The Peter Attia Drive - #268 ‒ Genetics： testing, therapy, editing, association with disease risk, autism, and more ｜ Wendy Chung, M.D., Ph.D.

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

00:00:16.580 my website, and my weekly newsletter all focus on the goal of translating the science of longevity

00:00:21.580 into something accessible for everyone. Our goal is to provide the best content in health and

00:00:26.780 wellness, and we've established a great team of analysts to make this happen. It is extremely

00:00:31.720 important to me to provide all of this content without relying on paid ads to do this. Our work

00:00:37.000 is made entirely possible by our members. And in return, we offer exclusive member only content

00:00:42.760 and benefits above and beyond what is available for free. If you want to take your knowledge of

00:00:47.980 this space to the next level, it's our goal to ensure members get back much more than the price

00:00:53.260 of the subscription. If you want to learn more about the benefits of our premium membership,

00:00:58.080 head over to peteratiyahmd.com forward slash subscribe. My guest this week is Dr. Wendy

00:01:06.440 Chung. Wendy is a board certified clinical and molecular geneticist and the new chief of

00:01:12.520 pediatrics at Boston Children's Hospital. Wendy earned her PhD in genetics from Rockefeller

00:01:17.900 University and an MD from Cornell University Medical College. She completed her residency in

00:01:22.760 pediatrics and her fellowship in molecular and clinical genetics at Columbia's New York

00:01:26.980 Presbyterian Hospital, where she then served as a professor of pediatrics and directed her 0.97

00:01:31.540 research programs towards the genetics of obesity, diabetes, breast cancer, autism, and rare diseases.

00:01:38.680 Wendy has received numerous awards for her research, as well as for her clinical and teaching

00:01:43.380 contributions, including being elected to the National Academy of Medicine. In my conversation

00:01:48.880 with Wendy, we focus on genetics from a variety of angles. We talk about what science and genetics

00:01:54.300 looked like before we could decode the human genome, as well as what we know currently when it comes to

00:02:00.320 whole genome and exome sequencing. This includes an understanding of the difference between clinical

00:02:05.380 genetic testing and what's available commercially. We also speak about genetics and newborn screening,

00:02:11.200 as well as a project that Wendy is involved in called the Guardian Study. We talk about genetics as it

00:02:16.720 relates to a variety of conditions, including PKU, which some of you may have heard of if you've ever

00:02:22.400 noticed on a diet soda can, it says if you have PKU, don't drink this, breast cancer, obesity, autism,

00:02:27.960 and cardiovascular disease. We ultimately talk about gene therapy, how it works, and what's required to

00:02:33.460 change a gene, and of course the future and ethics of gene therapy. So without further delay,

00:02:38.900 please enjoy my conversation with Dr. Wendy Chung.

00:02:46.720 Hey, Wendy. Thanks for making time to chat today. This is an especially busy day. As I learned,

00:02:52.680 you're literally in the process of moving from New York to Boston later today, no less. So I'll try not

00:02:58.720 to get in the way of that transition. But that's probably a good intro to kind of explain what it is

00:03:03.720 you do. You're moving from one prestigious institution in New York to another in Boston. Tell us where you're

00:03:11.340 going. Sure. I'm going to be the chair of pediatrics at Boston Children's Hospital and will be at

00:03:16.060 Harvard Medical School. You're both an MD and a PhD. How do you balance your time between,

00:03:22.760 let's not include the new responsibilities that will be administrative, but up until now,

00:03:26.800 how have you balanced your time between the lab and clinical practice? How do those split?

00:03:32.120 So they split about 20% clinical, 80% research, but truth be told, they're really together. So when I

00:03:39.140 think about things, I always say it starts with the patient and ends with the patient. So it starts with

00:03:43.300 the patient to me in terms of clinically sealing them. Many times the answers aren't obvious. And

00:03:48.840 so it becomes a research question. And at the end of the day, though, it has to go back to the patient.

00:03:53.520 So within this, that split, I think just signifies how much we have to learn and why research is so

00:03:58.780 important to improve clinical care. You didn't do a combined MSTP or MD-PhD program. You did your

00:04:04.720 PhD first and then went to medical school or did you do the combined program?

00:04:09.100 Yeah, I did a combined program between Cornell and Rockefeller. And since Rockefeller doesn't

00:04:13.460 have a medical school, we do that with Cornell.

00:04:15.860 I see. So presumably you knew you wanted to be a physician scientist as you went through training.

00:04:20.280 Yes, that's right.

00:04:21.220 What drew you to your current field of genetics? How would you describe to somebody what it is you do

00:04:26.360 in the lab?

00:04:27.320 I'll give you the short version of this, but I was fortunate enough early in my career to be exposed

00:04:32.200 to the National Institutes of Health and was able to spend as an undergraduate summers there.

00:04:37.500 And that was really when I was a biochemistry major as an undergrad, but had the ability to

00:04:42.120 work on phenylketonuria. And although it was mostly in the laboratory, was able to spend time at NIH at

00:04:47.500 the hospital and seeing patients and realize this whole paradigm, which to this day is really how I

00:04:52.900 do things about thinking about how these pieces fit together. And I couldn't really think about the

00:04:57.280 science without thinking about the patients. And I couldn't move forward and fill the gaps in our

00:05:01.840 knowledge for patients unless I did the science. I happened to, I think, be good at both. And so

00:05:06.500 it was a natural to me to do both. And I was young and not so worried about the number of gray hairs

00:05:11.500 that I would have by the time I finished. And so sat out on this relatively long training path,

00:05:16.500 but one that suited me extremely well.

00:05:18.720 Let's talk a little bit about PKU. It's maybe a good introduction to a genetic disease. Maybe tell

00:05:24.080 folks what it is and how frequently it occurs.

00:05:26.680 PKU stands for phenylketonuria. We may come back to it a little later in the show,

00:05:30.860 but it was actually how we started newborn screening. It was, to me, a paradigm in terms of

00:05:36.720 patients and families really working together to improve care and being very much partners in that.

00:05:42.320 And so that was even true for a condition that I started studying. As a biochemistry major,

00:05:46.900 I was interested specifically in the biochemistry of that. But I realized, and this was in the late 80s,

00:05:53.200 that a lot of what I was doing was doing genetic sequencing to understand the genetic basis of this,

00:05:58.600 what we call Mendelian or single gene condition. In this particular case, we did know the gene for

00:06:03.980 this condition, but there were so many other things that I was seeing that we didn't yet know

00:06:07.740 the genes for these underlying conditions. And it was coincidental, but really just, I would say,

00:06:14.060 fortuitous for me that the year I started my MD-PhD program was the year the Human Genome Project was

00:06:19.540 announced in terms of going forward. And I had, I think, the, I don't know, maybe foresight or good

00:06:25.700 luck to be able to see that it was going to be a brand new future when we would have that entire

00:06:31.360 encyclopedia of information to be able to think about human disease differently and thinking about

00:06:37.120 opportunities in the future and what one could do with the information when we had it. And so I

00:06:42.260 really planned my career in thinking about what I'd be able to do 10 or 20 years after I started,

00:06:47.920 and that ended up proving well. And so I've spent a lot of my time using that information and trying

00:06:52.340 to apply it to health. So what's the clinical manifestation of PKU?

00:06:57.020 So phenylketinuria is a bittersweet condition in the sense that can be associated with intellectual

00:07:03.120 disabilities if not treated, but if caught early and if treated with a diet that is restricted in

00:07:09.820 phenylalanine, one of the amino acids that we see in proteins. If we restrict that,

00:07:14.260 then even though individuals with PKU can't digest that and get rid of the toxic byproducts,

00:07:19.720 we can prevent those toxic byproducts from building up in the body and essentially poisoning the brain.

00:07:25.180 I say tragic because, and again, we may get to it, I've identified individuals who weren't picked up

00:07:30.980 through newborn screening and picked up through some of my research studies, for instance, as teenagers

00:07:35.420 and had irreversible intellectual disabilities. But yet we can prevent those types of problems if these

00:07:42.000 children are identified as newborns. And so that's what our whole newborn screening program was

00:07:46.640 originally predicated on, is that early diagnosis, early intervention, changing lives, improving lives.

00:07:52.220 And in fact, we do that extremely well for most individuals with PKU.

00:07:56.500 Is PKU a dominant gene and is it fully penetrant or is there any variability in that?

00:08:02.940 That's a great question. So it's a recessive condition, meaning it takes two to tango.

00:08:06.800 Both your parents are carriers for that condition. Within this, there is a spectrum. So we have

00:08:12.720 individuals that are what we call hyperphenylalanemic. So they don't technically have PKU. They're not

00:08:18.580 symptomatic. They wouldn't have problems with intellectual disabilities, but it's a spectrum

00:08:22.860 of severity. So beyond a certain threshold, you have too much of the toxic buildup, and that's when you end

00:08:28.640 up with the problems in terms of brain function. But there are some individuals who have what I'll call

00:08:33.480 subclinical phenotypes. That is, I could see it if I measured, if I bothered to measure the

00:08:37.980 amount of phenylalanine in their blood, but they have enough of the enzyme to be able to clear the toxic

00:08:43.180 byproducts. So that becomes one of the tricky things for us to do in screening newborns is to decide where

00:08:49.100 that threshold is to adjudicate this and figure out who needs treatment.

00:08:53.560 So is the screening that's done on newborns a genetic screen where you're looking for two copies of the gene?

00:08:58.840 The traditional way we do this in the very old days, believe it or not, was based on

00:09:04.020 looking at bacteria and how they would grow on an auger that was depleted in phenylalanine. And we

00:09:09.860 would take a heel prick from a baby and put a little dried blood spot punch of that dried blood spot onto

00:09:17.060 a bacterial lawn and see again where the bacteria would grow. So in the very old days, in the late 60s,

00:09:23.400 early 70s, we did the screen in that way. We got more sophisticated and had ways of being able to

00:09:28.720 directly measure phenylalanine. And we now do it with a process called tandem mass spectrometry,

00:09:33.400 but we still use that dried blood spot to be able to do it. Interestingly enough, with a program that

00:09:38.620 we've recently started called Cardian, we also have an orthogonal way of being able to screen,

00:09:44.100 which is based on looking at the DNA. And so that we've got two different, essentially,

00:09:48.280 data streams coming in to help us with the adjudication of what I was describing before.

00:09:53.220 Who really needs treatment? Where's that threshold? And being able to be

00:09:57.060 better in terms of our test parameters, both sensitivity and specificity, so that we can

00:10:03.080 really identify with greater certainty the babies who need treatment.

00:10:07.280 That's interesting. So just to make sure I understand, the reason you don't just rely on,

00:10:12.020 say, a genome sequence at the moment, even though you know what gene to look for, it would be targeted,

00:10:16.460 you wouldn't need to do a whole genome sequence, is because that wouldn't tell you about the phenotype

00:10:20.840 fully. And the phenotype is just as important as the genotype may be more important as you make

00:10:26.140 a clinical decision about dietary restriction.

00:10:28.760 Yep. So it's very insightful, and I'm going to repeat it back just because it might be a

00:10:33.100 subtle thing for some people. Reading out the DNA sequence, we're good at, we're not perfect at.

00:10:38.860 And so being able to decide based on your DNA sequence alone for phenylalanine hydroxylase,

00:10:44.780 the relevant gene, we're not able to perfectly make that one-to-one correlation about whether or not

00:10:50.600 you'll be beyond a threshold of disease in terms of phenylalanine levels. So we do need,

00:10:55.840 in this case, I'll call the phenotype, the level of phenylalanine levels in the blood.

00:11:00.020 We do need to have that phenotype before we get to the phenotype of intellectual disabilities,

00:11:05.900 which is what we're trying to prevent. So you're right that in this case, we use two different

00:11:10.820 data streams to come in. The other reason why the phenotype alone is imperfect is you can imagine,

00:11:16.320 depending on what you've eaten, your phenylalanine levels fluctuate during the course of the day.

00:11:20.980 And so we do it as a cross-sectional one time, and you might happen to get a baby at just the wrong

00:11:26.560 time or just a, you know, sort of higher level. And so being able to have both of those data streams

00:11:31.900 come in allow us to be even better in terms of the accuracy.

00:11:35.560 Now, it's interesting. If you look at a can of soda or anything that has aspartame in it,

00:11:40.000 it always presents this warning and says, if you have PKU, beware. I always find that kind of

00:11:45.440 interesting because the absolute amount of phenylalanine in a minuscule amount of aspartame,

00:11:51.640 which is found in a Diet Coke or something, seems really small. Is that clinically significant?

00:11:56.520 And if so, wouldn't that suggest that even milligrams of this amino acid could be problematic

00:12:02.020 to those afflicted?

00:12:03.400 So for those afflicted, they have to be really careful. If you actually have PKU, then we have

00:12:08.080 a very special diet for you. You're not taking diet sodas. You're not doing anything with aspartame.

00:12:13.100 In fact, we may have you on what's effectively a pretty not-so-fun diet to be on. And in fact,

00:12:19.280 many people don't want to be on that diet for life because it's not the tastiest diet.

00:12:23.580 But for women in particular, when they become pregnant, it becomes not only their own body

00:12:28.360 they're influencing, but that of the fetus developing inside. And so we have to be quite

00:12:33.920 careful with women with PKU when they're pregnant as well. And because of that, from a labeling point of

00:12:39.180 you, we want to make sure they're aware of what's in different food products that they might be eating

00:12:43.500 because they won't feel the effect right away. It's not as if they get a headache or something

00:12:47.460 like that. And we don't want to see the effects on the developing fetus later on. 0.64

00:12:51.520 You know, given that amino acids show up in all sorts of places, I mean, and it's not just like,

00:12:57.100 well, I'm eating eggs, so therefore I'm just eating methionine. I mean, are there actual protein

00:13:02.360 sources that are completely void of phenylalanine? I mean, there must be if you're able to subside on

00:13:07.620 some sort of phenylalanine-free diet. What types of foods are excluded completely from this diet?

00:13:14.540 We don't exclude 100% phenylalanine, so it's a low-protein diet in general. So we still need

00:13:20.660 some protein, and you still have some essential amino acids for your body to be able to grow.

00:13:25.160 You need to make muscle, especially as a developing child. There's a lot of growth that's there,

00:13:29.400 so we don't completely restrict. And it's actually a titration, if you want to think of it that way.

00:13:34.320 That is that we make dietary interventions, and we check, and then we go back, and we diddle,

00:13:38.480 and we go back and forth to be able to get it just right. So it's a lifelong treatment in that way,

00:13:43.360 and it becomes even more critical for young children as both their brain is developing,

00:13:48.180 their body is developing. We have to get it just right. So not trivial to do, but on the other hand,

00:13:53.500 when we're good about it, children grow up very healthy.

00:13:56.160 But it is lifelong treatment. In other words, just getting through adolescence is not enough.

00:14:00.000 If you come off the diet later in life, will you still suffer cognitive changes,

00:14:04.960 or are you most sensitive to those during development? And by development, I mean sort

00:14:08.580 of adolescence and childhood. You're most sensitive during childhood, for sure, when the brain is

00:14:12.920 growing and when you're making those synapses and connections and being able to develop all of

00:14:17.980 the things you're learning to do, you're definitely most sensitive then. I find that some people,

00:14:22.720 adults in particular, will tell me about differences that they have in terms of clarity of their thinking

00:14:27.000 and other things if they're just totally off the diet and not restricted whatsoever. But it is

00:14:31.960 different in the sense that you don't crash and burn instantaneously. It's more subtle in terms of

00:14:36.380 what you feel, how your body feels. You know, there are certain diseases, we'll probably talk about

00:14:40.500 them like sickle cell anemia, where they're recessive conditions, but having one copy of the gene and

00:14:47.240 therefore not having the full phenotype poses an advantage. And that's at least in some part explains the

00:14:53.280 propagation of the gene. Is there any such analogy to be made here? Are there benefits in having one copy

00:15:00.000 of this gene? And obviously, there are huge detriments to having two copies.

00:15:05.060 So not that we know of. I will say that I don't think that we know everything, but it's not so obvious

00:15:09.420 in terms of the frequency of these mutations. There are, I think, historical reasons why we see it in certain

00:15:15.100 parts of the world versus others. But as far as we know, no selective advantage.

00:15:19.800 And is there an ethnic distribution?

00:15:21.980 There are. So we do tend to see this, for instance, more in, for instance, Ireland. It tends to be more

00:15:26.720 frequent than we see in other parts of the world. Sub-Saharan Africa is an example. And, you know,

00:15:31.680 that has to do with historical reasons in part and where variants occurred, migration patterns of how

00:15:37.240 peoples migrated around the world. But we do see it really for all four corners of the world. PKU is seen 0.99

00:15:43.320 everywhere. And in fact, in newborn screening, pretty universally screened throughout the world for places that have

00:15:48.720 newborn screening programs.

00:15:50.240 Just to paint the contours of it, in Ireland, what's the frequency that a child is born with this?

00:15:55.620 It's a good question. I'm going to hazard a guess, although I'd have to say I'd need to fact check

00:15:59.440 myself. We're probably in the order of one in 5,000 or so.

00:16:02.520 Okay. And in the U.S., less than that, but by what mark?

00:16:05.240 A little bit less than that.

00:16:06.040 Okay.

00:16:06.300 Maybe one in 10,000.

00:16:07.560 So still pretty common condition, relatively speaking.

00:16:10.540 Yep.

00:16:10.700 I suspect we're going to talk about this in much greater detail as we go on. But before we do,

00:16:15.080 I'll want to build up a foundation so people understand the basics. But just before we leave

00:16:19.220 PKU, is this something that you think in your career will be a target of gene therapy?

00:16:24.440 It's a great question. And I have to say something I think a lot about. I think we have yet to see

00:16:29.520 certainly many inborn errors of metabolism. So things that are conditions like phenylketonuria,

00:16:35.080 but other things that have to do with the way the body digests, processes, metabolize foods,

00:16:39.820 are, some would say, easier targets for gene therapy. And we can go into more or less detail

00:16:45.720 about this. But in part, many of these genes are expressed in the liver. So the liver is kind of

00:16:50.980 the, if you want to think about it, the metabolic brain of the body or the metabolic clearinghouse.

00:16:55.660 The brain, or rather the liver, is a relatively easy place to target in terms of gene therapy.

00:17:00.860 There are ways that the liver is clearing things and some of the vectors and the delivery systems we

00:17:05.600 use are relatively easy to get to the liver. And for certain conditions where these are recessive

00:17:11.620 conditions with loss of function, you have to add back the missing enzyme or protein,

00:17:16.960 but you probably don't have to get to 100%. And just for some of your listeners will be astute.

