The Peter Attia Drive - #323 - CRISPR and the future of gene editing： scientific advances, genetic therapies, disease treatment potential, and ethical considerations

00:00:00.000 Hey, everyone. Welcome to the Drive podcast. I'm your host, Peter Atiyah. This podcast,

00:00:16.540 my website, and my weekly newsletter all focus on the goal of translating the science of longevity

00:00:21.520 into something accessible for everyone. Our goal is to provide the best content in health and

00:00:26.720 wellness, and we've established a great team of analysts to make this happen. It is extremely

00:00:31.660 important to me to provide all of this content without relying on paid ads. To do this, our work

00:00:36.960 is made entirely possible by our members, and in return, we offer exclusive member-only content

00:00:42.700 and benefits above and beyond what is available for free. If you want to take your knowledge of

00:00:47.940 this space to the next level, it's our goal to ensure members get back much more than the price

00:00:53.200 of a subscription. If you want to learn more about the benefits of our premium membership,

00:00:58.020 head over to peteratiyahmd.com forward slash subscribe. My guest this week is Feng Shung.

00:01:07.880 Feng is a professor of neuroscience at MIT, as well as an investigator at the Howard Hughes

00:01:13.360 Medical Institute and a core member of the Broad Institute of MIT and Harvard. He earned his bachelor's

00:01:19.540 degree in chemistry and physics from Harvard University, after which he went on to earn his

00:01:23.260 PhD in chemical and biological engineering at Stanford University, where he worked with one

00:01:28.040 of our previous podcast guests, Carl Deseroth, in developing the technique of optogenetics.

00:01:34.480 From there, he returned to Harvard as a research fellow before starting his own research lab and

00:01:38.760 professorship at MIT in 2011, where he subsequently contributed mightily to the development of the CRISPR-Cas9

00:01:45.620 system for gene editing. Feng has earned numerous honors and accolades for his work,

00:01:50.700 including being selected for membership in the National Academy of Sciences, the National Academy

00:01:55.820 of Medicine, and the American Academy of Arts and Sciences. He is also a fellow of the National Academy

00:02:01.840 of Inventors. In this episode, we explore the origins of CRISPR and discuss Feng's early work

00:02:08.400 in optogenetics at Stanford. We discuss the foundations of gene editing, discussing the challenges

00:02:13.800 and breakthroughs in the field, and how CRISPR revolutionized the process. We talk about the

00:02:18.480 practical implications of CRISPR, such as its potential to treat genetic diseases, the importance

00:02:23.600 of delivery methods, and the current success and limitations in targeting cells like those in the

00:02:29.440 liver and eye. We discuss the ethical considerations of gene editing, touching on the debate surrounding

00:02:34.940 germline modification. Finally, we reflect on Feng's personal journey, the significance of mentorship

00:02:40.680 and education, and where he sees potential for the future of science and genetic medicine.

00:02:45.920 So without further delay, please enjoy my conversation with Feng Shung.

00:02:55.260 Hey Feng, thank you so much for detouring your trip and coming through Austin. I've been really

00:02:59.860 looking forward to sitting down with you for, frankly, about a year. So this is a topic that I don't think

00:03:07.180 there's anybody who's heard this podcast who hasn't heard the term CRISPR, but I think very few people

00:03:12.640 can actually explain it and explain what a powerful tool it is. But I do think that before we get there,

00:03:18.180 it would be really helpful to kind of understand a little bit more about your journey on one hand,

00:03:22.460 and then the journey of gene editing as a parallel. Let's start with yours. You and I overlapped a little

00:03:29.280 bit because, I mean, not temporarily, but you were at Stanford. Were you a postdoc in Carl Deseroth's lab?

00:03:34.660 Well, first of all, thank you for having me be on this podcast. I've listened to your podcast

00:03:39.100 on and off, especially when I'm running around or exercising, and it's always, I learn a lot

00:03:44.700 from listening to the podcast. Thank you for having me here. So back to Stanford, I was there as a

00:03:50.480 graduate student. I was in the lab of a researcher named Carl Deseroth. I was there for five years.

00:03:56.580 That's right. You did your PhD there with Carl. Now, as you know, Carl and I were classmates.

00:04:02.260 Carl's been on this podcast. And so maybe folks who either didn't listen to that podcast or who did,

00:04:08.540 but have forgotten, give us kind of a quick summary of the type of work that you and Carl did.

00:04:13.960 When I was working with Carl Deseroth, we developed a technology called optogenetics.

00:04:18.600 And it's a way of studying brain cells in the brain, how they are connected together and how they

00:04:25.680 mediate memory, mediate different types of physiological function. The way it works is that

00:04:31.600 we took a gene from a green algae. And this is a gene that senses light and converts it into

00:04:39.500 electrical current in a cell. So we can put this gene from the green algae right into the brain cells

00:04:45.660 in a mouse. And we can shine blue light or a yellow light and control the brain activity in

00:04:51.740 these mice. So for example, if you wanted to study sleep, you can put this gene into different groups

00:04:58.840 of cells in the brain and stimulate them. And you can find out which ones of these are controlling

00:05:04.540 wakefulness or which ones are causing the mouse to become more sleepy. So if you do this systematically

00:05:10.220 one by one from one type of cell to another type of cell, you can gradually start to

00:05:15.380 put together a picture of how the brain is wired together. And then also what are the different

00:05:20.660 components that govern all sorts of behaviors from sleep and wakefulness to thirst and hunger

00:05:27.860 to memory and even to motivation and happiness. So it was really fun to be at Stanford and working

00:05:33.400 with Carl.

00:05:34.420 To me, the thing that always stood out about the technique was just the resolution. I don't know

00:05:38.460 what a great analogy would be, but it was resolution at the level of the word rather than the page.

00:05:44.300 If you were thinking about a book, for example.

00:05:47.000 Right. What is incredible about these algal proteins is that they are very, very fast.

00:05:53.020 So you can show the way the brain cells are able to signal to each other at the action potential

00:05:59.560 level. So action potentials are these individual signals that they're basically like the phonemes

00:06:05.620 of the speech that one neuron speaks with another neuron. And you can control it at every single

00:06:11.300 phonemes level. And that is pretty cool.

00:06:14.560 And since we're going to be talking a lot about gene editing, what was the technique that you

00:06:18.280 guys used to insert those algal genes into the brains?

00:06:21.780 The way that you would put a gene into the brain is usually by using a virus. So this is a virus

00:06:28.740 that exists in nature, but we have engineered it by removing everything that is pathogenic about

00:06:34.200 the virus and then replacing those pathogenic genes with the gene that we're trying to put

00:06:39.320 into the brain. So in this case, it's the gene from the green algae. And by injecting the virus

00:06:45.280 into a brain area that you want to study, the virus will infect all of the cells in that region and

00:06:51.940 then make those cells begin to produce this algal protein. So once the neuron starts to carry this

00:06:57.360 algal protein, it becomes light sensitive. So you can turn blue light on it and be able to stimulate it.

00:07:04.080 Yeah. Amazing. What year did you finish your PhD?

00:07:07.240 In 2009.

00:07:08.720 Okay. And then you went to MIT or do you went to the Broad?

00:07:12.260 Then I went to Harvard. I was there for about a year. And then I went to MIT and Broad right after

00:07:17.880 that.

00:07:18.780 Okay. And then your attention there sort of turned. You obviously worked some on optogenetics,

00:07:23.940 but what else did you pivot into?

00:07:25.740 As I was working on optogenetics, and especially toward the end of graduate school, I began to

00:07:32.000 realize that one of the biggest bottlenecks facing optogenetics is our ability to insert the algal gene

00:07:41.180 into specific places in the genome. And the reason for that is because in order for us to study

00:07:47.000 different types of brain cells, we need to have very precise targeting of different types of brain

00:07:52.700 cells. Brain cells are not just one type. Neurons is not a single type. There are probably

00:07:56.620 hundreds of different types of brain cells. The way that they're defined is based on their

00:08:01.960 molecular property. So each brain cell, even though they all share the same genome,

00:08:07.080 they have different sets of genes that are turned on. That's why brain cells that control pain

00:08:12.360 sensation versus brain cells that are involved in Parkinson's disease are different. So the way

00:08:18.220 that you would target one or another type of brain cell is by figuring out what are the molecular

00:08:23.960 signatures of that cell. So if you know that gene A is turned on in that brain cell and not in another

00:08:30.960 type of brain cell, then you can insert this algal gene into the region that's controlling gene A.

00:08:38.160 That way it will only get turned on in the first type of neuron. And the way to insert this gene

00:08:44.080 into that precise place in the genome required gene editing. And it was really hard to do at the time.

00:08:51.940 And so I thought maybe if I wanted to get optogenetics to become even more powerful and useful,

00:08:56.920 we need to make gene editing more easy to use. So by the time I went to Harvard, I began to focus more

00:09:03.820 on trying to figure out how do you more easily be able to modify the genome?

00:09:09.780 Yeah. So this is to me where the story gets so interesting because you stumble across a problem

00:09:15.100 that you're trying to solve because of your passion. You've already alluded to why this is so difficult.

00:09:21.140 And I guess maybe just go back and explain something so that the listener understands.

00:09:25.140 Why was it easy to do what you did in Carl's lab, relatively speaking, where you're putting

00:09:31.640 an entire gene into presumably an adenovirus and letting the adenovirus infect the neurons and stick

00:09:38.840 a whole gene in? Why is that a different problem than the one you just described that you started

00:09:43.440 to solve at Harvard? Make sure everybody understands that distinction.

00:09:46.500 The work that I was doing when I was a graduate student with Carl Diceroth is that we were simply

00:09:50.920 trying to insert a gene into brain cells.

00:09:53.220 Meaning you don't get to care where it goes specifically.

00:09:57.100 We can get it into the rough area in the brain, but there are many different types of cells there.

00:10:02.280 And so we weren't as precise in our ability to target those cells. And we also developed some tricks

00:10:09.360 to be able to get it into a specific type of cell, but that was only limited to mice because we can

00:10:15.680 genetically modify mice. And it will take a long time. It will take a year or two years to be able to

00:10:21.480 make those mice available to engineer them, but it wasn't generally applicable. And so especially if

00:10:28.800 you think about how to turn optogenetics into a therapeutic to use in a human, we certainly

00:10:35.640 couldn't go in and use those transgenic technologies to make it work in the human brain. So this was a

00:10:42.020 problem. Okay. So let's keep that story on the side for a moment and let's tell in parallel another

00:10:49.240 story, a story that started long before you were even in college, right? When you were a young kid.

00:10:55.160 Right. And it stems from an observation that grew out of a discovery in sequences that existed in

00:11:02.120 bacterial DNA, a certain type of repeating structure. They had all these interesting

00:11:06.900 characteristics. So let's go back to the eighties and tell a little bit of that story. And

00:11:11.360 obviously I wouldn't be going through this if it wasn't going to quickly converge with your life,

00:11:16.100 change the direction of your life, but let's go back to the eighties.

00:11:18.400 I was born in the eighties, but what was really interesting that you're pointing out is that back

00:11:24.420 in the eighties, there was a group of Japanese researchers who were just looking at DNA sequences

00:11:30.380 of bacteria and they were looking at E. coli. And what they found is that within some of the genomes,

00:11:37.180 the DNA sequences of these bacteria, there are these regions that are very repetitive. So it would

00:11:42.520 just repeat over and over and over again. Normally genome sequences are not repetitive because

00:11:48.240 they encode genes and different genes. But here they found that there are these repeat sequences

00:11:53.420 that are all grouped together. So they're clustered and they are not tandem repeats. So it's not repeat

00:12:00.420 one next to each other, but they're interspaced by a short fixed length gap. And so it's basically A,

00:12:07.920 B, A, C, A, D, A, E, A, G. And so it just continues to repeat itself. But in this

00:12:17.900 regularly spaced pattern, and when they first found it, they had no idea what this sequence

00:12:23.580 was all about.

00:12:24.580 And there's something else about those repeating segments that was quite interesting

00:12:28.080 as well, which is in each direction, whichever way you read them, they were the same.

00:12:33.080 Right. So they're called palindromes. DNA is double stranded. So there's a top strand

00:12:38.420 and there's a bottom strand. And so this is why they look like a double helix. And so they twist

00:12:43.080 it in turn. What is interesting is that when you read these repeat sequences from the top

00:12:47.420 and you read in the reverse way on the bottom, they're almost the same. So they're a palindrome.

00:12:53.220 So you had these palindromic clustered repeats that were interspaced. And if you say that in the

00:13:00.280 right way, you get a few letters. C-R-I-S-P-R. CRISPR. So the name CRISPR, that term was actually

00:13:10.000 coined nearly 40 years ago, right?

00:13:13.000 So CRISPR is really a brilliant acronym. And so C-R-I-S-P-R stands for exactly how these repeats

00:13:20.280 look. Clustered, regularly, interspaced, short, palindromic repeat. So CRISPR. It's really

00:13:29.020 brilliant and it's very catchy. But this name wasn't the name that was given to these repeats

00:13:34.400 back in the 80s. In fact, it had many different names.

00:13:38.440 C-R-I-S-P-R. Did that name not come along until the 90s?

00:13:40.320 C-R-I-S-P-R. It didn't come until the early 2000s.

00:13:42.160 C-R-I-S-P-R. Oh, really? Okay.

00:13:42.820 C-R-I-S-P-R. Yeah.

00:13:43.280 C-R-I-S-P-R. Before Francesco Mojito.

00:13:46.160 C-R-I-S-P-R. Mojito came up with that name, but I think that was in the early 2000s.

00:13:49.820 C-R-I-S-P-R. Okay.

00:13:50.000 C-R-I-S-P-R. Now, if my reading of the history is accurate, understandably, scientists in the 80s

00:13:57.400 90s focused on the repeating segments. Obviously, any good scientist, I think, would look at that

00:14:04.340 and realize this is a very interesting observation. How do we figure out what it is? But the focus was

00:14:08.980 on the repeating segments, what you described as the A in the A, B, A, C, A, D. And I believe it

00:14:17.320 was Francesco who was the first to observe that actually what's interesting is not the repeating

00:14:22.540 segments. It's the seemingly uninteresting segments in between them that are different.

00:14:28.960 Exactly. So these CRISPR repeats, they have this conserved A sequence. And this A,

00:14:36.160 of course, repeats many, many times. So it's the most obvious thing. And usually when we are

00:14:41.300 looking at things, we look for things that have the strongest signal. So when you have 20, 30 repeats

00:14:46.680 of the same sequence, that's the strongest signal there is. But it turns out that is not an interesting

00:14:51.820 part. The interesting part is actually the non-repeating sequences that's interspaced

00:14:57.560 between pairs of these repeats. So for Francisco Mojica, he's a Spanish researcher. He's been looking

00:15:04.280 at bacteria and looking at sort of weird sequences for a long time. And what he did back in the early

00:15:10.780 2000s is that he took these non-repeat sequences and he just searched against viruses in the bacterial

00:15:19.200 world. A few of them he found matched virus sequences. And so that was really a breakthrough

00:15:26.100 because it started to highlight that maybe these non-repeating sequences are foreign to the bacteria.

