Nobel Prize lecture Drew Weissman Nobel Prize in Physiology or Medicine 2023

I now have the great pleasure to introduce this year's Nobel Laureate in Physiology and Medicine, Dr. Drew Weissman. He was born in Lexington, Massachusetts. He received his MD-PhD degrees from Boston University in 1987 and did his clinical training at Beth Israel Diakonis Medical Center at Harvard Medical School. His postdoctoral work was performed at the National Institutes of Health with Dr. Anthony Fauci. He then moved to Perelman School of Medicine at the University of Pennsylvania in 1997, where he met Dr. Karikou. He's the Roberts Family Professor in Vaccine Development and Director of the Penn Institute of RNA Innovations. When Dr. Weissman was working late in the lab, which was so often the case, his wife was sometimes asked by her friends what her husband was doing. She then often sarcastically answered, oh, Drew is saving the world. As it turned out, he eventually did. Dr. Weissman, we are very much looking forward to your Nobel lecture entitled Nucleoside Modified mRNA LNP Therapeutics, please. Thank you very much. I need to thank so many people. I'll do that throughout my entire talk, but I need to thank the Nobel Foundation, the Nobel Selection Committee, and all of the scientists here for being present for these talks from Katie and I. Most of all, I have to thank Katie, because Katie is the person who introduced me to RNA. I was a dendritic cell scientist from the NIH who had lots of dreams. Katie supplied the RNA, and together we made those dreams come true. I wanted to start by talking about vaccines, because that's where I started when I came to Penn. I had an interest in vaccines, mostly because dendritic cells, which are antigen-presenting cells, are the principal cell to target with the vaccine. So I wanted to develop ways of loading dendritic cells with the view of making better vaccines. I had every method I could think of except mRNA, and that's where I luckily met Katie at the copy machine. There are many different types of vaccine platforms, and everybody in this room has had many of these, maybe not all. They started off hundreds of years ago with live viruses, attenuated viruses. The first one was by a physician, Edward Jenner, in England, who noticed that milkmaids were not developing smallpox. When he looked at the cows, he noticed the cows had pox lesions that looked a lot like human smallpox. He cut open those pox lesions and injected them into people, and they were protected. That was likely the first vaccine ever developed. After that, a number of other platforms developed of inactivated viruses. Those are the influenza viruses that us old folks get. We also have protein vaccines like tetanus toxoid, viral vectors. These are the adenoviral vectors that were developed for COVID-19, and DNA and RNA. What I wanted to talk to you about first was the Nucleus Modified mRNA LNP vaccine platform. We started this about 10 years ago, and we used a couple of model systems. The ones I'll show you now are for influenza. We chose influenza because it's a yearly infection that sweeps the world, and it changes every year, which means you have to change the vaccine. Unfortunately, it means we have to guess what next year's viruses are going to be. Sometimes we get it right, and often we don't. Influenza in the United States leads to about 30,000 to 60,000 deaths per year. Very large number of people get infected with the virus and are out for days to a week with symptoms. When we first started doing this work, this was the literature. These were what were reported for responses to vaccines. Vaccine manufacturers always compare to what the old vaccines did when they develop a new vaccine. Live virus, inactivated virus, and other delivery routes gave neutralization titers around 1 to 100, and that was considered pretty good. 1 to 40 is protective. We took the approach, well, let's encode the hemagglutinin protein, which is the major surface protein of influenza, as an mRNA. We immunized animals with that, and we compared that to two other types of standard vaccination. You'll see in the middle, that's inactivated virus. On the left are live virus in the nose, but the mRNA blew us away. I did this with a collaborator who ran an influenza lab. It took him a month to get these results because he kept having to repeat them and dilute them more and more. He came storming into my office saying, what did you do to these samples? It's impossible to have a titer this high. He was accusing me of giving him falsified samples to make his life miserable. I explained to him, no, these are the real results. The titers were 50 times higher than inactivated virus. They were five times higher than a live virus vaccine. Typically, when you come up with a new vaccine platform, if you double the previous vaccine, biotech and pharmaceutical companies are happy. When they saw a 50-fold increase, you can imagine their joy. As an immunologist, I wanted to understand why we were getting such enormous titers. This is a cartoon of how B cells react to viruses, antigens, vaccines, proteins, anything that's foreign. Follicular