00:17:22.360 You had alluded to carriers for these conditions, recessive individuals who have one copy of the gene

00:17:28.920 that's working just fine, but one copy of the gene that's not. They tend to be fine. That was the

00:17:34.680 point, essentially, of what you were saying. Do carriers have any advantage or disadvantage? And

00:17:39.140 basically, they're indistinguishable, is what I would say. Which means for us, in terms of a gene

00:17:45.060 therapy or gene addition or gene replacement strategy, you don't have to be perfect. You have

00:17:50.360 to get some in there. You have to get enough in there, but you don't have to get 100%. And so for all

00:17:54.960 those reasons, these types of conditions are interesting in terms of gene therapy targets. And as you alluded

00:18:01.820 to, given that the treatment is lifelong, that kind of stinks. And so for something that could

00:18:07.460 be transformational in terms of quality of life, many of these metabolic disorders certainly are

00:18:13.300 interesting in terms of genetic therapies. So we may come back to it, but I will not be surprised if

00:18:18.920 within our lifetimes, people will be trying genetic therapies for PKU.

00:18:23.280 I definitely want to come back to this, both from a historical context and then also to talk about the

00:18:27.960 future. But before we do, it might make sense now to pause and kind of go back to some of the basics.

00:18:32.340 I know that our listeners are quite sophisticated in general, but I always think it helps to just put

00:18:37.300 some foundational knowledge in place. So you talked about how you began your PhD. I'm just doing the

00:18:42.580 math, but it sounds like you began your PhD in the early 90s. And this was about a decade before the

00:18:47.880 human genome was sequenced. So at that time, you know, obviously we understood the structure of DNA.

00:18:53.540 We understood that it was a double helix. We understood that DNA was a template that was used

00:19:00.460 to make RNA and that RNA was then used to make protein. And that's essentially the axiomatic

00:19:06.940 principle of life. Although there, maybe we can talk about some edge cases there. How back in that era,

00:19:13.920 like how did you do genetic work? Maybe explain the differences between sequencing, protein gels,

00:19:19.760 and what the state of science was a decade before the human genome was sequenced. And also,

00:19:25.600 if you can remember, speculate on what was believed to be the outcome of the human genome sequence and

00:19:31.720 how that differed from what was actually found. So I'll give you a couple examples that'll bring

00:19:36.440 a smile to some graduate student's face out there somewhere. So when I first started my PhD,

00:19:41.820 eventually PCR, which people know about polymerase chain reaction, is a molecular Xerox machine that we

00:19:47.620 use to amplify DNA and use it for sequencing and other things. We relatively soon after I started

00:19:53.700 graduate school had automated thermocyclers, but within PCR, one had to change the temperature for

00:19:59.720 different stages of this amplification process. We had some where you'd have to denature the DNA,

00:20:04.720 so you'd heat things up. Others where the enzyme, the polymerase, would work at a different temperature,

00:20:09.600 so you'd have to bring the temperature down. And so there were three different temperatures that

00:20:13.260 you'd cycle at. In the very early days, we didn't have machines that would cycle between these

00:20:17.140 three different temperatures or go through 30 different cycles of this. So you can imagine

00:20:21.560 the cheapest labor as a graduate student. So you'd have an ice bucket, you'd have a heating block,

00:20:26.180 and you'd have, you know, sort of a bath, a water bath in terms of a different temperature

00:20:30.320 with a timer in which you'd be literally moving samples and you'd be essentially a robot being able

00:20:35.440 to do this in the early days. And we'd work with radioactivity to do the DNA sequencing. We'd have

00:20:41.100 these gels where we'd be reading out these ladders of sequence. It was all very manual and not very high

00:20:46.180 throughput. Certainly, I did that, as I said, reading out the phenylalanine hydroxylase gene

00:20:50.760 to be able to see all of this. And those were the early days. But if you think about scale and what

00:20:55.780 was necessary to do this for 3 billion base pairs, there was no way that that could be scaled. And so

00:21:00.700 whole industries evolved in terms of being able to do this in a more automated way to be able to do

00:21:06.700 this. And really the whole world organized itself around ways to do this massive project. In the early

00:21:13.760 days, we had different chromosomes that were designed to different areas. So Columbia used to

00:21:18.120 be the chromosome 13 center of the universe in terms of being in charge of that. That meant that

00:21:23.320 chromosomes are ordered by largest with the smallest number. So chromosome 1 is the largest chromosome.

00:21:28.080 So you'd have some groups that had bigger jobs than others. But we would spread these around and

00:21:32.560 different groups would come together from around the world for a chromosome 13 meeting, for instance,

00:21:37.340 and try and compare notes. We would have things that we called yeast artificial chromosomes in which we

00:21:42.420 literally under a microscope dissect out these chromosomes and put these into these constructs so

00:21:47.660 that we could make more of the DNA and be able to go through and sequence these. There have been

00:21:52.220 transformational technologies that have allowed us to go through in terms of higher throughput,

00:21:57.060 greater processivity. But one of the things that the Human Genome Project did that I think was really

00:22:01.860 important was in the early days, data sharing, data access, really being able to make the world come

00:22:07.740 together by allowing large groups of people to work together. It wasn't every person for themselves.

00:22:13.820 It really was a scientific enterprise collectively. And that's, I think, a fundamental principle in which

00:22:20.320 many of us as genomicists believe very firmly in, in terms of data sharing, privacy, and protecting

00:22:26.180 individuals, individual patients, individual participants, but yet being able to make data freely

00:22:31.380 available as immediately as we can so that we can all use it and get smarter together and learn from each

00:22:36.760 other and be able to advance science as quickly as possible.

00:22:40.520 I just can't help but want to go back even a little bit further. I remember one of the most

00:22:44.160 interesting books I read, oh, probably in medical school, was The Double Helix, which is, of course,

00:22:50.240 the relatively short but completely fascinating and gripping story of the discovery of the structure of DNA.

00:22:56.900 Do you want to just briefly explain? Because I sort of think right now we are so far removed

00:23:01.840 from how much ingenuity was required to figure out that structure and how that laid the foundation

00:23:08.540 for all that came. And refresh my memory. So this was, what, 1953? Is that right?

00:23:13.760 Yeah, I think that's about right.

00:23:14.840 Yeah, in that ballpark. So up at that point, did we understand that there were 23 pairs of chromosomes?

00:23:20.540 We already knew that, correct?

00:23:22.320 Believe it or not, there were arguments in the early days about whether it was 46 or 48 total

00:23:27.500 chromosomes and pairs. And it was hard to visualize. Eventually, you know, we got the counting down.

00:23:31.900 We were able to separate them out enough by size and eventually banding pattern. But

00:23:35.700 in the early days, even controversy about that.

00:23:38.820 So what was it that the four individuals, typically just gives credit to Watson and Crick,

00:23:43.660 but really there were four people that played a pretty pivotal role in this.

00:23:47.320 What was the breakthrough that they had that allowed them to understand

00:23:51.140 the structure of this molecule?

00:23:53.440 So I won't say that this is my super subspecialty, but in terms of the crystallography

00:23:58.620 structure, being able to do that, get a high enough resolution and really have, I think,

00:24:03.220 the insight in terms of being able to imagine this was all of these things coming together,

00:24:08.060 quite technical, but also, as you said, I think some unsung heroes in this story as well.

00:24:13.920 And really, it was this remarkable ability to look at 2D images that were captured and understand

00:24:20.480 the mathematics and the picture that this had to be a double helix. And it's interesting when you go

00:24:26.400 back and look at some of the other proposed ideas, each one of them had a shortcoming. Each one of them

00:24:31.620 made sense until you realized, nope, this wouldn't project in this way or that way.

00:24:37.200 So what was the first human gene that was identified? How long after the structure of DNA? I'm curious as to

00:24:44.980 what the gene was. And more importantly, I guess, what were the methods used to identify a gene,

00:24:50.240 you know, long before we had sequencing?

00:24:52.480 So there were biochemical things that were done. So you mentioned sickle cell disease as an example.

00:24:57.920 So we had proteins, we knew about proteins, protein electrophoresis, being able to see that. And so

00:25:03.900 conditions like sickle cell disease and other hemoglobinopathies, we knew at the protein level

00:25:08.320 well before we knew at the DNA level. So that was something pretty characteristic. Other cases we

00:25:13.740 knew based on enzymatic activity. And so we could see biochemically in a test tube, if you will, what

00:25:19.040 the reaction that was run. So we knew about many of those things before we ever knew the exact DNA

00:25:23.600 sequence or exactly what the genetic variants were that caused those conditions.

00:25:27.700 And by the time you were a graduate student, so still pre-human genome sequence, so right around the

00:25:33.400 time that Rudy Leibel is figuring out what leptin is and things like that, how much resolution did

00:25:39.000 you have into what a gene looked like at that time? So it was pretty painful at the time. We would use

00:25:44.940 these things called linkage maps to try and figure out what chromosome, something, a condition was on,

00:25:49.880 be able to get closer through linkage analysis to the right neighborhood, the right zip code,

00:25:54.120 eventually the right address. When we did this, we didn't have great signposts to be able to even

00:25:59.080 figure out where we were within this. We didn't have things like structures of genes, references.

00:26:05.480 So as you were doing this, you were sequencing not just one person with the disease, but also you had

00:26:09.760 to sequence quote unquote normal people or average people for comparison. We didn't have that as

00:26:14.080 something you could just look up online. We didn't have the internet as an example at the time to be

00:26:19.100 able to see, you know, have investigators work together from around the world is what's much more

00:26:24.020 sort of old school passing papers back and forth and, you know, meeting in various locations.

00:26:30.160 So it was a lot slower, a lot more laborious. And with this, I have to say, and you had mentioned

00:26:35.020 Rudy Leibel, he did have this big, bold idea in terms of cloning a gene for obesity. And the first

00:26:40.780 time he put in a grant, putting out that idea, people thought it was just totally ridiculous. The

00:26:45.480 idea that you could, we called it positional cloning, but identify a gene solely based on its position

00:26:51.060 within the genome understanding and not requiring any understanding of the biology or physiology,

00:26:56.920 but just purely based on genetics and genetic mapping. People thought it was impossible to do.

00:27:02.280 So, and of course, subsequently, you know, a whole generation of disease or a whole generation of

00:27:06.560 scientists found diseases that way. If you can imagine that process was often a decades long process

00:27:12.960 or longer. I mean, this was not something you did, you know, very quickly. So I often tell people,

00:27:18.240 you know, the first gene that I cloned took eight years. The last gene I cloned took eight hours. So,

00:27:23.080 you know, this is just absolutely astronomically different in terms of how we now find disease

00:27:27.600 genes. Yeah, it's very interesting. I looked at a graph. The speed that you're describing even

00:27:32.880 exceeds Moore's law. It's on a Moore's law trajectory with an enormous step function when high throughput

00:27:38.480 sequencing came along, which we can probably get to later because I think it's important for folks to

00:27:42.780 understand that. What was the first organism for which we had a whole genome sequence?

00:27:46.620 Within that, it was certainly a microorganism. I don't remember if it was E. coli or something

00:27:51.000 similar in terms of a bacteria, but definitely a very small organism. Yeast were another important

00:27:56.420 part of our library. And so being able to understand those small organisms, certainly much easier.

00:28:02.980 You know, even when it comes to the complexity of the human genome, some of us would say that,

00:28:07.720 well, it's been announced repeatedly that the Human Genome Project has been finished,

00:28:11.400 but it's only been recently even that, as we say, telomere to telomere,

00:28:15.220 we've been able to see the sequence, really the entirety, including some of the

00:28:19.600 cryptic portions that are hard to read through. So there's still things we have to discover

00:28:23.980 even within the human genome. So before the human genome was

00:28:28.700 sequenced, which I think was around 2000? The first announcement, yes.

00:28:32.720 There was an expectation that humans would have how many genes based on the understanding of how

00:28:38.000 many genes these far, far simpler organisms had. So there were estimates for some people as many as

00:28:43.400 100,000 genes. By comparison, I think current estimates are about 20,000 genes in terms of the

00:28:49.160 complexity of people, humans, that is. But yet the complexity at the individual gene level is probably

00:28:55.200 more complicated than we appreciated. Our ability to have different what we call isoforms or versions of

00:29:01.800 the way genes are cut and pasted together or how they're utilized in different ways over time and space

00:29:08.340 by different organs or cell types. Any one gene could be made into a dozen or more different gene

00:29:15.560 products. And so some of that complexity was not at the level of the individual gene, but how that gene

00:29:22.180 is reused in slightly different ways. And so anyway, it was shocking the first time I think we appreciated

00:29:29.080 that it was about 20,000 genes in the genome. There were definitely people that were surprised by that.

00:29:34.340 Yeah, it seems like such a small number given the variation between individuals. In fact,

00:29:40.700 outside of identical twins, we have, what, 8 billion people on this planet, all with distinct genomes.

00:29:46.820 And yet the homology between us is how strong? Like, in other words, how similar are we all genetically?

00:29:52.880 So we're all 99.9% the same. About one in a thousand base pairs is what you and I probably differ

00:29:59.080 by on average. So as humans, we're pretty similar to each other. And most of the genetic variants

00:30:05.220 differences that we have are not meaningful. You know, they don't cause any differences in terms of

00:30:09.400 the way our bodies function. On the other hand, you know, as something as subtle as one in three

00:30:15.340 billion base pairs can be the difference between life and death, can be the difference in terms of

00:30:20.140 the way the body or the brain functions. So small nucleotide differences can be profound

00:30:25.420 depending on what genes and when those genes work. So basically 3 billion base pairs, 20,000 genes,

00:30:33.060 46 chromosomes is sort of the hierarchy of organization. At what point, well, maybe just

00:30:39.020 explain to folks the difference between coding and non-coding portions of the gene.

00:30:44.440 So when you think about all of those A's, T's, G's, and C's that you talked about,

00:30:49.320 it's a relatively small portion of that information that gets moved from the DNA to eventually the

00:30:56.300 protein. So the portions of that that are made into the proteins are about, I don't know, let's say

00:31:02.220 for round numbers, about a percentage and a half of all of that DNA sequence. That other, say, 98.5%,

00:31:09.220 there's a lot of it that, to be honest, we have no idea what it does. There are certain portions we do

00:31:13.660 understand. They're very important for regulation to know where and when and how much that gene is

00:31:19.860 expressed. There may be other things that are subtle in terms of being able to attract binding

00:31:25.360 factors, transcription factors, other things that may modify the DNA, biochemical changes to the DNA

00:31:31.740 itself, which may affect expression. And there are also what people have called junk DNA as well in

00:31:36.800 there. Repetitive sequences that probably don't do anything positive for us, but, you know, get carried

00:31:42.400 along in the ride. But there's a lot that we also don't know. We don't know everything clearly about

00:31:47.880 this. And there may even be disease-causing variations that are in that space that we

00:31:52.180 haven't even recognized yet. To your point, it's a small minority that actually encodes ultimately

00:31:58.140 what we think of as most of what forms the body, physically forms the body in terms of proteins.

00:32:03.980 And in 2000, when the Human Genome Project results were announced, what fraction of those

00:32:10.600 3 billion base pairs were identified? So within that, I don't know, we'll say round numbers about

00:32:17.020 70% or so. Interesting. And did that include all of the coding segments? Or was it not yet understood

00:32:23.140 at that time what fraction of those were coding and non-coding? The majority of the coding was

00:32:28.280 identified at that time, to answer your question. There were a few portions of the genome that were

00:32:33.460 hard to read out for various reasons or hard to map and put the pieces together. One of the things,

00:32:38.820 just for the listener to realize, is that there was a bit of a jigsaw puzzle. When we did, and when we

00:32:44.160 do do the sequencing in many cases, we're not sequencing, I use the term telomere to telomere,

00:32:50.320 or end to end along the chromosome. So it's not as if we get one continuous strand of the DNA sequence

00:32:56.040 that comes off the sequencers, or we can just read through it and know it comes together. In many cases,

00:33:01.520 we have pieces of it, and we have to informatically put the pieces back together and put it back into the

00:33:08.480 right order. And in some cases, that's because we have overlap between those pieces. And so we can

00:33:13.340 see based on overlap, this is the first piece, that must be the second piece, that the third piece

00:33:17.540 based on the overlaps. And so we make these things called contigs or contiguous sequences of DNA and

00:33:24.100 put the puzzle and the pieces together in that way. There are certain regions of the genome that

00:33:28.720 are complicated. They're what we call repetitive sequences. And so they may not be unique, and it may be

00:33:34.180 hard to even sequence through those regions. And so putting those pieces back together in the right

00:33:39.040 order, in some cases, has been challenging to do. And so sometimes you'll hear some of us as geneticists

00:33:45.480 say, there's a dark matter, or there's some cryptic regions of the genome that we haven't been able to

00:33:50.600 really dig into. And that's because of some of these complexities of the sequences there and our ability to

00:33:55.980 sequence through them. So even despite what you're saying in terms of knowing the genes, or even knowing

00:34:00.900 portions of the sequences, we didn't necessarily know that it was all part of one gene or that we

00:34:06.120 had all the pieces or put it all together yet. And so some genes and some diseases were easier to crack

00:34:11.360 than others as a result. So today, if somebody goes out and gets a commercial genetic test, what's the

00:34:18.260 difference between someone who goes out and gets a whole genome sequence and, say, someone that goes to

00:34:25.400 one of the over-the-counter sequencing services like a 23andMe? What's the difference in the

00:34:30.680 analysis? And what's the difference in the information? This is an important question. And

00:34:35.800 if people are listening, this is time to perk up and listen closely. So there's a big difference,

00:34:41.240 and not all genetic testing is the same. And I'm not being critical of any of the companies that do

00:34:46.000 this, but just to realize they're trying to serve a different purpose. So 23andMe is an example,

00:34:50.740 or Ancestry.com is another example. Those are more things that are not medically sort of targeted.

00:34:57.020 They're not trying to answer a specific medical question of, do you have an increased risk of

00:35:01.460 breast cancer? Do you have an increased risk of heart attack? They're really not getting at that

00:35:05.600 level of detail. Just as an example, Ancestry.com is very good at being able to understand your

00:35:11.380 heritage. You're literally where your family is from, where your ancestors are from. It's quite

00:35:16.720 detailed at this point in terms of being able to say what part of the world your family comes from.

