Scientist Stories: Patrick Hsu, Modifying Genomes with CRISPR

Hi, good morning. So you guys have heard so far about reading genomes, about synthesizing them. I'd like to tell you today about how we can modify genomes inside of living cells. So we've known for quite a long time now that DNA really underlies the basic building blocks of life and cell function. So having ability to rapidly and precisely manipulate this genetic information would be really broadly useful for a number of fields, such as in fundamental biology research. For example, looking at the function of ion channels that might underlie neural function, the cellular learning and memory, or even in pathological states, the development of diseases like autism or schizophrenia. If genetic mutations also seem to be able to lead to disease, There are these tantalizing possibilities for therapeutic applications as well, where you could directly correct disease-causing mutations with healthy DNA sequence. Finally, being able to change and manipulate DNA or introduce non-native DNA would allow us to endow biological systems with synthetic function for biotechnology. For example, we could reprogram or refactor simpler life forms for useful tasks, such as factories for producing biofuels or genetically engineered plants for improved agriculture. The first steps that molecular biologists took to really realize this vision in many ways began in the early 1970s when Herb Boyer, then at UCSF, working with Stanley Cohen at Stanford, first reported the construction of functional organisms that were able to combine and replicate information from different species. The development of what are called restriction enzymes was a watershed moment for biology. For the first time, we scientists were able to manipulate DNA molecules and harness them to create novel medicines and biotechnology, such as synthetic insulin, synthetic growth hormone, and so on. But restriction enzymes typically recognize a specific 6-base-pair sequence. Okay, so how could we have tools like restriction enzymes that would be able to recognize any DNA sequence and do cut and paste of DNA inside of living cells. So around the same time, around 1975, there's this growing realization that DNA damage can stimulate DNA repair. This was first discovered partly on the basis of UV irradiation experiments, also known as something that happens when you're tanning on the beach outside maybe my lab in sunny San Diego. About 10 years later in the late '80s, there's this experiment that was done by Jim Haber then at Brandeis who showed this really interesting phenomenon. If you could create a targeted double-stranded break in a special place of the yeast genome, you would stimulate repair only at that site. So putting all these disparate pieces together in the mid-'90s, Srinivasan and Chandra Segaran created the first targeted molecular scissors. By fusing these molecular scissors to programmable zinc finger proteins. And this idea of making targeted breaks in the genome to stimulate DNA repair is really the main underlying concept behind sort of modern genome editing technologies. So the idea here is that targeted DNA breaks can facilitate alteration of the genome. If you take some stretch of DNA, it's broken, the cell's going to attempt to stitch its DNA back together, okay? And it can do those through one of typically two major repair pathways. One is known as non-homologous end joining. And now that's quite a mouthful, but what that really means is the cell can summon a host of proteins that have high affinity to these broken DNA ends, and it'll try to stitch the DNA back together. Usually it does so perfectly, but oftentimes this can happen in an error-prone way where you can introduce insertions or deletions at that break site. If this happens in the right place inside of a gene, that can scramble the protein or truncate it, leading to loss of function. So you can knock out the function of that gene and essentially turn it off. The second possibility is actually much more powerful, and actually how the cell can choose between one versus the other pathway is still really under active investigation. That pathway is known as homology-directed repair, and it takes advantage of a very special property of human cells. That for every copy of a gene, we actually have two. So when you break one copy, oftentimes the cell can look to the sister gene for repair. And as bioengineers, we can trick the cell by introducing in synthetic repair templates that look a lot like the sister gene, but in fact carry the particular mutation that you want to introduce. This can allow us to introduce mutations that may be found in disease to try to study them. We might even be able to correct mutations for therapeutic applications. So this idea of making these targeted breaks, you can first have programmable DNA binding domains that could target anywhere in the genome. That was the real challenge. We've known about, basically, proteins that can cut DNA for quite a long time. So how could we target and also easily retarget desired regions in the genome? So we looked to nature, and one of the first solutions for this were actually zinc finger