B cells recognize and pick up particles, antigens, adjuvants, and get activated. They get help from a CD4 helper cell. They become something known as a germinal center B cell. The germinal center B cells are critical because those are the cells that rapidly proliferate, affinity mature, and produce long-lived plasma cells and memory B cells. We set up an assay where we measured these three different types of B cells. We did it in an antigen-specific manner. What that means is we took HA, hemagglutinin protein, and fluorescinated it. Then, we added it to a flow cytometry panel where we could identify each of these subpopulations of cells. We could count how many B cells in this mouse spleen or lymph node recognize the antigen in the vaccine. We saw, not surprisingly, that for all three, there were about 50 times more antigen-specific B cells with the RNA-LNP vaccine. That explained the incredibly high titers. It didn't explain where they came from. We forgot about some mice. Actually, Norby, he's here somewhere, forgot about some mice. We found them about 13 months later and decided, well, let's look and see what's going on. We took bone marrow out of these mice and we measured long-lived plasma cells. The long-lived plasma cells are what makes antibody that appears in your circulation. They can be there for your lifetime. We measured antigen-specific hemagglutinin plasma cells in the bone marrow. The numbers were enormous. 0.05% of all nucleated cells were making hemagglutinin antibody. That's higher than any progenitor cell in the bone marrow. It's an enormous number. We had another question, which is, is the antibody response any good? Some vaccines are great at making antibodies, but they're not good at neutralizing antibodies that are protective. In the influenza field, there's 18 different subtypes of influenza A and typically two of influenza B. The main ones that infect people and that you get vaccinated with currently are H1N1 and H3N2s. There's many different families. These families reside in different animal species. Avian influenza, which can possibly cross over into humans and it occurs all the time versus pig influenza, which crossed over in 2009. The current H1N1 was a pig influenza. These are new pandemic viruses that cross over from animals and change the world of influenza. They're always a fear. Every year, we guess what the influenza is going to be. Sometimes, we're right and you get decent protection. None of these vaccines protect against a pandemic crossover virus that can appear at any time. What we did is we took these animals, and they were immunized with an H1, and we challenged them with a very distant H5. It shares about 60% homology with the vaccine. We challenged these animals that had been immunized with an H1HA, and they were completely protected. This told us not only were we making enormous antibody titers, but we were making enormous neutralization and protection titers. These vaccines are now in phase one clinical trials in people being developed as universal influenza vaccines. These are vaccines that will prevent infection from mutated virus, from crossover viruses, from birds or pigs, or any source possible. It's the hope for the future. Instead of having to make a new vaccine every year, we make a universal vaccine that protects people for 10 or 20 or some unknown number of years. As an immunologist, I had a problem. Our vaccine did not make sense. The issue was, Katie showed you that nucleoside modified mRNA has no adjuvant activity. It doesn't induce any inflammatory cytokines. Vaccines require adjuvant. Adjuvants stimulate the immune system and say, hey, this is a foreign antigen. You need to do something. As far as we could see, there was no adjuvant in this vaccine. The lipid nanoparticles had not been described as binding any adjuvant stimulating receptors. We were confused. Typically, most vaccines have an adjuvant. They say, oh, they're Th1, Th2 biased. It doesn't matter. They're both inflammatory. There are many different types of CD4 helper cells. We focused in on one particular type known as T-follicular helper cells. T-follicular helper cells, and this brings us back to the germinal center B cells, they're required to form a germinal center. In the germinal center, they supply help to the germinal center B cells to get them to affinity mature, to get them to mature and become isotype switched, long-lived plasma cells, long-lived memory cells. Without TFHs, T-follicular helper cells, you get a poor antibody response. We investigated, was this vaccine somehow inducing TFHs in a previously undescribed adjuvant activity? We turned to the monkey model. We did this because we could measure antigen-specific TFHs. It also allowed us to do something else. We compared the RNA LNP vaccine to a standard protein vaccine with a very potent TFH-inducing adjuvant, double-stranded RNA. We measured TFHs, both total and antigen-specific. As you can see, the mRNA LNPs just blew the double-stranded RNA out of the park. We went back and calculated. On average, most adjuvants—ALM, MF59, others, double-stranded RNA—about 5% of the CD4 helper cells