00:35:21.040 If you might be adopted, as an example, not know about your heritage or your ancestry,

00:35:25.300 be able to give you some of that. And I'll also say for better or for worse, if you're trying to

00:35:30.680 find this out, you may identify some of your blood relatives, some people who you would know from a

00:35:36.120 family reunion, and some people you might not know for some reason. And people sometimes find out about

00:35:40.880 that. Like I said, even people who are adopted, I've known to find some of their, actually their birth

00:35:45.760 parents that way. So that's one type of thing, but that's not really for the intention of identifying

00:35:51.480 information for a medical purpose. And so I just want to warn the listeners that if you get something

00:35:58.040 back, or more importantly, if you don't get something back from those tests, it doesn't mean

00:36:02.020 an all clear for your health. It doesn't mean that you're free of cancer or won't have any increased

00:36:07.540 risk. So on the other hand, there are other tests that are really designed for a medical purpose to

00:36:13.100 answer a question. And you didn't ask about this specifically, but many of the listeners will know

00:36:18.040 that they would have gotten a test, for instance, if they were thinking about having children,

00:36:22.480 planning a family, wanting to know if they are children, or if they were at increased risk of

00:36:26.780 having a child with something like Tay-Sachs disease or cystic fibrosis, one of those other

00:36:31.260 recessive conditions that we alluded to. And so that's not what the attention for personal health

00:36:37.280 so much, as I said, thinking about future children or families. And let me be clear, this is not necessarily

00:36:42.980 about abortion, but this is about being able to care for a child long-term and think about

00:36:47.540 reproductive options. So that's another use case and a very common use case in terms of what people

00:36:52.980 will do. Another common use case is for thinking about cancer risk. And so some people may have a

00:36:58.300 family history of cancer. Some people may say their particular heritage is such that, for instance,

00:37:03.600 if they happen to be of Jewish ancestry, they may be concerned because they know there's a higher chance

00:37:08.320 of having a BRCA, sort of called breast cancer 1 or BRCA 1 or 2 mutation. And so some people will do a

00:37:14.660 very targeted test. And I'm emphasizing targeted, very specific clinical question. And again,

00:37:21.520 it's answering that question. It's not necessarily giving a genetic clean bill of health for everything.

00:37:27.100 It's very focused. On the other hand, you alluded to what I'll call a genomic test. And I'm going to

00:37:33.120 make a distinction between genetic and genomic. And what I mean by genomic when I'm saying this is it's

00:37:39.880 really including, as we talked about, all genes. So it's not focused on just a handful of genes.

00:37:45.940 It's really focused on the genes in the genome, those 20,000 genes. You can look at just the coding

00:37:51.880 regions that we talked about before, those looking at the protein sequence that we call that in the

00:37:57.620 aggregate, an exome, because those little pieces that code the genes are exons, E-X-O-N-S. And when you put 0.94

00:38:06.600 them together in the aggregate, we call it the exome. Other individuals are interested in knowing all

00:38:12.620 3 billion of their base pairs, their entire genetic sequence, and we call that a genome. And that will

00:38:17.700 include everything, both the coding and the non-coding regions. I think of that in terms of

00:38:23.120 genome sequence as being in some ways the T-H-E genetic test, right? It's all-encompassing. One can blind

00:38:30.740 yourself to look at very focused subsets of genes based on a clinical indication, or you can look

00:38:36.560 at everything because you want to look at everything about your health, or because maybe you don't know

00:38:41.140 all of the genes for your particular symptoms, and you have to be all-encompassing in terms of that.

00:38:46.680 And we may get to some of those use cases, but there are many conditions that are genetically

00:38:50.680 heterogeneous or have many different genes that can cause them. And so we want to be all-encompassing

00:38:56.080 in terms of looking at that. Even though we can sequence all that data right now, we can't

00:39:01.060 interpret it all. So as an example, out of those 20,000 genes, we have now assigned functions with

00:39:08.240 disease for about 7,000. But that still means that for over 50% of those genes, we don't know of a gene

00:39:16.400 disease association. And even, I will say, because this happens to me with fair frequency, even when I

00:39:22.920 think I know about an association of a gene with a disease, if we study it further, we'll realize that

00:39:29.400 they don't map just one-to-one. There may be more than one disease associated with a gene, and so

00:39:34.240 there's still things that we're figuring out about what those genes do.

00:39:37.700 So Wendy, let's just talk technically about those different options that you laid out, and let's start

00:39:41.600 with the most comprehensive. So I've had a whole genome sequence done, and I believe it was done off

00:39:47.280 a couple of tubes of blood, maybe even just one tube of blood if I recall, but no more than two.

00:39:51.360 So 10 cc of blood at the most. So I sent that over to, it was a university that did it, it was part of a

00:39:56.500 clinical trial. What did that university do with those two tubes of blood to extract the insight,

00:40:03.640 to read the 3 billion base pairs that make up my whole genome?

00:40:09.880 So I'm guessing, but I'm guessing that what they did is, in the white blood cells, in that tube of

00:40:15.440 blood, they opened those up to extract what we call the genomic DNA. So the DNA included in each of the

00:40:21.580 nuclei of those white blood cells. From that, they, depending on how they did this, they may have

00:40:27.760 captured out specific fragments and then read through all of that using what we call short read

00:40:33.420 sequencing. Again, my guess in terms of this. That short read sequencing was probably less than 100

00:40:39.580 nucleotides for each little fragment. As they did that, they then had to do that informatic

00:40:45.420 computational step of putting all of those pieces together. And in most cases, they were probably

00:40:50.680 able to do that because they had a reference sequence. They knew what the average person

00:40:55.460 looked like, and they put your sequence right on top of that. And to the extent that those mapped

00:40:59.780 uniquely, they could put that all together. There may have been, though, some of your sequence where

00:41:04.840 they didn't know where it fit. And so it got sort of put aside in an area that they didn't even

00:41:09.720 analyze. But there were some of your sequence that probably got put aside that they didn't know

00:41:13.480 where it fit, how the pieces fit together. And in fact, the reason I say this, and this is very,

00:41:19.120 very highly technical, is that we're not perfect at doing this. And so there are times when there's

00:41:23.720 information over here at the side where we haven't used. As they did that, they're then able to read

00:41:28.700 out. And depending on the purpose of what your analysis was, they could read out and they could

00:41:32.580 say, well, for instance, if you are interested, I don't know anything about your own medical situation.

00:41:37.380 But if they said, well, we're interested in knowing whether or not you have PKU, they could look at

00:41:41.780 your phenylalanine hydroxylase gene. They could read out all the sequence and warn you, as I said,

00:41:46.760 one in a thousand base pairs, there's going to be a difference in this. And so they see a difference,

00:41:51.680 then they have to do an interpretation. And the interpretation is actually a lot more sophisticated

00:41:57.480 than one might imagine. Because again, there are literally tens of thousands of genetic variants in

00:42:03.480 your genome and what they mean and whether or not they do anything whatsoever is hard to know.

00:42:08.680 Each of us has what people think of as mutations or genetic variants that are associated with disease

00:42:14.960 and cause a problem. Each one of us has those, some that are very, very powerful, some that are

00:42:20.780 kind of wimpy and they infer some very small risk. But in the aggregate, you put together a lot of these

00:42:26.920 little wimpy genetic variants and it may amount to something more substantial. So depending on when your

00:42:32.280 genome was sequenced, when it was most recently interpreted, you might've gotten really profound,

00:42:37.660 profound, powerful information in terms of taking care of yourself or you might've gotten the sort

00:42:41.780 of, eh, we don't see much here. You know, sort of you go on your merry way. And for the average person,

00:42:46.840 my guess is if you were middle-aged when this was done, for the average middle-aged person,

00:42:51.400 it's mostly, eh, we don't see much here because you've survived, hopefully as a relatively healthy

00:42:56.600 person to this point, that you've essentially made it through some of the most devastating things we

00:43:01.760 can see in the genome. One more question before we leave the whole genome sequence and talk about

00:43:06.960 the whole exome sequence. One thing that they learned in me that was quite interesting was that

00:43:12.260 I was mosaic for a certain gene. This was only realized because as part of this clinical trial,

00:43:18.560 everyone in my family was sequenced and one of my children had a full copy of the gene,

00:43:24.760 which they got from me, but I was mosaic for it. So I didn't have very much of it. Can you explain

00:43:30.340 what that means? So this can come up in a couple of different ways, and I can talk more or less

00:43:35.520 about this if you're interested, but you would think we're the same, every cell in our body,

00:43:39.760 exactly the same. But in fact, that's not true. And your viewers or your listeners, rather,

00:43:44.880 will realize this when we think of cancer. Our genomes are not stable over our life course from

00:43:49.080 the point of conception to the point of death. And when you think about every cell division,

00:43:53.180 if you have to copy over 3 billion letters, we're not perfect. And we have spell checkers to try and

00:43:58.240 catch these things, but our body doesn't always catch them. And over time, mutations can accumulate

00:44:03.920 in the body. And of course, as they accumulate, this may lead to aberrant cell growth, which is

00:44:08.380 essentially what cancer is. And so cancer is at the heart of a genetic disease, but oftentimes not

00:44:14.820 from the genes you're born with, but for the changes that happen over your life course.

00:44:19.240 Now, in certain individuals, and I'm guessing this was the case when you just described your family,

00:44:24.500 this will be true, not just of your skin cells, for instance, if there's too much UV damage to

00:44:29.500 your skin in the summertime, but this can happen in your germline as well. So it can happen in the

00:44:34.240 egg and the sperm. And when this happens, again, as what we call somatic mutations or mutations over

00:44:40.040 the life course, again, if they're in the germline, they can be passed down to the next generation.

00:44:45.340 And so you can be what we call a germline mosaic. You can be a mosaic and mosaic, just like a tile

00:44:50.660 pattern that you see in a bathroom or something, right, where you have some color tiles, one color

00:44:56.260 and some other tiles another color. They're a different pattern because some cells have the mutation

00:45:00.960 and some don't. And so in the same way, you can have what we call gonadal mosaics, meaning that the

00:45:05.920 germline is affected. In some cases, you can see those gonadal mosaicism, that mosaicism in the blood as

00:45:12.320 well. So what you were talking about in terms of getting a blood sample, you might see that a certain

00:45:16.160 fraction of the cells have those mutations in the blood. And then if you see them in the next

00:45:20.720 generation as well, you'll know that it was transmitted through the germline. I'll share an

00:45:25.740 interesting factoid for individuals. The number of those mutations in the germline actually increases

00:45:33.540 over the life course. And so in particular, if you think about the biological process for

00:45:39.820 spermatogenesis with men, those sperm continue to divide over the life course and those mutations

00:45:45.360 can continue to accumulate over the life course. And so in fact, some of the conditions that are

00:45:50.680 associated with de novo mutations or new mutations, we see the frequency of that being greater for

00:45:56.620 parents, for instance, who are older parents at the time of conception than for parents who are

00:46:00.960 younger at the time of conception. And it's not, you know, that it's like astronomically exponentially

00:46:05.840 higher. It's a linear process for those types of mutations, but we do see those increasing over the

00:46:11.140 life course.

00:46:12.160 Historically, we would assume that women are more susceptible to that via age. I mean,

00:46:16.380 the rate of either aneuploidy or mutation seems to rise more sharply with women earlier,

00:46:22.740 starting probably in mid to late 30s. You're pointing out that the same is true of spermatogenesis.

00:46:28.660 Am I correct in saying that the egg seems more impacted by the sperm? And if so,

00:46:33.840 why is that? Is it a more complex division?

00:46:36.060 So what you're bringing up is that meiosis in the two sexes is different, and it is susceptible to

00:46:43.880 different underlying biological processes. So as you're saying, for women, if you look at the curve 0.99

00:46:50.040 in terms of problems, aneuploidy or sex, or not sex, but rather chromosome differences, namely Down

00:46:56.320 syndrome is what many people think about, increases with advanced maternal age. And that has to do

00:47:01.520 with the stickiness of the chromosomes at meiosis and the ability to separate or not. And so the

00:47:07.960 curve that is associated with that, which many people learned at some point, is that there's an

00:47:13.640 inflection. It's not a linear relationship with maternal age, but as you said, in the mid-30s,

00:47:18.780 it starts to increase more significantly. And as a result of that, there's a whole sort of medical

00:47:24.960 way that we can follow women when they're pregnant to try and pick up, if they're interested, those 1.00

00:47:29.980 particular chromosome issues. The difference when it comes to what we call DNA sequence differences,

00:47:37.500 so again, not whole chromosomes, but single letters, is that that process of being able to have the cell

00:47:43.700 divide and replicate and copy over that information happens at every single cell division, and there's

00:47:49.260 a certain probability that that will happen. And so that's a linear relationship in men.

00:47:53.480 And based on the biology, men, obviously, from a reproductive point of view, may have children

00:47:58.440 over a larger period of time. So we can see greater differences across men as they're reproducing. So

00:48:04.880 biology is a little bit different between the two sexes.

00:48:08.460 We have a clear sense why you see more of these meiotic differences with age. In other words,

00:48:15.220 what is the fundamental characteristic of aging that is driving that? Is it sort of like evolution says,

00:48:22.700 well, I don't care because I don't want you to reproduce after a certain age? You know, again, 0.96

00:48:27.740 not to anthropomorphize evolution, but therefore, I'm not going to preserve the integrity of your

00:48:34.220 genome beyond a certain age? I mean, I'm curious as to, see, to me, that strikes me as an explanation,

00:48:39.300 but not the reason. The reason must be something more fundamental at the level of an aging hallmark

00:48:45.760 process.

00:48:46.420 I hadn't thought of the question quite that way. But I do think if you think about throughout all

00:48:52.420 of human history, the age at which people were having, reproducing and having children

00:48:57.760 has skewed much younger than it is in terms of current society. So I think we have pushed the 0.97

00:49:02.380 boundaries to a certain extent in terms of what the biology historically has been. I don't know if

00:49:07.820 we had a, you know, use before date in terms of ovaries and gonads before, but that I think just

00:49:13.760 historically has been what's happened. So within that, the fundamental biology for women has been 1.00

00:49:19.340 proteins that are responsible, as I said, in terms of meiosis and separating of the chromosomes and

00:49:25.280 being able to have, if you remember when meiosis starts in females, it actually is starting 0.99

00:49:30.480 way, way early in gestation. So even during fetal life. And so those proteins have to be working in

00:49:37.500 intact from before a fetus is even born, before a child is born, lasting all the way through whenever

00:49:44.300 that pregnancy is conceived, essentially, when those gametes are finally dividing. So that could be

00:49:49.220 30, 40 or more years for that process to have to work. And that's kind of asking a lot when you think

00:49:55.360 about the biology. For men, I'll just throw in another interesting factoid that for some of these

00:50:01.220 mutations that arise during the process of gametogenesis, there are even what we call selfish sperm

00:50:07.120 mutations. That is that certain mutations may even give a selective advantage to those spermatocytes

00:50:13.280 where they may have a reproductive division advantage, for instance. And so we may see more

00:50:18.280 of those in terms of the biology of what we see in the next generation because of the effect they have

00:50:23.140 even in terms of directly on the sperm. So lots of biology at the root of that.

00:50:29.000 So let's go one step down from the whole genome sequence to the whole exome sequence. So

00:50:33.620 if as part of my blood test, I only wanted the whole exome sequence, are they actually doing

00:50:38.980 the whole genome sequence and just reporting out the exomes or are they actually doing something

00:50:43.460 technically different?

00:50:44.780 So to answer the question, you have to read the fine print on your genetic test report or

00:50:48.840 your clinical trial consent or whatever it is. So I don't know the answer, you know, without knowing

00:50:54.220 that. It certainly was the case that a few years ago, not that long ago, it was so expensive

00:51:00.360 to sequence a genome that we rarely did it. It was really just cost prohibitive. The exome was a

00:51:06.000 really good shortcut because we didn't know what a lot of the other information meant anyway. So we

00:51:10.860 were kind of, it would have been just throwing it away. Increasingly, we have a better sense of what

00:51:15.440 the non-coding regions do. We have better ability to interpret and recognize genetic variants. And so

00:51:21.920 I would say as the sequencing costs have been coming down, there's more of a shift to going to

00:51:27.240 genomes rather than exomes. But truly most of what is medically used at this point is in the coding

00:51:33.700 space. So even if you're sequencing a genome, it's still focused on the coding regions. On a research

00:51:39.560 basis, though, very different. And of course, we have to do research before we can apply it clinically.

00:51:44.640 A lot of the research now is understanding what all of those regions do and being able to

00:51:49.500 eventually use that information clinically. So I do think within the next decade, we're going to see

00:51:54.120 a shift. And we may even shift from this, what we call short read sequencing of these hundred base

00:51:58.540 pair fragments, even to things that are much longer in terms of accuracy, able to read through some of

00:52:04.020 the greater, the areas that are more complex and probably that we've been missing out on before.

00:52:08.980 And a short read, you said, is about a hundred base pairs?

00:52:11.240 Plus or minus, yep.

00:52:12.080 So what's the technical limitation for making that longer?

00:52:15.800 So as you do this, there's just lower and lower fidelity for each base pair that you do. And so at some

00:52:22.000 point you start getting errors within your readout, then you don't want to have too many errors because

00:52:26.580 you can't distinguish between what are true biology versus what are artifacts of what you're doing in

00:52:32.200 the laboratory.

00:52:33.500 How are you correcting the errors? If you're getting even one error every in one short read sequence,

00:52:39.380 that would, given that we only differ from each other by one and a half percent of those base pairs,

00:52:44.080 that would be, for example, you wouldn't be able to use this information in court.

00:52:46.800 How do we preserve the fidelity of this to make such bold claims as, hey, we found this blood at

00:52:53.260 the scene of the crime and we absolutely know dispositively it belongs to this individual?

00:52:58.420 We do it not just once is the answer. So in fact, when we read this out, we have a term called

00:53:03.960 read depth. It's how many times we look at any one nucleotide in the genome and we don't look at each

00:53:09.440 position just once. Depending on how much we want to spend doing it, we might look at the read depth of

00:53:14.760 30x. So we might look at every one position on average 30 times. And if you see out of the 30

00:53:20.800 times 29 of one nucleotide, one of the other, you say, oh, that one, the odd ball there, that's

00:53:26.920 probably just a sequencing error. That's an artifact of our method in the lab. The other 29 are the ones

00:53:32.340 that we want to pay attention to. In some cases, again, I bring in cancer. It becomes very important

00:53:37.720 to know for those somatic mutations that weren't there from birth that they may be there in a small

00:53:42.580 number of cells, but cells that can be very, very dangerous for cancer. And so we may increase the

00:53:47.740 read depth to be very, very high. We may go up to 1,000x in terms of read depth to make sure we've

00:53:52.760 got the accuracy to see something reproducibly even in 10% of cells. But again, we're 10%. You need 10%

00:54:00.120 of a large number to have the assurance that what you're seeing is in fact representative of the

00:54:05.280 underlying biology, not an artifact. Let's talk a little bit more about an example there. So you brought up

00:54:10.680 BRCA1 and BRCA2. So typically the genetic tests, that would be done off a whole genome sequence

00:54:17.000 or even a targeted sequence because let's say a person says, I want specifically to be tested for

00:54:22.060 breast cancer genetics, but you're still testing on the actual DNA taken from white blood cells,

00:54:28.320 correct? So these days we do things lots of different ways. And just again, for some of the

00:54:33.140 listeners, we've tried to make everything we do more accessible. This is a fundamental sort of core

00:54:37.880 value for me. So as an example, we can even now do things from cheek swabs, from saliva samples,

00:54:43.720 from blood samples. So again, we try and make it less invasive, easier to do, even potentially from

00:54:49.140 home. And so for some people who have done something like 23andMe, they may have even done it that way.