proteins. Which you could think of like Lego blocks. These 3-base Lego blocks then recognize 3 bases in the genome at once. And just like beads on a string, you could string together these 3 Lego blocks together to recognize 3, 6, 9, and more DNA bases. However, in order to recognize any DNA sequence, you would need a lot of different kinds of these modules. So how do other organisms solve these problems? So if you look in nature, actually, it turns out a breakthrough came in 2009 from plant pathogens. Where there are actually these little module proteins that were able to recognize one DNA base at a time: A, C, T, and G. So now you'd only need 4 of them, and again, you can mix and match them to recognize any sort of DNA sequence that you might want to target. You could take these zinc fingers or tail proteins, fuse them to molecular scissors, and make targeted DNA breaks. So that sounds great. It seems to work, except it actually takes really laborious expensive and specialized protein engineering. It was just kind of difficult to use, kind of like a technology many of you may have used in the past. So we started to look for simpler solutions, and we were inspired by, basically, bacterial adaptive immune systems. So this is something that allows bacteria to defend themselves against invading bacteriophages, which are the bacterial forms of viruses. CRISPR is a system that essentially is like a detective in your favorite cop show. It can take fingerprints at a crime scene, put those into a fingerprint database, and use that to ID criminals during later sort of things that they might do. The CRISPR system similarly has a host of proteins and associated small RNAs that allows it to capture molecular fingerprints from the invading foreign phaged invader and actually put that into a special region of bacterial genome that we call the CRISPR locus. And so that allows it to essentially fingerprint the foreign invader to have a molecular memory. So let's take a closer look at how that happens. If you take some stretch of DNA, shown here in blue, Cas9, which is the targeted nuclease of the CRISPR immune system, can recognize these short sequences called PAMs throughout the genome. They're essentially these targets. You can think of it like stop-and-frisk police tactics, okay? So it finds these PAMs throughout the genome and basically opens it up, okay? And it interrogates this sequence for complementarity between that target DNA and a short guide RNA shown here in red. So you've just heard about how we can synthesize DNA. You can change this roughly 20-base-pair sequence from something that targets the phage DNA to something that targets the human genome itself, okay? And when this happens, Cas9 is able to swing in the two blades of its scissors, rough C and H and H, to create a double-stranded break in the DNA that can then lead to error-prone and precise repair to do the actual genome editing, okay? So if we have these kinds of technologies for modifying genomes, one of the first applications that comes to mind is gene hunting. So let's start at the top of the wheel. We have people who have genetic disorders. You can use technologies like the Illumina HiSeq or Oxford Nanopore in order to sequence people to try to find candidate disease genes that are associated with the disease state. We take a bunch of people like, let's say, this healthy audience. Apart from natural human genetic variation, most of you will look pretty healthy, right? But if you start to sequence people that have a very specific kind of disease, you'll start to find these mutations shown in red scattered throughout the genome. Which ones of these are disease-associated? Which ones are actually causal? Using tools like CRISPR, we can actually get from correlation to genetic causation. So back to the wheel, we have these candidate disease genes. Using CRISPR, we can now start to create disease models to try to figure out if we recreate these genetic mutations, do we see disease phenotypes, okay? And for example, if we're trying to study ion channels, we could, for example, mutate those ion channels in brain cells and see, do they fire differently? Do they wire differently? Do they talk to each other differently? Do we see the kinds of problems that we see actually in patients? But oftentimes when we're modeling disease, you want to look at something at the physiological or behavioral level, right? And to do that, you can make animal models such as rodent or non-human primate models to better get a sense of things like cognitive disorder. Now, if you have this genetic causality though, the tantalizing possibility is, could we actually go ahead and try to fix these directly in humans? Doing these kinds of genetic surgeries is a really exciting vision where you could imagine taking the CRISPR components, the Cas9 and the guide RNA, and directly putting it into a patient, either systemically to target diseases that affect an entire organ or tissues throughout the body, like muscular dystrophy, or in a more targeted way, such as for treating hemophilia, where your liver is not producing enough blood clotting factor, so you could go directly into the hepatocytes of