauses the intervals, and takes real smoke through this device right to the lumen of one of those chips, not cigarette extract. When you do that, it's quite impressive. You can see that normal chips, normal patient lung-derived chips, smoke doesn't do anything, for example, to the inflammatory cytokine IL-8, but the COPD chips, it doubles. And we now do transcriptomics, multiomics, proteomics, and glycomics on chips. This is transcriptomics on the right. The 3 lanes at the right are chips of patients that were healthy that we exposed to cigarette smoke as on the left. The 9 at the left are from a clinical study with patients who are otherwise healthy but smoke cigarettes. What you could see is that the top quarter versus bottom 3/4 are pretty similar in terms of their expression patterns, but the reviewers said, your chips are all very, very similar and these are also variable, it should not be published. Then we realized that in a way we can do this study better than a clinical study because this is matched comparative modeling, of these patients before and after the stimulus were interested in smoke. So these genes are actually related to smoke exposure, whereas these patients have different life histories, work histories, smoking histories, etc. And so this eventually was accepted, and then we're asked to do a little mini-review on matched comparative modeling in vitro of human patients. We continued on, we developed an intestine chip. We initially started with human Caco2 derived intestinal epithelial cells, which came from a tumor that normally grow as flat cells used by pharma, grown in transwells to look at intestinal barrier function, and it's poor. In vivo, the cells are columnar and they form villi. We put the same cells in the same medium used for 50 years. All we do is we give them peristaltic-like deformations and trickling flow like in your gut, and they spontaneously form intestinal villi. Now, you heard last week from Hans Clever, and he's done this beautiful work on intestinal organoids and many other organoids, and, um, and they're wonderful for studying cell differentiation and even some tissue morphogenesis, but, um, you can't get access to the lumen. You can't measure transport, you can't measure absorption, you can't culture microbiome for extended times. So we do is we now use patient-derived organoids, his method, we break them up into cells or aggregates. We put them on the top part of the membrane, and we put endothelium on the bottom. And now you get beautiful intestinal villi, proliferative cells at the base. They differentiate in all the layers. We've done duodenum, jejunum, ileum, and colon. And if you look at transcriptomic analysis— and this is for defense response, drug response, digestion, proliferation response to nutrients— and you compare the duodenum in a patient, the organoid that you make from the duodenum, and the chip you make from the organoid, you can see that the intestine chip is much closer to the actual in vivo situation. And again, this was asked in the beginning. This is one example of a benefit of chips versus organoids alone. We've also made a colon chip, and this is a histological section through the chip that looks just like in vivo, and villi at the top, and goblet cells, and so forth. What you can do here is you can look at mucus production and accumulation over time, which you really can't do easily in an organoid, and we could turn the chips on the side, use darkfield, and you can see the mucus in real time accumulating over days in these chips. And we could actually confirm that we recreate the human mucus structure for the first time in vitro of having a penetrable layer and a non-penetrable layer of similar thickness as seen in human colon. But what is probably most interesting is we now can culture human— complex human gut microbiome, anaerobes and aerobes, for extended times on these chips in direct contact with the epithelium and the mucous layer. So, to do this, we created a hypoxia gradient where we had a chamber where we have chips that have oxygenated medium going through the vascular channel, and then we have an oxygen-free environment above— I'm sorry, And we have inline oxygen sensors on the chips, and we can show that we can maintain oxygen levels below 0.5% in the lumen, which is needed for anaerobic bacteria. But you get a gradient so that the endothelial and epithelium remain alive and functional. When we put anaerobes on these chips, we actually find they grow. And interestingly, barrier function gets better better having microbiome, healthy microbiome microbes on these chips. This is apparent permeability. So essentially going down in this case is better barrier function. Now we've taken human stool specimens of microbiome, so complex microbiome, and we can keep for at least 5 days in vitro over 200 operational taxonomic units. This is 200 different types of bugs from 11 different genera with similar complexity. As in stool, in direct contact with human intestinal epithelium. We have now— this is a movie of how dense those little speckles are bacteria around the villi, and this is happy, functional barrier intact. We're now doing this with diseased microbiome, we're doing this with vagina chips, we're doing this with cervix chips, we're doing it with lung chips. And so this to me is something you cannot do with organoids and is a major advance in the field. We've also done another chip, was a human kidney glomerulus chip. In this one, we didn't use organoids, we had to develop our own iPS technology to induce kidney podocyte formation, because there were no available sources of kidney podocyte, which are the specialized cells on one side of the glomerular basement membrane that have foot processes that impact on it, and glomerular endothelium on the other side, and this is where you have you know, urine filtration. I won't go into the method, but this is all published, but we can get highly differentiated podocytes combining matrix and actually putting them on the chips. But what was interesting is that again, we gave 1 beat per second of deformation to mimic the expansion of the glomerulus with each heartbeat, which is known to happen when you visualize in vivo. And only when we give that mechanical distortion, do we see foot processes go from podocytes to touch the basement membrane of the endothelium, and only then do we mimic urinary clearance in vitro with inulin versus albumin of, of seen in vivo in the glomerulus. We've also used iPS to develop a human blood-brain barrier on chip. Those of you involved in brain cancer, some of the big problems is developing large, you know, therapeutic antibodies across the blood-brain barrier, We now have chips that have endothelium, human brain endothelium on one side, astrocytes and pericytes on the other. And now the first time we get barrier function with a transepithelial electrical resistance in the thousands of ohms centimeters squared rather than the tens to 100. And we could also reconstitute drug shuttling functions like transferrin receptor antibodies that Genentech has been developing for drug shuttling. Now, in terms of cancer, you know, we had these different chips, and I had worked in cancer since 1978 as an undergrad, and I knew that one of the big advances in the, in the mouse field was going from subcutaneous implantation of cancers to orthotopic models, where you put a breast cancer in a mammary fat pad or prostate cancer in the prostate gland, and they looked a little bit more like patients. So, we had human organ microenvironments, so we developed human orthotopic lung cancer chips here. We did this with a human non-small cell lung cancer adenocarcinoma model. We used H1975 tumor cells that pharma often uses for drug development. And this type of cancer emerges in the distant bronchiole, the small airway, but it almost always grows clinically in the alveoli. When looked at at autopsy. So we put these cells, we either plated them in the alveolus and the airway, we plated them at a low density at the same time as the normal cells and let them grow out. Or in the case of the airway, we could actually let the airway form, it's so thick that we could inject with a needle at a later time point. And what you see on the right is if you grow these cells in conventional dishes with conventional medium, that's the gray circles, they grow really rapidly. If you grow them in the medium we use on chips, which is unique, there's no growth in a plastic dish. So all the growth I'm going to show you is due to the local microenvironment. If you grow them in a small airway where they, where they emerge in vivo, but grow, but don't really