00:54:54.320 Let's pick one of those examples, a cheek swab or a blood test. And you're going to do this enormous

00:54:58.900 depth of readout to really make sure that there are no copies of the BRCA gene or any of the other

00:55:04.640 genes that you're looking at. How many genes, by the way, when we do a breast cancer genetic screen,

00:55:10.560 how many known genes are we looking for? So interestingly, it's a little bit of a la carte.

00:55:16.860 So what I mean by that is it's your choice. It's your choice either as a doctor ordering the test,

00:55:21.740 it's your choice as a patient that's getting the test. In some cases, you know you have a family

00:55:26.860 history of a BRCA mutation. And so within that case, we may know the exact address to go to,

00:55:33.040 and it's a very simple plus minus readout. And we don't have to do a whole genome sequence for that.

00:55:37.420 We just need to look at one gene and say, yay or nay, it's there or not there for that particular

00:55:41.440 variant. Beyond that, there are even, I alluded to this before, but there are certain variants that

00:55:46.800 are seen in certain communities. So as an example, if you happen to be Ashkenazi Jewish, there are three

00:55:52.300 different spots in BRCA1 or 2 that account for the vast majority of all mutations in those two genes.

00:55:58.160 And if you know that, we can take a shortcut and we can basically say for literally a small fraction

00:56:03.640 of the cost of sequencing a genome, boop, boop, boop, we look at those three spots, we get yay or

00:56:07.980 nay, and you've got most of the information that you need. To your point, there are some people who

00:56:13.040 come in with a family history of breast cancer and they say, but I want to be careful. And so in that

00:56:18.640 circumstance, we may do a panel of 50 different genes, 5-0 different genes that'll cover most of the

00:56:24.780 genes that we see for hereditary, not just breast cancer, but ovarian cancer, colon cancer, the most

00:56:31.460 common cancers that we see that are driven by germline or inherited genetic factors. So for round

00:56:38.280 numbers, 50 is a good number when you're trying to be really comprehensive. If you said, just give me

00:56:42.660 the focused breast cancer things, it might be more like 10. You've said this twice now, but I think it's

00:56:47.500 helpful for folks to listen. When last I checked, maybe 5% of cancer was accounted for by germline

00:56:55.200 mutations. 95% of cancer is accounted for by somatic mutations. Is that still accurate, would

00:57:01.120 you say? I'd say I'm going to modify that just a little bit, not to be a contrarian, but for the

00:57:06.200 genes that I'll call monogenic, highly penetrant, let me unpack that a little bit. Monogenic, single

00:57:13.340 gene, highly penetrant, high probability that over the life course you'll develop cancer if you have

00:57:19.720 this particular genetic variant. So when you limit yourself to that, yes, about 5% of cancers are due

00:57:26.100 to those powerful single genes, high probability of cancer. Now, on the other hand, over time, we've

00:57:32.220 realized that there are additional genes that I'll call moderate risk genes. Many of those genes may

00:57:37.460 confer something like a two to threefold increased risk as opposed to something like a tenfold increased

00:57:43.020 risk. So there's another probably 5% or so that are due to those. And then there's this other thing

00:57:49.180 that we call polygenic risk. Poly meaning multiple, genic genes, polygenic multiple genes. And the number

00:57:57.280 of genes we oftentimes look at in those circumstances may be anywhere from 100 to hundreds or even in some

00:58:03.900 cases thousands of genetic variants all mathematically sum together to understand what the risk is associated

00:58:10.880 with that package. All of us have genetic variants that go into that polygenic risk. And part of the

00:58:18.400 question is along a distribution, are you at the high end of that risk curve or are you at the low end

00:58:25.300 or at the average end? And so within that, this is now something that is not clinically being utilized

00:58:31.280 routinely, but we are on a research point of view trying to understand clinical implementation for now

00:58:37.560 those polygenic risks for cancer. Assuming that that amounts to, I don't know, we'll figure it out,

00:58:42.880 but that might amount to 10% of cases. You'd say, well, look, 20% of cancer has a genetic component as

00:58:50.800 opposed to, and it's broken down into those three categories of monogenic, highly penetrant. I think

00:58:57.400 the second category, was it monogenic, not highly penetrant or not monogenic? I'd call it monogenic,

00:59:03.480 moderate risk. Moderate penetrant and then polygenic. Right. And those would be the three.

00:59:08.360 And then going back to this case of say the breast cancer example, a woman says, I just want to do a

00:59:13.600 deep dive on breast cancer. I don't know which genes it is because all my family's deceased, but four

00:59:19.740 women in my family have died of breast cancer. We're going to do this cheek swab and you're going to look 1.00

00:59:24.000 at 10 genes that are associated plus whatever the polygenic genes are. Why is it that you don't need to

00:59:30.960 look directly at breast cells? Why is it that we can infer that what we see in a cheek cell,

00:59:39.020 an endothelial cell or an epithelial cell rather in the cheek or in a monocyte in the blood is also

00:59:46.640 captured in mammary tissue? Good question. And the answer is it's probably not. So what you're doing

00:59:52.420 when you're doing the cheek sample, the blood sample is you're really getting at the germ line. So you're

00:59:56.980 mostly getting at what you were born with, what that inherited susceptibility is. As we talked about

01:00:03.320 though, your genes are changing over your life course. Your cells are changing over your life

01:00:07.720 course. And cancer doesn't happen overnight. You don't go from a normal cell to a cancer cell

01:00:11.980 overnight. There's a progression in terms of going through this. And so there are other ways that

01:00:17.420 people have thought about that I'll call it a liquid biopsy. So this idea that you might be able to,

01:00:23.400 and it's a slightly different test than what I was describing before, but where you would look for

01:00:28.340 these somatic mutations, you described this before, but when you're looking for that needle in a haystack,

01:00:33.680 if you've got a tumor that's going to slough off some of that DNA into the circulation, you might be

01:00:40.220 able to see some of that fragmented DNA floating around, and you might be able to pick up some of those

01:00:45.980 mutations that might be reflective of that mammary cell that's either gone awry and is a cancer,

01:00:52.060 but maybe not something that you're detecting on mammography or something else. And so this is,

01:00:57.900 in some ways, been the holy grail of being able to do cancer screening. It's not quite ready for

01:01:03.460 prime time yet. And people think about it more, I would say, right now for thinking about recurrence

01:01:08.880 of cancer. So how do you monitor someone who's had a previous cancer diagnosis? You think they're all

01:01:13.780 clear in seeing whether or not they've had a recurrence. The other use case people have thought

01:01:18.840 about it as someone who might be at high risk of cancer. So someone who's identified in whatever

01:01:23.340 reason, based on an exposure, based on a genetic profile, but it's not ready yet for population

01:01:28.940 screening in terms of being able to pick up cancers at an earlier stage. We're still relying on other

01:01:33.700 things to do that. So let's now talk about what happens at the level of the 23andMe's and the

01:01:39.840 ancestries and companies that are obviously doing something far less than a whole genome sequence or even

01:01:46.680 a whole exome sequence, just on the basis of the cost at which they can offer these things.

01:01:51.400 What are they technically doing with the epithelial cell of your cheek or the saliva or the white blood

01:01:58.640 cells that they get? Again, and I'll say, read the fine print of what you sign on the consent form.

01:02:04.060 Number one, it may change over time and I don't represent any of those companies. So I don't want to

01:02:07.960 misspeak in terms of what they're doing. They are in general though, number one, not trying to detect

01:02:12.880 cancer. So any of what I talked about, not the purpose of what they're doing. They're in general,

01:02:18.020 not trying to read out the genome, at least not for the purpose of getting you medical information

01:02:22.660 for what I call news you can use to manage your own healthcare. They're largely doing it in a way

01:02:28.380 that I'll call more recreational. And so with doing that, for any of you who have done 23andMe,

01:02:33.760 you may find out something about, for instance, if you were to eat asparagus, what your urine might smell

01:02:39.040 like, or what your earwax might be like, or if you're lactose intolerant, they are things that

01:02:44.140 are related to how the biology of your body works. They are related to genetic variants. So those two

01:02:49.260 things go together, but they're not telling you based on your earwax, if you're going to have major

01:02:53.680 problems with hearing loss down the road or, you know, cancer risk or things like that. So that's why

01:02:57.960 I use the term recreational in that way. But what are they technically doing? Depending on the company

01:03:02.940 and depending on what they're doing, they're oftentimes reading out what we call single nucleotide

01:03:07.760 polymorphisms or so-called SNPs. So they're not reading out the entirety of your genome. They're

01:03:13.400 not reading out all 3 billion base pairs. They are selectively going in and saying, at this exact

01:03:19.220 address, do you have an A or do you have a G? At this other address here, do you have a C or do you

01:03:24.200 have a G? And based on that, they may selectively look at those particular variants and say, with your

01:03:29.860 profile, I know that your family, you know, originally came from Egypt or wherever it is, you know, in terms of

01:03:36.860 being able to look at and stress-free. Or they may say, based on looking at this, I know this particular

01:03:41.380 genetic variant may predispose you to be lactose intolerant. I would expect that you're going to

01:03:45.600 have problems in terms of eating ice cream for dessert tonight. You know, that's generally the

01:03:50.600 type of thing they're reporting out. Depending on, again, the company and the terms of the agreements,

01:03:55.400 there may be differences. But generically, that's what many of them are doing.

01:03:58.500 And so if a gene differs by more than one nucleotide, a SNP is not of much use, unless you sample two SNPs

01:04:07.680 in the same gene.

01:04:08.880 It gets tricky. So technically, a single nucleotide polymorphism could be a genetic variant that has a

01:04:16.440 big, big effect on a gene and could, from a medical point of view, be very, very impactful. So it's not just

01:04:21.360 the size that matters. It's, you know, as some people would say, location, location, location. It's like real

01:04:25.980 estate. So it all matters which variant we're talking about. But in general, the ones that

01:04:30.700 they're looking at are not the ones that are medically impactful. They're just normal variants

01:04:35.140 that are innocent bystanders, but help us understand where our ancestors came from. For the most part,

01:04:40.360 that's what the companies are doing.

01:04:42.260 Although they do comment on some important ones. So you mentioned isoforms as normal variants. So the

01:04:49.080 APOE gene has three isoforms, the 2, 3, the 4 isoform. All are relatively common. I mean, the 3 being the

01:04:57.740 most common, then the 4. The 2 is not that common. But all would be considered, quote-unquote, normal

01:05:02.380 variants. One obviously comes with a much higher risk of neurodegenerative disease. A 23andMe test does

01:05:09.900 read out that prediction. Presumably, that tells us that those genes only really differ by one base pair,

01:05:16.960 correct? The 2, the 3, the 4 only differ by one base pair. That's therefore the only place they

01:05:21.060 need to sample. But I've seen in 10 years several instances of a missed call, meaning the snip read

01:05:30.700 ends up not matching with the more rigorous exome sequence. Why do you think that's the case? Does

01:05:37.400 that surprise you?

01:05:38.540 I will say that, you know, in doing this, I'm not someone, you know, that has been asked to go in

01:05:44.420 QC or do quality control or anything like that for the laboratories. I have known of things as

01:05:50.840 simple as these are done in plates that oftentimes are 12 by 8 plates. And if you flip the plate the

01:05:57.260 other way, you can have sample switches and you've got a different person being read out for a

01:06:01.500 different thing. It can be something as simple as a logistic like that. And I have seen that lab

01:06:05.580 error before. There can be sample switches at multiple places. And at the end of the day,

01:06:10.860 I will say that if you were doing something from the recreational side to something where you were

01:06:16.480 going to make a major health care decision, you were, as a woman, for instance, going to go through

01:06:21.000 and have a mastectomy, get a second opinion. Like, be sure that this is really you and it's really the

01:06:26.720 result you think it is before you do anything irreversible or, you know, go out and buy that big

01:06:31.220 life insurance policy. And I would say that in general, you know, if you were getting a second

01:06:35.000 opinion about cancer diagnosis.

01:06:36.460 You mentioned the Guardian study earlier. Can you say a little bit more about what that study is?

01:06:41.720 So the Guardian study, as you can imagine, based on what I started out the conversation with

01:06:46.180 phenylketonuria, is I've always been wanting to be able to get information that people could use to

01:06:53.080 be able to maximize health and being able to just be the best person they could be. And I started out

01:06:59.140 in 1996. Again, I had started out in the space of PKU decades ago and had been studying a disease

01:07:06.000 called spinal muscular atrophy for about a decade with colleagues of mine. And this is a neurodegenerative

01:07:11.740 condition and used to be the most common genetic cause of death for children less than two years

01:07:16.400 of age. And I realized starting in 2016 that we were just at the cusp of potentially a treatment that

01:07:24.500 might slow down or stop the neurodegeneration. Yet tragically, if we didn't identify babies before

01:07:31.160 they started showing symptoms, it would be too late. That is, we'd have this window of opportunity.

01:07:36.140 So we started out a newborn screening program for SMA, and then babies that were identified through

01:07:41.640 that had the option, if they wanted to, of going into a clinical trial. That ended up being quite

01:07:46.160 synergistic in the sense that we did identify babies who would have been predicted to have the

01:07:50.520 most severe type of SMA. They did get into early clinical trials right away. They did benefit from

01:07:56.120 those that helped in terms of the ultimate evidence that was necessary to show the efficacy of those

01:08:01.420 treatments. And because of that, and because we were able to show that we could do it technically

01:08:06.000 and that people wanted it, SMA has been added to the recommended universal screening panel for babies

01:08:11.860 across the United States. And so now 4 million babies born each year in the United States are

01:08:16.360 screened for SMA. And we have three FDA-approved treatments, including a one-and-done gene therapy.

01:08:21.820 So babies can now be identified within the first week or two of life, get a one-and-done IV infusion

01:08:27.680 of the gene, and go on to have much, much better life, if not be quote-unquote normal, at least as

01:08:33.040 far as we can see so far. So that got me thinking about doing this on larger scale. We've since done

01:08:38.940 a newborn screening study for Duchenne muscular dystrophy, and just recently actually was FDA-approved

01:08:44.660 a treatment for DMD, Duchenne's muscular dystrophy. But I'm, I don't know, I'm getting older and I'm

01:08:50.640 getting more impatient. And I didn't want to do these one by one. I started thinking about how

01:08:54.920 could we do these at scale for population health, but not just for one condition at a time, but how

01:09:00.100 could we do it for dozens or hundreds or potentially even more conditions? And so the Guardian study

01:09:06.540 actually stands for something. It stands for Genomic Uniform Screening Against Rare Diseases

01:09:12.920 in All Newborns. And if you put the letters together from that, it spells out Guardian.

01:09:17.880 And the idea behind that is to take that same newborn screening dried blood spot that we use

01:09:22.520 already for PKU, that we talked about, sequence the genome. We don't need to read out everything

01:09:28.620 in the genome. We only read out the genes that we consent people to read, that they consent to in

01:09:34.380 the study. And those genes are genes that have, I call it, news we can use, information that has

01:09:39.540 treatments immediately available. And in planning this study for almost four years or just over

01:09:45.820 four years with families, we had many, many iterations about what they wanted, what they

01:09:51.080 wanted us to screen for, what should be, what should enable them to be the best parents and give their

01:09:56.020 children the best chance at a healthy life. And we thought about also this dynamic change in what

01:10:01.920 we'd have treatments for. The fact that the world is changing rapidly and we wanted the flexibility

01:10:07.180 that if a new treatment became approved tomorrow, boom, we could instantly change the screen and be

01:10:12.320 able to implement that. We wouldn't have to wait a decade to gather the evidence to do that. We

01:10:16.860 wanted to be nimble and flexible. So the reason for using the genome as the backbone is it gives us

01:10:22.460 that infinite flexibility to be able to adapt and to be able to move the field forward. So we've been

01:10:28.080 doing the Guardian study in New York City since September 2022, and right now have screened just over

01:10:33.800 3,000 babies. And it really has been remarkable to me in terms of being able to see just the broad

01:10:40.160 support from our community in doing this. In New York City, you should realize, most of the listeners

01:10:45.180 probably do realize the wonderful diversity we have in New York City. That is that of the people who

01:10:50.560 participate, it's not just white folks, it's not just people who are from Ireland. We were talking

01:10:54.940 about Irish individuals in PKU, but it's people from around the world. We have about a quarter of people

01:11:01.120 of European ancestry, of Latina ancestry, of Black ancestry, of Asian and other ancestry. So

01:11:08.400 it becomes really, I think, representativeness or representative, basically, of the world. And it

01:11:14.220 also is geared to leave no baby behind, because newborn screening is kind of this one universal

01:11:19.560 thing where everyone goes through the health system in the same way. And by making this free and being

01:11:24.780 able to allow everyone to enter, if they so chose, we can really see also what information people want

01:11:31.880 and what they don't want. Within this, I guess one of the things that's been refreshing to me is to see

01:11:38.200 that about 74% of parents that we approach and offer this to decide they want to do this. And that's an

01:11:44.380 important number to me. When we did this for SMA, the number was 93%. When we did this for Duchesne

01:11:49.340 muscular dystrophy, it was 84%. And so within this, it's not 100% of people who want any of these

01:11:55.880 genetic screens, and that's perfectly fine. I'm not trying to force anyone to do anything. But it's

01:12:00.540 also not 10%. The majority of parents are saying, yes, if there's something I can do to ensure that I

01:12:06.180 have a healthier child, give it to me. Help me be a better parent. Why would I not want to do this,

01:12:11.440 is mostly what we hear. Within this, I also appreciate, and we do this with regular newborn screening,

01:12:18.180 just the traditional newborn screening. And we realize that traditional newborn screening isn't

01:12:22.120 perfect. I never thought it was. Nothing ever is. But we realize that adding this additional

01:12:27.260 dimension, and we've even done it for PKU within this study, this additional dimension helps us to

01:12:32.640 do a better job. And so just as an example, we've also identified part of newborn screening identifies

01:12:38.740 some children with severe combined immunodeficiency, a problem where you can have an overwhelming

01:12:44.960 infection and die from this. But a treatment is available, including a bone marrow transplant.