00:15:33.780 They came from somewhere else and somehow bacteria acquired them into this repeat pattern. And that

00:15:41.100 really started to launch the CRISPR revolution because that observation and that inference allowed people to

00:15:48.940 start to realize that maybe this has something to do with how bacteria and the viruses are interacting

00:15:55.320 with each other.

00:15:55.920 Yeah. What's interesting when you again look at that story is that when he tried to publish that

00:16:01.880 finding, it was rejected by virtually every significant journal out there. They viewed this as either

00:16:07.880 incorrect, uninteresting, whatever. It was ultimately published, although I don't recall the name of the journal

00:16:12.680 that first published that finding, but it was many rungs below nature, science, et cetera. The irony of science

00:16:18.820 sometimes. Maybe folks, we can remind them just how the human immune system works because just like bacteria,

00:16:25.560 we are also encountering viruses all the time. Viruses do nothing good for us just as they do nothing good for

00:16:32.360 bacteria. We have a pretty negative relationship with viruses. All they want to do is use us as hosts to replicate their

00:16:38.500 genomic material. And in the process, they tend to make us sick. So obviously when a virus infects

00:16:42.640 a bacteria, it's just using its genetic machinery to replicate and it's going to kill the bacteria. So

00:16:47.840 as you said, the bacteria needs a tool to fight back. Now we do it through the creation of antibodies,

00:16:53.180 but how did this story continue to unfold? What were the bacteria doing or what more to the point,

00:17:00.320 what did Francesco realize was happening with this artifact left behind of viral DNA?

00:17:07.600 In the early days of CRISPR research, there were actually several different converging lines of

00:17:13.920 work. So there's Francisco Mojica, who's looking at these repeat sequences within the bacterial genome,

00:17:21.360 but then there were also other researchers who began to zero in on a group of genes. So these are

00:17:27.900 things that are telling the bacteria to make certain types of protein. There are these certain genes that

00:17:33.280 are right next to the repeats. And it took a while for people to begin to associate the genes and the

00:17:39.600 repeat to be together as one single system. But the people who are studying the genes realized that these

00:17:46.960 genes were carrying nucleases. So nucleases are proteins that usually go and cut up either DNA or RNA.

00:17:55.600 And so they initially thought that maybe these genes were involved in DNA repair. So for example,

00:18:02.080 Eugene Koonin, who is a really brilliant bioinformatician, he's been studying these genes,

00:18:06.640 and he really started to zero in on what biologically these genes may be doing. But the linkage with these

00:18:13.840 repeats took a little while longer to get associated. But it was really when the discovery that there are these

00:18:21.040 viral sequences in the repeats and that there are these genes that are associated with the repeat

00:18:27.280 that are involved in, say, DNA cleavage, that started to really put together a framework for thinking

00:18:35.280 that maybe this is a system where these viral sequences are working together with the DNA cleaving

00:18:43.200 proteins to go and recognize viral sequences and try to cleave viral sequences.

00:18:48.080 And we can put some names on these things so that people can start to see where the thing is going.

00:18:53.040 So you talked about these genes that were near, but at a distance from both the palindromic repeats

00:19:01.920 and the interspaced segments that we now realize were copies of viral DNA. And as you said, they coded

00:19:09.920 four proteins or enzymes called nucleases, which cut DNA. And similarly, we have helicases,

00:19:17.920 so you have certain enzymes that can unwind the DNA so that they can go in and cut. And these were

00:19:22.560 referred to as CRISPR-associated proteins, correct? Correct.

00:19:25.680 Which is abbreviated as Cas. So was Cas1 and Cas2 the first that were identified in that regard?

00:19:33.360 That's exactly right. So these genes that are right next to the CRISPR repeats,

00:19:37.840 they're called Cas proteins. Although they probably underwent many, many different renaming

00:19:43.440 over the course of two decades. But eventually, the community researchers were studying CRISPR proteins

00:19:51.120 and CRISPR-RNA, they came together and started to really curate these different genes. And so Cas1,

00:19:58.480 Cas2, Cas3, Cas4, et cetera, these are things that are kind of numbered based on, in part, the order

00:20:05.920 that they were discovered. So the most popular protein, or the most widely used protein now,

00:20:11.280 is called Cas9. And this is one of the Cas proteins that are found among an array of many,

00:20:16.880 many different CRISPR proteins. So yeah, so these Cas proteins work with the CRISPR-RNA. And the CRISPR-RNA

00:20:23.360 refers to these repeats, which are encoded in DNA, but they are made by bacteria into RNA. And those RNA

00:20:31.360 are called CRISPR-RNA. They don't encode protein, they simply are a short guide sequence that directs

00:20:39.440 the Cas protein to find the target virus sequence. So Cas protein, CRISPR-RNA, they together form a

00:20:46.960 complex that go and provide a defense function for the bacteria.

00:20:50.960 And whereas our defense against a virus is going to be making an antibody and or activating another

00:20:59.920 type of immune cell with an antigen receptor on it called the T-cell, the defense of the bacteria is

00:21:06.480 simply to cut the genetic material of the virus to kill the virus. Obviously, a much more directed

00:21:13.040 approach. Let's walk through two scenarios, first infection, reinfection, and explain the difference

00:21:21.600 between how the bacteria defends itself with special attention to the use of the CRISPR system

00:21:27.200 and the Cas9 enzyme. First infection now, out of the gate, you're in E. coli, I'm a bacteriophage,

00:21:33.280 you've never seen me before, I come along, I've just injected my viral DNA into you.

00:21:39.120 So CRISPR is an adaptive immune system. So it means this system can evolve with the bacteria to be able

00:21:47.360 to accommodate many, many virus infections. So when the virus, which is called a bacteriophage,

00:21:54.080 first infects the bacteria, it will inject its genetic information into the bacteria.

00:21:59.760 The virus is usually very powerful and very potent. It will probably wipe out

00:22:04.800 most of the bacterial population. But for a very small number of cells, maybe one out of a million,

00:22:11.040 the CRISPR system will successfully recognize a piece of the DNA of that virus

00:22:18.160 and begin to insert it into this repeat area in the CRISPR system.

00:22:23.360 And how long a piece is that typically? How many nucleotides in that piece?

00:22:27.120 It's usually 30 letters long. And that is enough for the bacteria to uniquely recognize the virus.

00:22:34.240 So during the first infection, most bacteria die. A very, very small number of them begin to acquire

00:22:40.960 a snippet of the genetic information of this virus. And they insert it into the CRISPR system.

00:22:46.560 So those bacteria that survived have now acquired immunity against these viruses.

00:22:52.160 And in the process of surviving, what do they have to rapidly do to fend off that first viral infection?

00:22:59.840 So the bacteria has many different defense systems in addition to CRISPR. So in fact,

00:23:04.880 CRISPR is not the first line of defense. There are other things that are also very powerful technologies

00:23:11.200 now that are called restriction endonucleases. These are proteins or enzymes that bacteria use.

00:23:17.680 They don't adapt, so they don't evolve, but they recognize fixed letter sequences. And sometimes

00:23:24.320 these will always get activated first and try to fend off the virus. But if it doesn't,

00:23:29.040 then there are a host of other defense systems. One of these will eventually work.

00:23:34.480 But unfortunately for the bacterial population, these things don't come in quick enough. And that's

00:23:39.440 why most of the cells die. But the few where these things were able to keep up with the virus,

00:23:44.720 that's what allows the CRISPR systems to begin to acquire the genetic information.

00:23:48.720 So now let's talk about that subsequent infection. So we've obviously, through a very Darwinian mechanism,

00:23:55.040 selected a subset of the E. coli in this case that indeed are able to not just survive with their

00:24:01.280 first and second line defense against the bacteriophage. But now they've also developed

00:24:05.440 the memory, so to speak, the way we would use the term. So now it's a month later, the same phage

00:24:11.040 comes along, infects you, inserts its genetic material. But now you actually have, interspersed

00:24:17.520 between your CRISPR repeats, you actually have the 30 nucleotides that match a sequence within that

00:24:24.560 virus. So now how do you spring into action to resist this infection?

00:24:28.320 So after the first infection, in a way, the bacteria has been vaccinated against this virus.

00:24:34.960 So the second time when this virus comes around, it will inject its genetic information into the

00:24:40.000 bacteria. But now the bacteria in the CRISPR repeat area has a signature of this virus. So the repeat

00:24:48.560 area will get turned on and it will start to make CRISPR RNA that carry a 30 base pair long or 20 base

00:24:56.320 pair long guide that's able to recognize the incoming virus. So the cast protein will bind to this CRISPR

00:25:04.080 RNA. They will go and try to search along all the DNA sequences in the bacteria. When it finds a match

00:25:12.080 in the virus's DNA, it will activate the nuclease and it will cut the DNA.

00:25:17.280 And approximately how many base pairs are in viral DNA?

00:25:21.280 Depends on the virus. They can be small, maybe 10,000 letters, or they can be long,

00:25:26.160 100,000 or even longer.

00:25:28.000 Which again, just for context so people understand, we have about 3 billion. So we're still talking about

00:25:33.760 tiny, tiny, tiny amounts of DNA. And by the way, does it differ if it's an RNA virus versus a DNA

00:25:40.080 virus? Is the process identical? It's very similar.

00:25:43.440 So the reinfection, they tend to make pretty quick work because now once the bacteria is able to make

00:25:51.040 the cast protein, which it can make really quickly, in between the CRISPR segment, it makes the RNA

00:25:56.640 segment to guide it and match it. And really that's kind of a CRISPR RNA plus a truncated other version

00:26:03.200 of an RNA that holds the RNA in the cast protein. We call that whole thing the guide RNA.

00:26:08.480 How quickly can that Cas9 enzyme holding the guide RNA, how quickly in actual time does it go through

00:26:17.600 the entire sequence of viral DNA until it finds its place to land and cleave?

00:26:23.040 That process is probably pretty fast. I don't know exactly how quick it is, but also it's important

00:26:29.120 to recognize that in a single bacteria, there are many, many copies of the Cas9 along with the guide RNA.

00:26:36.960 And so that means once the virus comes in, there are many, many copies of Cas9 that are simultaneously

00:26:42.480 in parallel searching against the virus DNA to see whether or not there's a match. And this is why

00:26:47.920 the system is so powerful because it's able to very quickly, in a parallel fashion,

00:26:53.840 find the match and then inactivate the virus within minutes.

00:26:58.080 And there's something else that is pretty unique about where those cuts take place. When we bring

00:27:04.960 in the viral memory, it always begins with three particular nucleotides. What's the significance of

00:27:11.840 that?

00:27:12.080 Right. So what you're talking about is what CRISPR scientists call PAM sequence, P-A-M or protospacer

00:27:20.160 adjacent motif. Just a jargon. What is significant about that is that it is a sequence that is only

00:27:26.960 found in the bacterial virus's genome, but not in the bacteria's genome. When the bacteria acquires a piece

00:27:34.880 of the virus's sequence and sticks it into its own genome, one question is, couldn't the CRISPR system

00:27:41.360 go and recognize the bacteria's genome?

00:27:43.360 It's sort of like, couldn't you turn the weapon on yourself and cut your own DNA and kill yourself?

00:27:48.880 Exactly. Yeah. So how does it avoid this self-targeting or autoimmunity against itself? And this is where

00:27:56.160 the PAM sequence comes in. The PAM sequence is in the viral genome. It's right next to the sequence

00:28:02.960 that is acquired into the CRISPR system, but it itself is not acquired into the bacterial genome. And so

00:28:10.000 what you see in the bacterial genome's CRISPR repeat is just the recognition sequence, but no PAM. And so

00:28:16.960 Cas9 requires the PAM to activate recognition and cleavage. And so without the PAM, it doesn't cleave

00:28:23.440 itself, but it's still able to target the virus. So, Feng, does this bring us more or less up to

00:28:29.440 speed with the state of the art when you turned your attention to how can I create a finer resolution

00:28:37.840 for gene editing for my optogenetics problem? Is this about where the state of the art was?

00:28:42.480 Yeah. So I started to work at MIT and the Broad Institute in 2011. And so when I first started,

00:28:50.320 I went to a scientific presentation and they were talking about CRISPR. And they mentioned that

00:28:56.000 CRISPRs are nucleases. Because I was thinking about nucleases and gene editing at the time,

00:29:00.960 when I heard that word, it just got me interested. And so I went on to Wikipedia and looked at what CRISPR

00:29:07.840 is. And Francis Mojica and Siobhan Moinu and Rodal Berengu, they had just published

00:29:14.000 the very early studies on the Cas9 system. Back then it was still called Cas5. It was only later

00:29:20.800 on renamed to be Cas9. But there was all these papers that if you read them, there aren't too many

00:29:26.800 of them. You can piece together the information and you can get a sense that this is an RNA-guided

00:29:32.880 DNA-targeting and cleaving enzyme. And at the time, there were other gene editing systems that

00:29:38.080 were being worked on by researchers in the field, something called zinc finger nuclease or TALEN.

00:29:44.640 And these are systems that use proteins to recognize DNA, but not using RNA to recognize DNA.

00:29:50.400 Let's talk a little bit about both of those, because they were the state of the art

00:29:54.640 up until 15 years ago. And I want to understand both how they work and why they may not have been

00:30:02.640 sufficient for your application. In other words, why were you looking for something beyond two

00:30:06.400 techniques that already existed? When I first became interested in gene

00:30:10.000 editing in the late 2000s, there were people already developing gene editing technologies.

00:30:16.240 In fact, there were multiple iterations of technologies that came around. There's something

00:30:21.440 called mega nuclease, which was very, very early. And then what got me excited was a New York Times

00:30:27.040 article. I think it was 2008 or 2009. It talked about a system called zinc finger nuclease. There's

00:30:33.280 a company in California that's called Sangamo Biosciences, and they were already developing zinc

00:30:39.680 finger nuclease for gene therapy to be able to go and edit DNA and treat disease. But zinc finger was a really

00:30:48.720 challenging system for scientists to adapt and use because it required very sophisticated protein

00:30:55.120 engineering. The way it works is that zinc fingers are protein domains. They're just one glob of protein.

00:31:02.880 Each finger, each zinc finger can recognize three letters of DNA. And they occur in nature,

00:31:09.440 so you can find them in naturally occurring DNA binding proteins called transcription factors. And they

00:31:15.760 allow transcription factors to go and recognize different genes in the genome to modulate their

00:31:20.720 activity, either turn them on or turn them off or change how much they are expressed in the cell.

00:31:26.400 But how do you get specificity when you are only recognizing three nucleotides? The probability of

00:31:32.160 those three nucleotides showing up seems pretty likely across the genome.

00:31:36.320 Exactly. So nature has solved this problem by forming zinc finger arrays.