grow very well, there's no growth for 10 days, and then they grow slowly. If we inject them after the chip has been formed, they're just dormant. In contrast, if we grow them on the alveolus chip where they grow in vivo, they grow rapidly. And what was amazing is that when we looked plus or minus breathing, we found that breathing suppressed growth by 50%. We then could look at invasion. It's very nice with these pores, you could watch GFP-labeled tumor cells invading through the pores and then invading into the vasculature. Breathing also inhibited invasion by 50%. And we could show that the mechanical breathing motions regulated therapeutic responsiveness as well. So, for example, if you look at growth without breathing, they grow rapidly, and drugs like rilucitinib, which is an EGF— it's a tyrosine kinase inhibitor, first-line drug— inhibits growth. If you have breathing, they grow slower, like I said, but there's absolutely no effect of the drug. They just go slowly right through it. We actually could show that this is regulated mechanically at the level of EGFR receptor signaling. I think this is important because in cancer, patients come to the clinic, they're treated with a big cancer, it shrinks back, they're sent home, and then there's always residual tumors that then come back again. This is the type of model now that you can used to study dormancy or residual tumors. Now, I'm funded the last couple years by a Cancer Grand Challenge grant with Thea Tilsey as the, the PI, and we've been developing organ chip models of chronic inflammation-associated cancers, starting with esophageal cancers. And the goal was basically to explore the microenvironmental— particularly the effect of epithelial-stromal interactions, which is something I've worked on again since I was a graduate student. The idea that that the stromal environment regulates the, the growth and development and progression of cancers, in part through the extracellular matrix. And so we are now getting clinical specimens from Lorenzo Ferri's lab in Montreal. We're getting healthy epithelium, fibroblasts, endothelium, and immune cells. That's the ultimate goal, um, from, from either, uh, for example, they'll take out a surgical resection of a a patient who has fulminant cancer, but in that same tissue organ removal, there'll be dysplasia, Barrett's, metaplasia, and adjacent normal. And so we are now building chips with these. I'm just going to show you a preliminary result. If we grow what was from a cancer sample, they grow rampantly on these chips, and you can see them overgrowing. This is 14 days on chip, this is in a transwell, so much higher growth rate, and they, they're growing all over each other in a disorganized way. We grow them in a differentiation medium and an expansion medium, and the cancer cells grow rapidly regardless. We then took a sample from an, quote, healthy adjacent, healthy region from this patient, and What we found is that in the differentiation medium, they form columnar cells that look almost like stomach, which is sort of a metaplastic type of a phenotype, whereas in the expansion medium, they grow rapidly. And so actually, this looks like it might be a first model of Barrett's esophagus in vitro, and just a glimpse of what's to come. But the ultimate goals of our work is really to show that we could recapitulate human pathophysiology and predict human responses to drugs using clinically relevant dose exposures. And I say this because almost everything all of us do with cells in a static dish, whether it's cells, organoids, or multi-physiological systems, is we bathe them statically in a drug. Where in our body, drugs are going through our vessels, going through an endothelium, into the tissue, drug levels are going up when we take our pill and then going down as it clears, and it goes up and down during the day. And this is known as drug pharmacokinetics, how the drug levels change in time, which is absolutely critical for drug activities in vivo. So, first example is a human bone marrow chip we published a year ago. This chip, we take cells from patients, we take CD34 cells, we have bone marrow-derived stromal cells, we put them in a matrix gel in one channel, and then we put endothelial cells in the other channel. And over a month, we get blood formation of all different lineages. This, for example, are neutrophils isolated on chip by FACS, and this is in vivo, and this is the erythroid lineage. Look, Giemsa stain showing all different cell types. We maintain the progenitor cells and stem cells for— we still have to quantify the percent of stem cells, but we certainly have progenitor cells for at least a month. This is much higher than in classic suspension culture. I'm funded by the FDA on looking at radiation countermeasure drugs, and one of the problems with radiation studies is that animal models don't mimic the dose sensitivity. Excuse me. Here you can see that we get similar dose sensitivity to patients when we have these chips exposed to radiation at 1 to 2 gray. Each colored dot is a different patient. 5 patients showing you the reproducibility from patient to patient, and we also see the appropriate progenitor cells being affected most. We then looked at cancer drugs. So this is now 5-fluorouracil, and again, in vitro, people can do bone marrow cultures, they show toxicity, but it takes 5 days. In patients, the drug's infused over 2 days at a concentration around 4 micromolar. When we infuse our chips with drugs at different doses, we see at 4 micromolar, we can see toxicities right here that matches what you see in patients, whereas a static gel or a suspension culture, you do not see this. Well, what was most impressive to us was AstraZeneca, the pharmaceutical company, reached out. They had a drug in clinical trials that had very peculiar regimen-specific toxicity in that The same dose over 2 hours gave you neutropenia and anemia, and over 48 only gave you neutropenia. And they could not mimic this in vitro very well or in animals. So, they had the pharmacokinetics measured in patients. That's shown at left in gray. So, there in patients, 2 hours goes high and then goes low, 48 is high for extended time and then goes low. We could mimic this on the chip because we have flow. Again, you can't do this in static culture or organoid culture. We mimic it precisely. We precisely mimic the drug toxicity seen in humans. And we could also mimic recovery, which is very hard for them to study in animals, and this is not seen in suspension culture. This is what was with an Aurora kinase inhibitor. We also actually collaborated with clinicians who were studying a rare genetic disorder in kids called Schwachmann-Diamond syndrome, which has aberrant blood cell formation. We got CD34 cells from these rare patients, put them on chips, and we see the abnormal blood cell formation on chips, and we discovered that it's due to— we could see the— it was involved in maturation of the blood cells. But what was interesting was we also discovered a new mechanistic insight not seen before, which is that we found an abnormal low CD13 to CD16 ratio by FACS, which was a, a neutrophil maturation not seen before or described before in the syndrome. And we went back and we looked at 8 of these patients and found out that there's a subgroup that 4 of them had this and 4 didn't. So, I mean, this is useful for getting some insight into these rare disorders where it's hard to get patients, but more importantly is it's very hard to do clinical trials on these patients. And this suggests that you might be able to make chips from rare patients from around the world optimize drugs, deliveries, pharmacokinetics, and then maybe, you know, do a much more efficient trial later on. Now, in the beginning, I mentioned about biologics and non-human primate limitations. And this New York Times two weeks, a month ago was that there really are not enough monkeys to do vaccine testing, let alone biologic testing. And so there's got to be a better way. And so we have developed what we call a human lymphoid follicle chip. And lymphoid follicles are in lymph nodes and tertiary organs, and this is where the immunization, the vaccination responses really happen. So what we do is we take peripheral blood from patients, we isolate B and T cells, we put them at high density in a, in a Matrigel-type gel in one channel, and we feed them through the second channel And what we find is that they spontaneously organize into clusters, and when you give antigen, you self-assemble germinal centers, which is where this all happens, which— and they expand with stimulation, the number and the volume. We find