01:12:50.300 And so because of that, we have as part of newborn screening, a way to screen and identify some,

01:12:56.400 but not all, children that have that. We've added this now genome sequencing to enrich and improve that,

01:13:02.500 and in fact, identified a baby that was missed by our traditional newborn screening for SCID,

01:13:07.620 yet is at increased risk in terms of this overwhelming infection, but yet with the opportunity to

01:13:12.720 intervene at an early stage when a bone marrow transplant will be most effective. And so there

01:13:18.100 are numerous examples where we've identified, whether it's Wilson's disease, whether it's

01:13:22.480 severe combined immunodeficiency, whether it's achondroplasia, but other conditions that are

01:13:26.900 treatable that we just needed to identify those babies. And as we've done that, the number of children

01:13:32.520 that, and I just know because I've been practicing in New York City for 25 years, I know sort of how people

01:13:38.280 navigate the system and how they get through. And we've been able to really get to many of the

01:13:42.920 people who are usually unfortunately left behind, either because they're immigrants, they don't speak 1.00

01:13:47.780 the language, they don't have the same health insurance. But individuals that we're seeing come

01:13:52.280 out positive for this are very, very different in terms of reflecting our community than the people

01:13:57.960 who navigate the healthcare system and get in to see us. And we realize based on other studies that

01:14:03.300 we've done that most of the children that would have been diagnosed, if ever they were diagnosed,

01:14:08.620 on average aren't diagnosed until somewhere between 8, 9, 10 years of age. And so we're

01:14:13.960 able to identify them literally a decade earlier before a lot of damage has been done to their body.

01:14:21.060 So it's just the beginning. You know, 3,000 is great. I think it demonstrates that we can do this.

01:14:26.300 I think it tells us what our community wants out of this. It shows us some pitfalls in terms of how

01:14:31.740 it's hard to do and what we need to do to do it better. But I do fully believe that this

01:14:36.340 both de-risks this in terms of being able to also have groups that are working on therapies,

01:14:42.400 be able to realize that this is something, there's an opportunity now for treatment for this,

01:14:46.920 and is a powerful way of moving forward health equity, at least for children for the next generation.

01:14:52.960 Is this something that's done only at Columbia, or is it a multi-center New York hospital endeavor?

01:14:57.460 So right now, this is done through our New York Presbyterian Hospital System. So it's not just

01:15:02.140 Columbia, but it's through this hospital network. It's only so far in those hospitals. But based on

01:15:07.820 our success for this, we are figuring out how we can be able to expand this more broadly and really

01:15:12.740 think about this, as I said, as integrating within the public health infrastructure. Not trivial to do

01:15:18.340 this on scale. As an example, doing this in New York State, if we were to do this for every baby,

01:15:23.080 we'd need to do it for about 210,000 babies a year. So no small feat, but something that we're

01:15:29.140 gaining the experience to know what the pain points are and how to solve for them.

01:15:32.880 Is this all funded by NIH?

01:15:35.120 So within this, as you can imagine, this is not inexpensive to do. So in fact,

01:15:39.760 none of this is funded by NIH. NIH, I won't go into all the details, but NIH is able to fund

01:15:45.440 programs that are about this big. People may or may not see this if they're just listening to me,

01:15:49.580 but very small amount. This ends up being about two orders of magnitude larger in cost than anything

01:15:55.780 that NIH can fund. And so it's a challenge in terms of, as you think about big, bold,

01:16:00.880 new transformative ideas, how do we as a scientific community accomplish these? And so we've done this

01:16:07.000 by putting together many different stakeholders. I don't think any one group could be able to take

01:16:11.760 this on. And truth be told, we're not completely there with the funding. I think we needed to

01:16:16.060 demonstrate that we actually could do this in this first feasibility stage before gaining the

01:16:21.160 resources to do this with what I hope will be at least a hundred thousand babies to get to the 0.99

01:16:26.180 sample size we need to see some of these rare conditions and to know what the outcomes are and

01:16:30.740 that we really can screen for them. So what's the actual cost of doing the sequence for each baby?

01:16:37.360 So to look at those 250 some odd conditions, what's the bench cost?

01:16:41.640 So as we started out doing this round number, $1,000 per baby. So thinking about generating the

01:16:47.600 data, interpreting the data, getting it back to folks, cost of about $1,000 per baby. The goal

01:16:53.680 is to be able to do this and get it down by an order of magnitude. Can we get it down to $100 per

01:16:58.900 baby as an example? And in doing that, can we think about the economic impact? Most importantly,

01:17:04.320 the health impact for the baby. But as we think about as a society, how to be able to afford doing this,

01:17:10.480 we are doing the economic analysis to understand. But the good thing is sequencing costs are

01:17:15.360 decreasing. Analysis interpretation costs are decreasing. More of this can be done in automated

01:17:21.000 ways as we understand what normal variation is for people around the world. And that's one of the

01:17:26.120 critical factors is doing that around the world. Now that $1,000 is a fully loaded cost. That's the

01:17:32.280 interpretation. That's the overhead. That's the PI time and such, right? The sequencing cost must be

01:17:36.560 significantly less than that, given that Illuminate could do a whole genome sequence for $1,000 now,

01:17:41.120 right?

01:17:41.300 Even over the course of this study, the sequencing costs have come down, if you can imagine it. And

01:17:46.420 we just started it, like I said, September 2022, less than a year ago. But already the sequencing costs

01:17:51.940 have come down. I expect they'll continue to come down in terms of this. And so data generation certainly

01:17:57.660 can be done for well less than $1,000 now. But as you said, part of it is the interpretation. And

01:18:02.960 we have study staff that explain the study to everyone, explain results to apparently. So,

01:18:07.960 yeah, includes multiple pieces.

01:18:10.440 So at the outset of this discussion, you mentioned that SMA actually has a successful gene therapy. I

01:18:15.560 was not aware of that.

01:18:16.720 Yes.

01:18:17.160 How many of these single gene, highly penetrant conditions that children are born with, be it

01:18:23.900 inborn errors of metabolism or neurodegenerative diseases, et cetera, how many of them have FDA

01:18:28.160 approved gene therapies already?

01:18:30.540 Not very many of them. So really the shining example is SMA or spinal muscular atrophy. There are very

01:18:37.040 few gene therapies that are now approved. There are now, for instance, I mentioned Duchenne muscular

01:18:42.620 dystrophy, hemophilia. There are a few other conditions, but it is still literally a handful

01:18:48.400 for gene therapy, literally gene therapy. Others have treatments available. And I think for many of

01:18:54.220 those, I'll just give one example that we've had through the Guardian study, a condition called

01:18:59.500 Wilson syndrome, for instance, that leads to ultimately liver failure and need for liver transplant.

01:19:05.020 The treatment that we give children for this is zinc. So it's something that it doesn't require

01:19:10.520 gene therapy. It doesn't need anything that fancy. We can simply use zinc to out-compete

01:19:15.340 copper and make sure that we don't end up with a copper overload situation. And we have treatments

01:19:20.280 that are very well tolerated, pennies a day in terms of doing this, we hope, extremely effective

01:19:26.020 long-term. So although gene therapy is wonderful, I want to underscore we don't always need gene

01:19:30.940 therapy as long as we have that early diagnosis. If there were three diseases today that you see

01:19:37.360 in a pediatric practice that would be most amenable to gene therapy in terms of some aggregate score of

01:19:45.320 the technical nature of doing the gene therapy and the lack of alternative therapies elsewhere and the

01:19:52.580 number of kids afflicted. If you were sort of to take that as your triple proxy, what would be,

01:19:58.080 if you could wave a magic wand, what would be the three diseases you would want to put

01:20:02.460 high on the list of gene therapy targets?

01:20:05.600 That's a great question. I don't think I've ever thought that through exactly in that way.

01:20:11.000 So there are a lot of neurological conditions. Let me start with that in terms of places that we are

01:20:15.660 woefully behind in terms of treatments. And I'll give you one shining example of Tay-Sachs disease

01:20:22.120 as an example. Many individuals, the way they've dealt with this because there's no treatment is

01:20:27.860 simply not having children or not having children together or going through other reproductive options.

01:20:33.840 But things like Tay-Sachs disease is just a terrible condition and it has relatively high population 0.68

01:20:40.020 prevalence to one of your points and has really nothing available in terms of treatment today.

01:20:45.340 And so those children are born healthy, normal children and die within the first few years of

01:20:50.780 life with a degenerative course oftentimes associated with epilepsy. So a condition Tay-Sachs or a condition

01:20:57.360 like that, I think, fulfills all the criteria that you're talking about. Other similar conditions like

01:21:03.360 that, although we still need to understand treatability, are things like Fragile X, another condition that we

01:21:09.380 see, in this case, X-linked by the, you can tell from the name Fragile X, similar in terms of high frequency.

01:21:16.140 Really nothing in terms of treatability at this point and perhaps the ability, it's a, we can get into it or not,

01:21:22.700 but the gene therapy for this is a little bit trickier. Different strategy from a technical point of view than

01:21:28.080 one would take for Tay-Sachs. So whether it's gene therapy or gene editing or other types of things, there would be a

01:21:33.940 different technical strategy. And then I would say they're just a large group, I won't pick one, but large

01:21:39.920 groups of inborn errors in terms of liver disease that primarily affect the liver. So whether you're

01:21:44.840 talking about something like maple syrup urine disease, propionic acidemia, but something like that, that

01:21:50.660 we have good programs in place to identify those children's and our treatments, they're just not up to

01:21:57.440 snuff yet. They're not quite as good as they should be.

01:21:59.720 Well, let's talk about how genetic therapy works. So God, probably 23 years ago, we had a tragedy in

01:22:07.200 one of the most highly publicized examples of early gene therapy. The tragedy, of course,

01:22:13.740 really didn't have to do directly with the gene therapy. It had to do with the vector that was

01:22:18.920 used to deliver it. And the young man who received that gene, I can't remember what it was for. It was

01:22:24.080 for an inborn error of metabolism as well, wasn't it?

01:22:26.000 Yeah. Urea cycle defect called ornithine transcarbamylase deficiency.

01:22:30.960 Yeah. And Jesse, what was his last name?

01:22:33.600 Jesse Gelsinger.

01:22:34.600 Gelsinger, yeah. And this was at Penn at CHOP, right?

01:22:36.880 That's right.

01:22:37.500 About the year 2000, if my memory is correct.

01:22:40.200 I think that's about right.

01:22:40.940 Maybe 99.

01:22:41.140 I may be off by it.

01:22:41.840 Yeah.

01:22:42.280 Yep.

01:22:42.680 And so let's talk about that. Let's talk technically about that. So they used an adenovirus,

01:22:47.280 but explain what that means and what was the state of the art 25-ish years ago? And let's

01:22:54.100 contrast that with what's being done today.

01:22:56.080 For those of you who are listening, adenoviruses obviously causes the common cold. And so whether

01:23:00.580 it's adeno or adeno-associated virus, we oftentimes use that as the vector or the delivery vehicle

01:23:06.580 to deliver genes. Within this, the viruses are manipulated. They're engineered, so to speak,

01:23:12.440 so that they're not going to be contagious. So even though you might get a cold or pass it

01:23:15.940 along to someone, you're not going to do that with gene therapy. Yet there are problems

01:23:20.100 with this because the common cold is common. And so people may have been infected with adenoviruses

01:23:25.740 and their body may try and mount an immune response when it's infected, as it would be

01:23:31.560 with a cold. And that's where a lot of the mischief comes in. And unfortunately, it hasn't ended with

01:23:37.760 Jesse and Jesse's death. There have been other deaths in the gene therapy space with others that

01:23:43.840 have had a response to the vectors, oftentimes an immunological response to that. In Jesse

01:23:49.900 Gelsinger's case, it was the vector and the gene therapy was targeted at the liver. We've talked

01:23:55.600 about a little bit of that already. But sometimes there can be an overwhelming response from the

01:24:00.720 liver where the liver starts to fail, where there's an immune response that goes on. And so, as I said,

01:24:07.320 this has not been solved completely at this point. There have been other genetic therapies,

01:24:12.200 other diseases, not just liver diseases, but where there have been similar responses. And I think one

01:24:17.800 of the things that we've learned from this is we have to be very careful with people with underlying

01:24:21.760 liver disease when it comes to this, because a fragile liver can get tipped over, especially with

01:24:27.120 adenovirus. We also have to be careful about who's been exposed to those viruses. So sometimes we do

01:24:31.820 screens to be able to see who might have one of these responses. But ultimately, and partly because of

01:24:37.440 that, people have also been trying to figure out other delivery systems, other vehicles,

01:24:41.480 other ways of being able to get those genes into cells that may not be as toxic or problematic.

01:24:47.200 Help me understand a little bit. So an adenovirus is very common. Presumably, in the case of Jesse,

01:24:53.160 he'd already been exposed to some antigen that was related to this. And so he already had memory B

01:24:59.360 cells and memory T cells that were ready to mount a healthy immune response should he have been exposed

01:25:04.680 to that very common adenovirus again. The fact that he had such a harsh response to the gene therapy

01:25:10.960 was that because of the dose of adenovirus that he received? Or does that imply that he would have

01:25:17.620 had some sort of catastrophic multi-system organ failure had he just had another exposure to that

01:25:23.700 exact adenovirus as a cold, and that it, as you said, was a function of his underlying liver health?

01:25:29.800 So it's a very tricky situation. And I say this because there will be some either patients or doctors

01:25:35.640 advising patients about genetic therapies or genetic therapy trials in the future.

01:25:40.880 It's tricky in the following sense. You're right that there's a dose response in terms of the

01:25:46.240 immunological response. And so you don't want to go too high on the dose because you don't want to

01:25:50.760 have too big a response. On the other hand, with these gene therapies, you oftentimes get one chance

01:25:56.100 at this in terms of doing this, because once you've given the therapy, the body is going to mount an

01:26:01.880 immune response to that and would neutralize that same therapy if you were to give that again.

01:26:06.940 And so the tricky thing about this is you don't want to go too high and you don't want to go too

01:26:11.360 low, because if you underdose it and if you don't get enough in and that's your one shot on goal,

01:26:15.920 you've burned it. And so within this, it's a tricky situation to figure out how to get it just right.

01:26:21.220 And as it is with many clinical trials, when you're first in human, you don't know, right? It is first

01:26:26.460 in human. Oftentimes there are non-human primates in terms of trying to figure out as much as you can,

01:26:31.340 but it's still not a person and each person is unique. And so as you're doing it, it is a tricky

01:26:36.580 situation in terms of getting it right. When you do the first person, you learn and you figure out

01:26:41.000 from there whether you're going to go higher or lower, but there is someone who's going to be

01:26:44.440 the first person. And there's no way to engineer the adenovirus to make it invisible to the immune

01:26:50.320 system while still able to insert its DNA package into the cell. So there's certainly things that we do

01:26:57.720 to try and do this better. As I alluded to, one of the advantages for someone, for instance,

01:27:02.800 with SMA, where we're dosing them at a week or two of life is they haven't had the common cold.

01:27:08.020 They haven't been exposed to these things. They've got a fresh immune system. So the likelihood they're

01:27:12.560 going to have a response like this is much lower. So we haven't been seeing those types of things

01:27:16.400 with newborns that we've been treating with SMA. There are other vectors that are not just vectors,

01:27:21.660 other delivery vehicles that we use that are not viruses. And so that's something else that in

01:27:26.580 terms of developing new technologies, things that may not pose the same problems. I don't want to

01:27:32.560 guarantee that that's going to be the case, but certainly with the experience we've had,

01:27:36.820 even from the COVID vaccine, from mRNA vaccines and having delivery systems that were basically

01:27:43.100 lipid nanoparticles to deliver nucleic acids to cells, we've learned a tremendous amount from

01:27:49.020 millions of people who have been treated with that. And some of those technologies may prove to be

01:27:54.040 helpful with other delivery systems. Maybe this is a good time to explain what CRISPR is and how it

01:28:00.000 factors into gene editing. Although I guess before we do that, explain what is required to change

01:28:08.300 a gene. So I don't know, pick someone with sickle cell anemia. So if you wanted to use gene therapy

01:28:16.100 to fix quote unquote sickle cell anemia, my vague recollection from medical school biochemistry is

01:28:24.000 that was a one amino acid change, correct? Correct. Was valine one of them? Yep. So the eighth amino acid

01:28:30.680 exactly in hemoglobin beta is that there are two different sort of variations or flavors that that

01:28:36.400 amino acid you can change, but that causes sickle cell disease. Yes. Okay. So if you want to

01:28:41.760 permanently change that, it's not enough to do it in the red blood cells that are floating around in

01:28:47.160 the bloodstream because they're going to be trashed in the spleen a couple of months from now. You must

01:28:52.300 change the DNA of the stem cells in the marrow, correct? That's exactly right. You have to get those

01:28:58.900 progenitor cells from which the future generations of red cells will be derived. And let's assume you had

01:29:06.020 the correct gene sequence. That's not hard to do. We know what that is. And we can put that into a virus

01:29:12.020 and maybe just explain to people why viruses are great vehicles. What is it about a virus that makes it an

01:29:17.780 ideal candidate here? Well, the nice thing about a virus is it was designed by mother nature to infect

01:29:23.160 our cells, right? So it's pretty good at being able to do that. The viruses that you're talking to and the way

01:29:28.600 that I think about them are often helpful for gene addition. So where you've got a protein product that

01:29:34.540 hasn't shown up for work, it's not working, it's a loss of function, it's not present, you need to be

01:29:39.420 able to deliver or add back that gene. It may not be integrated into your genome. And in fact, there

01:29:45.860 are probably some advantages if it's not, but the virus can bring that in and bring it into the cells.

01:29:51.140 And to your point, and this is important, many times that means bringing it into stem cells so that

01:29:56.620 they can have the longevity of continuing to populate the body over time.

01:30:00.800 And so is it safe to say that the real challenge is that step? It's not just putting the corrected

01:30:08.840 version of hemoglobin, the gene for the corrected version of hemoglobin into the virus of choice.

01:30:14.040 It's figuring out how to get that to selectively infect a progenitor stem cell within the bone

01:30:20.580 marrow. Is that presumably why we don't yet have genetic therapy for sickle cell anemia?

01:30:25.760 Well, sickle cell is interesting in that there are a couple different strategies people think of. So

01:30:31.700 one is gene editing is the term I'm going to use. So it's fixing the gene in sort of where it is,

01:30:38.440 not adding a gene, but actually fixing the gene that's present. Another strategy that people think

01:30:43.980 about, again, for the aficionados who are thinking about that, as I mentioned, hemoglobin beta in terms

01:30:49.300 of the adult form of hemoglobin, there's also fetal hemoglobin that's made in utero. And so

01:30:55.640 with that, that actually has a very tight binding of oxygen because the fetus needs to be able to get 0.96

01:31:01.400 oxygen from the maternal blood. And it can substitute for adult hemoglobin actually quite efficiently.