00:31:40.960 So they tether multiple fingers together.

00:31:44.160 And all of them have to hit their target of three nucleotides.

00:31:47.920 Correct. That's right. So if you have an array of three fingers, they recognize as nine letters.

00:31:53.120 If you have an array of six fingers, then that's 18 letters. So in a complicated genome,

00:31:57.840 our genome with three billion letters, 18 would give us uniqueness. So 18 can allow you to find

00:32:05.280 or define just a single position.

00:32:06.880 And that's because four to the power of 18 is a big enough number?

00:32:11.120 It's bigger than three billion.

00:32:12.240 Yeah. Okay. So let's now talk about the talons. First of all, what is that technique and why was that

00:32:20.960 not necessarily going to serve your purpose?

00:32:23.520 So the challenge with zinc finger is that you have engineered the fingers to recognize

00:32:28.480 different combinations, three letters. And you have to make sure that when you tether them together to

00:32:34.240 make a zinc finger array, they can recognize what you intend to recognize in that multiple three

00:32:40.160 fashion. That turned out to be really cumbersome, and usually it doesn't work very well.

00:32:44.640 So then this other system that you mentioned called transcription activator-like effector,

00:32:49.600 T-A-L-E, nucleosideveloped. So these systems came also from bacteria. They came from a specific

00:32:57.600 pathogenic bacteria called Xanthomonas or Rezi. So it lives on rice plants,

00:33:02.960 and it's a major rice pathogen. The way these proteins work is that they're injected

00:33:08.320 by the bacteria that's trying to colonize the plant into the plant cell. And once it gets inside,

00:33:15.920 it homes into the genome of the plant cell, finds the gene, and then starts to turn it on. When it

00:33:22.960 turns it on, it allows the plant to become more susceptible to the bacterial colonization. And so it

00:33:28.640 allows the bacteria to be able to more successfully survive on this host plant. It's a major pathogenic

00:33:35.840 system. But what is really cool about talons or tails is that they recognize DNA in a very programmable

00:33:44.320 fashion. These proteins, they have repeat domains, just like zinc fingers, that form an array. But

00:33:51.840 individual domains recognize single DNA letters. And so you can find these tail proteins in the bacteria

00:33:59.680 that have 12 or 16 or even 20 different repeats. And so they can recognize long stretches of 12 or 16

00:34:08.560 or 18 or 20 DNA letters in the plant genome. The plant genome is also large. They are sometimes

00:34:14.400 two, three times the size of our genome. So recognizing long sequences is important for achieving precision.

00:34:21.520 That was the tail system. So what is really cool is that Ulla Bonas, who is a researcher in Germany,

00:34:28.080 and also Adam Bogdanovi, who is a researcher at Iowa State at the time, they discovered that there is a

00:34:34.240 specific code for how these tail proteins recognize DNA. It turns out that within each one of these repeats,

00:34:42.080 there are two amino acids letters that correspond with a specific DNA letter that it binds to. And so

00:34:49.200 if you take any tail protein and you just dial in different combinations of these two amino acids in

00:34:56.240 each one of the domain, you can specify what DNA sequence this tail protein is able to recognize.

00:35:02.560 So it turned out to be much more easy to use than zinc fingers.

00:35:06.000 So were you satisfied with that? I mean, obviously you weren't. You were sitting there looking up

00:35:12.000 Wikipedia for what CRISPR is when you heard about it. So something must have said to you, hey,

00:35:16.960 as good as the talons are, and as much as they're an improvement over zinc fingers,

00:35:20.560 they're still not good enough. Why was that? Why were they not good enough?

00:35:23.360 Zinc fingers were really hard to use. When I tried to engineer it, it was very hard, very difficult, almost

00:35:29.120 impossible for just a single researcher to get something that would recognize the DNA sequence

00:35:35.360 that you're trying to get it to recognize. So it was very hard to use. Talons, on the other hand, were

00:35:40.560 easier. But the repetitive nature of these proteins made it quite cumbersome to engineer new proteins to

00:35:48.400 recognize the sequence you're trying to edit. So for example, if you wanted to be able to modify

00:35:53.600 a specific gene, it could take you maybe several weeks or even a couple months to be able to

00:35:59.760 successfully make one of these tail proteins. And when you do that, usually it works, but not always.

00:36:06.080 And sometimes it's not very effective. Is that because it's so difficult to predict

00:36:13.120 the folding structure of the protein, even if you know its sequence? Why would it take that long?

00:36:18.000 It's long because from a technical perspective, these sequences are just very hard to work with.

00:36:24.400 Because they're very similar to each other, they're prone to recombination. So in order to get a tail to

00:36:30.000 work, you have to, in a very precise way, put every domain in the correct order. Because if you're

00:36:36.080 trying to recognize A, G, T, C, you have to put it together in A, G, T, C. You can't have A, C, G, T.

00:36:43.040 It just wouldn't work. So to put repetitive sequences together and line them up in exactly

00:36:49.520 the order you want, that is really challenging to do. It's possible, but it's very challenging.

00:36:54.640 So it's very interesting. I mean, I think it's worth just sort of taking a step back.

00:36:58.560 Most people can sort of remember what they were thinking 10 years ago, 12, 15 years ago. And

00:37:04.640 the human genome was sequenced nearly 25 years ago. And the promise of gene therapy

00:37:10.800 was hailed as right around the corner. And yet here we are a decade, more than a decade after

00:37:17.360 the sequencing of the human genome. And it doesn't appear that gene therapy through gene editing was

00:37:24.560 any closer than it really was from a practical standpoint a decade earlier. I think that's a bit

00:37:29.920 of a disconnect for people. I think most people who necessarily aren't in a lab would be surprised

00:37:34.880 to understand that simply knowing what the sequence of genes are. I mean, A, we've talked about this a

00:37:41.520 lot on the podcast. We still have no idea what most of these genes do anyway. We have no idea why there

00:37:48.080 are coding segments and the majority of the DNA is non-coding segments. And yet some of these non-coding

00:37:54.400 segments are where mutations exist that result in disease. I mean, all this stuff is still a bit of a

00:37:59.120 mystery. But here you're sort of getting at the central or maybe the most important or jugular

00:38:04.720 question, which is, if a person has a disease like cystic fibrosis or sickle cell anemia, where we

00:38:13.680 really know in unambiguous terms where on the DNA, where in the gene this lies, where in the DNA,

00:38:22.400 we know what the substitution is. We know which C was turned into a G or a T.

00:38:27.680 And all we need to be able to do is go in there and fix it. That was something we couldn't do 10

00:38:32.240 years ago. I think for many people, that's quite surprising. And I know that that's not the problem

00:38:36.840 you were trying to solve, but it's obviously the world you've created. So let's talk about those steps

00:38:42.120 down that road. So you figure out what CRISPR is. You bring yourself up to the state of our discussion

00:38:48.160 today. What's the next thing? The next thing in terms of CRISPR? Yeah, the next thing that you're

00:38:52.600 starting to now pursue, I mean, how are you now going about solving your problem for your good?

00:38:58.280 What point do you realize you're working on a problem that has a far bigger application

00:39:02.440 than just solving your problem?

00:39:05.320 Yeah.

00:39:06.040 I mean, I think that's a lot more quickly since I started work on gene editing, because

00:39:08.600 the human genome project was completed in early 2000s. And with the human genome having been

00:39:14.920 sequenced, and then also with DNA sequencing technology becoming cheaper and faster, scientists

00:39:22.200 were able to start to sequence many, many more genomes. And so they can start to make comparisons

00:39:28.040 between healthy individuals and also people who are affected by specific diseases to see what's

00:39:35.240 different between their genomes. And by doing that comparison, they can identify the differences

00:39:40.680 that may be causal for disease. And so to date, based on genetic analysis, researchers have probably

00:39:47.640 identified more than 5,000 genetic mutations that have a direct causative role in disease. And so these are

00:39:55.240 called genetic diseases. They are usually affecting a small population of individuals. They're not as

00:40:02.280 common as things like cancer or diabetes or what people call complex or complicated diseases. But

00:40:08.520 nevertheless, these are the ones where we know the exact genetic cause. And so the tantalizing idea is

00:40:17.160 then if you know the mutation in the genome, why not just go and fix it? And so that's where gene editing

00:40:22.120 comes in. And people have since the very beginning trying to realize this idea. They were trying to

00:40:27.960 work on it using meganucleases. They were trying to solve this using zinc finger nucleases. They were

00:40:33.720 certainly trying to use that talent to also treat diseases this way. But the challenge is that they

00:40:39.480 weren't very efficient. And it was also difficult to apply them to be able to treat the disease with

00:40:45.800 sufficient amount of efficacy. When CRISPR came along, especially with Cas9, it was much easier to be

00:40:53.480 able to design strategies to edit DNA. And that made it much more feasible for many, many groups to really

00:41:01.480 start to work on this idea. Were you finished with your postdoc and now starting your own lab?

00:41:06.760 Yeah. I had just started my lab at MIT.

00:41:09.240 Okay. So you've got how many PhDs in your lab?

00:41:11.880 I probably had maybe 10 PhD students.

00:41:15.880 Okay. And a couple of postdocs to boot.

00:41:17.800 Yeah.

00:41:18.200 And these people have all come to you presumably because they're interested in optogenetics.

00:41:25.000 Yeah. They were interested in different things.

00:41:27.000 So at what point do you come back into the lab and say,

00:41:31.000 I'm going to hit pause on the optogenetics problem. I'm going to kind of go down this CRISPR path for a

00:41:35.560 while.

00:41:36.680 When I first started my lab, I was already focused on the gene editing problem. So when students came

00:41:42.280 to me, even though they came to me wanting to work on optogenetics, I had to convince them that

00:41:48.040 there's this other problem that is also interesting and maybe we can try to work together and make a

00:41:53.720 difference there. And so I started to try to tell them about CRISPR, tell them about gene editing and

00:42:00.200 all the potential applications.

00:42:01.640 So keep going down the story now. So it's the early 2010s. What are the next steps that you take to

00:42:09.080 develop this technology?

00:42:10.200 So when I first started my lab, I was working on Talens. And then very quickly,

00:42:15.400 I had learned about CRISPR. And then I started to also get CRISPR projects going in the lab. And we

00:42:22.280 worked on both systems for a while at the same time. We pretty quickly realized that Talens were

00:42:28.920 difficult to use because of the cumbersome nature of how to make them. And then because of that,

00:42:35.400 the promise of CRISPR was much more apparent. So maybe I'll give another analogy. So we now have

00:42:43.720 a mobile phone. And on a phone, there are many different apps, apps that help you book trips,

00:42:49.000 apps that helps you send messages to your friends and family, apps that allow you to take photos.

00:42:54.360 You have a phone and the phone can do everything. You just load the app onto it. With Talens or Zinc

00:42:59.560 fingers, the analogy would be, you have to build a different device for each one of these functions.

00:43:04.520 Soterios Johnson Meaning you would need a different

00:43:06.280 phone for each app.

00:43:07.320 Soterios Johnson You need different hardware. Yeah. So for every gene you're trying to target,

00:43:10.280 you have to build a brand new protein to target that gene. And that is a very cumbersome and not

00:43:16.200 very effective process. With CRISPR, the promise is that CRISPR is like the smartphone.

00:43:22.360 Soterios Johnson You can load software onto

00:43:24.840 it to recognize different genes. And the software is the CRISPR RNA. These RNAs are very easy to

00:43:31.320 chemically synthesize. And you can define the gene by reading off of the sequence of the gene,

00:43:37.880 which is already completed through the Human Genome Project. So all of that was just a step function

00:43:44.440 improved over the Zinc finger and Talen technology. So we realized that if we can make CRISPR work,

00:43:52.040 not only in the bacteria, but put it into a human cell and get it to recognize genes in the human

00:43:57.560 cell, then we can have a much more powerful and much more democratized gene editing system.

00:44:03.800 Soterios Johnson So what was the

00:44:06.040 breakthrough or a set of breakthroughs that led to the

00:44:09.480 utility of this? And did it start with, let's just silence a gene first? I mean, let's figure out how

00:44:18.040 to go in and, with precision bombing, silence one gene. Was that the first problem before the,

00:44:25.000 let's actually take a strand of novel DNA and put it in?

00:44:28.920 Soterios Johnson

00:44:29.640 CRISPR is a natural nucleus. So in bacteria, it uses the guide RNA to recognize the virus DNA.

00:44:36.840 And then once it recognizes it, it will cleave the virus DNA. So make a double-stranded DNA break.

00:44:42.120 And so that's what we're trying to sort of make happen in the human cell. We try to program Cas9

00:44:48.280 with a guide RNA to go and recognize the specific gene in the human genome and then be able to cut it.

00:44:54.440 Soterios Johnson Which is valuable. I mean,

00:44:56.040 there are certain cases where overexpression of a gene is pathologic. And if we silence a gene,

00:45:01.640 we fix a disease.

00:45:03.240 Soterios Johnson Exactly. So there were a lot of studies done on how

00:45:08.360 breaks in the DNA would get repaired. So Maria Jason, Jim Haber, they had studied maybe a couple

00:45:16.040 of decades before that, how DNA repairs will get processed. And so what they found is that when

00:45:22.680 you make a cut in the DNA, so when you make a break, it will activate repair processes in our cell. So in

00:45:28.760 fact, our DNAs get DNA breaks all the time. And we have a robust process to be able to fix them

00:45:34.520 to prevent mutations. So what Maria Jason and Jim Haber found is that if you make a cut in the DNA,

00:45:42.280 that cut will activate one of two different repair processes. The first repair process

00:45:48.600 will glue the DNA together, usually correctly, but in a very, very small number of instances,

00:45:55.080 it will introduce a mistake. And that mistake will inactivate the gene. So it will no longer make

00:46:02.040 the protein product that it's supposed to make. And this is very useful if you wanted to inactivate

00:46:07.480 something in the cell. Sometimes there are mutations that are deleterious. And if you can

00:46:12.040 inactivate that deleterious mutation, then you can make the cell healthy again. And so that was a very

00:46:17.800 powerful method. The second repair process is called homology-directed repair, HDR. And this relies on a

00:46:27.480 template DNA that carries the sequence that you're trying to repair with. And so if you make a cut,

00:46:34.440 and you also provide a template DNA, then the repair process will copy whatever that's on a template

00:46:41.480 into the DNA break site. And this is a more powerful way to be able to change the DNA sequence

00:46:48.040 in a design fashion. Is there a risk when you cut the DNA and it repairs, that it repairs in a manner

00:46:54.760 that remains pathologic, even if it's distinct from the path that was already there? It is possible,

00:47:01.640 but the probability is much, much lower. What about the Holy Grail, which is to literally edit a new gene,

00:47:08.920 to put something in that didn't exist? What was required to take that leap?