that the flow is absolutely critical for this, so this would not happen in a static culture. They express the right biomarkers like CTLA-4, CXCL13. More importantly, They're primed for antibody class switching, and this is the increased expression of activation-induced cytidine deaminase and CXCL13. And when you give them a stimulation in CD40 ligand that's known to activate class switching, you induce plasma cell formation. These are the cells that make, make the antibodies, and you get IgG specifically as opposed to IgG on these chips. So we then got a commercial flu vaccine called Fluzone, and we vaccinated the chips. We got specific IgG against that Fluzone antigen, hemagglutinin. We also collected out of the blood, the vascular channel, cytokines, did analysis, and we happened to have cytokines in patients who were vaccinated with Fluzone, and we see very similar profiles. So we now are using this with Gates Foundation and drug companies to actually look at adjuvants. We can show adjuvant activity, and this may be a new way to not only study vaccines, but, um, immuno-oncology therapeutics as well. And last example is, uh, when we did this, we're focused on human, get away from animals, and the drug company said, will you develop liver chips that are human rat, and dog, because they explained that almost every drug company has to do preclinical toxicity, hepatotoxicity studies in dogs and rats because the FDA requires it, and they almost always get conflicting results, and then they have to kind of wing it, and that's one reason there's a lot of failure. So we now did this. We started in my lab, and it moved to Emulate, the company you heard about earlier, but this chip has primary human hepatocytes in a sandwich gel on one channel. It then has primary human— I'm sorry, human, dog, or rat primary hepatocytes, and similar for all three species, liver sinusoidal endothelium primary, as well as Kupffer cells and stellate cells. And when you do this, one example, these are— this is a drug that is known to have very different responses in human, dog, and rat. You can see inhibition in human, higher dose in rat, no response I'm sorry, in dog, no response in rat, and this is exactly what we recapitulated. This is published in Science Translation Medicine. We also recapitulated steatosis, cholestasis, fibrosis, multiple drugs. So we now, I think we're in the 15 to 20 different organ-on-chip level in my group, and others at the Wyss Institute like Kit Parker, Chris Chen, Jennifer Lewis, Sangeeta Bhatia, But when I first published a review in 2012, I had this figure where I said, because we have an endothelium line channel, you can imagine building integrated human body on chips. Imagine putting a drug orally through the gut chip, watch it be absorbed, watch it be metabolized by the liver chip, pass the fluid to the kidney chip, watch it be— can people turn their mutes on?
Yes, please turn the mic off, please. Okay, so I, I will— I will, uh, please turn the mic off, please. Okay, that's it.
Okay, you can continue. Yeah, so I was saying, you could, you know, you could basically watch the liver metabolism, kidney clearance, do you see heart toxicity, or, you know, and then what does it do to bone, or aerosol drug in the lung. And we have now done this We don't link by tubes, we use a robotic sampler because we can interface with pharmaceutical industry drug testing, but more importantly, we can pass a drop from one chip to the other and then take a drop and do mass spec to do drug levels or measure cytokines. We could change the order of the chips. We have one common medium, like a blood substitute, in all chips. And in this study that was published a year ago, we have gut, liver, kidney, heart, lung, blood-brain barrier, brain, neuronal networks in that case, skin, bone marrow, and pancreas. And this instrument can keep them alive and visualize them for a month in vitro. But the real goal of this was to be able to predict drug pharmacokinetic parameters in vitro, so you might be able to shortcut clinical trials design. To do this, we made a simplified model of gut, liver, and proximal tubule or kidney, so it was a first-pass model. And we have an arteriovenous reservoir to mix it so that it would be like taking a sample from that would be like taking a peripheral blood sample rather than sample right after your liver or right after your intestine. And what you want to do is by measuring the drug levels over time in these chips using computational scaling, and we did this in collaboration with CFDRC, which is a company that does for pharma human PK in patients, So they know about the volumes, the blood flow rates, and et cetera in vivo. So we're able to scale from these little chips to in vivo. But what you want to do is be able to predict key values like the Cmax, the maximum plasma drug concentration, or the T1/2, the half time that— the time it takes for the drug to reach half of its levels to give you a feel of timing. So we did this now with 2 drugs published in Nature Biomedical Engineering last year. We did oral nicotine through the gut chip, watched it be absorbed, quantified it there, then watched it be metabolized, then watched the metabolites be distributed, peed out by the kidney chip, and we measured the blood levels in the AV reservoir. We did the same with intravenous cisplatin, cancer drug. Here we had a liver chip and a kidney chip. We didn't need the gut because it was IV, so we put a bone marrow chip so we can measure pharmacodynamics, the effects on, on the cells. I'm just showing you the PK here, but this is nicotine. The dotted lines are what were predicted from mass spec analysis of changes in drug levels on our chips, and the dots with error bars are from a past publication with 3 different formulations of nicotine in human patients published in Sweden, where they have either a gum or 2 types of chewing tobacco they call snus. 3 different doses, 3 different formulations, and we predict them beautifully. Using the same software, we now do cisplatin, and this is— the dotted lines are predictions from the model, the dots with error bars are from another clinical study with two different infusion protocols. So this shows you that this is now actually possible. So in summary, I think our two-channel chips can faithfully recapitulate human pathophysiology. We mimic human responses to drugs and radiation using clinically relevant dose exposures, and we can quantitatively predict human drug PK parameters. And so we believe that these are now ready to be integrated into the drug development pipeline, and this is happening around the world through Emulate. I'll talk about later. Now, I recently published a review in the fall, and it was titled, Is It Time for Reviewer Number 3 to Request Human Organ Chip Experiments Instead of Animal Validation Studies? And this is something that we still see to this day, no matter what we do, even if we mimic clinical responses in humans, they want to have animal studies to validate it, even though we know 75 to 95% of the time they do not work. We talked about organoids, they're wonderful, they're higher throughput, we use them as part of our approach, but we feel that by breaking them up and putting them in chips, we could do things at the organ level with PK, with absorption, microbiome, that you can't do in 3D gels. Now, I, I often end by saying that, you know, the way big pharma does clinical trials now is they will spend tens of millions on drug development, they'll get it to the clinic, they'll do 10,000 patients, they'll spend millions, tens of millions more on that, they will almost always fail, then they will do statistical number crunching to see is there a genetic subgroup who maybe responded better, and if there is they'll then do a small limited trial, and if they're lucky, they get approved for a narrow application. What you can do with these chips with organoids and iPS now is, you know, choose 100 women who are asthmatic and, um, Afro-American who are ultra-sensitive to cigarette smoke and optimize the drug for them, and then use those 100 patients in a small clinical trial. And I think this is going to increase the likelihood of success, shorten it, decrease the cost, shorten the time frame. It could really revolutionize this field. And then other groups are already beginning to use these for personalized medicine, you know, your lung on a chip, your liver on a chip, your combined Crohn's disease intestine and liver to see drug efficacy and toxicity on chips. And this is something that hospitals are beginning to explore around the world. Now, I'm not sure how we're doing on time here. It's fine.