are TFH phenotypes. The rest are Th1, Th2, other types. With mRNA LNPs, over 50% of the CD4 helpers are T-follicular helper cells. We had enormous induction of these critically important CD4 helper cells. That's the reason why we get such potent antibody titers with the COVID-19 vaccines. The titer of antibodies with these vaccines is typically three to five times higher than what you get after infection with COVID-19. That's rare to have a vaccine that works much better than live virus or live pathogen infection. This gave us the mechanism for why we got such potent immune responses. The next question we investigated was that what about the LNPs or the RNA was responsible? We did a simple experiment. We took empty LNPs and mixed them with protein-antigen. Other people use ALM or MF59 or a variety of their adjuvants. We asked, could LNPs be the adjuvant? That's exactly what we saw. When we mixed the empty LNP with the protein, we got high levels of T-follicular helper cells and high levels of antibody production. It turns out that the LNP is an adjuvant. It's just not a typical adjuvant. We looked at what kind of cytokines the LNP induced. It didn't induce typical cytokines, which are usually type 1 interferons. It induced IL-6. IL-6 is a potent stimulus for T-follicular helper cells. The story started to come together that the LNPs are a potent adjuvant that induce T-follicular helper cells in part through IL-6 induction and by a lack of type 1 interferon induction. With that, we wanted to work on better vaccines. We've taken a number of approaches. We've developed probably close to 30 different vaccines. I'm working with a couple of researchers at Karolinska right now to develop different vaccines. We tried something unusual that we figured nobody else in the world would try because we were crazy and people knew that. We made a vaccine where we took one HA from every subtype of influenza. That's a vaccine that has 20 different RNAs mixed together. Nobody believed it would work. I didn't believe it would work, but I did it anyway. When we put this vaccine into animals, in both mice and ferrets, this is what happens when you put in only H1HA. You get a very potent antibody response to H1. You get a little bit of cross-reactivity to group 1 HAs and you get nothing to group 2. When we put all 20 HA RNAs together, we got equal antibody responses to every single protein. This surprised us first because we were getting 20 antigen responses in a vaccine. There was no antigen dominance. The critical fear of multi-component vaccines is one antigen will dominate and the others won't respond. We've done this with a couple of different RNA vaccines. We've never seen antigen dominance. We get a good response to all 20 HAs. I only say this when pharmaceutical execs aren't in the audience. In theory, I could see a day when we bring our children to the pediatrician for their every three-month vaccination. Instead of going at one month, 3, 6, 12, 16, 18, 24, up until they're 18 years old, we could be giving kids one immunization of RNA at one month, six months, two years, and be done. Instead of getting 20 vaccines per year for a number of years, they could get a few. It would greatly simplify parents' lives. It would drive pharmaceutical execs insane because they would lose all the vaccines they were producing. I don't know if it'll ever happen, but we can dream about it. We developed a new vaccine platform that we could encode just about any antigen. We could put as many antigens in as we wanted. We've done 20. We've got a vaccine with 75 going right now. I'll let you know if that works. It gave incredibly potent antibody responses. The two mechanisms were, first, Katie showed long-lived protein production. We see protein production for up to 10 days. Immune systems like that because you're constantly loading the immune system with antigen. A specific induction of T-follicular helper cells that drives the antibody responses. Now, I would often be asked questions at talks like this. If you look at the personalized cancer vaccines that Moderna and BioNTech are doing right now, and they both reported great results in melanoma and good results in pancreatic cancer, people ask, well, what's the difference between these two vaccines? The principal difference is BioNTech uses unmodified RNA, and Moderna uses nucleoside-modified mRNA. The other critical difference is Moderna uses lipid nanoparticles like the COVID-19 vaccine, and BioNTech uses a lipoplex, which is a lipid shell with an aqueous interior. That lipoplex has no adjuvant activity. They use unmodified RNA to supply adjuvant to their vaccine. Moderna uses the LNP adjuvant activity. When you compare these two together, we asked, who makes better T-cell responses? We compared using LNPs, 100% modified, gave decent T-cell responses, but unmodified gave much better, much more potent. That's why the unmodified likely worked better in cancer vaccines. The problem is when you make this number of T-cells with unmodified RNA, you don't make good antibody responses. That's why the clinical trial failed. We tried this in a monkey model for an HIV vaccine, and we again mixed different concentrations of modified and unmodified mRNAs. We saw the same thing. 