01:31:08.800 And so one thing one can do is actually turn up the amount of fetal hemoglobin expression and

01:31:15.420 essentially be the in situ version of your gene therapy. It's just manipulating the gene expression,

01:31:20.920 but for a gene that's already present in those individuals. And so there are ways of manipulating

01:31:25.460 gene expression in that case for fetal hemoglobin. So there are multiple ways to skin a cat, so to

01:31:30.580 speak, very technically different in terms of what we administer to people and what we're changing in

01:31:36.160 terms of the gene therapy. But I want to pick up on what you were talking about with gene editing,

01:31:41.420 because that's yet something else. That's really in situ.

01:31:44.920 Yeah. And I want to come back to that. Let's go back to one more thing on the gene addition.

01:31:49.540 If you were to add a gene for a corrected version of beta hemoglobin, would you actually run into a

01:31:57.320 problem now where you're making too much hemoglobin and half of it is appropriate and half of it is

01:32:03.360 sickle and therefore not? And you could argue you're creating more problems because you've

01:32:07.960 doubled the hemoglobin, but you still have the issue where the cells sickle and create all of

01:32:15.320 the distal ischemia that the person has. So is that the real reason that nobody's interested in doing

01:32:20.240 in addition therapy for sickle cell anemia? So that's exactly right. Just for the listeners,

01:32:25.220 we don't use the gene addition strategy for sickle cell anemia because we'd have to dilute out,

01:32:31.140 so to speak, so much of the hemoglobin with the sickling that physiologically we run into other problems.

01:32:37.360 Okay. So gene editing would hands down be the best solution for certainly a situation like that.

01:32:47.340 And probably many cases, could it be done with high fidelity and ease? So I guess let's talk

01:32:53.240 about gene editing in any way you see fit in whatever way you want to tell the story. I mean,

01:32:56.660 gene editing in and of itself is a whole podcast, I suppose, but what's the medium version of that

01:33:01.600 story?

01:33:01.920 So from a simplistic way of thinking about this, it's going in and in situ being able to correct

01:33:08.160 the genetic variant. Now, realizing we've talked about single nucleotide variants, those are the

01:33:13.460 easiest ones to edit. There's just one single base pair that needs to be flipped. It actually matters

01:33:18.600 what that base pair is. If you have to change an A to a G or a C to a T, believe it or not, there are

01:33:23.700 different base editors that can do different types of nucleotide switches. But there are many mutations

01:33:29.240 that are not just single nucleotides. There may be multiple nucleotides. There may be multiple

01:33:34.220 repeats, these sort of complex things that we talked about with Fragile X. There may be entire

01:33:39.700 chunks of chromosomes. They get very complicated. And furthermore, it may be when you think about

01:33:45.720 population genetics, you may have a mutation distribution across a gene that may be quite

01:33:51.720 heterogeneous. Sickle cell is an easy one. In the way that you mentioned, you're talking about the same

01:33:57.100 position for everyone with sickle cell disease. A couple different nucleic acids, but it's the same

01:34:01.900 position, essentially. Whereas other genetic conditions, it may be that almost everyone has

01:34:06.920 a different mutation. So you have lots of different things you need to edit, and that may be easier or

01:34:11.340 more difficult to do. So there are some nuances in terms of that. You mentioned CRISPR. So CRISPR-Cas9

01:34:17.860 is something that many will know because Nobel laureates were awarded for this amazing discovery,

01:34:23.300 which was, by the way, I will say, just really good science with creative women who are thinking

01:34:29.260 about other ways to use it. It wasn't fundamentally, the discovery was not made with the intention of

01:34:34.740 doing genetic engineering or genetic manipulations, but really smart people thinking about it.

01:34:40.280 The CRISPR-Cas9 system has, I think of it as an Achilles heel of a double-stranded DNA break.

01:34:46.340 So it fundamentally, in terms of being able to make the correction, has to cut the DNA,

01:34:51.680 cutting the two strands of the DNA to make the correction. And that fundamentally leads to some

01:34:57.560 instability as the cell repairs that process, which you use the word fidelity, which I like the use

01:35:04.180 of that term because it really is all about fidelity and potential off-target effects, where you

01:35:09.460 inadvertently introduce other than the intended correction, other genetic changes, and sometimes

01:35:16.260 those other genetic changes can cause mischief. We call them off-target effects. So things

01:35:21.640 where they may inadvertently destroy the gene, cause other changes to the gene, cause other changes to

01:35:26.900 other genes that you weren't even trying to target, but it can lead to problems of essentially promiscuity

01:35:32.360 or inaccuracy or low fidelity. This is, to me, in my opinion, the Achilles heel in terms of that

01:35:37.920 particular system. So there are others in terms of thinking about other technologies, other strategies,

01:35:43.900 that in terms of doing a double-stranded DNA break, will do a single-stranded DNA break. So it'll nick

01:35:50.300 just one of the two copies, which in terms of the process that the cellular machinery has for the repair

01:35:56.680 of that, ends up being a much higher fidelity system. So there are lower off-target, lower error rates.

01:36:03.120 It tends to be a more robust system. So this is still very, very early. I want to underscore this

01:36:09.060 very early in terms of doing this. There are lots of other complexities besides the machinery of what

01:36:15.480 I just described was prime editing, but other complexities in terms of the machinery to go in

01:36:22.000 and make the changes, what types of mutations can be repaired. Prime editing has strategies to be able

01:36:28.520 to do just not single nucleotides, but much more complex mutations to be able to fix. But ultimately,

01:36:34.880 it's a vehicle for delivery. It's getting in early enough before the damage is done to the body.

01:36:40.780 It's being able to get to the part of the body safely that you need to. It's sort of multiple

01:36:46.220 pieces of the puzzle that have to all be solved simultaneously to get the whole package to work.

01:36:51.120 So we're not quite there yet, but I'm optimistic that we are, as a lot of scientists working together,

01:36:56.320 realizing that this may be one sort of solution eventually when all the components are there

01:37:01.920 that may be scalable to deal with many different types of genetic conditions.

01:37:06.760 You were alluding to the differences in strategies between Tay-Sachs and fragile X syndrome. Do we have

01:37:13.000 enough information now in this discussion for you to explain the different strategies there?

01:37:16.860 I think so. So let's start with Tay-Sachs. Tay-Sachs is due to an enzyme that's missing. We talked

01:37:22.520 about recessives. It's a recessive condition. This is a degenerative condition. And so you want to be

01:37:28.560 able to get in early for all the reasons that we've talked about. And you could do a gene addition.

01:37:33.160 The gene, the enzyme is missing. So you can just pop it back in. It doesn't have to integrate. You

01:37:37.760 just need to get it early enough to do its job and not to cause any mischief along the way. So that

01:37:42.880 strategy would be a good strategy. The tough part is you need to get it into the brain. You know,

01:37:47.420 as you're thinking about the delivery system, brain is a complex organ. So you want to be able to get it

01:37:51.840 throughout the brain per function. What does that mean, Wendy? That means you'd have to introduce

01:37:55.940 like an intranasal virus. Is this something in glial cells, in neurons? Where does this enzyme

01:38:01.360 normally get made? Within this, as you said, there are multiple ways to access the brain. It's

01:38:06.380 obviously a protected space in terms of the blood-brain barrier. I doubt we're going to be

01:38:10.720 able to do it with intranasal, although, you know, there is some that you can get by putting this

01:38:14.780 in that way. Some cases we do intrathecal. So for women who've had an epidural, it's basically the

01:38:21.140 same way that we access the space for an epidural. In some cases, for anyone who's thought about

01:38:26.400 chemotherapy that we give for brain cancer, sometimes we actually have to do it into the

01:38:30.620 ventricle. It sounds a little bit barbaric, but we go through the skull and being able to do the

01:38:34.760 injections there. But my point is, it's not as simple as a simple intravenous infusion. It's not

01:38:40.320 like we can just give it peripherally and get it to the brain where we need to. So it's challenging in

01:38:45.180 that way. And as we think about it, in some cases, we need to get throughout the brain,

01:38:49.920 even into deep nuclei or different parts of the brain. So as an example, if you were to inject it

01:38:55.280 and you had a high concentration on the left, but it didn't get to the right, that would be a problem.

01:39:00.040 You need to be able to get even distribution as we're doing this. Anyway, with Tay-Sachs, though,

01:39:05.400 it is the case that, as I was alluding to before, you probably don't need to get to 100% of the

01:39:10.580 protein or the enzyme that's there. Even 50%, I'm sure, is enough. And it's possible you could get down

01:39:16.160 to 20%, and that would be just fine. And so that's one strategy. Fragile X is a little bit

01:39:22.240 more complicated. The actual mutation itself is what we call a trinucleotide repeat, a repeat that's

01:39:28.660 too big. And so we need to be able to make it smaller within doing this. It also has the same

01:39:34.140 problems in terms of being in the brain, but it's not just simply adding back some additional

01:39:38.940 Fragile X protein. So we can't just make it a gene addition strategy. We've got to really think

01:39:43.680 about the gene editing that I was alluding to, where you're fixing, shrinking the size

01:39:48.960 of that repeat back down to the normal size. What's unknown, and I think one of the things

01:39:54.000 we don't know until we do it in people, is what is that window of treatability? And just

01:39:58.920 to be provocative, to let the listeners think about this, is that window, even if Guardian

01:40:04.360 worked perfectly and we could identify these babies with this within the first week of life,

01:40:08.700 is that early enough? Of course, my hope is that for many conditions that will be.

01:40:12.660 It's possible that we'll need to go even earlier. And so there are some people that have thought

01:40:16.920 about even in utero gene therapy or genetic treatments for some conditions, not all of

01:40:22.100 them, but for some conditions where it might be necessary to get even during development,

01:40:27.000 fetal development, to have the maximal effect. So I'm not saying we're going there anytime

01:40:30.980 soon, but just to sort of think through that, there may be imperfect solutions unless one gets

01:40:36.800 to the right time and the right place, and to your point, the right cell type even.

01:40:40.580 And fragile X is also a recessive condition, so it can only impact women, presumably, because

01:40:46.720 you would need both copies of the X?

01:40:48.680 So fragile X, we do call it an X-linked recessive, but what that also means is that it's mostly

01:40:54.240 males who are affected, because males only have the one X. If they have that repeat expansion,

01:40:59.420 the males will be affected. Females can be affected, although it's much more unusual.

01:41:04.840 Because they would need to have got the X from their father as well.

01:41:07.760 I want to come back and talk more at the end of our discussion about the future of gene

01:41:13.780 editing and the ethics around it and things like that, which I'm sure is something you've

01:41:17.220 thought a lot about. Before we do that, I want to talk about some of the more complex diseases

01:41:21.000 that clearly have a genetic component, but they're probably much more polygenic. So let's start

01:41:27.780 with what your colleagues down the hall are doing with respect to obesity. How much do we understand

01:41:33.020 about the genetics of obesity? And does genetic therapy play any role there?

01:41:39.040 So I'll start out about talking with obesity, but I may switch gears at some point soon after that. So

01:41:44.680 obesity, we have ways of calculating heritability. It's to give us a scientific insight of how genetic

01:41:51.560 is a certain condition. So you can do this by looking at twins, for instance. You can look at

01:41:57.180 identical twins, you can fraternal twins, and you can see how similar they are in terms of body mass

01:42:02.240 index, adiposity, things like that as measures of obesity. And it does end up being highly heritable.

01:42:08.840 It's not the most heritable factor, but it is highly heritable. So that sort of points in one

01:42:13.260 direction that genes are important.

01:42:14.900 What's the heritability of obesity, by the way?

01:42:16.800 So the heritability running somewhere around 50% for round numbers or 0.5. Other conditions that are

01:42:23.100 extremely heritable, of course, are closer to 1. So as an example, and by comparison, type 2 diabetes

01:42:28.340 or non-insulin-dependent diabetes, more heritable, more strongly genetic in terms of that.

01:42:33.720 And type 1?

01:42:34.480 Type 1 diabetes, less heritable. A complex interaction, both of your immune system and the genetics that

01:42:40.640 govern your immune system, but also what you're exposed to early on in sort of the cross-reaction

01:42:47.000 your immune system has between self and non-self. So a little bit different model.

01:42:51.880 On the other hand, clearly there are, I'll call them environmental differences. And so you can

01:42:57.920 look at what's happened to the average body mass index of the average American over the last

01:43:03.000 generation. Our genes haven't changed, but on the other hand, by most measures, you can see that we're

01:43:08.200 more prosperous. In general, the average body mass index has increased. And there have been some

01:43:13.340 interesting studies looking at particular groups of Pima Native Americans, for instance,

01:43:18.920 that have genetically the same genes. They come from the same original community, but they live

01:43:24.280 in different environments. One in which it's more sort of a traditional environment in terms of

01:43:29.720 the amount of access to calorically dense foods and the amount of physical energy that's expended on

01:43:35.480 a day-to-day basis. And those same original groups, but in two different environments, you see much

01:43:40.980 more obesity in the one group that has access to, again, obesogenic foods versus the other that

01:43:47.020 doesn't. So again, strongly suggesting that it's not just the genes in terms of this.

01:43:52.900 Now, on the other hand, and this goes back to you were describing Rudy Leibel and his original work

01:43:58.280 in terms of identifying leptin and the leptin receptor through positional cloning methods,

01:44:04.600 leptin and the leptin receptor in terms of mutations in those genes do not account for the

01:44:09.400 vast majority of obesity. Very rare. I've certainly had patients with these conditions,

01:44:13.980 but that's just because of the nature of who I am, but very rare. And most people would not have seen

01:44:18.500 this. On the other hand, understanding the fundamental biology of how body rate is regulated

01:44:23.660 and governed, critical in terms of understanding those two molecules. And my point in this is,

01:44:29.240 I don't know if it's going to be for obesity, but it may be for other conditions like myocardial

01:44:34.280 infarctions or coronary artery disease, that knowing about the biology and sort of the final

01:44:40.600 common mechanism or the final common pathway through which the biology is regulated, one may

01:44:48.120 have ways of either pharmacologically, or some will say in terms of gene therapy, being able to make

01:44:54.040 permanent manipulations. So statins as an example, in terms of treatment for hypercholesterolemia,

01:45:00.560 there may be various different genetic mechanisms by which one has an increased risk for a heart attack,

01:45:05.540 yet statins seem to work for a lot of different people. And some have thought that, for instance,

01:45:11.440 rather than using that as a medication, would there be a way of genetically making a manipulation 0.97

01:45:16.560 so it's kind of a one-time and not having to require continued ongoing therapy. So I'm not saying that this

01:45:23.500 is exactly where obesity treatment is going to be going. And I am, in a good way, excited that we certainly

01:45:29.520 have better treatments for the first time, I think, for obesity now than we had five or 50 years ago.

01:45:35.680 So we may be going in a better direction.

01:45:38.700 We didn't really talk about epigenetics, but I think obesity might be a good time to do a little

01:45:44.460 bit of backtracking and explain what the epigenome is and how it changes, not just over a person's life,

01:45:51.420 but perhaps more importantly from one generation to the next. And the reason I'm asking the question

01:45:55.760 is, I wonder if it's playing a role in the propagation of obesity across generations,

01:46:04.260 even though, as you pointed out, we're not really experiencing much genetic drift in the period of

01:46:10.460 time that we're seeing an explosion in obesity. And so with all of that said, my question is ultimately,

01:46:17.660 do you think epigenetic changes could be explaining the increase we see in obesity

01:46:24.080 as a susceptibility to obesogenic environmental factors?

01:46:29.260 I think the bottom line is we don't know. But for those who don't know what the term epigenetics is,

01:46:35.320 break it down epi above and then genetics, the genes. And so there are chemical modifications that happen

01:46:42.400 to the genome, which are used to affect gene regulation. Some of those chemical modifications

01:46:48.200 include methylation, and those are dynamic. They can change over the life course. They can change by cell type.

01:46:54.080 And there are ways to be able to coordinate regulation of potentially groups of genes.

01:46:59.760 They're tricky to analyze. From a methodological point of view, scientifically, they're tricky

01:47:05.220 because they do vary over the life course, and they vary by cell type or tissue. And so using,

01:47:12.120 for instance, we've talked about this a lot, but using a blood sample as a matter of trying to get

01:47:17.440 the epigenetic profile for what's going on in your brain or your pancreas doesn't always work.

01:47:22.580 And it's hard to even know whether or not it works because it's not as if we're going in and doing

01:47:26.800 pancreatic or brain biopsies on most people. So, you know, you can do things in animal models. You can get

01:47:31.880 some indirect evidence, but it's hard to know for sure whether or not this is truly answering the question

01:47:37.480 you're trying to answer. So I'll say there's a lot of conjecture in the area of epigenetics and hard

01:47:43.580 to know for sure exactly what that is. On the other hand, I will say that we've known about something

01:47:49.540 called the agouti mouse, which is a mouse model for obesity. And depending on how much folate

01:47:55.080 you give that mouse, for instance, while the dam, the mother mouse is carrying her pregnancy, 0.58

01:48:01.440 her little mice, depending on the amount of folate in her diet does affect the epigenetics.

01:48:06.640 Folate is used in terms of methylation for the DNA. And so you can see a readout of the effect.

01:48:11.920 And depending on that, you can see a change in the coat color. You can see a change in the obesity

01:48:15.620 for these agouti mice and for their progeny. And there are things in terms of, as you said,

01:48:21.120 transgenerational. Potentially, we don't entirely know the mechanism of how that might be occurring,

01:48:26.400 whether it's epigenetics or other things. But there are things that we can see. But these are

01:48:30.360 complicated. So I'll just say I personally don't feel like scientifically we have all the evidence

01:48:35.200 to make definitive conclusions at this point. But one wonders about what many different contributors

01:48:40.680 could be, although I have to guess that this is not going to be the major, major driver.

01:48:44.840 Let's pivot to autism now. Autism is in the news all the time, it seems, and certainly appears as

01:48:50.920 though it's increasing in frequency. And it's unclear how much of that is due to an increase in

01:48:57.140 diagnosis and recognition versus how much of that is triggered by other environmental factors. But

01:49:04.600 there doesn't seem to be much confusion around the fact that there's a strong genetic component to it.

01:49:10.220 So let's start with that. Based on all of the twin studies, what is the heritability of autism?