We're fine?
Yeah, yeah. So the last 10 minutes, I just tell you, provide an example of what we've been able to do with this sort of technology just in the last 12 months for COVID-19. So I mentioned we have an airway chip, and we actually were funded by DARPA and NIH in the United States 3 years ago to develop, to leverage this model to study, to see if we could use it to accelerate drug development for potential viral pandemics. We were focused on influenza at that time. And so we developed a viral infection model. And what you see here is the airway chip I described. At the top right are GFP-labeled H1N1 influenza. And you could watch infection in real time in this device while it's happening, while air is going over it. You could put drugs through in real time as well. And so Using this, we can show that we could mimic viral strain-dependent virulence. H1N1 versus H3N2 versus H5N1. We also did a lot of alveolus chip versus airway chip. I'm not showing where H3N2 is actually more potent in the alveolus chip, and it's known in vivo, it produces more pneumonia than bronchitis. You can see that with virus, you get GFP virus, you get infection of the cells. You break the barrier by ZO1 junctions open up, you lose cilia, and the endothelium, which are not directly infected, actually are injured through tissue-tissue interactions. We could quantify barrier function, and we can see that here as well. But what's most important is that you can measure human host responses to infection as well as to drugs. So here I'm just giving you examples: IL-6, IL-10, RANTES, IL- 1-alpha MMP-1, MMP-8. And first thing is that you can see response to viral infection. This is with relative to control, this is just the epithelium alone here, the first bar. But then if you put the endothelium on the chip as well, you can see tissue-tissue crosstalk between why, for example, organoid would not show this sort of amplification by the presence of endothelium. And we could also show that the endothelium actually affects viral infection titers. When you have the endothelium, because there's a greater response, you actually have lower titers. We then could test drugs like Oseltamivir, which is Tamiflu, and you could see potent inhibition on chip. Now, this is where we were— I'm not going to show you, we've also modeled viral evolution on chips. We found drugs, repurposed existing drugs that can double Tamiflu therapeutic window. But we were here in February— January last year when COVID-19 hit. And my virologists on my teams, one day after they published this, the genome for SARS-CoV-2, developed our own SARS-CoV-2 spike protein expressing pseudovirus because we only have a BSL-2 lab. And the first thing they did is they tested drugs that they were aware of that worked in other viruses in an established cell line, human liver cell called HUH7 cell, that virologists often use for testing drugs. And one thing I'll say is that the drugs that— the cell lines that virologists use to test viral infection like Vero 6, HUH, lack the normal host response. They lack normal interferon response, for example. That's why viruses grow so well in them. In any case, we use these conventional lines. We tested 8 drugs. This is January of last year, and this has included hydroxychloroquine, chloroquine, and 6 other drugs. What you see in the white is a control virus and the gray are with SARS-CoV-2 spike protein. You can see that all 8 gave you good inhibition in the low micromolar range in these HUH7 cells. By the way, you will see many papers over the last year that have said we've repurposed existing drugs because the drug inhibits CoV-2 infection in Vero 6 or HUH7 or a similar cell line. Now, what we then did is tested these in our lung chips, and the first thing we did is we— and these often don't express the key receptors like ACE2 or TMPRSS2 in normal levels either, so that's another problem. So the first thing is we looked at our lung chips and we confirmed that they have— when they differentiate on chip, they have high levels of ACE2 and TMPRSS2. We took those same drugs, we now flow them through the chips, but now we're flowing them at a clinically relevant dose, this, the Cmax, or the highest level in patients. And now we find that hydroxychloroquine has no inhibition, chloroquine has no inhibition, Arbutol has no inhibition, amiodarone has no inhibition. All of these have failed in clinical trials. And then we have these 3 drugs that the most potent being amodiaquine, which is actually related to chloroquine. It's an antimalarial. But that we found to be effective. And so at this point, we were able to get additional funding from DARPA to leverage this pipeline to see if we could do accelerated drug repurposing. On top of the organs-on-chips at the Wyss Institute, we developed some computational approaches that we have that use transcriptomic data from patients, and then basically from healthy versus disease. We've done this in other diseases quite effectively, where you basically say across the whole transcriptome, thousands of genes change, but what genes would need to change to make a disease look normal? And then go through the data available over the web for all transcriptomic signatures for all FDA-approved drugs. And then basically prioritize what looks like might reverse the phenotype. We have another pipeline that does more of a network analysis approach that then can identify what potential targets are. And then we have a third pipeline that uses molecular dynamic simulation and medicinal chemistry targeting the spike protein itself. And all of these can find FDA-approved drugs that we then would go to higher throughput conventional assays with Native SARS-CoV-2, 'cause again, we can only work in a BSL-2 lab. So Matt Freeman at University of Maryland is an expert in coronaviruses, has a BSL-3 lab, has Vero 6 and now human ACE2-expressing A549 cells going. Validate these in organ chips that are now integrating recently into the BSL-3 labs, and then use a hamster model where we could do pharmacokinetics on the drug to know we'll get the levels we need. And then measure efficacy in the hamsters in Ben Tenover's lab at Mount Sinai. And just one example, that amodiaquine drug we identified in the airway chip to be effective, we confirmed with native CoV-2 in Vero cells that its IC50 is in the low micromolar range. In human lung A549 cells expressing ACE2, it's actually 10 times more effective in the nanomolar range, and you can reduce SARS-CoV-2 by 1, 2, 3 logs in vitro. Then we went to this hamster model. And here at the left, you can see that if we give it this drug 1 day before intranasal infection with SARS-CoV-2, we inhibit with amiodarone, and hydroxychloroquine has no effect. This is the histology. You can see a great reduction in viral load in brown. We also did animal-to-animal transmission studies. Like, we basically treated an animal and then put an infected animal in its cage. 100% of the time there is transmission and infection just being in the same cage without any intranasal injection of virus. And amadiquine inhibited this potently. And then finally, we did treatment mode. We gave this drug 1 day after infection, and we still saw inhibition at 3 and 7 days. So it means we're not just slowing the disease, we actually are inhibiting. And this drug is now in clinical trials, in 2 different trials in Africa, because amiodarone is, is actually not used in the, in the United States anymore or in Great Britain, uh, because there was in the '80s there was some toxicities that worried them. But it's been used for 40 years in Africa prophylactically in kids as an antimalarial without any problems. And so this is now in clinical trials all across Africa at 19 sites, at least 19 sites. Now, with our influenza project, serendipitously— well, serendipitously, we did a CRISPR screen to look at genes that might inhibit influenza. And it's a long story short, but as a result, we discovered a new class of immunostimulatory duplex RNAs that have a unique sequence motif and a 5'-monophosphate