100% modified gave relatively low T-cell responses. As you increase the modification, you increase the amount of T-cell response. We compared that to a chimp adeno vaccine. This is similar to what Oxford developed. The responses for the chimp adeno were actually a little bit lower, but statistically they're about the same. It tells you that depending on how you make the mRNA vaccine, you can change the characteristics. You can change the amount of antibody or the amount of T-cell responses. To address this, we tried a different approach. We asked, could we add cytokine adjuvants to change the vaccine? In this case, we're looking at IL-12. IL-12 is a potent T-cell-inducing cytokine. We did a simple experiment. We took a microgram of ovalbumin-encoding RNA and mixed it with a microgram of either empty or IL-12-containing RNAs and put them into mice that had 1,000 CD8 T-cells that had a T-cell receptor for ovalbumin. These weren't naive mice, but they only had 1,000 antigen-specific cells. We could easily measure their expansion. The addition of IL-12 enormously, about tenfold, increased the activation and expansion of these CD8-positive T-cells. They were present everywhere in the spleen and lymph nodes. Interestingly, they were effector T-cells. IL-12 induces effector T-cells. You can see here the KLRG1-positive, 127-negative. Those are effectors. Memories are 127-positive, KLRG1-negative. We greatly expanded the number of effector T-cells without affecting memory T-cells. We're using this now for new types of cancer vaccines that will be used in patients who have genetic deficiencies associated with cancer. BRCA is the most commonly one known. The idea here is that you treat people before they develop cancer. We know that it's five or ten years that cancer cells first start to appear before you've got full-fledged large tumors that impair function. If we treat these people maybe every five years with a vaccine that only makes effector T-cells, we'll clean out, clear away, kill all of the transformed cells and maybe completely prevent cancer from ever appearing in these patients. Another critically important thing is that adding IL-12, unlike unmodified RNA, doesn't lower and actually improves the antibody response. This is with IL-12 addition. It's about over half a log higher antibody titers. Now we've made a vaccine that gives better T-cell responses and better antibody responses. At Penn, we've got a bunch of different vaccines in clinical development. These include many different pathogens including malaria, TB, HIV, HCV. We have vaccines for food allergies like peanut allergies, have vaccines for environmental allergens like dust mites, and we've got a number of vaccines for autoimmune diseases that are all in development. Katie what I see as the future of mRNA therapeutics. If you ask anybody who develops nucleic acid or viral therapeutics, the biggest problem is targeting. It's getting the nucleic acid or the virus to the cells of interest. Lipid nanoparticles go to the liver. They also go to dendritic cells that make them a good vaccine. But what if you wanted to send them to the heart or the brain or the lung or the bone marrow? They don't go there. We developed a way of targeting lipid nanoparticles by adding targeting molecules to the surface. It's a little complicated chemistry that I don't need to go into, but we essentially add any targeting molecule to a peg lipid that sits in the lipid nanoparticle. It doesn't change the morphology of the lipid nanoparticles. It makes them a little bit bigger. The surface charge remains the same. All of those are critical. Most importantly, it doesn't impair function. In red, these are PCAM expressing cells. The LNPs are labeled with an anti-PCAM antibody. They bind very well. They make luciferase. When we add them to endothelial cells, only when you add PCAM, the endothelial cells take up the LNPs and make the RNA. Not only can they target, but they're functional. We injected these into a mouse and then analyzed where the activity was. In the absence of targeting, all the activity is in the liver. When they're PCAM targeted, we've switched activity to the lungs. There's about equal amounts of activity in the lungs. We switched where these LNPs are targeted to. Here, we can efficiently deliver to lungs. As an immunologist, I had an interest in modifying immune cells. T-cells, in particular, were interesting. They could be modified in a variety of ways. You could deliver antigen to them as a vaccine. You could deliver cytokines to change their function. You could deliver CAR molecules to kill tumor cells. The list goes on and on. There's a difficulty with T-cells. They don't have endocytic activity. What that means is that they're unable to take up nucleic acid particles. Our idea was we could combine targeting with getting into the cell by targeting a molecule that endocytoses after it's bound. That's exactly what we saw. When we added targeted LNPs to purified CD4 positive T-cells, in the absence of targeting, we had no uptake. With