00:47:14.680 There are different ways that you may want to change the DNA sequence. You may want to inactivate

00:47:18.920 something, you may want to delete something, and you may want to insert something. So to do each one

00:47:23.960 of these, you need to have a machinery that will allow you to do that. So the Holy Grail would be to

00:47:31.320 be able to insert a gene into anywhere you want precisely, and also very efficiently. And to date,

00:47:38.440 that ability is still not quite there yet. We're still working on, and many other groups are working

00:47:44.520 on developing technologies to make that happen with high enough efficiency. We can do it now with

00:47:51.880 very low level, maybe less than a percent, or maybe just single digit percent. But for a big gene that

00:47:59.320 we're trying to put in, we don't have a good way to do that yet.

00:48:01.800 Okay. So currently, is it possible to snip DNA anywhere you want with the current technology?

00:48:09.720 Yeah. With CRISPR, with Cas9, we can pretty much target throughout the genome and make cuts.

00:48:16.200 So what diseases are currently amenable to that type of gene therapy? And we'll put aside the

00:48:22.520 regulatory stuff, which we can talk about. But if we pretended that humans were laboratory animals,

00:48:27.480 where we could do an experiment and actually do this, what are some of the diseases that we could

00:48:31.640 directly affect in that manner?

00:48:33.800 So there are a lot of genetic diseases where there is a mutation that is pathogenic.

00:48:39.800 Overexpression. In this case, overexpression if you want to deactivate it, but some would be

00:48:44.520 underexpressed.

00:48:45.400 Yeah. So these are genes that are usually important for the body, but there's a mutation in it,

00:48:49.400 and that makes the resulting protein, the mutant protein, deleterious for the patient.

00:48:55.400 So if you can inactivate these genes, then you can treat disease. So for example, there are diseases

00:49:01.080 in the liver where there are certain proteins that cause amyloidosis and that can lead to serious

00:49:07.880 problems. So by using CRISPR, you can go in and cleave these genes, inactivate them so that they no

00:49:13.560 longer produce these toxic gene products. Huntington is another example where there are mutations that

00:49:20.200 are occurring in the gene that makes the gene deleterious. And so if you can go and try to

00:49:25.640 inactivate these deleterious mutations, it may be possible to treat the disease.

00:49:30.120 Is it generally the case that autosomal dominant diseases are an overexpression problem or an

00:49:36.920 expression of something harmful problem, whereas recessive diseases are the opposite, where you tend

00:49:43.560 to not be producing enough or something of that. And is that overly simplistic? Yeah, that's a good

00:49:48.200 explanation. So many people are probably familiar with Huntington's because it is one of the most

00:49:52.440 devastating neurodegenerative diseases imaginable, perhaps second only to Lou Gehrig's disease or ALS.

00:49:59.880 But unlike ALS, where we don't really know the etiology, we clearly know the etiology of

00:50:04.520 Huntington's disease, which is a gene. It's a genetic disease. It's an autosomal dominant gene.

00:50:10.520 And sadly, it doesn't present until later in life. So many times, individuals will pass this gene on

00:50:18.440 prior to the symptoms being manifest, and therefore they go on to suffer the fate of this disease,

00:50:25.560 having already passed it on. Tell us a little bit about Huntington's disease and why it may or may

00:50:30.120 not be amenable to this type of treatment. So Huntington's disease is caused by mutations in

00:50:35.480 the gene called Huntington. This is a gene that's expressed in the brain. And in mutated form, this

00:50:41.640 gene accrues an expansion of repeat sequences within the gene. And the longer the repeat is,

00:50:48.440 the more deleterious it is for the patient. And so the idea would be to try to shorten these repeats,

00:50:55.880 or maybe if you can reduce the amount of repeated sequence of the RNA that's expressed, you could

00:51:02.440 also get a cell to be healthier. The challenges are, the repeats happen within the coding region

00:51:08.840 of the gene. So the region that is important for making the protein sequence. So in order to make

00:51:15.160 the resulting edit successful, you have to do it very precisely. You have to delete exactly three

00:51:22.040 letters at a time from the repeat sequence. Otherwise, you will shift the frame.

00:51:25.960 And the gene, remind me, has how many base pairs?

00:51:28.920 The gene has thousands of base pairs.

00:51:30.760 Okay, so it's a huge gene.

00:51:31.800 Yeah. And these repeats can also be several hundred or maybe a thousand repeats long. And so you want to

00:51:38.680 be able to delete them very precisely. So that is one challenge. The other challenge is to be able to

00:51:44.120 deliver the gene editing machineries into the brain and get to enough cells. And there are some virus-based

00:51:50.360 technologies that are coming along. But still, we don't have probably the most suitable method yet.

00:51:56.840 So let's talk about delivery a little bit. How do you do this? We now have this idea. Hopefully,

00:52:03.480 people can wrap their head around this somewhat challenging idea of what CRISPR is. And maybe we

00:52:09.320 can again just sort of summarize it, right? You have a CRISPR vehicle, would be a Cas9 protein. And

00:52:16.440 you also have a guide RNA that is made up of both the piece that you actually want to put in

00:52:24.840 wrapped around another sort of tracer piece that holds it firmly in the Cas9 protein. That Cas9 protein,

00:52:34.360 by the way, does it require its own helicase to open that? No, it can do it itself. Beautiful. So

00:52:40.040 it's a one-man shop that runs up the host DNA, opening it, and waiting to find its match. And it

00:52:47.400 waits and waits and waits. And then it finds its match, holds the strands of DNA at the Cas9, and then

00:52:53.720 clip. Correct. And obviously, if you then put in something, it can... So how do you actually deliver

00:53:00.040 the Cas9 protein to an individual? That is really the big challenge. So right now, there are clinical

00:53:06.040 applications of CRISPR for treating different diseases, diseases in the blood, like sickle cell

00:53:11.080 disease, or diseases in the liver, diseases in the eye, and many other places. And so depending on

00:53:16.920 where you're trying to deliver CRISPR into, there are different technologies that people use.

00:53:22.040 Maybe the simplest might be in the blood? Sure.

00:53:24.040 Okay. So tell people what sickle cell anemia is and why it's amenable to this type of therapy.

00:53:30.280 Right. So sickle cell anemia is caused by a mutation that causes the red blood cell to sickle. So they

00:53:37.080 form a sickle form. And so they're not able to properly function, and sometimes they can aggregate,

00:53:43.400 and then this can cause occlusion in the blood vessel and can cause serious problems. And so these

00:53:49.160 red blood cells are made by progenitor or stem cells in the body that produce these red blood cells.

00:53:55.560 And it's a simple mutation. It's a single point mutation that changes one amino acid,

00:54:01.640 and that one amino acid based on one base pair change, I can't remember what it is. It's an

00:54:07.320 alanine to a glycine or something. It's quite trivial, is what leads to all of this downstream

00:54:12.200 badness you talk about. But now, current treatment for these patients is blood transfusions, right? I mean,

00:54:17.640 it's an awful disease, and these patients experience unbearable pain. As an aside, some might ask,

00:54:24.520 why does this disease exist? Why didn't Darwin get rid of this? Well, it turns out there's an

00:54:29.560 advantage to having the trait for malaria, right? So it turns out that if you have one copy of the

00:54:36.040 sickle gene, and the other one is normal, you have normal looking red blood cells, you don't get the

00:54:41.320 disease, but you actually get protection, as you said, from malaria. So that would keep this propagating,

00:54:46.760 particularly in a malaria rich area like Africa. This is why it's much more prevalent in a black

00:54:52.760 population than a white population because it offered some benefit. But if you have two copies

00:54:57.240 of the gene, you get the sickling. Those are not the people that are passing on their genes

00:55:01.640 historically. They would have perished before reproduction and also just perished in a great

00:55:07.160 deal of pain. But nevertheless, here we are today. So you have to be able to go into the bone marrow

00:55:12.680 to make this change because there's no point in doing this if you have to do this every week. You want

00:55:16.600 to do it one and done, right? That's right. One of the promises of gene editing is that

00:55:21.080 it can provide a single treatment that is a cure for the disease. And so in the case of sickle cell,

00:55:26.680 what happens is that the doctor will mobilize the stem cells, the bone marrow cells from the patient,

00:55:33.000 get them to come out and be able to harvest these bone marrow cells. And they don't necessarily do this

00:55:38.520 with a bone marrow aspirate. They do it by giving them medications that cause them to secrete more

00:55:43.960 progenitor cells into the plasma. That's right. This is the current practice in the medical field.

00:55:50.440 And so they will get the patient, harvest their bone marrow cells. And these cells are going to be

00:55:56.520 modified in the laboratory where researchers will take the messenger RNA for Cas9. So this will allow

00:56:04.840 the cell to produce the Cas9 protein and they will also take guide RNA. Maybe just tell folks really

00:56:10.040 quickly, sorry to interrupt, Feng, but maybe just explain really quickly for people the relationship

00:56:14.840 between DNA, RNA, messenger RNA, protein. It's the central dogma, but that way when you say what

00:56:20.440 you're about to say, they'll know why it works. So there are different ways to get a protein to the cell.

00:56:26.520 And the way that proteins are made in the cell is that they're encoded in DNA. And the DNA has to be

00:56:32.520 transcribed into messenger RNA. And that RNA is then translated by the ribosome into the protein.

00:56:41.080 And so if you can put into a cell, either the DNA, the gene for Cas9, or the messenger RNA for Cas9,

00:56:49.480 or the protein for Cas9, the cell will eventually have Cas9. Because if you put in the DNA,

00:56:55.960 the cell will start to make mRNA based on it. And then that mRNA will get translated into the protein.

00:57:01.160 But how do you put it in the DNA? Isn't that the whole problem that we're trying to solve?

00:57:04.840 Right. So there are different ways to put these things in. So for DNA...

00:57:09.240 Just use a virus?

00:57:10.520 You can use a virus or you can, if you're working with cells in the Petri dish,

00:57:14.680 you can directly electroporate the DNA into the cell. And this is done by zapping the cell

00:57:20.440 with the electrical current. It will rupture the membrane. When the membrane ruptures,

00:57:25.160 things can leak in. So if you have DNA that's outside the cell, when the cell membrane ruptures,

00:57:30.520 the DNA that's outside the cell will flow into the inside of the cell and get into the cell.

00:57:35.160 And this is actually how people treat sickle cell disease, except they're not putting DNA into the

00:57:39.960 cell. They're putting mRNA. So they incubate these bone marrow stem cells that have the sickle cell

00:57:47.400 mutation in a bath of mRNA for Cas9 and the guide RNA, separately.

00:57:53.640 And it's amazing that once those cells acquire both the mRNA for Cas9 and the mRNA

00:58:01.240 that is corresponding to the guide, it will translate into a Cas9 protein. It doesn't translate

00:58:10.040 guide RNA into a protein. It just stays there and they find each other.

00:58:14.120 That's correct.

00:58:15.720 It is so remarkable to me that anything works in this universe, but that's up there as one

00:58:19.480 of those things that kind of just amazes me at that.

00:58:21.160 Well, I mean, this is really the result of the biotechnology revolution, the molecular biology

00:58:27.080 discoveries that have really under sort of paved the biotechnology revolution.

00:58:31.480 And now what is the efficiency of that process? So once you go ahead and put those two bits of very

00:58:39.960 different RNA into the proximity of these bone marrow cells?

00:58:46.200 This can be quite efficient. In the laboratory or in these petri dish settings, this can be, you know,

00:58:52.680 approaching 100%.

00:58:54.520 Okay. So now, within a very short period of time, you have cut out and also reinserted.

00:59:03.880 It's a single base pair. So what are they doing? What's the exact thing you're asking Cas9 to do?

00:59:07.400 Actually, for sickle cell, the treatment that has recently been approved in the last year is actually different.

00:59:13.160 The way the treatment works is that for sickle cell patients, it's been found that for some individuals

00:59:21.080 who have the sickle cell mutation, if they also carry another mutation, or if they somehow is able

00:59:28.760 to express the fetal version of hemoglobin, then their sickle cell symptoms are much, much less.

00:59:37.080 And so for treating sickle cell patients with gene editing, the therapy actually goes and modifies

00:59:44.360 a different gene, modulates its expression, to then allow the fetal hemoglobin gene to turn on.

00:59:51.080 And why is that an easier solution than simply changing the one amino acid that's

00:59:57.160 broken in the first place?

00:59:58.280 Because the way it works is that it simply makes a cut and it doesn't require

01:00:03.800 template repair. That's right.

01:00:05.960 I see. So we're still at the point where in vitro, we're still better off with a cleavage,

01:00:13.160 just a straight cut of the DNA, than even a single base pair switch to fix one amino acid.

01:00:19.960 That latter is a harder problem.

01:00:22.120 The latter is a harder problem, but there's also very good progress on that front.

01:00:26.920 So for example, David Liu developed a methodology.

01:00:30.120 Also at Harvard.

01:00:30.520 Yeah, also. He's a colleague of mine, and he developed a technology called base editing.

01:00:35.240 And base editing allows you to use Cas9, you get rid of the DNA cleaving activity

01:00:42.440 of Cas9. So it simply goes and binds to DNA. So you use it as kind of a guidance system to direct

01:00:51.480 a different enzyme called a deaminase to be able to go and chemically modify a single base.

01:00:56.840 What's amazing to me, Fung, that that is easier than what we just wish we could do, which is

01:01:04.600 change that C to a G. Take out the C, make it a G. And that will give me the amino acid I want.

01:01:11.160 Like the fact that we're chemically having to modify it with a deaminase. What do you think is

01:01:16.680 necessary to take this next? Because we've already had, as you described it, a big step function from

01:01:22.760 where we were 10 years ago. What's going to be required for the next step function,

01:01:28.680 for the science fiction to start?

01:01:30.520 Yeah. So I think this really goes back to the division of genetic medicine.

01:01:37.000 Genetic medicine is very powerful. CRISPR is part of it. But it's really a two-component system.

01:01:44.120 There is the medicine itself, and then there's also the delivery technology. So you need to have

01:01:49.880 the right vehicle for delivery and the right payload to be able to treat the disease in the

01:01:55.480 right cell. And we're limited more on the delivery than the payload.

01:01:58.920 More on the delivery. So the payload technology has come a long way. We now have mRNA, we have Cas9,

01:02:04.600 we have base editing, we have prime editing, we have a lot of different types of editing technology,

01:02:09.800 and even epigenetic editing. But the bottleneck is, how do we put these really powerful

01:02:15.320 payloads into the right cells in the right tissue in the body?

01:02:19.480 Say those again. What were our tools? Cas9, base pair editing?

01:02:22.920 Yeah, we have Cas9, we have base editing, we have prime editing, we have different recombinases,

01:02:29.320 we have epigenetic modifiers that also are based on Cas9. We have mRNA, we have siRNA. We have a lot

01:02:37.320 of different things that can modulate and modify cells.

01:02:40.440 I want to actually come back and talk about the epigenetic modification using Cas9. It's not a

01:02:46.520 payload problem, it's a delivery problem.

01:02:48.040 Right.

01:02:48.520 Yeah.