or 5'-hydroxyl end, which normally do not induce interferons, type 1 interferons. But these are specifically induced type 1 interferon, which is the good interferon that fights off infection, viral infection, which is not induced effectively in COVID-19 patients, whereas the chemokine, the inflammatory cytokines, chemokines are. And that's why you get the cytokine storm and this imbalance. So what we have found is that this induced interferon beta, for example, very potently in airway epithelium as well as alveolar epithelium. They potently inhibit influenza H1N1 and H3N2 on our chips as well as in vitro. But they also inhibit SARS-CoV-2 very potently in vitro. MERS, and also— I'm sorry, this is SARS-CoV-2. I don't have it on here. SARS-CoV-1, MERS, and this is common cold virus, all by inducing interferon beta, type 1 interferons. And that now we've moved to hamsters, and that also potently prevents in a prevention mode as well as a treatment mode. And you can see the lungs are just dramatically different in terms of infiltration by immune cells. And then finally, this molecular dynamic simulation approach, we've been targeting a conserved region of the spike protein that interacts with the, the, with the cell membrane involved in fusion. And with molecular dynamics, we can model drugs that would have different binding affinity to our site. And you could see this would have high affinity, this would have lower. And we testing that against the native virus, we have— it scales beautifully in terms of their inhibitory activity. And we're targeting a region of spike that is not where antibodies from vaccines target. So we're trying to develop a more broad-spectrum antiviral that could be used in the future. And using this, we developed novel drugs, and we could also find existing drugs that target that site And we have two here that we could confirm inhibit the pseudotype virus entry in our lung chips. So progress so far, I mentioned that this amadiacrine's in two clinical trials in Africa. And again, this is all since, you know, first experiments were February of last year. We've screened over 550 compounds in the Vero6 assay. We've identified 65 with IC50 less than 4 micromolar. Remdesivir is about, 1 to 3 for comparison. We have 7 drugs below 100 nanomolar. With the ACE2 A549 human lung cells, we have 15 below in the nanomolar range. And many of these are not yet in clinical trials. And 88 of the 550 compounds predicted to be active by these computational algorithms are actually currently in COVID trials, not because of us, but just because other people came upon them in different ways. We are now focused on looking at combination— developing combination therapies with a good antiviral as well as a good host response modifier. And to end, you know, can't do this without an incredibly broad group of amazing people and expertise. The organ chips in the COVID-19 Long Long Sea in Haicheng Bay were the leads. That is in press at Nature Biomedical Engineering. It should be out in a month. I mentioned Matt Freeman and Ben Kenova. As a Harvard faculty member, I have to disclose that I'm a member of the board and the scientific advisory board and hold equity in Emulate, which is now commercializing this technology. So you basically can't believe anything I told you, but I'll just say that a lot of it was done before the company formed. And I invite you to our website. We've won 3 Webbys, the Academy Award of the internet for our website. I think you'll enjoy it. And thank you so much for the opportunity to share this with you. And I will end. Thank you.
Great, great talk, Don. So, I mean, amazing set of data that you presented. So may I start with maybe the first question? So what type of endothelial cells you use in your lung chip? Where they are? Are they from iPS?
What— Different chips, we use different sources, but for the most part, in the lung chips, we've been using commercial suppliers of primary cells, and those are lung microvascular endothelial cells, human.
Okay, and did you have also for your, let's say, brain chip— you did the brain chip, right? So did you try in your brain chip also brain tumors, glioblastoma, or something like this? We have not yet.
And Kit Parker's developed the neural network chip. We have a blood-brain barrier chip where we could do that as well. We've talked about it, we just haven't because of, you know, funding and limitation of resources, and COVID-19 is obviously slowing things, but it definitely can be done. Okay, great.
So then I have here the first, first questions. Just, I push this a little bit farther here. Okay, so there's Pierre Nassoir who said, "Wonderful talk, thanks. OOCs are mostly 2D or pseudo-2D with sophisticated physiological environment and are well suited for epithelial endothelia. But did you face failures for some organs due to the lack of physiological relevance of 2D in some cases?" For instance, in the case of liver, primary hepatocytes rapidly lose metabolic function in 2D in contrast to 3D. How come that the liver overseas works so well knowing that the metabolic functions are degraded also in the sandwich cultures?
So these are not 2D, these are obviously 3D. They mimic the structure of a living organ. They're just as I said, like a living 3-dimensional cross-section through them. We get intestinal villi, we, you know, we've made skin chips where we have a millimeter-thick stroma endothelium on one side and a full stratified epidermis on top. We have that vagina where we do the same, we've got cervix where we do the same, we've got bone marrow is obviously a 3D culture, the lymphoid follicles obviously 3D culture. You don't want to do 3D versus 3, you know, Look, I've been in this field since 1978, my first publication. 3D versus 2D is ridiculous, you know. It's functional. You've got to use clinical mimicry as your benchmark, right? I mean, and you should want to reduce the system down to the simplest system that recapitulates the function you care about so that you can get insight into how things work. And, you know, there was a time time in my career where people were all excited that Matrigel would give you epithelial differentiation, but collagen gels would not, or laminin on a dish would not. And I was, you know, basically my argument was it's not the molecule, it's the mechanics. Matrigel is flexible. If you cross-link it, you don't get the differentiation. Same with collagen gels. And I used 2D substrates. I had multiple papers in Science and other places, 2D substrates where I could basically get the same functionality with any matrix molecule, as long as they adhered. But I can control their shape, round, they differentiate, spread, they grew. In the middle, you know, I'm sorry, cupcake shape, they differentiate. Totally round, they go into apoptosis. And so, I would get away from this 3D, 2-day thing. You want to mimic in vivo. The lung is 2D, a flat layer, but it's 3D, but it's tissue-tissue interface. Now, in the intestine, they spontaneously, when you give them flow, the key part is flow and mechanical motions. The flow is why the liver does so well. It's not that it's in a sandwich gel. The person's right, they don't do that great in a sandwich gel. But giving them flow, even with different configurations of matrix, they are much more functional. They're highly metabolic. They need ins and outs continuously, and you don't get that in static cultures. So the key is the flow in these. The key is having air-liquid interface in some. The key is sometimes stromal cells, epithelial-mesenchymal interactions. The key is endothelial cells sometimes, but it's really giving them what they need to do their own thing. We don't change the medium usually, we use the medium that's been used before. It's just giving the right physical microenvironment that does it. And the liver, by the way, is more functional than almost any other liver system for a month. So, you know, it's not 2D versus 3D. And in the liver, they're plates, they are 2D plates, you know, so I hope that addresses that.