targeting, we had very high levels. Not only did we bind the cell, but we allowed the cell to take up the LNPs and translate the mRNA. We injected these into animals. We saw an increase in spleen activity with targeting. The spleen has about 20% CD4 positive T-cells in them. We did something interesting. We re-imaged the animals after we removed all of the organs on the previous slide. When we did that, all of a sudden lymph nodes lit up. These are periodic and inguinal lymph nodes. What this means is that we gave lipid nanoparticles targeted to CD4 T-cells intravenously into the circulation. They escaped from the circulation, went into the tissues, into the draining lymphatics, back to the draining lymph nodes where they targeted CD4 positive T-cells, were able to be taken up, and the RNA was translated. We could complete that entire process. Here, we purified the CD4s and showed that the activity is there. We were able to deliver nucleic acid therapeutics to a particular cell type in vivo. We used a model system known as the Cre-Lox system. What that involves is you put a stop codon in front of a fluorescent protein, and you put what are known as LOXP sites, specific DNA sequences. Then you deliver a Cre-recombinase mRNA. The Cre-recombinase cuts out the stop codon, and it makes the cells express the fluorescent protein. When we did that, we saw enormous levels of fluorescent protein in both splenocytes and lymph nodes. This is the fluorescent protein expression. There's almost nothing without targeting, very high levels with targeting. But from an HIV point of view, there was something more important. This axis is an activation marker. HIV forms latency in unactivated resting CD4-positive T cells. If you're delivering a therapeutic for HIV, it has to target, be taken up, and be translated in resting cells. That's exactly what we saw here. The resting cells had equal levels of gene recombination. This is now in a macaque experiment looking to cut HIV out of genome of latently infected cells. We saw very high levels of gene recombination, both in spleen and lymph node. We've looked at a variety of other tissues and see the same thing. Our conclusions were that we could figure out how to target specific cell types in vivo. We then went on to test another model system. If you come from Penn, Penn clinically developed CAR T cells. The first clinical trials were done there. The first two drugs were FDA approved at Penn. Penn has a close relationship to CAR T cells. What a CAR T cell is, it's an altered CD8 killer cell. CD8 killer cells recognize a peptide derived from a virally infected or oncogenic cell in association with the host MHC. They're very effective at killing. The problem is just about everybody in this room has a different peptide and many different MHCs. They're not translatable from one person to the next. What a CAR T cell does is it puts a piece of an antibody that recognizes a tumor associated antigen on the surface of the T cell. That way, you can make killer T cells that kill any tumor or any cell with that antigen. CD19 was the first one to be developed. The problem with CAR T cells, this is what it takes to make a CAR T cell. You start with a patient. You leukophores the patient. That's a machine that has a centrifuge in it that spins their blood, takes out the white cells, gives everything back. It does that over and over and over over a few hours, and it takes out a couple billion T cells. You then have to infect those T cells with lentiviruses in culture. You then have to stimulate them for 10 days to expand how many there are. Then you can give those back to the patient. Then about 70% of the time, they kill the tumor and cure the patient. The problem is it costs half a million dollars a dose because it's 10 days in a very fancy lab under GMP conditions. These sites are available in the U.S. and in Europe. There's one in China. There's, I think, one in South America. There's none in sub-Saharan Africa. They're very limited in where they are because of the cost and the expertise needed to do it. We had the idea, since we can now target cells in vivo, could we make CAR T cells in vivo? The thought was, well, if we targeted the LNP to bind to all T cells and then we put a CAR molecule in the mRNA, the LNPs would be taken up by the T cells, the RNA would be translated, and they'd put the CAR on the surface of the T cell. We chose a CAR that recognized activated fibroblasts and a model of cardiac fibrosis. I apologize for the echocardiography readings, but what you need to realize is the gray are normal animals. These are just different measures of heart function. As an internist, this is what I look at. This is how much blood is pumped out by the heart. In a normal heart, it's about 70 percent of the blood is pumped out. When a heart becomes fibrotic due to hypertension, that reduces. The lower it goes, the worse off the patient is. We treated these mice with a single dose of LNPs that targeted T cells, and we returned their heart function to normal. Their ejection fractions returned to normal, the size of their ventricles, everything returned to normal. We stained