01:02:49.080 And it's also a biological problem.

01:02:51.240 Right, because many diseases will not be amenable to this. So I'm sure everybody thinks,

01:02:55.240 well, Peter, we've heard you say that cancer is a genetic disease. Does that mean that once we solve

01:03:00.920 the delivery problem, we solve cancer? Why is that not necessarily true?

01:03:04.680 That is because cancer is caused by many different risk mutations in the cell. And so it's difficult

01:03:11.720 to treat cancer by correcting the mutation because you really have to be able to correct

01:03:17.960 at a very, very high efficiency. If you have a few cells, a few cancer cells that have not been

01:03:25.000 corrected, those cells will continue to divide and replicate and form tumors and even metastasize.

01:03:31.480 That is really challenging. But people are using gene editing in the cancer therapeutic context.

01:03:37.880 What they're doing is that they are using gene editing to engineer immune cells so that the

01:03:43.720 immune cells are more potent at recognizing and killing cancer cells. And this is part of what's

01:03:49.160 called immunotherapy.

01:03:50.200 Yeah. Everybody's talking about AI. Does AI enable this any better on either side,

01:03:57.240 on the payload side or on the delivery side?

01:03:59.320 AI is very powerful for protein engineering. In the past few years, there's been amazing breakthrough

01:04:06.520 in the use of AI for predicting protein structures. Each protein is made of a unique sequence of amino acids,

01:04:15.160 and the unique sequence allows the protein to fold into a specific shape. And one of the holy

01:04:21.160 grail problems for a long time has been, how do you take just the letters of a protein and predict

01:04:28.120 what the shape of the protein looks like? And this is something that many, many scientists have worked

01:04:33.880 on for a long time, but hasn't been able to come up with a good solution. But it was really in 2020,

01:04:39.480 2021, the use of AI by this group called DeepMind that was able to come up with a solution called

01:04:46.760 AlphaFold2. And AlphaFold2 is an AI-based system that has learned from all of the structures of

01:04:53.640 proteins that scientists have experimentally determined. And when they solved those structures,

01:04:59.800 there's a huge database of them. And they were able to use AI to look at all of them and learn

01:05:04.840 from that large database to then come up with a prediction system called AlphaFold2.

01:05:09.960 It's amazing to me. Maybe just explain to people again,

01:05:14.200 pick your favorite protein. How many amino acids are in it in the primary sequence?

01:05:17.800 We can pick the green fluorescent protein from jellyfish, maybe 300 amino acids.

01:05:22.840 300. It's a relatively small protein.

01:05:24.600 Yeah.

01:05:24.840 Okay. So it has a primary structure, which is like, what are the actual amino acids?

01:05:30.280 Correct.

01:05:30.680 It has a secondary structure, which is like, well, when does it actually form a helix? When does it

01:05:36.200 form a sheet? It has a tertiary structure, which is kind of like how it starts to bend. And then it

01:05:43.080 has this quaternary structure, which is how the whole thing fits together in complicated three

01:05:47.480 dimensional folds. Exactly.

01:05:49.720 What you said a moment ago is if a scientist wants to make a protein and they know what it

01:05:55.720 needs to look like, they kind of know what the quaternary structure is. I got to get this protein.

01:05:59.720 I got to design a molecule that fits in that receptor. It's almost a trial and error problem to

01:06:05.560 go from the primary sequence to that. Right.

01:06:08.360 You're saying that AI is really good at doing this thing where it knows the relationship

01:06:16.360 between primary sequence and final quaternary structure. Is that just literally linear regression

01:06:23.640 at a level we've never understood it because humans can't do it, but is it basically just solving

01:06:28.440 the world's most complicated linear regression problem?

01:06:30.840 I think at a very fundamental level, that is what is happening. But the human brain, we can't process

01:06:36.840 so much data very effectively, but with AI, you can have these massive neural networks

01:06:43.880 that can really process this. Let's go back to animal models for a moment.

01:06:48.200 You touched briefly on the idea of a transgenic mouse. And again, people who have listened to this

01:06:52.600 podcast for years are no stranger to the transgenic mouse. So many amazing breakthroughs in science

01:06:59.400 have come through these. And frankly, just through genetic understanding of mice, making changes to a mouse

01:07:05.560 gene. You know, we recently had Dina Dubal on the podcast. We talked about Clotho, to me,

01:07:10.280 one of the most interesting proteins out there. And of course we learned the story of how Clotho came

01:07:15.880 about. Silencing a gene found this thing. Overexpress the gene found this thing. Okay.

01:07:21.880 How does CRISPR enable that today? Has it changed the ability of people working on totally other problems

01:07:30.040 to get there quicker using laboratory animals?

01:07:33.960 Right. So the transgenic mouse technology really revolutionized biology. And when the transgenic

01:07:40.840 mouse technology was developed, the way it worked is that you would start with stem cells for mice,

01:07:46.440 you would modify these stem cells, and then you put the stem cells back into an embryo, and then you

01:07:53.320 transplant the embryo into a mouse so that the embryo can develop into a fetus and then be born as a new

01:07:58.760 mouse. That is a very long process. And then once the mouse is born, usually that mouse does not have

01:08:05.960 100 percent of that genetic modification. So the mouse is called a mosaic. And then you have to take

01:08:12.760 that mouse and breed the mouse again with another mouse and hoping that the mosaic part is the one

01:08:19.560 that gets expressed and you basically try to concentrate the mosaic portion of it.

01:08:24.520 Exactly. So when you use this methodology, it can take probably a year or maybe even longer

01:08:32.120 to be able to generate a specific transgenic mouse. So it used to take a long time to do this. But now

01:08:38.120 with CRISPR technology, what people can do is they can directly inject the gene editing Cas9 and guide

01:08:44.760 RNA into an embryo, into a single cell embryo, and then modify it there.

01:08:49.560 So again, this is so brilliant because we can't do this in humans, of course. I mean, we could,

01:08:54.840 but that's a whole ethical discussion we should talk about. But we can directly do that. So we

01:08:59.000 can make a transgenic mouse in one go.

01:09:00.760 Exactly. So the mouse gestation period is 21 days. So after 21 days, you have a transgenic mouse.

01:09:06.200 What has this done to the field of biomedical research?

01:09:09.240 It has accelerated biomedical research dramatically. So imagine yourself as a graduate student,

01:09:14.840 and usually a PhD will take five years.

01:09:17.880 And a lot of that time is the rote work of transfecting mice and waiting for, yeah.

01:09:24.120 Yeah. So if you have to wait two years to get a mouse so that you can begin your experiment,

01:09:28.280 that is a long time. So now with CRISPR or gene editing, you can get a mouse in two, three months.

01:09:35.720 So let's play devil's advocate for a moment. When you did your PhD, you had to slog through the

01:09:40.760 old-fashioned way. Were there hidden benefits of that? In other words, did it give you more time

01:09:46.120 to read, more time to be curious, more time to fail? Again, it's a tangential discussion, but do you

01:09:52.120 worry that with this remarkable precision tool that budding scientists are missing out on an experience

01:10:00.680 that you and your entire generation and everyone before you had? Or do you think that that's a

01:10:05.480 relatively small price to pay for the pace of development?

01:10:09.400 I'm not too concerned about it. I think if we can accelerate the accumulation of knowledge,

01:10:15.960 the acquisition of data, I think that will really help science move a lot faster. I think in the future,

01:10:22.520 especially with higher throughput technologies, CRISPR, DNA sequencing, and many other things,

01:10:28.920 together we're going to be accumulating new data for biology at an exponential pace.

01:10:35.640 And we're not only going to be relying on our own ability to analyze data, but we're going to have

01:10:41.000 AI to help us. These large systems that can draw much, much larger regression analysis. And with that,

01:10:48.280 I think biology discovery and disease treatment development will really accelerate. I think that's

01:10:55.400 a really exciting future. Now we've talked a lot about Cas9, but you've also specifically done quite

01:11:01.640 a lot with another protein, another CRISPR-associated protein called Cas13. What's the difference and

01:11:09.160 how does this potentially impact future work?

01:11:11.560 CRISPR is a bacterial immune system and there are many, many different types of CRISPR. And so in nature,

01:11:19.400 bacteria are invaded by DNA viruses, by RNA viruses, all sorts of different viruses. Cas9 protects bacteria

01:11:26.840 against DNA viruses. But then there also needs to be a CRISPR system that protects against RNA viruses.

01:11:33.080 So Cas13 is the RNA analog of Cas9. It uses a guide RNA to recognize the RNA

01:11:40.120 virus and then cleave the RNA genome. And what is really interesting about Cas13 is that unlike Cas9,

01:11:48.280 it not only cleaves the recognized RNA, but once it recognizes a piece of RNA, it also turns on

01:11:55.640 almost a suicidal function. It goes and cleaves any other RNA that's in the bacterial cell.

01:12:01.320 And so in the infection cycle, you can think of this as an altruistic system where when the Cas13

01:12:11.080 recognizes my cell has been infected by RNA virus, I'm going to shut myself down, kill the cell and

01:12:18.120 save the population. So hang on, why is that the case? So why does the cell not choose to do that

01:12:23.720 with the DNA virus? Are RNA viruses necessarily more lethal to bacteria?

01:12:28.280 Yeah. RNA is usually more abundant. Once the RNA gets produced, there are many copies. And so it's

01:12:36.360 difficult to shut down every single copy of RNA. Whereas with DNA, there's usually just one copy of

01:12:41.560 DNA. So if you can shut it down. So when a DNA virus infects a bacteria, let's just talk about a very

01:12:48.600 specific bacteria phase, latches on the outside of an E. coli, it's going to shoot in a piece of DNA.

01:12:54.280 We've already talked at length about how that works and the role Cas9 plays in that.

01:12:59.160 When a different bacteria phase comes along, you're saying it inserts a lot of RNA or just the RNA

01:13:06.740 replicates much quicker.

01:13:07.800 RNA replicates quicker.

01:13:08.440 Got it. Okay. So it inserts a small amount of RNA, but remind me why a small amount of RNA will be

01:13:14.980 amplified much quicker than a small amount of DNA. Is it because it's one step less? It's already made

01:13:20.240 the machinery to go in front of the translating ribosome.

01:13:23.760 Exactly. Because DNA has to go to RNA to make proteins that go back to make DNA.

01:13:28.960 So you're saying Cas13 also has a suicide feature.

01:13:32.560 And this suicide feature is actually quite useful for developing diagnostic technologies.

01:13:38.240 And so especially during COVID, we use Cas13 to develop a way to detect coronavirus RNA.

01:13:46.160 And it provided a way to have a simple and rapid detection method.

01:13:51.200 So say more about that. Was that really the first application?

01:13:54.000 That was, I think, one of the first applications. Yeah.

01:13:57.520 How would that have been done before? And what was the speed and accuracy of it being done in that way?

01:14:02.560 So the most widespread method for diagnostics is using PCR. So it checks for the nucleic acid sequence.

01:14:09.280 But PCR is a laboratory-based test. It requires a machine called a PCR machine, which is complicated

01:14:16.800 to run. And you have to be in a laboratory environment to go through the test. What Cas13 provided is

01:14:23.120 something that's more similar to an antigen test, where you can run it at a point of care or even

01:14:30.000 potentially at home. And it was simply required to take a swab to have the sample. It makes the sample

01:14:37.360 into a buffer. And then Cas13 would react in that buffer to be able to detect the virus sequence.

01:14:44.080 Then you load it onto a paper strip. It will run. And then you see whether or not a band shows up.

01:14:49.440 When we think now about gene editing going forward, and again, we've already established that the

01:14:55.920 payload is not the problem. It's the targeting and the delivery. What are the relative advantages

01:15:02.080 and disadvantages of Cas9 versus Cas13? And are there other Cas or CRISPR-associated proteins out

01:15:08.880 there that might even be better than both of them?

01:15:11.680 So Cas9 and Cas13 both have therapeutic applications. But one of the challenge is that

01:15:18.880 they are large proteins. Cas9 is 1,300 amino acids long. And Cas13s are usually around 1,000 amino acids

01:15:26.960 long. So they're fairly large proteins. How long was Cas9?

01:15:30.560 1,300 amino acids. And so in order to get Cas9 into a cell, you have to be able to fit Cas9 and

01:15:37.600 the guide RNA into your delivery system. And if you're using viral vectors to deliver, they are

01:15:43.920 usually very compact and you can barely fit Cas9 in. So then the nice thing is, what if you have

01:15:50.560 something that's smaller? And so we and also many other groups have worked on trying to discover new

01:15:56.080 proteins that are more compact and more easily packageable. And there are some out there, but

01:16:03.280 sometimes and usually they're not as specific or not as active as Cas9. And so there are trade-offs

01:16:10.080 to using some of those systems. And we're working on engineering them and there's a lot of good

01:16:15.040 progress turning those systems into a specific and comparably active systems as Cas9.

01:16:20.720 So what would be, if you could wave a magic wand, how big a Cas protein would you tolerate such that

01:16:28.160 your Cas protein plus your guide RNA would easily fit into your delivery vehicle? Would you want to be

01:16:32.960 half that size and be 500 amino acids? Do you need to be smaller than that?

01:16:36.400 If you could shrink it down to 1000 base pairs, so 300 amino acids, that would be ideal,

01:16:42.480 but I think that's challenging to do. Let's maybe make sure I understand why.

01:16:46.640 You're saying, look, at the end of the day, I need a protein that has the structural integrity to hold

01:16:52.720 the guide RNA in place, make its way down into the nucleus, open up the DNA. So it's basically two

01:17:00.560 enzymes. It's a helicase and a nuclease, and it has to be able to march along the DNA,

01:17:05.520 recognize while holding, again, it's a mechanical problem if you really stop to think about it,

01:17:10.400 just on the smallest level, and then hold, cut, boom, insert eventually. And you're saying,

01:17:15.600 at some point, we just abut the limits of physics. I need a certain amount of amino acids to make that

01:17:20.800 structure. But it sounds to me like you're saying, you don't believe that that exists in nature.

01:17:26.480 You're not out there looking for a 300 amino acid Cas protein anymore. You're now going to say,

01:17:32.960 let's use AI to help us build one.

01:17:35.440 There are natural forms of proteins that are small and are like Cas9. In one of our projects,

01:17:42.240 trying to look for a small Cas9, we thought, let's look at the evolutionary origin of where Cas9 came

01:17:49.440 from. By tracing the evolutionary history of Cas9, we found that there is a very large

01:17:56.240 family of protein called ISCB that is the ancestral form of Cas9. ISCB is a very small protein. It's

01:18:05.920 only about 450 mL, so it's about a third of the size of Cas9. And it does exactly what Cas9 does.

01:18:13.040 It's got a helicase activity, it's got nuclease domains, and it also works with the RNA guide to

01:18:19.840 recognize and cleave DNA. But what is different between ISCB and Cas9 is that the guide RNA for

01:18:26.160 ISCB is much, much larger. So you rob Peter and you got to pay Paul.