the liver. However, it turns out going in vivo, directly into the body, is actually really a challenging task. It's one of the major central themes of the drug delivery field that's been evolving over the last few decades. However, taking cells out of the body to do ex vivo therapy is another much— also very exciting near-term opportunity where you can imagine taking cells out of the body, modifying them in a dish, and putting them back into the patient. This is really combining CRISPR with some of the most exciting things in biotechnology today, which is called cancer immunotherapy, where you could take T cells out of the body, delete the inhibitory signals that a cancer can use to try to shut down the immune system of the host, and essentially supercharge them in addition, put them back into the patient. So these kinds of ex vivo therapies or in vivo therapies are really somatic modifications. They're modifying basically organs or tissues that already exist in an adult. So if they happen at the DNA level, they would be permanent changes, but they wouldn't be heritable. And so you could also imagine potentially using CRISPR to make heritable changes directly in an embryo. So you can imagine the kind of controversy that this has caused. How do we draw the line between what is a disease mutation or one— what's another trait, right? How do we actually define a disease? Genes also exist in a very complex sort of environment, society within the cell. How can we actually predict the changes that we're going to make? This is really sparking a very vigorous and healthy international debate today. Another thing you can imagine is starting to try to make better models of very uniquely human diseases, to start to look at, for example, cognitive disorders or neuropsychiatric disease like autism or schizophrenia in nonhuman primate models. Another exciting thing you can do with CRISPR is actually not to just manipulate individual cells or organs, but actually to try to manipulate entire populations at once using concepts like gene drives. You can use CRISPR cassettes that are able to spread themselves throughout genomes to break the barriers of Mendelian inheritance so that you can quickly spread mutations or genes throughout a population. Could we use this to, for example, wipe out malaria-carrying mosquitoes? Could we predict the consequences or results of doing these kinds of population-level changes? These are all really interesting questions for the future, but what I'd like to sort of close with is not really the genome editing, but actually what we can do beyond modifying DNA, making non-permanent changes within a cell, what I like to call genome engineering. So let's go back to this stretch of DNA that we want to target, okay? You could, for example, target it with the Cas9 nuclease and make it break, stimulate repair, and make the particular type of edit that you want. But a second, also very powerful possibility is actually neutering the molecular scissors so now it can bind to the targeted region, but not cut. And if you could do that, you could turn the scissors into essentially a molecular GPS that will allow you to bring some sort of function to that region specifically. So this effector domain will have some sort of desired function. It could be a nuclease for cutting, but it really could be anything else. And this is really driving the development of this new idea of proximity enzymatics, where you might want to target some function to a locus of interest. So why is that interesting? If we could fuse Cas9 to fluorescent proteins, for example, to illuminate the inner life of the cell, this will allow us to start to track things, for example, sense things in living cells to look at cell division or how a cell may respond or behave in response to its environment. We could also not just look, but we can also touch, okay? We can manipulate dynamic cellular responses by changing how the cell can convert its DNA to RNA and eventually to proteins to orchestrate cellular logic. By manipulating these dynamic cellular responses, we can really start to understand the sort of overarching function of cells and how they can sort of respond to their environment. Finally, you could also imagine connecting CRISPR with epigenetic enzymes that can change how a genome is architected, how it's folded, how it's accessed and regulated. That's actually what makes a cell a heart cell or a lung cell or a liver cell or a neuron, as opposed to the DNA, which is actually largely the same throughout your body. Okay. So if you could imagine doing these kinds of more intelligent genetic surgeries, would we be able to endow our bodies with synthetic capabilities? Okay. Imagine a future where your cells have conditional preprogrammed logic where you could, for example, activate anti-inflammatory pathways during an allergic reaction or stimulate inhibitory neurons to quiet epilepsy. Having these kinds of intelligent synthetic functions by modifying genomes is really, I think, some of the future of what we can do with genome engineering, and I really look forward to collectively exploring these ideas and executing on them in the future. Thank you very much.