Okay, thanks.
Oh, and you asked about failures. Yes. We have not had a failure. We've not been able to not build an organ that we tried. The only one that was tough for a long time was the kidney glomerulus, because we couldn't get podocytes, but then we had to do iPS, and then we did, it worked. But we have not, I can't think of one that we were not able to do.
Okay, thank you. So the, So there's also one other thing that I think it's important is in fact the stiffness, you know, the matrix stiffness. How you incorporate this?
Again, I've worked in this field before people talk about matrix stiffness. I'm one of the people who showed matrix stiffness, but the stiffness, it's a balance between the cells' contractile forces and the matrix. And each cell, it's like a bow and a bowstring, it's isometric tension. If you're, you know, I published tensegrity models years ago, you can read about this. But when the matrix stiffness allows the cell to pull it into the right state of tension, what we do is by giving them the forces that they experience in vivo, we let them put out their own matrix. And then they come to the right balance to do their own thing. The cells are putting out their own matrix. Even when you have a matrix coated on a dish, the cells put out their own matrix. But because it's a rigid dish often, they're putting the matrix out in the incorrect way. And so they're in that mode of a growth configuration. But we basically provide the microenvironment, the physical microenvironment. So the cells put out their own matrix of the right configuration, and they're turning it over over time. So people have asked that, are you using stiff PDMS membranes? But the PDMS membrane is coated with matrix. A good portion of it, 30%, 40%, is pore-only matrix. And then the cells put their own matrix on top of it and make it their own. That's the answer there.
Yeah, OK.
OK. I will add, though, as we do work with systems that have a 3D gel like the bone marrow and the lymphoid follicle, in that case we do explore about the matrix stiffness and density and so forth.
You're right, I think it's an important parameter. For instance, when you look on the brain, on brain tumors, for instance, the brain tumors, the stiffness is much lower in the brain tumor versus the normal brain. Oh, it's a factor 3. Around, you know.
And the tumors are, you know, less contractile than normal. And, you know, exactly, endothelium in tumors are— have aberrant— we've shown aberrant contractility, and they sense the stiffness of the environment differently. These are all interdependent. This, you know, the cells are making the matrix and adding to it and degrading it. They're feeling its stiffness, which affects how much matrix they put out and how much they degrade. But they're trying to come to the state that, that, that calms them, the right level of tension.
And each cell is programmed differently, right? Right. So, so, so, you know, surely the— so the work of Peter Friedl with the increased matrix stiffness that in fact has an effect on the invasive process and how they, how they invade, in fact. Of course, yes. So maybe Beautiful work. So, so it's a link also with this, all this. Okay, so there's another question here. Karine Dementon. So, amazing talk, thank you. From all your data, what level complexity would you like to reach in terms of environmental surrounding cells you may add to your chips, if getting closer to in vivo possible? Is it possible? So may I finish? What is still needed with organ chips to convince Reviewer 3 from requiring systematic animal models before publication?
Okay, let's start with the beginning, which is that I didn't go into, but like, we've integrated dendritic cells into the intestine by just flowing the cells in the bottom, and they spontaneously integrate into the epithelium. Epithelium and put out processes to explore the top. We've integrated stromal cells, fibroblasts into the lung alveolus chips to look at fibrosis. I mentioned skin where we have a thick dermis with collagen gel and a millimeter-thick skin tissue. But, you know, you really want to start progressively integrating complexity. Sometimes you discover you don't need cells you thought you needed, like the immune cells in the lung pulmonary edema, but in other cases you absolutely do. And, um, but I think the cancer project is really a great example. I mean, we start with the epithelium, we're adding the, the stroma, we're doing tissue recombinants of if you have a, you know, a tumor stroma and you add it to a normal epithelium or a pre-malignant, you know, a metaplastic, what happens? And it's, by the way, it's interesting what happens. Um, But then we want to have the endothelium there too, and then we clearly want to have the immune cells there as well. We've largely done, you know, PBMCs, but we have done some work, you know, obviously with the lymphoid follicle chip, but we now are linking the lymphoid follicle chip to other chips and to explore, you know, whether we can see, you know, also almost like an immune follicle in the intestine, for example? Can you see tissue-specific differentiation? We've put monocytes in the lung alveolus and they differentiate into lung macrophages with lung phenotype. So creating these microenvironments, you can add that kind of complexity. But I would say stromal cells and immune cells are really the biggest. Some systems you might want— obviously, neuronal cells, the brain, We are working now, we have a grant on Alzheimer's where we're making blood-brain barrier with iPS from patients, and you can make the astrocytes and the pericytes and the endothelium as well as neuronal cells. So patient, all from the same patients, another thing you'd want to do. In answer to the last question about the reviewer number 3, I really think it's more people getting educated about the capabilities of these chips. And I was asked to speak to Congress a month 2 weeks ago, where there is a proposal from congressmen to start a new institute at NIH on animal alternatives. And I'm not sure it's that necessary, but what may be really useful is more people getting funded and being aware. It's not just our chips, there are many chips, there's organoids, there's many other systems. The more people that are aware that there are other alternatives, I think there'll be less reviewers demanding it. But it's going to be a time because science is so mouse-focused the last 40 years, it's going to take a long time.