01:18:30.880 Exactly. And RNA is not as stable, not as robust. It's more prone to degradation.

01:18:37.840 Yeah, you don't want to go any bigger on guide RNA. You need the best of both worlds. You need

01:18:42.880 a Cas-associated protein that is small, and you want to be able to keep the guide RNA small.

01:18:49.840 Right. That's the holy grail.

01:18:51.600 Okay. And is it your prediction that that will have to be developed synthetically?

01:18:56.880 I think through engineering. It's not clear that such a compact system has been developed by nature,

01:19:03.440 but we can start from nature's Cas9 or ISCB as a scaffold and then begin to engineer.

01:19:09.440 So what kind of race is on for that? Because I can't even fathom the commercial value of that.

01:19:15.360 If there was a new protein called the Cas-alpha or Cas-omega, whatever, the synthetic version

01:19:22.720 that was small enough to now easily deliver payload, this is a trillion dollar product.

01:19:29.600 A lot of people are working on it. I think it's certainly going to be very, very useful.

01:19:33.360 Is the lion's share of this work being done in universities or is it being done inside of biotech

01:19:38.960 companies? It's in both places now.

01:19:40.880 Is this one of the biggest races going on in CRISPR biology today?

01:19:46.400 I'm not sure this is the biggest race. There are a lot of other capabilities that we want to realize.

01:19:51.120 Okay.

01:19:51.600 For example, how do you insert large genes into the genome precisely and efficiently? I think that

01:19:58.000 is just as important of a problem. Cas9, even though it's large, we can deliver it

01:20:03.120 to some cells already. So there's already-

01:20:05.600 In vivo or only in vitro?

01:20:07.200 In vivo.

01:20:08.000 What cells can be delivered in vivo using Cas9 currently?

01:20:11.280 So liver.

01:20:12.320 Okay.

01:20:12.720 Yeah. Liver is a place where we can use lipid nanoparticles to deliver Cas9 and guide RNA into.

01:20:19.600 The COVID vaccine is made using mRNA and lipid nanoparticle. And so it's a very similar approach

01:20:26.720 where you formulate these lipids with the Cas9 mRNA and guide RNA.

01:20:32.320 And what is it about the liver that makes it amenable to a lipid nanoparticle?

01:20:36.480 There's a lot of lipid recycling happening in the liver. And so we have a lipid nanoparticle,

01:20:41.120 they get bound by these recycling proteins and they get taken up.

01:20:44.720 Interesting. Does it matter the manner in which it's given, intravenous, intramuscular,

01:20:49.920 do all lipid nanoparticles end up there provided they're not ingested orally and presumably digested?

01:20:55.760 A lot of it goes to liver. Yeah. Because liver also is one of the areas in our body that filter

01:21:00.640 out all the toxins. Yeah.

01:21:02.160 So everything gets trapped there.

01:21:03.280 So in other words, if you were thinking about one of the rare inborn diseases of metabolism,

01:21:09.280 these are the really rare diseases where children are born without a particular enzyme that allows

01:21:14.000 them to either metabolize a certain protein or glucose even for that matter. Do we think that

01:21:19.760 we're on the cusp of being able to use, I don't want to suggest that Cas9 is primitive,

01:21:24.960 but in the context of our discussion, we've realized it has some limitations.

01:21:28.720 But do we think that the CRISPR-Cas9 system is already good enough and sufficient

01:21:33.920 to cure some of those diseases?

01:21:36.000 For some of them, potentially. It will depend on the mutation, whether or not we can use Cas9 or

01:21:41.200 base editing or prime editing to be able to fix the mutation.

01:21:44.160 In other words, you have to figure out the payload problem.

01:21:46.160 Right. Yeah.

01:21:47.040 But if it's in the liver and it can be addressed by targeting hepatocytes in the liver,

01:21:51.840 then the delivery problem is already addressable.

01:21:55.280 Okay. What about genetic diseases of the eye? That's also a place that's easy to reach

01:22:00.720 and you don't require systemic administration. What are the genetic conditions of the eye that

01:22:05.760 might be amenable to this treatment as it stands today?

01:22:09.120 Yeah, there are different eye diseases that are affected by single gene mutations.

01:22:14.160 For example, LCA10 is one of the eye diseases that there's already been a CRISPR strategy being

01:22:20.480 developed for. The way that these diseases in the eye are treated is by designing Cas9 and a guide

01:22:27.440 RNA to be able to knock out the gene that is causing degenerative conditions in the eye,

01:22:33.440 and then using a viral vector to deliver the Cas9 gene and also the guide RNA intraocularly

01:22:40.400 into the patient. And so once the virus gets into the cells in the eye, they will make Cas9,

01:22:45.440 they'll make the guide RNA, and then that will carry out the modification.

01:22:48.480 So in that sense, it's a bit of the old meets the new.

01:22:51.520 You're taking the oldest trick in the book that was the original, original gene therapy

01:22:56.720 vehicle, the virus, and you're combining it with a far smarter payload,

01:23:01.440 which is Cas9 and guide RNA. Where do these stand in clinical trials right now,

01:23:06.240 both the liver and the eye, which would obviously be the leading edge of this?

01:23:10.400 So the eye is the first place where, in the US, a gene therapy was developed and approved.

01:23:15.520 And so this is a drug called Laxterna. Laxterna puts a gene into the eye to be able to treat the

01:23:21.600 disease called LCA2. And so basically, the virus provides a gene that is missing in the cells in

01:23:28.240 the eye. And that is able to allow the patients to regain some light sensitivity. And so that was

01:23:34.320 the first gene therapy.

01:23:35.360 And these are patients that are completely blind without it?

01:23:38.320 That's right.

01:23:38.800 And then how much light sensitivity do they get once this gene is inserted?

01:23:42.320 They get some so that they can move around in a room with large obstacles.

01:23:46.240 Wow.

01:23:47.040 Yeah. And this is all done just through a single injection in the eye.

01:23:50.720 A single injection. Wow. What are other ocular targets?

01:23:54.960 So LCA10 is another one that was being developed by Editas Medicine. And the way this disease

01:24:02.800 developed is that there is a mutation that causes degeneration in the eye. And the idea is to use

01:24:10.080 Cas9, use the same viral vector system to deliver it into the eye and have Cas9 inactivate this mutant

01:24:17.200 gene so that it can slow down or stop degeneration.

01:24:20.800 So LCA2, you have to put an active gene in that was missing?

01:24:27.440 Right.

01:24:27.840 LCA, was it 9 or 12?

01:24:30.560 10.

01:24:30.640 10, you have to deactivate a gene.

01:24:32.560 Correct. That's right.

01:24:33.280 Okay. Where is that in clinical trials? What phase is that in?

01:24:36.240 So it went through phase 1, 2.

01:24:38.000 Okay. What was the efficacy in phase 2?

01:24:40.240 There was some efficacy.

01:24:41.520 What's the phenotype of this patient without gene therapy? Is it total blindness as well?

01:24:45.360 Also blindness or deteriorating vision.

01:24:47.840 So blind by what age?

01:24:49.520 Probably 30s.

01:24:51.280 Okay. Wow. And then what did the phase 2 find? How much eyesight were they able to restore?

01:24:56.320 They found that there's some improving vision, but I think it wasn't as robust as they had help for.

01:25:01.920 Okay. Why do you think that is? Why do you think that these therapies are not fully

01:25:07.200 restoring vision? In the case of the LCA2, is it because by the time you treat these people,

01:25:14.000 the neuroplasticity part of it has lost its window of development? In other words,

01:25:19.440 is the problem that if you treat these people late in life, the part of the brain that receives

01:25:25.200 the visual signal isn't developed? Or is it literally that we're just not fixing the eye?

01:25:30.080 The retina doesn't regenerate. So if it's already degenerating, you have to deal with what is left in

01:25:36.240 the retina. And so what you can restore is really capped by whatever that's already left there.

01:25:43.360 And plus, there's also inefficiencies in the delivery systems. You're not restoring 100% of

01:25:49.040 the cells. And if you're layering on gene editing on top of it, which also has some less than 100%

01:25:55.360 efficiency, when you multiply that all together, that's why you're not getting full restoration.

01:26:01.120 So what would have to be true to fully restore vision in those patients? What set of things

01:26:08.240 would have to be true therapeutically? I think it's only very hard to fully restore

01:26:12.560 vision. If you really want to fully restore vision, you probably have to regenerate the retina. So you

01:26:18.400 have to replenish cells that are missing. Could that be done epigenetically? What do we know about the

01:26:24.880 epigenome of the retina in the adult versus the infant? Do we know that if we could simultaneously,

01:26:34.720 or even in two treatments, correct the genetic defect, but then return the epigenome to the milieu

01:26:42.720 of embryogenesis or early life, could that fix the problem?

01:26:48.400 That could potentially work. Although this is not my expertise. So I don't know

01:26:53.920 all the processes involved there. If you can recapitulate development in the eye and allow

01:27:01.280 cells to redevelop the retina as it was developing during development, then potentially you can

01:27:08.320 regenerate the eye. On the side of the liver, what has been the success rate of the Cas9 delivery

01:27:15.120 system there? The lipid nanoparticle delivering the Cas9 payload to be more exact.

01:27:19.200 Yeah. In the liver, it's quite robust. You can probably get 80, 90% in the liver.

01:27:23.760 And that's sufficient to correct certainly any underexpression, right?

01:27:28.400 Right. In the liver, there's also an interesting development where scientists are trying to target

01:27:34.400 a gene called PCSK9 to be able to treat cardiovascular disease.

01:27:38.480 So they're going to use a lipid nanoparticle. You're going to put in a new

01:27:43.040 PCSK9 gene that's inactive, or you're just going to deliver a vehicle that paralyzes PCSK9?

01:27:49.040 Yeah. The strategy is to paralyze PCSK9 so that you can inactivate PCSK9 and then reduce cholesterol.

01:27:55.680 Any idea how much a treatment like that would cost?

01:27:58.640 I don't know. Initially, maybe tens of thousand dollars.

01:28:04.160 Well, I mean, to be honest, that would be really cheap because the drug

01:28:08.400 that inhibits PCSK9 inhibitor, and there are three of them out there, when they came out,

01:28:14.160 they were $15,000 a year drugs. Now they're $6,000 a year drugs. So call it $10,000 a year

01:28:20.960 as the current drug cost. So you're saying that- Maybe $50,000 or-

01:28:24.400 Oh, it might be that much more. Okay. Again, still relatively inexpensive compared to the whole thing,

01:28:28.640 to a lifetime of therapy. Is there any particular reason it costs that much if you actually just look

01:28:34.400 at the drug and don't try to bake in the cost of the last 10 years of developing it? In other words,

01:28:38.480 if a company today came along and said, I'm interested in developing a drug that one and done

01:28:44.000 treats PCSK9 and lowers LDL cholesterol indefinitely, how much would it actually cost

01:28:50.240 to make that drug under GMP conditions at scale? I don't know the exact cost of good,

01:28:55.680 but it's probably much lower than that. But in order to get to that, we need to get the processes

01:29:01.440 developed. And so I think over time, we're definitely going to be able to get down the cost.

01:29:06.560 Large-scale mRNA production, large-scale formulation of nanoparticles. I think those are the processes

01:29:12.800 that people are making really good progress on. So what is driving the cost right now? What is

01:29:18.000 making it expensive today? I think there are a lot of factors, including development cost and also

01:29:24.880 the fact that these drugs is the first time developing these modalities of drugs. And so

01:29:32.640 the manufacturing processes need to get developed.

01:29:35.680 And what specifically is it? Presumably, we're just making synthetic Cas9 and Cas13 all day long,

01:29:42.160 right? In the laboratory, at laboratory scale.

01:29:44.960 I see. But commercially, not.

01:29:47.120 Right.

01:29:47.680 So you make your own Cas9 and Cas13 in your lab?

01:29:50.800 We do. Yeah, we do. You can also buy it from commercial companies. But the scale is usually

01:29:56.560 for laboratory use. If you think about it, a mouse is 10 grams. A human can be 40 kilograms. So that's,

01:30:04.800 you know, a thousand times larger in body weight. So you need a thousand times more material. And I think

01:30:10.800 it's developing the process to scale that to that level to be able to treat human beings. That is where

01:30:17.680 the expensive process is. Is there any foundational roadblock to doing this? Or is there a Moore's

01:30:24.880 law aspect to this where it's just going to get significantly cheaper over time in the same way that

01:30:31.600 transistors just get significantly cheaper over time?

01:30:34.560 I'm not an expert on this, but I think there will definitely be efficiencies from scale. Enzymatic

01:30:40.640 processes for making these RNAs is getting much better. Purity is improving. So all those things,

01:30:47.360 I think they will compound and that will result in significant cost reduction.

01:30:51.680 How much do you follow slash pay attention to the regulation of gene editing? I know that,

01:31:00.080 gosh, sometime in the last six, seven years, there was a very controversial case in China with a

01:31:06.160 scientist who had, it seems somewhat nefariously from my reading of the story, edited the embryos of

01:31:15.360 kids to render them on the surface. This sounds like a great idea, but rendered them immune to HIV.

01:31:23.760 And I believe one of their parents was HIV positive. This was an IVF case, but it turned out that there

01:31:30.240 was an enormous amount of backlash from the scientific and medical community, because I guess one,

01:31:37.040 he did it without the full consent of the medical community. And also the belief that the risk of

01:31:42.800 transmission of HIV was quite low to the child in the first place. Can you say more about that case

01:31:47.360 in particular, and then also what the current state of genetic manipulation is from a sort of ethical

01:31:54.560 legal standpoint? So this was back in 2018. There was a couple who wanted to have children. The father was

01:32:01.600 HIV positive. And so they wanted to make sure that their children are not infected with HIV. And so scientists

01:32:10.400 made a mutation in the embryos that is a gene called CCR5. A CCR5 mutation is a naturally occurring

01:32:19.040 mutation in the human population. Single-digit percent of individuals carry this mutation,

01:32:24.240 and these individuals are naturally immune to HIV infection.

01:32:28.160 This is the magic Johnson mutation.

01:32:30.240 Right. They don't have this specific protein on the surface to allow HIV to bind and enter the cell.

01:32:36.240 And so the scientists edited the embryo, human embryo, to remove the CCR5 gene. But it turned out that

01:32:43.440 that edit wasn't complete. Earlier we talked about mosaicism. So the result of his edit was actually two

01:32:51.680 girls who were mosaics for the CCR5 mutation. So the editing happened, and embryos were implanted in the

01:33:00.800 mother and carried to term. And so two baby girls were born. And when the genetics of these baby girls

01:33:07.680 were analyzed, it turned out that they were highly mosaic, suggesting that editing wasn't very efficient.

01:33:13.840 And the edit was done at how many cells, in theory? Was it one?

01:33:18.480 I don't know exactly how many cells, but very early on, post-fertilization.

