Assembling an Egg

Vineeta Agarwala, Justin Larkin, Judy Savitskaya, and Lauren Richardson

On this episode of the Bio Eats World Journal Club, we explore the very compelling question of whether we can use our understanding of developmental biology to create oocytes (aka eggs or female gametes) from stem cells in the lab. If possible, this could be on par with the development of in vitro fertilization in terms of extending fertility. But creating an oocyte from a stem cell has some unique and high-stakes challenges. Host Lauren Richardson is joined by a16z general partner Vineeta Agarwala and deal partners Judy Savitskaya and Justin Larkin to discuss the research article “Reconstitution of the oocyte transcriptional network with transcription factors” by Nobuhiko Hamazaki, Hirohisa Kyogoku, Hiromitsu Araki, Fumihito Miura, Chisako Horikawa, Norio Hamada, So Shimamoto, Orie Hikabe, Kinichi Nakashima, Tomoya S. Kitajima, Takashi Ito, Harry G. Leitch and Katsuhiko Hayashi, published in Nature, which makes a big step towards this goal. The conversation covers which aspects of oocyte biology the authors were able to replicate, which they were not, and where we think this field might be heading.

Show Notes

  • A brief discussion of the biology [3:24] and terminology [5:27]
  • Potential uses for oocyte biology [7:34], what it would take to convert cell types [10:36], and the development of the oocyte [12:36]
  • Limitations of this study [16:59] and the future of oocyte biology [19:25]


Hanne: Hi, I’m Hanne.

Lauren: And I’m Lauren, and this is the Bio Eats World Journal Club, where we discuss breakthrough scientific research, the new opportunities it presents, and how to take it from paper to practice.

Hanne: So, Lauren, you’ve titled this episode “Assembling an Egg,” but I’m going to guess we aren’t discussing your favorite breakfast recipes today.

Lauren: Hanne, you know me well, nope. Today, we are exploring the very compelling question of whether we can use our understanding of developmental biology to create oocytes, AKA eggs — you know, the female counterpart of sperm — from stem cells in the lab.

Hanne: Okay. So, refresh my memory. Do we have stem cells in our adult human bodies that could be used to turn into eggs?

Lauren: Yes and no. Adults have stem cells, but their ability to turn into other types of cells is limited. For example, we have stem cells in our bone marrow, but they can only produce the different types of blood cells, but scientists now know how to turn some of our cells back into stem cells, which are called induced pluripotent stem cells, or iPS cells. If possible, this adult cell to iPS cell to egg cell transformation could be on par with the development of in vitro fertilization in terms of extending fertility, but creating an oocyte by this path is tricky. So, I’ve gathered some of our colleagues, a16z general partner Vineeta Agarwala, and deal partners Judy Savitskaya and Justin Larkin, to discuss a recent research article published in Nature, by Hamazaki et al., that makes a big step towards this goal. We discuss what aspects of oocyte biology the authors were able to replicate, which they were not, and where we think this field might be heading. We start with Vineeta describing the state of innovation in fertility. 

Vineeta: Fertility, even just within the U.S., is approaching a $10 billion industry, and the majority of innovation that we see happening in the space is related to care delivery. How can we expand access to fertility treatments to couples who need them? That’s an important problem, but at the same time, it makes us wonder whether there are biological breakthroughs that could change the face of the industry even more than care delivery technology might.

Justin: Yeah. For so many aspects of reproductive therapies and infertility challenges, the egg or the oocyte is often the critical limiting step, both from an ability to actually get to a healthy pregnancy, but also in the process of the patient journey. Having seen numerous people go through the IVF process and other kinds of iterations of IVF, it’s often just so hard from a patient journey perspective — literally being painful, takes lots of time, it’s expensive. And so, the thought of being able to fundamentally change access to oocytes, and eggs that potentially could be grown in vitro in ways that cut the process down and make it more affordable, has the potential, I think, to not only open up reproductive technology and therapies to a broader population, but to make that care experience just fundamentally different for patients. And so, seeing the early innings of that potentially be suggested in this paper was fascinating to me.

Judy: Vineeta and I were actually pretty surprised the other day when we realized that the most expensive part of IVF, the single biggest line item, is the drugs that are required for egg retrieval. So, this is a way to just get around that entire process.

Overview of oocyte biology

Lauren:  Right. Could you unpack that? How is this solving the problem?

Vineeta: Women are born with a couple of million potential eggs, and as we age, we lose those eggs, and we don’t make more. And this paper provides this suggestion of a way to engineer, from pluripotent stem cells, oocytes that can potentially be fertilized and give rise to a new being, without being dependent on that very constrained, constantly dwindling supply of oocytes that we’re born with as women.

Lauren: Right. I think that is such an interesting idea, that when your baby is still in you, it has its full complement of eggs that it’s ever going to have. Females don’t make additional eggs after birth, so you have to work with what you have. As Judy said, it’s very expensive to retrieve eggs from a woman. So, this paper asks a really tantalizing question of — can we make new eggs from cells after birth, from your own cells after you’ve been born?