the heart muscle for fibrosis. The normal animals, you see very little. The fibrotic animals, you see all this red fibrosis. One treatment with LNPs essentially cured these animals. Instead of a half a million dollars, 10 days, specialized facilities, now it can be an off-the-shelf drug. Somebody comes in with cardiac fibrosis. Heart disease is the number one killer in the world. They get an injection. They go home. Their heart is better. This can also be expanded to many other diseases and many other applications. The other thing I wanted to tell you about is bone marrow stem cell targeting. To me, this is one of the most critical applications for RNA LNPs. Currently, there are thousands of genetic diseases of the bone marrow. These include all of the immune deficiencies, SCID. They include the thalassemias. Most importantly, it includes sickle cell anemia. Sickle cell anemia has 300,000 people a year are born with the disease. They're mainly in sub-Saharan Africa, but they're present throughout the world. In the U.S., they're about to approve, FDA approve, a gene therapy for sickle cell. The problem is, is it costs about a million to four million dollars per person. You multiply a million times 300,000, that bankrupts the world. We need better ways of treating genetic diseases of the bone marrow. The problem with bone marrow stem cells are their rarity. This is the number of cells in the bone marrow of a mouse. This is the number of repopulating bone marrow stem cells. It's a tiny fraction. Hitting those rare cells is very difficult. We took the same approach that we did for CD4. In this case, we used a marker known as CD117 or CKIT. With a single treatment of mice and we're delivering a fluorescent protein, we were able to hit about 90% of bone marrow stem cells. Incredibly efficient. We went back to that Cree lox mice and we followed them over time. The animals got a single treatment here. Then we followed their blood over time, looking for a fluorescent protein from the Cree enzyme. After a few weeks, 100% of the red cells, white cells, and granulocytes were fluorescent. We had 100% efficiency of gene editing the bone marrow of these mice. For stem cell transplant people, they only believe stem cell results if you do a secondary transplant. What that means is you take the bone marrow from these mice and you take it out. You irradiate a new group of mice to kill their bone marrows. Then you put the new bone marrow cells in. When we did that, within a couple weeks, 100% of red cells, white cells, and granulocytes were all gene edited. That meant we had 100% efficiency at gene editing bone marrow stem cells in vivo. That's critical for certain diseases. For diseases like SCID and other immune deficiencies, if you fix half a percent of the stem cells, you cure the disease. For sickle cell anemia, you have to fix 25% of all of the repopulating bone marrow stem cells. With an efficiency of 100% targeting, then however well we can gene edit will likely give us potentials for cure. This is now a cure where we can line people up on the street, give them a single injection of RNA LNPs, and cure their disease. No million dollars, no fancy lab facilities, just off-the-shelf injections. It can also be used for a variety of other things. We do bone marrow transplants for a number of different types of cancer. You give people high doses of chemotherapy and sometimes radiation to prepare them for this. There's a mortality rate of around 5% from that treatment. Now we can deliver toxic genes to bone marrow stem cells and selectively kill them. Mortality will be much, much lower. We can deliver therapeutic proteins to stimulate granulocytes or platelets or other needed cell types. We can deliver really any kind of protein. We're also able to target a variety of other cells and tissues. We can target brain. Katie talked about treating strokes. We're now setting up to do studies on large animals, delivering therapeutic proteins targeted to the brain. We can deliver to the heart. We can deliver to lungs. We can deliver to T-cells. We can deliver to any immune cell. We continue to expand what we can target. In theory, someday gene therapy might be as simple as walking into a doctor's office, having them take a vial out of the fridge, injecting it into a person, and curing the disease. RNA therapeutics has enormous potential for vaccines, for genetic diseases, for therapeutic protein delivery, for the treatment of a variety of diseases. I joke with my lab people, we haven't thought of everything that we can do with RNA. That's their job. I've done my job. The future of RNA, I think, is really going to be enormous. I need to thank all of the people involved. Of course, Katie, who I started all of this work. Norbert Pardee and Hamida Parhiz in my lab were the leaders for vaccines and targeting technology. People from Acuitous Therapeutics made the lipid nanoparticles. Vlad Muzhikantov's lab developed the targeting technology that we use for all of our in vivo targeting. I have to mention John Epstein because he's the CSO of my university. The many other labs that we've worked for. Thank you very much.