01:33:22.160 But to have guaranteed no mosaicism, would you need to do this edit at a single cell

01:33:29.840 and then proliferate only that cell and discard the rest of the blastocytes?

01:33:34.320 In theory, that's the way to do it.

01:33:36.160 And do we know what this person did?

01:33:38.000 I don't know exactly how much of the technical details were released.

01:33:42.960 So these girls are born mosaic, but what's their phenotype with respect to CCR5?

01:33:48.480 So some of their immune cells have CCR5 and some don't. So that means that they can be

01:33:53.680 infected by HIV for those cells that are CCR5 positive.

01:33:57.200 But it probably means that they'll never die of AIDS because they will always maintain a

01:34:02.080 population of T cells that are not susceptible. They could be HIV infected, but they'll never

01:34:07.440 have a T cell level that falls so low that their CD4 count gets below whatever the threshold is for

01:34:14.640 AIDS.

01:34:14.880 Right.

01:34:15.280 Okay. But the world really responded harshly to this, correct?

01:34:19.920 Correct. That's right.

01:34:21.120 What was the fallout for this scientist?

01:34:23.280 The reason that there was a huge backlash is because there really wasn't any...

01:34:27.680 There was no medical need to do this.

01:34:28.960 Correct. That's right. It's not justified. So the scientific and also medical community,

01:34:34.320 everyone voiced their concern about this ethical issue that just occurred. And I think the scientist

01:34:39.920 was put under house arrest for a while. And this incident also really motivated much more

01:34:47.920 ethical discussions around gene editing. There was ethical discussion before, but it really

01:34:53.120 focused the issue. There were multiple working groups that were established, multinational working

01:34:58.480 groups between the different national academies, US, China, UK, as well as the WHO had several

01:35:06.560 working groups. And I think in the US there is legal regulation that prevents a germline modification,

01:35:13.520 but it's certainly not the case internationally.

01:35:15.760 What do you think is being worked on outside of the US in places where there's no regulation with

01:35:22.640 respect to germline mutation? So for example, are people trying to basically say, look, we're going

01:35:28.720 to get rid of APOE4 isoform and make it APOE3 or 2 isoform? I mean, what is the extent or ambition

01:35:37.200 of people? Are we going to delete LPA genes so that there's no LP little a phenotype? I mean,

01:35:43.280 you could make a case that a lot of these things would unidirectionally improve human health.

01:35:49.280 What are people working on? And then let's talk a little bit about the ethics.

01:35:52.560 I don't know what is happening internationally because people are not publicizing their work.

01:35:58.960 But I think what is known is that there is international consensus that we don't want to

01:36:05.760 do these types of embryo germline editing right now. Even for things where it really matters. So the

01:36:12.480 APOE4 and the LPA don't, you don't have to do that. But what about Huntington's? What

01:36:19.120 about inborn errors of metabolism where children are going to probably die during infancy? Cystic

01:36:26.560 fibrosis, things for which we have no viable treatments. Is the international consensus still

01:36:33.440 that this type of gene editing will not be pursued?

01:36:36.880 I think there's a lot of discussion going on. So it's certainly not a settled issue. But what I think

01:36:42.560 scientists have all come to terms with is that the technology still needs much more validation

01:36:48.880 and development before it's ready for application in this germline editing setting. The efficiency,

01:36:56.000 the specificity, both need to be optimized further in order for there to be any chance of

01:37:02.720 a germline therapy having the intended effect.

01:37:07.200 And how much of that do you think is the main reason for the hesitation here versus the slippery

01:37:14.640 slope argument, which says, yes, of course, those applications are absolutely worth pursuing. But if

01:37:21.360 we do that, how far are we from doing APOE4, which may be a little bit gray? And then what about

01:37:29.440 identifying genes that are, frankly, not remotely related to lifespan or medical necessity, but

01:37:37.680 instead relate to kind of the stuff that people talk about in science fiction? Hey, this is a gene

01:37:42.720 that might make my kid a little bit taller. This is a gene that relates to intelligence or some other

01:37:48.640 physical characteristic that is completely irrelevant to their health. Is it more about this is dangerous,

01:37:54.640 we don't know how to do it? Or even if we figure this out, we don't want to open up a pen

01:37:59.360 doors box. I think this is very much the debate that's happening right now. There are patient

01:38:04.560 groups who are advocates for using this, using gene editing in a germline setting to treat disease.

01:38:11.760 Assuming the technology is safe and efficacious, then we should do it because it will alleviate

01:38:16.720 suffering. So there is that argument. There are also people arguing the slippery slope argument,

01:38:22.640 which is if we allow for X and Y, then eventually we're going to be getting into uncharted territories

01:38:30.080 with designer babies and so forth. So that is very much the debate that's happening. I think while that

01:38:35.680 debate is happening, I think it's important to recognize that science is also progressing on other

01:38:40.960 fronts. There are likely advances in science that will achieve the same outcome without gene editing.

01:38:49.120 And so all those things I think are a competition with each other in the ethical debate. And that's why

01:38:54.400 this is such a complicated issue.

01:38:55.840 You're one of three or four people on the planet who not only know more about this technology,

01:39:02.720 but have been personally responsible for it. As such, I can't imagine you get to sit quietly

01:39:10.320 during these debates. Even though you're not an ethicist or a philosopher, I'm sure people want

01:39:16.000 to understand what you think. So how do you think about this? Not from a technical standpoint,

01:39:22.560 which we'll continue to talk about in a minute, but from an ethical standpoint, where do you think

01:39:27.440 the line should be drawn in what we as a scientific or medical community permit with respect to gene

01:39:34.720 editing in the germline, which is obviously the area where we're, I guess we should have made that

01:39:39.440 point a little more clear at the outset. We're talking about making an edit that will persist in perpetuity.

01:39:44.320 Yeah. I've thought about this issue a lot. The things that are very easy to agree on

01:39:51.600 is that if there is an obvious and important medical benefit, so especially for inborn errors

01:39:58.720 of metabolism or other non-genetic mutations, if the technology is there, it's something that is

01:40:04.560 definitely doable. And I think it's okay to improve the lives of those patients.

01:40:08.640 Is that a mainstream view? Or do you think that there is still a lot in the establishment who have

01:40:13.600 not come around to that yet because of this slippery slope argument?

01:40:17.040 There are certainly people who have not come around to that. And I think when making decisions

01:40:22.560 like that, when making a medical decision like that, it's also important to consider what other

01:40:27.120 alternative methods there may be. For example, pre-implantation genetic testing is another method where

01:40:34.480 you might be able to screen out embryos that have those mutations.

01:40:38.400 Which for IVF, we certainly can. Obviously for IVF, all of those diseases can be checked.

01:40:44.640 Right.

01:40:45.120 So then it gets to maybe a more complicated question, which is,

01:40:50.960 is there a role for genetic engineering within in utero, for example? Or is that just so technically

01:40:57.360 challenging that it's much easier to focus on just the IVF side of things?

01:41:01.680 Dr. So certainly on the disease treatment, monogenetic genetic disorder, IVF, I think that

01:41:07.200 that is probably the most likely application. Further down the road, there are some obstacles to

01:41:13.600 solve. One, we don't really understand biology enough. So if you wanted to make your kid 30 IQ

01:41:19.520 points smarter, we don't really know how to do that. And so the biology needs to catch up. But I do

01:41:25.840 think that as society continues to develop, if the science is there and the technology is also there,

01:41:34.880 I think people will opt to do that.

01:41:37.200 Sorry. Meaning let's just assume, and by the way, I think this is personally,

01:41:41.760 Fung, my view is I think these problems are way harder than people think. I think these polygenic

01:41:48.800 nuanced traits like intelligence, athletic performance, resilience, happiness, all of those

01:41:56.720 things, I think are infinitely more complicated than we believe, and probably just as environmental

01:42:05.600 as they are genetic. So even if you give somebody the right genetic template, if they're not in the

01:42:10.960 right environmental surrounding, they won't necessarily even develop the way you hope.

01:42:16.480 Absolutely.

01:42:17.520 You might give somebody 30 more IQ points. It doesn't translate to a person being demonstrably

01:42:22.160 more intelligent if they aren't in the right environment. That's my prediction is that that

01:42:27.040 stuff is way, way, way harder. And the biology is complicated, but it's more complicated than any

01:42:34.640 lay person is thinking about.

01:42:36.240 And also another thing is biology has a lot of compensation. You might make some mutations that

01:42:42.080 make the person smarter, but that could also increase cancer risk or impair athletic capability,

01:42:48.240 make a person live shorter. Those complicated trade-offs are things that I think people maybe

01:42:54.480 don't fully appreciate.

01:42:55.680 Yeah. For me, there's a reason I'm not an ethicist. Things seem much more obvious to me sometimes,

01:43:03.280 which means I'm probably not a nuanced enough thinker on these regards. But when it comes to a

01:43:08.000 disease that's going to kill you with no other treatment, it seems like a no brainer. Genetic

01:43:13.280 editing and genetic engineering should absolutely be considered a part of the equation. At the other

01:43:18.800 end of the spectrum, when we're talking about things that are truly superfluous, like intelligence,

01:43:24.000 height, athletic performance, eye color, any sort of trait, those seem clearly like a horrible reason,

01:43:30.640 not for the ethical reasons like, well, what does that say about disparity? Forget all that. It's

01:43:38.720 the reasons you just said. We have no idea how complicated the system is. If you're going to

01:43:43.760 take a huge risk, there has to be a huge payoff. The asymmetry of that strikes me as too much.

01:43:50.320 And then there's the gray area. And I guess on the gray area, I do feel a bit nuanced, right? Which is,

01:43:55.440 should we just get rid of LPA genes? Well, maybe, but then to your point, we have antisense

01:44:01.040 oligonucleotide inhibitors that are just around the corner that effectively will get rid of LP little

01:44:07.600 a. They shut the gene off. So why would we risk trying to eradicate this from the population when

01:44:14.320 we can just give you a drug that does the same thing with fewer side effects and less permanence?

01:44:19.600 Exactly. Or you can do it not in the germline, but just in the somatic cells. That could also be

01:44:25.540 beneficial. And then there's yet the other area, which is, look at something like autism. Autism is

01:44:32.200 a genetic disease, despite what many people will have you believe. But the heritability of autism

01:44:36.880 is so high that we know it's largely a genetic disease, but very polygenic, very subject to other

01:44:44.640 things. And do we really want to get rid of it? I don't know. Maybe. I mean, certainly when you see

01:44:51.080 a child that's devastated by autism, who's nonverbal, who's harming themselves, it would be a very logical

01:44:58.360 thing to say. But when you start to move further from that end of the spectrum towards the end of

01:45:03.380 the spectrum, where people are quite functional with autism, you might say, well, I don't know.

01:45:08.360 It actually comes with some benefits as well. And do we want to create a homogeneous society

01:45:14.680 where everybody is the same?

01:45:16.800 I don't think we want to.

01:45:17.880 This to me is the crux of sophistication. Yeah.

01:45:20.520 Yeah.

01:45:21.040 Right. Whether it be mood.

01:45:23.180 Because that diversity is important for the human race as a whole. That diversity

01:45:28.160 is what will allow the human race to be resilient in the long run.

01:45:33.100 So I want to talk a little bit about your story going back. So you grew up in China. You came

01:45:38.240 to the middle of the country, right? Where'd you grow up?

01:45:41.740 I grew up in Iowa. So I moved to Des Moines, Iowa when I was 11 years old.

01:45:45.900 What brought your parents here?

01:45:47.440 My mother came to the U.S. first. She was an exchange scholar in computer science. She had a

01:45:53.140 chance to visit one of the schools in Iowa. And she saw that the educational system in Iowa was much

01:45:59.540 more hands-on as opposed to job memorization. And because of that, she thought I would maybe

01:46:05.100 benefit more from that type of instruction. So she decided to stay in the U.S. and immigrate here.

01:46:11.680 When did you start to become aware of your intellectual abilities, for lack of a better term? I mean,

01:46:17.520 was school always so easy for you? Were you sort of breezing through high school, acing every test

01:46:22.380 put in front of you? Or what was it like?

01:46:24.340 I don't think I'm all that smart. I've always been interested in science. I've always been very

01:46:30.540 curious. So I think I've just always kind of followed my curiosity to do something that is

01:46:36.560 enjoyable and fun. What is really interesting is I didn't like biology at all.

01:46:42.100 You preferred physics and chemistry and mathematics?

01:46:44.240 Yeah, and math. It was more logical. Things that I can understand and build a mental model for

01:46:48.920 and then be able to predict things. Biology was very much enjoyed memorization. So when I first

01:46:54.840 moved to Iowa, in seventh grade, there was life science in middle school. I remember it was really

01:47:00.560 about memorizing trees and dissecting frogs and identifying anatomical parts. It wasn't so much

01:47:06.780 based on logic. It was during seventh grade that I attended a Saturday enrichment class. It was

01:47:14.020 on molecular biology. I didn't know what biology had to do with molecular things. So I

01:47:18.900 went to it. And in that class, that's when I started to learn about the advances in modern

01:47:25.480 biology. The teacher was very passionate. He taught us about DNA, RNA, protein, the central dogma.

01:47:32.940 But he also had us do fun experiments like extracting DNA from strawberries and putting an

01:47:40.660 antibiotic resistance gene into a bacteria so that it can survive on ampicillin. He also, at the same

01:47:46.800 time, showed us this, I call it a documentary, but it's actually a Hollywood movie, called Jurassic Park.

01:47:53.000 Because if you put yourself in my shoes at a time where you're learning all these molecular biology

01:47:59.280 fundamentals, gene splicing, all that, and then watching Jurassic Park to see all those theories

01:48:07.220 being so tangibly there to make a dinosaur. You know, the movie really felt like a documentary,

01:48:14.760 but it was also just so inspiring and exciting.

01:48:18.200 Because that's about when Jurassic Park came out. It would have been the mid-90s.

01:48:21.840 Yes, that's right.

01:48:22.200 Which is when you were in middle school.

01:48:23.420 Yeah, it was 94. So anyway, so I saw that and I became really interested in learning more about

01:48:29.240 molecular biology. We also learned about gene therapy and the promise of using molecular biology to build

01:48:35.600 rational design medicine. So that really captivated my imagination. The teacher was amazing. His name

01:48:42.740 is Ed Pelkington.

01:48:44.260 Was he only the teacher of the enrichment class or did he also have a regular class and he taught in

01:48:49.820 high school or middle school?

01:48:50.760 He was a consultant for the school and he was working with kids for an interest in science and

01:48:55.840 technology. But he remembered that both myself and also a few other students were really captivated by

01:49:02.880 molecular biology. And so by the time I started 10th grade, he came to us and he said,

01:49:09.280 there's this really cool opportunity that you guys may want to take advantage of.