Yeah, it will take a long time to re-educate people, right? Okay, so that's the— that's, that's it. So now the, the other question is on the SARS-CoV. So your, your chip clearly, uh, clearly showed that some drugs are working in your— some drugs that are working in normal systems, or are claimed to work in standard systems, 2D systems, are not working in your 3D system. And I think it's a pretty neat thing that you showed, because you know, we had in France here high claims about hydroxychloroquine, that it would be extremely efficient and stuff like this. And your work clearly showed it's absolutely, let's say, ineffective. Every study around the world has shown that. I mean, it's, it's pretty clear. So how— so now, how do you see the interaction between the epithelium, between the epithelium, between lung epithelium and the blood vessels? What is going on with the virus? Is the because it has been claimed in some bioRxiv papers from Krista Betzold and others that in fact, that surrounding that, not the endothelial cells get infected, but maybe SMCs or the pericytes can get infected also by the virus. Have you looked into that a little bit or not?
We are just getting the native SARS-CoV-2 infection on chips going in BSL-3 labs, so we haven't looked. But Vivek Thacker, I think is his name, in Lausanne has— I was in a meeting where he reported that he's used the Emulate lung alveolus chip. And interestingly, he doesn't get high levels of infection because they don't have high ACE2. But he sees them infected via neuropilin-1, which is another receptor people have identified. And he sees that that infection results in activation of the endothelium and procoagulation. So he can model the hypercoagulation in the lung chip seen in patients. And it's not a direct infection, I don't think. I think it was tissue-to-tissue signaling.
Yeah, but the neuropilin is also expressed in endothelial cells. You have the neuropilin-1.
Yes, I mean, maybe it was infected endos. I can't, I can't remember. Of course it is, I work with Mike Clagsburn for years, but I can't remember whether it was direct or indirect, but you could model that in the chips. Now, multiple centers including the Institut Pasteur now have emulate systems in BSL-3 labs, Public Health England has it, NIAID, the military. And so I think you're going to start seeing more experiments with these chips. But that's— and we have done small airway, just the first studies to get potent infection on chips with SARS-CoV-2, my collaborators, now that they've got it in their labs. But we haven't yet done any mechanistic studies. The first one I saw there was his work, but that's the type of question you could begin to ask.
Okay, and the lung, again, and the lung endothelial cells are— so from where they are, they are derived from iPS too? No, they're primary.
So they're primary from patients? These are ones we use are commercial. I mean, they're from patients.
Okay, great. Okay. So they have another question from Julie Deschanets-Merville. Did you try to play with age of cell donors? Has SOOCs helped to study aging? That's a great question.
I mean, one thing we are doing now with funding from the FDA is we're studying male versus female. Because clinical trials often are dominated by male, and that I think will be interesting. We've talked about, could you model pediatric trials by using chips with young cells? We do note in our intestine chips when we have organoids from different age patients, but, but we have not gone through in a systematic way. We've talked to cosmetics companies about using different aged skin cells and endothelium because they know they have different phenotypes even in transwells. But that's, you know, those are the types of studies that would be really interesting to ask.
Okay, thank you very much. And so what, what else, I mean, for the cancer field, so what, what, what other cancer type you, you aim to study now?
Well, we're doing colon as well as part of that program. But, you know, it really is the ability to do orthotopic studies, whatever cancer cell type you want. You know, we're limited by capacity. You know, I do have to say these are not high-throughput systems. They're sort of high-value, high-content. It would be like, you know, we even do transwells or organoids, you know, or cell culture first work out doses, get insights into how to get the system going, and then we do the ultimate studies on these instead of doing animal studies. So it's, you know, it's like— it's more like an animal study but it's human. And you can see— I always like to say there's sort of a window on molecular-scale activity inside living human cells inside a tissue and organ context. You know, we have integrated electrodes to measure transepithelial resistance in real time. We've integrated oxygen centers. We are just developing ion channel sensors. So, you know, you can measure GFP or reporter constructs. And so you can go back and each time you know exactly where to look to look at the epithelium, endothelium. It's not like an organoid that's constantly changing every day. So the real advantage of this from an analytical perspective is that window perspective. Okay, great, thank you.
So now, um, so I have a question about the Wyss Institute. So let's, uh, can you explain a little bit how, how did you get the funding? How, how did you build this institute? And what is now the, uh, let's say the new, uh, what is the future of the institute?
So, um, it started with the provost of Harvard asking me to co-chair a committee to envision the future of engineering across the university and its affiliated hospitals. We developed this vision that basically was engineering had transformed medicine and the environment by leveraging engineering principles and trying to solve specific problems. But we felt that everybody focuses on how little we know, but the fact is we've uncovered a huge amount of information about how nature builds, controls, and manufactures from the nanoscale up over the last 50, 75 years. And we basically felt that we're now at a point where we could leverage biological principles to develop new engineering innovations. We also— I felt personally that this institute should not focus on medicine alone. It should be environment, manufacturing, non-medical as well as medical, because there's a lot of know-how in medicine that you know, in biology that could be useful for the environment, for manufacturing. Synthetic biology is a great example, right? Or, you know, dental adhesives, or etc. But there's also things in the industrial world that could be useful for medicine, and, you know, one is 3D printing. I did a startup in 1998 on 3D printing, including, you know, medical devices, because I saw that, like, wow, this has got to be useful in medicine. It was too early company died in 2001, but that's now happening, you know. So that was the vision. The donor was somebody who went to Harvard Business School and had shown interest in something in bioengineering through Harvard. So I was introduced to him through Harvard, but we did develop a good relationship. He loved the vision, and we were able to get that big gift. And then it was a 5-year gift. It's not an endowment. That's all operating funds. And so it was live or die in 5 years. So we had to be successful. So that puts a huge pressure on you. And the same for the second 5-year gift and the third 5-year gift. So I brought some of the most amazing people in the Boston-Cambridge region together. It's a consortium of Harvard, MIT, BU, Brigham and Women's, Mass General Children's, all the major hospitals. We have some partner institutions in Europe as well, based on people who've been visiting scholars and collaborations needed to continue. We've recruited 45 people from industry because— with extensive experience, because we're focused on translation, and our measures of success focus on intellectual property patents, corporate alliances, licensing deals, startups. We want to see near-term impact, yet we've had one Science or Nature paper every month on average for the last 12 and a half years with between 11 and 18 core faculty, all of whom keep— almost all have their home labs and they still have labs in their home departments as well. So it's quite an amazing place. We have— we're responsible for almost 25% of all of Harvard's IP and startups each year with this small group of faculty. Amazing staff. So, um, I don't know if that answers the question, but that's kind of where we're at.
Yeah, that's, that's really, really very impressive. So the people, the PIs that you have, they have also another lab elsewhere?