01:49:16.880 I will say to your point, autism is even within the name, a spectrum. So it's not just one

01:49:23.720 condition, it's a spectrum. And it goes from severe, what some people will call profound autism,

01:49:30.200 and can be associated with intellectual disabilities to other individuals at the mild, quote unquote,

01:49:34.720 milder end, who are quite talented in many ways, yet have social challenges. So within that entire

01:49:40.860 spectrum, if one includes everything within that, the heritability is estimated to be approximately

01:49:45.660 0.8, although some individuals will say even as high as 0.9. The point within that, though,

01:49:51.440 is that it's not 100%. And in fact, we do know of times over the life course, in particular,

01:49:58.060 prenatal and early childhood that are important to the developing brain, and where changes in exposure

01:50:04.620 beyond the genes can play a role. So as an example, prematurity is one of the more common,

01:50:10.200 if you will, exposures. But in terms of what happens to the developing brain,

01:50:14.380 and if you are born when you're 26 weeks old, much higher probability of autism than if you're

01:50:19.880 born at term at 40 weeks. And so there are other factors beyond just the genes that are involved.

01:50:25.280 But clearly, the other point that I'll make about heritability is one calculates heritability

01:50:30.620 as a measure of the inherited genetic factors. But you mentioned it once already, one of the factors

01:50:37.320 in autism is that there are de novo, or new genetic variants that occur for the first time in the

01:50:43.400 individual with autism. And those individuals aren't captured in that measure of heritability,

01:50:48.640 because heritability is fundamentally trying to get at transmitted genetic variants that are going

01:50:53.920 from parent to child. And those de novo genetic events are new in the child. And so there are

01:50:59.520 genetic aspects not included in heritability, if that makes sense.

01:51:03.840 Yeah. So what are the genes that seem to be responsible for autism?

01:51:08.040 So depending on who you ask and how you want to define this, there, I think everyone would agree

01:51:13.580 there are at least 100 genes that have been identified with high confidence as being associated

01:51:18.280 with autism. Depending on how rigorous you want to be about this process, you know, some people would

01:51:24.160 say that we estimate that there are at least a thousand genes, and we probably, you know, are about

01:51:29.860 a third of the way there in terms of having some sense of those genes. Not surprisingly, those genes are

01:51:36.120 genes that are in the brain. They're expressed in the brain, they function in the brain, not surprisingly.

01:51:40.740 And many of those genes are especially active during development. And so what I mean is intrauterine

01:51:46.960 fetal development within the brain specifically.

01:51:50.740 And what do they code for? I mean, how many of those genes would be genes in the exome versus the intron?

01:51:56.820 Most of the ones we recognize, underscore the ones we recognize are in the coding sequence,

01:52:01.700 but that's a limitation of what we recognize. We do realize that statistically, we see that there is

01:52:08.260 a signal in the non-coding space, but we have less evidence to implicate specific genes or specific

01:52:14.500 genetic variants individually in the non-coding space, because the effect size or how powerful

01:52:20.440 they are is somewhat reduced compared to those coding sequences. The other issue is not just where

01:52:26.640 in the genes, but what genes are involved. And so the genes that are involved fundamentally can be

01:52:32.540 genes that function at the synapse. So the connections between brain cells and communicate

01:52:37.160 between brain cells, that happens to be one thing that's quite important. They can be cells that are

01:52:42.360 rather genes that are important in regulation of genes and gene networks. So many of them are

01:52:47.400 transcription factors, histone modifiers. We talked even about epigenetics, some of those genes that may be

01:52:53.200 responsible for those epigenetic changes, but they often, I think of them as having multiple

01:52:58.880 downstream genes that they affect. So it's not having a very, you know, sort of focus, it's more

01:53:03.620 a universal effect that they have. Those genes that have that more global effect oftentimes have a more

01:53:10.080 global effect on brain function and cognition. So it may not be that it's autism only, but they may also be

01:53:17.100 associated with intellectual disabilities. They may be associated with epilepsy. They may be associated with

01:53:22.080 more sort of global effects in terms of brain function. And to the extent that that term autism is used across the

01:53:29.000 spectrum, there are oftentimes those individuals described as profound autism. So there can be different

01:53:34.820 things. There can also be, I'll just put as an example, we mentioned, I'll go back to PKU, believe it or not. So I

01:53:41.360 happened to run a very large autism study called SPARK. And within SPARK, we identified a teenage young man who

01:53:48.240 actually has his autism as a responsible of undiagnosed PKU. Even autism can be caused by,

01:53:55.160 you know, full circle, an inborn error of metabolism where there are toxic things that build up in the

01:54:00.540 brain and then cause that dysfunction of the brain. So not everything is a sort of primarily in terms of

01:54:06.940 the brain, but things that can diffuse to and have an effect on the function of the brain.

01:54:11.200 But to be clear, autism is a clinical diagnosis in the same way that familial hypercholesterolemia

01:54:18.020 is a phenotypic diagnosis. It's a diagnosis in the case of FH where LDL cholesterol has to be above

01:54:24.680 190 milligrams per deciliter off treatment. And it's incredibly heterogeneous in terms of the genes

01:54:31.920 that are responsible. To my last count, I think there were more than 3,500 genes that could produce

01:54:37.120 that phenotype of high LDL cholesterol. So autism is the same, right? The diagnosis is clinical. It's

01:54:44.020 a phenotypic defined disease and maybe up to a thousand genes involved in that or a thousand

01:54:50.460 different ways to get there or more, right? Exactly right. So it is a DSM diagnosis in terms

01:54:55.780 of clinical behavioral criteria. I know this gets confusing for people, but one can have a gene that's

01:55:01.460 identified as causal, but the diagnosis is still a behavioral diagnosis, simply a gene associated with

01:55:08.180 that. And as you said, not just one single gene, it doesn't map one-to-one. In fact, no one gene or

01:55:13.680 genetic factor accounts for more than 1% of individuals who have that clinical diagnosis of

01:55:18.460 autism. So incredibly heterogeneous. And what's the approximate prevalence of autism today?

01:55:24.040 Round number's 2%. You were alluding to it before, but this number has fluctuated over time,

01:55:29.360 whether it's for all the reasons you said, but about 2% today.

01:55:33.500 Does it just seem like it's more, or is this really a function of greater awareness?

01:55:37.900 So I think it's a function of several things. It doesn't help that the definition has changed over

01:55:42.000 time. So literally the DSM diagnostic criteria have changed over time. And so for that, not surprising,

01:55:47.680 the prevalence has changed over time. In a good way, there is greater recognition and diagnosis,

01:55:53.360 as you alluded to. We've seen this in particular for underserved individuals that are more frequently

01:55:58.100 diagnosed now. So I think the disparities are decreasing, and I think that's a good thing.

01:56:02.700 But there are also, I'll say, maybe there are things that are changing in terms of society,

01:56:08.060 changing the biology. I don't know. We haven't been able to put our finger on that, but there

01:56:12.260 are possible contributing factors with that. And then there's also a motivation to a certain extent

01:56:17.280 in terms of the way our society works to be able to access resources. And so people that may not

01:56:22.560 have been motivated to get a label per se, they may still have known it, they may have, you know,

01:56:27.240 thought it to themselves, but they didn't necessarily seek a diagnosis or a label, except

01:56:31.800 that there were resources, educational resources, support resources that were important. And we want

01:56:36.320 to make sure those individuals get those resources.

01:56:39.700 What are some other, both neurologic and non-neurologic sequelae of autism, or call it

01:56:44.900 comorbid conditions with autism?

01:56:47.120 So I think that's a good way to phrase it. Comorbid conditions is one of the things that I think

01:56:51.600 about. So as an example, some individuals will have epilepsy associated with their autism. For

01:56:56.900 some individuals, that epilepsy will be recognized very early. For some individuals, it won't come

01:57:01.180 until the teenagers or adolescents, but that can be incredibly important. Within this, their behavioral

01:57:07.140 co-occurring diagnoses, for instance, of anxiety is quite frequent. ADHD or attention issues, again,

01:57:13.440 quite frequent. And I think some things we're just beginning to understand, although I think it's

01:57:18.580 incredibly important, is that most of what we know about autism is based on individuals below the

01:57:23.840 age of 20. Those are the individuals who've been studied most. And I think there's a whole lot we

01:57:28.780 don't know about adults with autism. And I can say I do know some conditions that are associated

01:57:34.020 as degenerative conditions as well. That when people are adults, there may be particular subtypes of

01:57:39.760 autism that are neurodegenerative because the genes that are involved are responsible for

01:57:44.400 neuromaintenance, being able to sustain the brain and continue functioning. And when they're not

01:57:49.060 functioning at some point, start having things associated like Parkinsonism. Some subtypes that

01:57:54.800 may be associated with increased risk of, we mentioned obesity. Believe it or not, some of these same genes

01:57:59.800 may also predispose to obesity, and especially with some of the medications we use to treat some of

01:58:04.920 these behavioral conditions, even increase the effects, the metabolic effects and weight gain and

01:58:10.200 diabetes associated with that. And there may be other things as well. But to a large extent,

01:58:15.220 I would say it's under-recognized and we have a lot of more gaps in our knowledge. But many people

01:58:20.860 who continue to need those, that understanding, and I think earlier you used a term of precision

01:58:25.480 medicine. I don't mean it to sound like a cliche, but you can imagine that it's a large percentage,

01:58:30.640 2% of the population, great heterogeneity. Everyone doesn't need to have the same sort of rules that

01:58:36.620 they're following the same rule book or the same management guidelines. And how do we get greater

01:58:40.620 specificity to not overburden people, but yet to be able to also allow them to achieve their full

01:58:46.680 potential and lead their healthiest lives? You mentioned that most of what we know about autism

01:58:51.060 is based on studying people who are up to, but below, typically 20 years old. Does that suggest that

01:58:58.020 prior to about the year 2000, there was nobody really studying this? Because presumably if we were,

01:59:04.480 we would know about what people look like later in life today.

01:59:08.280 So many of the adults with autism, number one, were not diagnosed as having autism. They may have had,

01:59:14.140 you know, some of these challenges, but things have just changed over time. And so having a label,

01:59:18.840 having a diagnosis has changed with society. Other individuals who were studied 20 years ago have not

01:59:25.200 been followed longitudinally. And that's hard. Although there have been some epidemiological studies

01:59:30.740 like Framingham that have followed individuals over long periods of time, it's hard to be able to do

01:59:36.500 that. People move, people, you know, investigators lose funding, people die. I mean, you know, lots of

01:59:41.860 things that happen. And so just knowing what someone looked like at two and that same person at 22,

01:59:47.820 there are very few studies in terms of in children, what that looks like. And so I think that's been a

01:59:52.840 large part of the problem as well.

01:59:54.020 I mean, I'm shocked to hear that the heritability is as high as it is, 0.8 to 0.9. What is it for

02:00:00.180 other DSM conditions such as bipolar disorder and schizophrenia?

02:00:05.560 So bipolar disorder and schizophrenia, certainly much lower, especially in even things like

02:00:11.580 depression, major depression, lower still. So this is actually one of the highest heritable factors in

02:00:18.040 terms of behavioral health or psychiatric conditions.

02:00:20.520 And just approximately, what are the heritability factors for those other conditions you just

02:00:25.000 mentioned?

02:00:26.120 So more in the neighborhood of 0.5 to 0.6 or for something like major depression, even lower,

02:00:30.980 maybe more like 0.3.

02:00:32.440 Wow. What is the implication, by the way, if the heritability is 0.8 to 0.9, that almost implies

02:00:41.000 for certain that a person with autism will have an autistic child, doesn't it?

02:00:45.260 Well, that's only half the equation, right? So it takes two to tango in terms of making

02:00:50.040 a child. So both individuals are contributing genetic factors as we're thinking about this.

02:00:55.800 And so, and it's the combination of those factors together within that combination.

02:01:00.920 So if you're talking about-

02:01:01.900 Even if we're not talking about polygenic

02:01:03.760 combinations. So in other words, of those, I think you said, 100 to 1,000 genes that seem

02:01:09.220 to be implicated in autism, some individuals with autism may only have one of those genes,

02:01:14.160 correct?

02:01:15.100 Oh, absolutely. So some of them may have only one gene that's the predominant sort of contributor

02:01:20.400 in terms of this. Some will be more of that polygenic combination of factors.

02:01:24.260 But a single gene individual has a 50% chance of passing that gene onto their offspring. And

02:01:29.220 assuming it's fully penetrant, they would have effectively a 40 to 50% chance of transmitting

02:01:34.460 autism to their offspring.

02:01:36.380 That's correct. Now, for the types of genes that you mentioned, many of those individuals

02:01:41.020 actually won't go on to have their own families. And so one of the factors within this is that

02:01:46.060 those highly penetrant single gene factors, many individuals, for instance, you can imagine if

02:01:51.780 they're not living independently, if they're not verbal, you know, they won't pass those genes down.

02:01:57.200 Let's pivot for a moment to cardiovascular disease. There are a couple of things that stand

02:02:01.480 out from a genetic perspective. One I've already mentioned briefly, which is FH. The other is

02:02:06.620 LP little a attached to the LPA gene. I believe that LP little a is the most prevalent atherogenic

02:02:13.860 condition that is genetically associated. Roughly one in 10 people haven't elevated that. That would

02:02:18.660 be another great example of how gene addition therapy would be of no use because you'd want to be gene

02:02:25.200 editing that. What else do we know about cardiovascular disease beyond those two special cases of elevated

02:02:33.040 LP little a and familial hypercholesterolemia? How much of the rest of ASCVD appears to be heritable?

02:02:41.040 Well, the interesting thing is, and we alluded to this a little bit, when we think about myocardial

02:02:45.780 infarctions, hyperlipidemia, in that sort of cardiovascular health, there may be genetic factors that are

02:02:53.860 numerous, but from a therapeutic point of view, we may not target all of those genetic contributors. There may be

02:02:59.600 final common biology. And so I think that's the theme that I see most in terms of thinking about coronary

02:03:05.560 artery disease, myocardial infarction, things like that. Within the cardiovascular disease space, though, I do

02:03:11.640 think they're interesting. I'll just give one other example in terms of a relatively common but rare disease, that of

02:03:18.300 cardiomyopathies. And so when you think about genetic cardiomyopathies, they affect on average about one in 500

02:03:24.760 individuals. And so that's not one in a million. It's also not 10% of the population, right? We're

02:03:30.160 somewhere in the middle. And I will say that I was waiting to see the data come out, but we've known about many of

02:03:36.520 those genes for a long time. We've known about the mutations. We've known about frequency. We've known about

02:03:42.060 natural history and waiting to see whether or not genetic therapies would be effective for those

02:03:47.780 conditions. And it's not yet in people, but I will say the early data are looking promising in terms of

02:03:53.520 animal models to be able to reverse this or prevent this. I still think the heart is a tricky organ to

02:03:59.760 be futzing with. I'll just put it that way. It always makes me a little bit nervous because, you know,

02:04:04.720 electrical things can happen very suddenly and it can be quite dangerous. So, you know, that always

02:04:09.240 makes me a little bit nervous to think about genetic therapies for the heart. But like I said, some of the

02:04:14.200 early data from the Seidmans in particular look like there could be promising roads ahead. And like I said,

02:04:20.060 it's a common condition that otherwise oftentimes we treat with transplant, you know, when we, when the heart

02:04:25.660 finally fails. And so it'd be lovely if we didn't have to wait for hearts for transplant.

02:04:30.880 Yeah, it's interesting. So going back to what you said about the ASCVD side, it almost sounds to me like you're

02:04:36.840 saying that if genetic therapies and treatments are going to be deployed against conditions, especially like FH,

02:04:45.420 you would really do it to more mimic the drug than you would to try to correct the defect. And that

02:04:52.180 makes a lot of sense in the case of FH because of the heterogeneity, right? You know, to have 5,000

02:04:57.640 different gene therapies for the 5,000 different genes that can be altered in the result of hyperlipidemia

02:05:05.080 is a bad idea. Whereas if you can simply knock out PCSK9 as a gene, you basically take care of everyone.

02:05:14.640 And so let's talk about what that means technically. So presumably there are genetic therapies that are

02:05:20.700 already in the works at looking at targeting PCSK9. PCSK9 inhibitors as a class of drug

02:05:27.220 have been perhaps the most exciting drug class introduced in the last decade. And the results

02:05:33.460 have lived up to the hype. I mean, when Helen Hobbs first made the discovery of the individuals that

02:05:39.740 were both hyper and hypofunctioning PCSK9, I still remember reading those papers 15 years ago thinking,

02:05:46.620 this is too good to be true. This will not pan out was my, that was my dumb prediction. This will not 0.95

02:05:54.420 pan out. And I was wrong. I'm delighted to be wrong. So what does that look like now? What does the gene

02:05:59.800 therapy look like to silence a gene in this case, but without creating some unintended consequence?

02:06:04.800 It's going to start out with not just, you know, the average person who might have a higher risk. It's going

02:06:10.000 to start out at the extreme. For someone, we talked about Jesse Gelsinger, who's going to be willing to be

02:06:14.780 the first in and try this and be that brave first person. And so I won't claim that I've designed the

02:06:20.900 clinical trial that goes with this, but just to say it's going to start at that extreme. And there are going

02:06:25.740 to be a lot of complexities. I will say, and I'm sure many listeners are thinking of this, the cost with

02:06:31.220 gene therapies is prohibitive right now. It's not as if we could, if any single gene therapy were $3

02:06:37.780 million, which is not uncommon for current gene therapies, we can't afford to spend $3 million

02:06:43.160 per person with the number of people who are at risk in the U.S. population.

02:06:47.620 That SMA single gene therapy is about a $3 million treatment?

02:06:51.340 I'll say the range of genetic therapies right now runs between about $1 and $3 million. So depending

02:06:56.860 on which ones are out there, and it gets complicated. Sorry, just to say one thing about

02:07:01.860 that, it is worth keeping that in perspective, right? I'm not going to sort of advocate one way

02:07:06.780 or the other, but it's not uncommon to spend a million dollars on chemotherapy at the end of life

02:07:13.540 for a year's worth of life extension. And if you contrast that with a million dollar gene therapy

02:07:20.620 in infancy that gives 80 or 90 years of life extension, it at least puts those two treatments

02:07:28.320 in context. I agree with you 100%. Health economists have been trying to get at that in terms of the

02:07:34.780 value, you know, if we're talking about true value in terms of this, and I'm not going to pit one

02:07:39.780 against the other in terms of this, but you do have to think about, and I would argue not just the

02:07:44.420 healthcare system cost to the person, but it's the societal cost to the family, to the community,

02:07:50.320 there's a lot of cost that goes into this if you do the economic analysis accurately.

02:07:55.620 On the other hand, you know, to say that that might be worthwhile in one case,

02:08:00.520 you do have to think about scaling, and I'll just throw out a number for you. 10% of the U.S.

02:08:05.940 population has a rare genetic condition. They may or may not know it, but that's just true in terms

02:08:10.880 of these monogenic factors that we've been talking about. So that's 10% of the population.