01:49:14.120 The Iowa Methodist Medical Center had just opened a gene therapy lab and they have a volunteer program

01:49:20.680 at a hospital. And maybe you guys can apply and specify that you wanted to volunteer in the gene therapy lab

01:49:27.160 and maybe you get to go there. So we applied, me and three other kids, and we got admitted to the gene

01:49:33.300 therapy lab. They taught us how to do experiments. As you remember, the very first experiment I did

01:49:39.080 where the scientists had me put the gene for a jellyfish protein called green fluorescent protein.

01:49:46.980 This is a protein that makes jellyfish glow at night. I put this gene into a human cancer cell.

01:49:52.040 Just using an adenovirus?

01:49:54.420 Using just lipid. We wrapped it up with a bit of fat and then it got absorbed by the cell. And so

01:50:00.120 I did that the afternoon, the day before. And then the following day, I went to the lab and the

01:50:06.640 scientists took me into this dark microscope room. There was a beam of blue light coming out of the

01:50:11.560 microscope. We put the petri dish with the cells on it and he told me to look into the eyepiece.

01:50:16.720 And I saw a field of green cells that were fluorescently.

01:50:21.820 That had taken up this gene.

01:50:22.720 It looked alien to me.

01:50:24.140 And you were in 10th grade.

01:50:25.280 I was in 10th grade. And that moment just made me feel so inspired about what we were doing.

01:50:33.580 And this idea that we can use our knowledge to then start to engineer rationally new medicine.

01:50:40.340 From that moment on, I thought science is really cool and I want to do more experimental science.

01:50:44.680 Do you ever just reflect on the direction your life took in response to something as seemingly

01:50:51.060 arbitrary as where your parents chose to move? Your good fortune to be in that enrichment class

01:50:58.200 and then your good fortune to have a very special teacher who not only did this incredible work

01:51:04.120 with you, but then stayed as a mentor, got you to apply to this program as a sophomore in high

01:51:10.680 school. I'm sure you've heard the arguments for the remarkable set of circumstances that allowed

01:51:16.740 Bill Gates and Paul Allen to be in the right place at the right time from the right high school that

01:51:22.620 had a computer lab that took these people who were naturally quite brilliant, but more importantly,

01:51:28.240 put them in an environment where they could leapfrog ahead of the time. Do you see a parallel there?

01:51:34.500 I think I've been very fortunate. And I think there are a couple of things that I over time have

01:51:40.980 developed more and more gratefulness and appreciation for. Number one, I think is the importance of

01:51:47.520 teachers. I've been very fortunate that throughout my life, I've been just met with many teachers who

01:51:54.600 care about their students and teachers who really want to find as many opportunities as possible

01:52:01.580 to help their students develop. They really care about nurturing the next generation.

01:52:06.460 And I think having that and having experienced that was really a great fortune of mine. The second,

01:52:12.420 I think, is really about America. I think having been able to immigrate to the U.S. and to go through

01:52:18.280 the American education system where there's really a merit-based system where you can work hard and be

01:52:26.480 able to learn and develop and have access to opportunities, that is really special about

01:52:33.320 America. So I think education and having a system where it nurtures people who want to develop and

01:52:40.640 work hard, I think, has been my great fortune.

01:52:43.680 Do you think that had you grown up in China, it would have been different? Not because obviously people

01:52:49.720 in China don't go on to do amazing things. It also produces great scientists. But just again,

01:52:54.740 speaking to the randomness of this beautiful situation with this amazing teacher in this

01:53:00.280 program and all of the things that followed, do you think you could have still stumbled onto the

01:53:05.920 direction you would have? Maybe even in a different field. I guess what I'm getting at is you're very

01:53:10.460 modest, but obviously you're very special in what you've done. I wonder if you think you would have done

01:53:16.340 great things regardless, or do you think that no, sometimes you have to be in the right place at

01:53:22.000 the right time? I think fate is important. There was a lot of opportunity in China too. China from

01:53:28.620 the 1980s all the way to now has developed exponentially, but I don't know where I would

01:53:35.080 have ended up. There were a lot more people in China. Things were competitive, and the education

01:53:39.700 system is different. I like to tinker with things, and in China, it's much more memorization-based.

01:53:45.180 I don't know, but I'm very grateful for having had all of the teachers and opportunities here.

01:53:51.540 What did you study at Harvard for your undergrad?

01:53:53.480 I majored in chemistry and physics.

01:53:55.200 And you did that obviously knowing you were interested in genetics, but did you feel that

01:53:59.520 you wanted a greater grounding in the physical sciences because of your ultimate ambitions,

01:54:04.880 or why did you choose that as opposed to biology or molecular biology?

01:54:08.220 Exactly. I knew I wanted to study biology and engineer biological systems, but I felt that

01:54:15.320 biology was developing so rapidly, and new information was accruing on a weekly basis with

01:54:24.400 new papers and new studies. That is probably more beneficial if I study something that's not

01:54:30.580 going to change very much from the time that I enrolled in college to the time I graduate.

01:54:35.380 Right. And so chemistry and physics were much more established fields, and it provided a

01:54:41.060 scientific foundation that I think has been very tremendously helpful as I continue to work

01:54:46.860 in biology. And at the same time, I did work outside of classes in a biological lab where

01:54:53.460 I was practicing biological experiments and reading the latest literature, and I thought that was

01:54:58.840 a nice combination.

01:54:59.560 And when you selected Stanford for your PhD, did you do so because of Carl, or was there

01:55:06.440 another reason, and then you ultimately, after a year or so of classwork, found your way into

01:55:11.720 Carl's lab?

01:55:12.920 Carl hasn't started at Stanford yet. When I was growing up, especially after moving to Iowa,

01:55:18.320 I read about people like Steve Jobs and Bill Gates and internet revolution that was happening.

01:55:25.440 And so I had a special feeling about Silicon Valley. I was very intrigued by it. And so

01:55:32.400 when I applied to graduate school, I thought Stanford being in Silicon Valley would be a really good

01:55:38.260 place to be. I made the decision that way.

01:55:41.360 How did you then make the decision to go back to the East Coast for your postdoc, back to Harvard,

01:55:46.580 MIT area, as opposed to stay in the Silicon Valley and pursue your passions there in closer

01:55:53.720 approximation to industry? Although, of course, everybody realizes Boston would clearly be

01:55:58.680 not too far behind the Silicon Valley for biotech industry.

01:56:02.680 I think at the time, it was probably the most interesting opportunity. I should mention that

01:56:07.240 Carl Dicerath is a phenomenal mentor. I've learned a tremendous amount from him, not only about doing

01:56:13.060 science, but also how to just be a good contributing member of the scientific community with sharing of

01:56:20.500 information and reagents and the transfer of knowledge to as many people as possible.

01:56:25.420 And he was also a great mentor where he just gave me opportunity to try things.

01:56:30.100 So during graduate school, for many months, I was working on a biofuel project in Carl's lab.

01:56:36.740 What year was this?

01:56:37.800 This was 2007, 2008. So it was an energy crisis, oil crisis. Biofuel was an important thing.

01:56:44.840 That's what Alex Hravines and I were working at, at the exact same time.

01:56:48.240 Right.

01:56:48.420 Around that time, I thought this was an interesting problem. So I explored that in Carl's lab. We

01:56:54.700 thought maybe we could even go and raise venture capital and if we had something working out well.

01:57:01.140 But it turned out that around 2008, the financial crisis also hit. And so a lot of those opportunities

01:57:07.380 at that moment seem to have just vanished. But at that same time, I was contacted by someone from

01:57:13.780 the Harvard Society Fellows. And they said, we liked your work as a graduate student. Would you be

01:57:19.260 interested in coming to the Society Fellows and just explore science here? Thought that was interesting.

01:57:25.540 They said, we'll give you a stipend and you're not expected to do anything specific. You can just

01:57:31.820 come and hang out and think about interesting things. So I applied and went there. That's where

01:57:38.100 started to explore the gene editing technology.

01:57:40.380 So you're barely 40 years old. You're one of the most prominent scientists in your field. Are you

01:57:48.020 optimistic about the state of science? And I'll point out two things that are negative that you

01:57:52.840 don't need to comment on. So I don't want to put you in hot water, but there's been a clear attack on

01:57:57.800 meritocracy in the United States. So a lot of the opportunities that served you well might not have

01:58:03.660 served you well today. They might not have been available to you today for various reasons,

01:58:07.140 including your race. And similarly, I think despite all of the remarkable scientific advances during

01:58:13.600 COVID, there was also a loss of public faith in science during COVID, where some of the lines

01:58:19.360 between science and advocacy got blurred greatly, in addition to a whole bunch of other things.

01:58:24.980 So in other words, I guess what I'm saying is there are a lot of things today that don't look as

01:58:28.900 promising as they did maybe 10 years ago, vis-a-vis the field and just sort of the state of the art.

01:58:35.300 Despite those things, do you remain as optimistic as ever, or do you have concerns?

01:58:40.340 I am an optimist. I think that's the only way to be, because if you're not optimistic,

01:58:44.760 then there's only downside when you are like that. But I am optimistic about science. I mean,

01:58:50.020 I'm so excited about all of the rapid advances that are happening. Knowledge about biology, about

01:58:56.820 the human body, physiology is accumulating at a very rapid pace. Every week, there are interesting,

01:59:03.460 exciting discoveries that are being reported. And our technologies for studying biological systems

01:59:09.600 are also accelerating, with sequencing technology development, to protein mass spectrometry,

01:59:17.560 to gene editing, to new microscopy. That is allowing us to collect new and more rich data,

01:59:26.020 cheaper and faster. And then there's AI. With the advance of AI and larger computing platforms,

01:59:33.660 we can analyze and learn and build models around those data that are much more powerful and much

01:59:41.440 more predictive and generative than ever before. And that combined with future advances in robotics,

01:59:49.160 where we can have automation of experimentation, and maybe even building a closed-loop system where

01:59:55.740 AI and with human intervention can design and iterate our experiments rapidly, I think it's

02:00:04.160 going to accelerate science and medicine and human health beyond our imagination. So I think that's

02:00:10.260 really exciting. At the same time, I think we need to continue to motivate people's interest in

02:00:16.320 science. We need to make sure that the best talent are going to science.

02:00:20.560 Do you worry that we've lost the focus on STEM? I mean, obviously, people talk about the heyday

02:00:26.120 of science, right? During the 1960s, when kids growing up saw the space program, saw the Cold War,

02:00:33.080 the best and the brightest wanted to do science, they wanted to do engineering. Do you feel that's

02:00:38.960 slipping away? And if so, what do you think is necessary to bring it back? How do we get the

02:00:43.200 absolute best people into STEM? I think we can do more for sure. I think we need to get kids more

02:00:49.920 excited about science. We need an education system to support kids who are curious and who have special

02:00:57.360 interests in the science and technical things, and to give them opportunities like the ones that I had

02:01:03.640 to really explore those curiosities and interests. When they're young, they are optimistic and they have

02:01:11.100 energy. And if we can nurture that and help them maintain that curiosity and energy, I think that's

02:01:19.080 how we get the best people to continue in science. Yeah, I couldn't agree with you more. And this is

02:01:24.560 not an investment that pays off in a year. You've got to be doing this today, and then you'll get your

02:01:29.600 payoff in a decade, in two decades. And sometimes those investments are really hard for society to make

02:01:35.220 because it's hard to say, I've got to pay money now for something and I don't get paid back for 10 or

02:01:40.080 20 years. Those are hard things to accept. But I hope that people listening to this podcast appreciate,

02:01:45.860 certainly in your case, what an enormous value it was for that investment. In many ways, it's a great

02:01:52.280 story of lots of things. It's a story about immigration. It's a story about science education.

02:01:56.560 It's a story about curiosity. It's a story about the American dream. So anyway, I'm really glad we

02:02:02.880 finally got a chance to sit down. It was absolutely worth the wait. I'm excited to follow your undoubted

02:02:08.260 continued success over the coming decades. Thank you so much for having me here today.

02:02:13.600 Thank you for listening to this week's episode of The Drive. It's extremely important to me to

02:02:18.240 provide all of this content without relying on paid ads. To do this, our work is made entirely

02:02:23.220 possible by our members. And in return, we offer exclusive member-only content and benefits above and

02:02:29.860 beyond what is available for free. So if you want to take your knowledge of this space to the next

02:02:34.160 level, it's our goal to ensure members get back much more than the price of the subscription.

02:02:39.360 Premium membership includes several benefits. First, comprehensive podcast show notes that detail

02:02:45.440 every topic, paper, person, and thing that we discuss in each episode. And the word on the street is

02:02:51.280 nobody's show notes rival ours. Second, monthly ask me anything or AMA episodes. These episodes are

02:02:59.160 comprised of detailed responses to subscriber questions typically focused on a single topic

02:03:04.020 and are designed to offer a great deal of clarity and detail on topics of special interest to our

02:03:09.380 members. You'll also get access to the show notes for these episodes, of course. Third, delivery of our

02:03:15.280 premium newsletter, which is put together by our dedicated team of research analysts. This newsletter

02:03:20.660 covers a wide range of topics related to longevity and provides much more detail than our free weekly

02:03:26.900 newsletter. Fourth, access to our private podcast feed that provides you with access to every episode,

02:03:33.760 including AMA's sans the spiel you're listening to now and in your regular podcast feed. Fifth,

02:03:40.720 the qualies, an additional member-only podcast we put together that serves as a highlight reel featuring

02:03:46.900 the best excerpts from previous episodes of the drive. This is a great way to catch up on previous episodes

02:03:52.640 without having to go back and listen to each one of them. And finally, other benefits that are added

02:03:57.440 along the way. If you want to learn more and access these member-only benefits, you can head over to

02:04:03.000 peteratiamd.com forward slash subscribe. You can also find me on YouTube, Instagram, and Twitter,

02:04:09.940 all with the handle peteratiamd. You can also leave us a review on Apple Podcasts or whatever podcast

02:04:16.520 player you use. This podcast is for general informational purposes only and does not

02:04:21.720 constitute the practice of medicine, nursing, or other professional healthcare services, including

02:04:26.100 the giving of medical advice. No doctor-patient relationship is formed. The use of this information

02:04:32.220 and the materials linked to this podcast is at the user's own risk. The content on this podcast is

02:04:38.420 not intended to be a substitute for professional medical advice, diagnosis, or treatment. Users should

02:04:43.700 not disregard or delay in obtaining medical advice from any medical condition they have, and they

02:04:48.760 should seek the assistance of their healthcare professionals for any such conditions. Finally, I take

02:04:54.360 all conflicts of interest very seriously. For all of my disclosures and the companies I invest in

02:04:59.840 or advise, please visit peteratiamd.com forward slash about where I keep an up-to-date and active list

02:05:07.300 of all disclosures.

02:05:13.700 Thank you.

The Peter Attia Drive - October 28, 2024

#323 - CRISPR and the future of gene editing： scientific advances, genetic therapies, disease treatment potential, and ethical considerations ｜ Feng Zhang, Ph.D.

Episode Stats

Length

Words per Minute

Word Count

Sentence Count

Misogynist Sentences

Hate Speech Sentences

Summary

Transcript