So yeah, so, you know, George Church, Jim Collins, uh, you know, Pam Silver in the syn bio area, Peng Yin, uh, Dave Mooney in, in immunoengineering, myself. I mean, yeah, I still have my lab at Children's Hospital. George Church has a big lab at Harvard Medical School. We don't have many faculty offices. It's, it's, it's the guys who do the work, the staff, the postdocs, the students. That's most of the people on site. And we've, we've been between 300 to 400 staff, and we, we're not going to grow past that in terms of future. We want to keep this sort of startup culture. You know, there are institutes that, like the Broad, that have 2,000 people or 1,600 people. And we don't want to do that. We want to stay small and be agile. Yeah, that's fantastic.
And how do you recruit your people? I mean, how do you recruit them, for instance, at the postdoctoral level, and what is the source?
You know, we, you know, at the student level, we could take students from any one of the collaborating institutions We take international PhD students. I had a thesis defense in Germany yesterday who come and visit sometimes. It's the way everybody else does, just write to us, you know, write to individuals. We also post ads as well. Yeah, so I mean, it's, you know, we're both— we're a mixed model of academia and sort of commercial. Yeah, each of us has to teach and do the committees and do all our academic department requirements. We all have our home academic departments. This institute is really focused on technology development and translation, not education. Yet we all teach, you know, in our— from our home departments.
But in your home departments, you have also your own— you have also projects that are in fact independent of the WISS. Project.
Yeah, we all have, you know, we all have grants. I mean, as well. Yes. And sometimes we have grants that have subcontracts to our home department so that they can be in both. We can be in both. But we, we allow all the staff of the faculty member in the home departments to access the resources, the, the fabrication facilities, the imaging, uh, or eventually develop projects if they're going to take them towards translation at the So it's an interesting model. We published a paper, if you look at my— I'm the last author in Bioengineering Translation a few years ago on the model itself for anyone interested.
Okay, great. Thank you very much. So there's just one final question. If— so it's another question. If I may ask another question, are the limitations of multiomics or biochemical analysis due to the limited number of cells on the chip? Is there detection?
We have, you know, you obviously can pool samples from multiple chips if needed, but we've been able to do transcriptomics, proteomics, glycomics, and what other omics have we done? I'm missing one, but, you know, on single chips. The one place that has actually led to new innovations is sometimes— generally we can do cytokines, but in our lymph node chips, we're trying to detect cytokines and antibodies that are very low level. So we had to use more sensitive detection like Quanterix Samoa system, or we had to develop our own electrochemical sensors that can have online sensors with small, you know, with femtogram per mL sensitivity for off these chips. So there are times where it can be on the limits of detection, but for the multiomics, it's been pretty doable.
Okay, thank you very much. So maybe if the students— if our students in the room— so are there students in the room? So do you want to ask a question, students, please? It's now or Never. Students, you are there or not? I have no— in the conversation thing, I don't have any input. Let's wait a few seconds.
Are they all muted by you, or can they turn their mute off?
No, no, they can write it down. Let's say quickly. And then, okay, so there's, there are no questions up to now. So maybe they are, they have some, some other things to do, in fact. So let's, so then let's, let's close this session. So I would like to thank you. It was very, really incredible talk. It was I mean, outstanding, outstanding. One question. Oh, there's one question. Okay, there's— okay, I want to put my camera. Um, so thank you very much for your very interesting talk. I had a question which is not related to science, but yet, um, so what is your opinion about public and private sector? If we, for example, for our students, if we want to for our future if we want to look for postdocs and everything and for our future career? What is your opinion about this?
It's a great question. It's interesting at the Wyss Institute because we are translation-focused. My scientific advisory board once told me that when they visited, they wanted to meet with the students alone, and they asked them, how many of you wanted to go into academia when you applied when they applied, came, how many wanted to go into industry? It was about 50/50. And then, but then they asked like, you know, how do you feel about it now? And it was about 50/50, but they were totally different people. So I, you know, this week I've had 3 people in my group who originally wanted to do academia who tell me they want to do industry now. It's quite interesting. We've had people who are our staff who we hired from industry number of them who now are faculty at MIT, UCLA, and other places because they got Science and Nature papers and all of a sudden said they want to do the academic group. So the world's a lot different than when Andreas and I started, where it was really— well, first of all, there was no— it was, you know, it was either or. Um, one thing I will say is that in my experience, that if you want to go into industry, uh, it's much better, you know, if you go in as right after you're a student, or if you're after a postdoc, or if you're a faculty, you're on a different trajectory in terms of salary, responsibilities, and so forth. If you go into industry, you know, after a postdoc, this— you have to view it more as a path that's not just science, but you're now going to manage a team, and if you really want to move in an organization, you're going to need to grow and take on, you know, new responsibilities over time, whereas in academia you tend to be kind of doing the same thing. You'll build the team, but you're kind of doing the same thing over time. Um, there are different types of gratification. It's really a personal decision. I think it's nice to know now that you can change direction if you're not happy. You know, startups are different than big pharma, and, you know, in the startup world, people are always terrified about taking a chance and it failing, but there's nothing more attractive when you're hiring in a startup than somebody who's already been in a startup, whether it failed or not. It's called, you know, serial entrepreneur. They have experience, they've been through it. So, um, you know, I, I do think you, you want to get your PhD, and probably it's— you'd want to get a postdoc, um, because again, it, it, it really is different if you go out just as a student or a master's or something. You're more on a technical route than a really professional route, but we really have a pretty equal split of people going both ways. And obviously academia oscillates with funding and faculty openings, and that affects things, and biotech and pharma, they oscillate as well. And also, so it depends where you are and so forth, but having the experience in both, I think is a great thing. I've stayed in academia, but I've done multiple companies. Bob Langer, who you may have heard of, told me early on, I was thinking of leaving academia at one point in a company I had started, and he said, well, you're so successful, why not stay in academia and you could have a foot in both camp? And it's right, that's where I fit in. Other people don't wanna deal with the grants and the administrative stuff. And the reviewers and the papers and, um, you know, other people have a very clear view of what they want. So you got to look into your— what you get excited about and follow that path. You know, it's just follow the— if it— if you're uncomfortable, don't do it because someone told you you should do it.
That's one thing I will say. Okay, so is there any other questions, or is this okay like this? It was a very interesting talk and even this great advice.
Thank you very much. Thank you.
Good luck. Good. So then finally we can close. So we had one student. Good question too. Yeah, yeah. She's good. So she's a good student. But so I would like to thank you very much for this really fascinating talk and fascinating data. And so we will send you the wine, so just send me the address so you will get it. And maybe once we— so we can meet again. So it would be, let's say, a pleasure to see you again. Thank you so much.
Thank you all.
Bye-bye. So thank you very much. Bye.
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