02:08:15.940 If you now think about the, you know, percentage of the population that's obese or that has

02:08:20.240 obesity or type 2 diabetes or some of these other common conditions that someday might be treated

02:08:25.620 by these one-and-done types of things, then it becomes, in terms of the society, what can we

02:08:31.500 afford to do? What are the competing other healthcare costs or other societal costs in

02:08:35.820 general that we have? And, you know, how are we going to right-size these? I will say that I'm

02:08:40.440 confident that as we have more ability to understand how to do this, you know, there's a lot of fixed

02:08:47.460 costs, but the marginal cost is not nearly as high. I think you know what I mean by that in terms of,

02:08:51.800 you know, both what it takes to design the clinical trials, to do the manufacturing,

02:08:55.820 to do the monitoring, you know, all of these things, they won't scale linearly. And so I do

02:09:00.700 think we'll have some cost realization that we can recoup. But, you know, I think those are the big

02:09:05.500 questions that we think about is how are we going to afford this and what are the key things that we

02:09:09.720 need to do to enable doing this on scale? To your point, what are the competing alternatives? If it is,

02:09:16.400 you know, a medication that you're taking every day, if you can't really get, you know, to a good

02:09:20.920 point in terms of the MI risk or anything in terms of heart health or stroke health or other things.

02:09:27.260 So all of those will go into it and eventually a health economist will price this out and figure

02:09:31.600 out, you know, what a reasonable fee is to charge for such therapies. You know, again, just going back

02:09:36.160 to what it costs today on the gene therapy side, it still seems like, I don't say this to be

02:09:41.720 disparaging of the field at all because the field is remarkable, but I just say it is more of an

02:09:45.420 observation of where we are relative to, say, gene sequencing. We're still in its infancy, aren't

02:09:50.400 we? Oh, absolutely. I mean, as sad as it is to say, given what you said about Jesse Gelsinger's

02:09:57.000 case was more than 20 years ago, we are still at our infancy in terms of being able to realize all

02:10:02.700 the potential. I don't even think we have all the technologies or the vehicles or delivery systems

02:10:07.540 that we're going to eventually be effective. Again, thinking it through that way, in the year

02:10:12.260 2000, it cost $1 billion to sequence a human genome. Today, it costs $1,000. So that's a six-log

02:10:23.960 reduction in cost, in part through Moore's law, in part through the step function change of high

02:10:30.100 throughput sequencing. We don't need a six-log reduction in the cost of gene therapy to make

02:10:36.240 it readily available. Quite frankly, a two-log reduction would make this a game-changer. Does

02:10:44.200 that strike you as something that's feasible in the next two decades? I would say, yes, there are

02:10:48.940 going to be certain catalytic transformative breakthroughs that will make and enable those

02:10:54.280 changes that you're talking about. So again, I'll go back to what we did with the COVID vaccines.

02:11:00.660 It was incumbent upon everyone to be able to come up with solutions. And the solutions

02:11:05.460 that allowed for the adaptability and even changing the vaccine on the fly were remarkable to me, at

02:11:12.560 least from a scientific point of view. And I know some people may push back on me, but that is my true

02:11:16.800 belief. If you could think about the same way with delivering an mRNA vaccine and doing the same thing

02:11:22.740 and realize that it's different for genetic gene addition and doing it, but it's not entirely different

02:11:28.400 in terms of how to do this. And so as you're talking about, I call it rinse and repeat, but it's being

02:11:33.660 able to do this and having the infrastructure, the delivery system, the regulatory system, the

02:11:39.660 manufacturing process, all of those things. Once you get this and get this down, there are ways to scale

02:11:44.940 this and to be able to, if the gene fits, if it's a certain size, if there's certain mutations, to be

02:11:49.540 able to do this repeatedly. So I do have that hope, but there are going to be the function I think we're

02:11:54.320 going to see. I call it a step function, right? So you're going to see it's not going to be linear.

02:11:58.600 It's not going to be, right? So that's what we'll see.

02:12:01.220 So last year I read Walter Isaacson's biography of Jennifer Dadna and the story of the discovery of

02:12:06.880 CRISPR. I thought it was a fantastic book. I can't recommend it highly enough to people who are listening

02:12:11.160 or watching. I'm sure you've read it. There was a fantastic discussion of the ethics of this. And once you

02:12:17.760 realize the power of gene editing, you very quickly start to pivot away from the discussion we're

02:12:25.260 having today, which is a child is born with Tay-Sachs disease. This child is going to be dead

02:12:31.040 in a couple of years and it's a very ugly death. There's really nobody in their right mind that

02:12:36.620 wouldn't be in favor of a therapy there. We could go through all the examples we've talked about and

02:12:41.980 there's nobody I can imagine that's going to say, if a woman is born with a BRCA mutation and

02:12:47.660 she has the choice between a gene edit to fix it or a mastectomy to remove her breast, we'd probably 1.00

02:12:55.780 prefer the former. If for no other reason, then it ends the gene there and her daughter won't get it 1.00

02:13:02.400 or her son won't get it, et cetera. How can people think about the next layer of complexity, which is,

02:13:09.060 well, should we be able to take a person who has an APOE4 isoform and turn that into an APOE3 isoform

02:13:19.640 or an APOE2 isoform, which would actually come with significant protection against neurodegenerative

02:13:25.080 disease? That's a slightly different case because, of course, the penetrance and the risk profile is

02:13:32.920 different. So how do you, as a scientist, think about this? Because I think that both ethicists and

02:13:38.680 scientists need to be a part of this discussion. So I agree completely. And I have to admit,

02:13:44.020 so one thing I think we universally agree on, I hope, is that we're not doing gene editing,

02:13:49.940 gene manipulation to affect the next generation. So we're not looking for things that are transmissible

02:13:55.100 in the germ line. We're not trying to create superhuman where we're trying to, you know, fiddle

02:14:00.760 with the genes to either correct them or to be able to enhance them. That's the term that I'll use

02:14:06.100 that's transmissible. I think that's a line scientists and ethicists.

02:14:10.040 But sorry, would that be true even with something like Tay-Sachs or cystic fibrosis or things like

02:14:15.620 that? That's what the consensus among the scientific community is, is that, again, that you would treat

02:14:20.920 the soma or the body of the person that might be at risk or have those diseases, but you wouldn't try

02:14:26.880 and do manipulations that would be transmissible to the next generation.

02:14:30.360 I see. So my argument for BRCA is only partially correct. I argued that the gene therapy would be

02:14:36.800 favorable because it would spare the woman a mastectomy and spare the risk of transmission. 1.00

02:14:42.300 You're saying, no, you would only do it in the somatic cells, not the germ line. She could still

02:14:46.960 transmit that gene to her daughter. 1.00

02:14:48.720 That's correct. That's the current consensus scientifically. The other part of it is this

02:14:54.020 tricky thing that you're talking about, which is enhancement in terms of it's not correcting

02:14:59.300 something that's problematic. It's enhancing the body the way it is or trying to, in some

02:15:05.180 sense, prevent a disease process. But I will say the enhancement, once you get to the point

02:15:10.520 of enhancement, that's a trigger word for we shouldn't go there. And part of this is that

02:15:15.840 we're not as smart as we think we are. You know, we can think that something's not going

02:15:19.740 to have off-target effects, that it's not going to disrupt a gene inadvertently. But I think

02:15:24.640 in the short term, what saves us is that the risk profile given the uncertainty in the long

02:15:30.280 term is so high that there aren't going to be either scientists or people, I think, who

02:15:34.360 are going to, I hope, go for things that are trivial in terms of enhancements. I'd say

02:15:39.400 that's the first part of this. But to your point in the APOE234 situation, I think is one

02:15:46.040 that people think about. I also think that although probably the average listener here is more

02:15:51.100 sophisticated about this, I think the average person, you know, walking down Broadway here

02:15:56.060 is not going to think about this in quite the same way. And so they're going to think

02:16:00.120 about things that will be enhancements, whether it's being able to increase your earning potential,

02:16:05.680 being able to be taller, more athletic, you know, funnier, you know, there are multiple

02:16:10.280 dimensions in which people would, and people have been surveyed to figure out, you know, how

02:16:14.040 they would value certain attributes, you know, would definitely be thinking about doing

02:16:18.660 that. I think at this point, it has to be just enhancement is a line that people are,

02:16:23.200 I hope, not crossing, that it really is about disease and being able to make people healthier

02:16:27.380 as we're doing this. But to the extent that there are certain medical industries that are

02:16:32.740 not covered by insurance, that people that are not regulated, there are certain parts of

02:16:37.120 the world that do things differently, where people can go and seek certain things. I do hope

02:16:42.680 that there is a consensus scientifically about places we don't go, because otherwise there

02:16:47.020 will be ways that people will find to do things. I mean, is it worth just explaining the story of

02:16:52.380 kind of the initial blow up around CRISPR and what took place in China and how that brought the

02:16:58.860 scientific community in some ways closer around this consensus? So the circumstance in China was

02:17:05.620 something I have to admit was, I thought was a little unusual. So it was a circumstance in which

02:17:12.020 the CCR5 gene was manipulated to try and prevent children from getting infected with HIV. By

02:17:19.380 manipulating that gene, it's not as if that was curing a disease, to your point. It's not as if it

02:17:24.220 was preventing a disease that was a foregone conclusion. It was preventing transmission of

02:17:30.420 infectious agent that those children were at increased risk for, but it was not a foregone conclusion that they

02:17:35.820 would have been HIV or HIV positive. And so... Wasn't it also the case that there was very

02:17:40.540 little chance they were going to be because these were children born of IVF and the sperm are washed in 0.99

02:17:46.220 that situation? And therefore, because in this case, I believe the father was HIV positive, the mother

02:17:51.540 HIV negative. But it was a situation where it was almost entirely possible to prevent the transmission

02:17:57.820 of HIV to the offspring. Is that correct? Right. That was my point in terms of, I don't think heroic

02:18:04.180 measures needed to be made to be able to go. There were perfectly, I think, reasonable ways that are

02:18:10.160 very effective, not just reasonable, but very effective. So it was an odd sort of selection of

02:18:15.360 a use case, I guess, is the point that I'm making scientifically. If I were going to do something,

02:18:19.240 I would have done it for this use case. But regardless of that, the scientific community also

02:18:23.520 just, I think, rallied around that a line had been crossed, right? That this was essentially a form of

02:18:29.300 enhancement. It wasn't something that was saving a life, saving a high probability of

02:18:34.080 something for which there were no other treatments, anything else. But I think the other point that

02:18:39.180 I made is true, which is that this is, the world is global and scientists are in all sorts of places

02:18:45.340 and it only takes one person to be able to do something that's crossing a line. And there are

02:18:50.860 people that there is tourism, medical tourism in places and people I've seen go to where they feel

02:18:57.220 that there's something that they think is important that they'll seek. And so I think it is important for

02:19:01.920 hopefully people to uphold certain ethical standards and for us to, you know, have those guiding

02:19:07.360 principles so that people know where those lines are. Do you think it would have been different if

02:19:11.280 the first use of CRISPR in humans was to do something that we could all agree on would be a

02:19:16.940 great use case? In other words, would the field be in a different place today? So what year was this

02:19:21.400 that that, and I'm blanking on the name of the scientist who did this, but I know he's been

02:19:25.360 basically, if not put in jail, certainly put out of science. I can't recall the year either. But I'm

02:19:32.760 trying to think of what would have been that perfect use case. What were you thinking?

02:19:37.480 Sickle cell. Let's say you took a kid or Tay-Sachs for that matter. You took a kid who had a disease 0.65

02:19:42.980 that was going to significantly impair the quality of their life. You used gene editing to fix the gene

02:19:49.800 and produce a phenotype that could have a normal life expectancy. So again, same technology,

02:19:57.600 but far better application. I think what most people in the field, and this is usually possible,

02:20:04.140 is that rather than trying to diddle the genes, if you're at the stage of an embryo with in vitro

02:20:10.080 fertilization, you can simply do selection of an embryo that doesn't have that genetic risk. And

02:20:16.820 therefore, you're not increasing the risk of something off target, and you're still accomplishing

02:20:21.260 what it is that you would set out to achieve. I have seen some ethicists make the argument that

02:20:26.600 if there were a very limited number of embryos, and that were not possible to do, would that be,

02:20:31.600 you know, a circumstance in which, like you said, for the purposes of direct therapy for a very bad

02:20:37.460 disease in which that couple had no other alternative in terms of having a biological child, would that

02:20:43.160 be possible, knowing that that would affect the germline. So as opposed to what we were talking

02:20:48.360 about before would be something. And that is what I would say is gets us close to the edge in terms of

02:20:54.700 thinking about that, sort of the use case that you were talking about, but not as a matter of routine

02:20:59.780 and not a routine goal in terms of the intention to do in a universal way.

02:21:06.020 It's incredibly fascinating. If you had a crystal ball, you know, and you could look into the future

02:21:10.560 in 2040. Probably you're coming to the end of your career, you're thinking about retiring,

02:21:16.340 but you're looking back at a 40 plus year career in a space that has probably seen more transformation

02:21:24.200 than most fields in all of medicine. What would you not be surprised to see happening in 2040

02:21:33.420 in this field? I would not be surprised that diagnosis is trivial. I do hope we're getting to that

02:21:40.220 point. I think there are not just with the cost of sequencing decreasing, but with machine learning,

02:21:46.860 artificial intelligence, the ability to ingest huge amounts of information, genomic information,

02:21:53.420 clinical information, other information, the diagnostic ability in medicine in general. And I won't say this

02:22:01.240 is limited just to genetics and genomics, but is especially true in this field. Diagnostics will be

02:22:06.660 hopefully trivialized. And I hope, although I can't guarantee it's going to happen, I hope from an equity

02:22:11.920 point of view will be more accessible to more people around the world just because that cost barrier will

02:22:17.080 decrease. What I'm not sure of is on the therapeutic point of view to what we've been talking about, how much of

02:22:22.920 that we will have realized within 20 years. While I'm optimistic we will have gotten a lot of that done, I'm not sure how

02:22:30.020 how much it will penetrate either parts of society in the United States or globally in terms of what

02:22:36.300 we'll be able to do. And this is just, I think it's very hard for me to predict timing. And I think Bill Gates

02:22:42.260 has a quote similar to this. You know, I do think within some points in the next hundred years, we'll get to this

02:22:47.980 point, whether it's the next 20 years or not, and exactly what percentage of individuals we will have been able to

02:22:53.980 serve. I'm not sure. Because there are going to be, as I alluded to, these kind of step functions, these

02:22:58.620 transformative things that it's just very hard for me to predict, given that I feel like we got stuck for

02:23:05.180 the last 20 years, but I do feel like we're gaining momentum. But there's, it doesn't take much in society

02:23:11.160 and in science to get us stuck. So I think that's just a word of caution. I would not have predicted,

02:23:17.000 I hate to make things about COVID, but I would not have predicted in society what's happened in the last 10

02:23:21.520 years or some of the trajectories we've had. So I will not use my crystal ball and I'll say we'll be

02:23:26.600 much farther ahead, but I don't know, big confidence interval in where we're going to be in 20 years.

02:23:31.780 Well, Wendy, congratulations on your new role at Boston Children's Hospital. Obviously for folks

02:23:36.940 not aware, that's certainly probably one of the three most preeminent children's hospitals in North

02:23:42.720 America. And maybe you would argue the single most, but anyway, that sounds like a wonderful

02:23:46.840 opportunity. I assume you'll be able to stay involved in Spark and Guardian as a PI as those

02:23:52.940 tend to expand. So I especially want to thank you for making time on the day that you're literally

02:23:57.080 moving to Boston to make time to sit down for so long. So congratulations and thank you again very much

02:24:02.060 for your input.

02:24:02.920 Thanks. It's been a lot of fun.

02:24:04.300 Thank you for listening to this week's episode of The Drive. It's extremely important to me to

02:24:09.740 provide all of this content without relying on paid ads. To do this, our work is made entirely

02:24:14.720 possible by our members. And in return, we offer exclusive member-only content and benefits above

02:24:21.140 and beyond what is available for free. So if you want to take your knowledge of this space to the next

02:24:25.680 level, it's our goal to ensure members get back much more than the price of the subscription.

02:24:30.200 Premium membership includes several benefits. First, comprehensive podcast show notes that detail

02:24:36.960 every topic, paper, person, and thing that we discuss in each episode. And the word on the street

02:24:42.580 is nobody's show notes rival ours. Second, monthly ask me anything or AMA episodes. These episodes are

02:24:50.680 comprised of detailed responses to subscriber questions, typically focused on a single topic

02:24:55.540 and are designed to offer a great deal of clarity and detail on topics of special interest to our

02:25:00.900 members. You'll also get access to the show notes for these episodes, of course. Third, delivery of

02:25:06.640 our premium newsletter, which is put together by our dedicated team of research analysts. This newsletter

02:25:12.180 covers a wide range of topics related to longevity and provides much more detail than our free weekly

02:25:18.420 newsletter. Fourth, access to our private podcast feed that provides you with access to every episode,

02:25:25.300 including AMA's sans the spiel you're listening to now and in your regular podcast feed. Fifth,

02:25:32.220 the qualies, an additional member only podcast we put together that serves as a highlight reel featuring

02:25:38.420 the best excerpts from previous episodes of the drive. This is a great way to catch up on previous

02:25:43.600 episodes without having to go back and listen to each one of them. And finally, other benefits that

02:25:48.600 are added along the way. If you want to learn more and access these member only benefits, you can head

02:25:54.000 over to peteratiamd.com forward slash subscribe. You can also find me on YouTube, Instagram, and Twitter,

02:26:01.480 all with the handle peteratiamd. You can also leave us review on Apple podcasts or whatever podcast

02:26:08.040 player you use. This podcast is for general informational purposes only and does not

02:26:13.240 constitute the practice of medicine, nursing, or other professional healthcare services, including

02:26:17.620 the giving of medical advice. No doctor patient relationship is formed. The use of this information

02:26:23.740 and the materials linked to this podcast is at the user's own risk. The content on this podcast is not

02:26:30.120 intended to be a substitute for professional medical advice, diagnosis, or treatment. Users should not

02:26:35.680 disregard or delay in obtaining medical advice from any medical condition they have, and they should

02:26:40.420 seek the assistance of their healthcare professionals for any such conditions. Finally, I take all conflicts

02:26:46.760 of interest very seriously. For all of my disclosures and the companies I invest in or advise, please visit

02:26:53.100 peteratiamd.com forward slash about where I keep an up-to-date and active list of all disclosures.

02:27:05.680 Peteratiamd.com forward slash about where I keep an up-to-date and active list of all disclosures.

The Peter Attia Drive - August 28, 2023

#268 ‒ Genetics： testing, therapy, editing, association with disease risk, autism, and more ｜ Wendy Chung, M.D., Ph.D.

Episode Stats

Summary

Transcript