Vineeta: I would think of it as a replacement for egg freezing, is one way to contextualize where this could, hypothetically, fit into the industry. Today, women are freezing eggs in their 20s so that they may access them in their 30s and 40s or beyond, and this would provide an alternative to having to undertake that process of retrieval, freezing, and hoping that the eggs are usable.

Justin: Often in those retrievals, you’re only able to get — you know, let’s call it 1, 2, 3, sometimes if you’re lucky, more than that — but a limited number of oocytes, and so you have a limited number of shots on goal on potentially having a successful pregnancy. Where I think what got me excited about this paper is this idea of potentially having more opportunities [for] a successful pregnancy that aren’t as constrained by the cost and kind of care experience today that drives the typical IVF process that we alluded to earlier.

Lauren: We’ve thrown a couple different terms around. We’ve thrown around oocyte, we’ve talked about eggs. Let’s define this. What is an oocyte, exactly?

Justin: Yeah. An oocyte is not a static thing, and it goes through a number of different developmental phases, starting from an undifferentiated cell, ultimately to being a true oocyte, which is what we would call a gamete or female sex cell. 

The journey to get there is quite a fascinating one. As you alluded to, it starts, actually, in utero. Then as it gets further along in its development, it starts out as what we call a primary oocyte. And as a primary oocyte, it still has the deployed genetic material. Then it starts meiosis, which is the process of going from being deployed and having two copies of chromosomes, to being haploid or just having a single copy. And in that process, it actually arrests and will stay in that arrested state until that female hits puberty, and then that full mitotic process isn’t completed until after ovulation and even with fertilization. 

So, it’s highly dynamic, and I think the critical piece here is that there’s a lot of gene regulation that’s going on throughout, which opens up the opportunity for research studies like this to really define what are the regulators of that process. Can we recapitulate that in vitro setting to potentially have some of the applications that we talked about earlier?

Lauren: Yeah. I definitely was doing my background reading in preparation for this conversation, and really appreciating all those different steps that happen during — in utero. What happens before puberty, what happens during puberty, what happens during ovulation, what happens at fertilization — there’s all of these different steps, all these different developmental changes that all of them seem to carry the name oocyte.

Justin: One other nuance that I think is important, too, is just that the oocyte in isolation isn’t necessarily enough, right? There has to be what they call a follicle, which is a set of cells that surrounds it. It’s very hormonally active, a lot of cell signaling and transcription factors are produced in that process. And we start to see early recapitulations of that in this study. But in vivo, inside of the female, there’s a lot of other, kind of, cellular activity that goes on around the oocyte cell itself that has a lot of hormonal impact downstream.

Potential uses in healthcare

Lauren: Yeah. That’s a really good point, that it’s very dependent and has a very complex interaction with the ovary and the cells of the ovary. So, on previous episodes of “Journal Club,” we have talked about converting one cell type to another. For example, we’ve talked about converting stem cells into the cells of the pancreas that produce insulin. But what are some of the particular hurdles in creating an oocyte that you wouldn’t have to deal with in just converting a stem cell into, say, a pancreatic cell?

Vineeta: One way to think about the second-order challenge here is that you have to not only get to a primary oocyte, which has grown and has the right differentiation to at least have started on the path of egg generation, so to speak. But beyond being a primary oocyte, you have to get to a secondary oocyte. And the difference is that you have to undergo the whole process of meiosis, which is how we generate genetic diversity. 

And so, that secondary oocyte is a pretty complicated thing to make. Not only because its genome has to have halved in a very unique way, by a very unique process, but a lot of epigenetic signaling that then goes on to determine gene expression networks in the subsequent fertilized embryo are thought to be driven by a program set in that oocyte. So, some genes are expressed off of DNA you got from your mom, some genes are expressed off of DNA you got from your dad, and not vice versa. A lot of that epigenetic programming is thought to stem from a pattern that’s encoded in the secondary oocyte.

Judy: Another major difference between this and other cell types that you could potentially reprogram is just the morphology. So, oocytes are the largest cell that we have in humanity. The size and shape is just really different from what you would expect for other cell types.

Lauren: Right. So, in my previous example of turning a stem cell into a pancreatic cell, there might be morphological differences between those two cell types, but they aren’t necessarily at the scale that you would see in a stem cell to an oocyte. And, fundamentally, you aren’t changing the genome when you’re doing this transition from a stem cell to a pancreatic cell — the genome stays exactly the same. You’re just changing what genes are expressed. 

When you’re creating an oocyte, you’re halving the number of chromosomes that you have in a cell. So, a mature oocyte only has half the number of chromosomes that a normal cell does, and that’s because it meets with a sperm that has half the number of chromosomes that a normal cell does. And now, when they fuse and fertilize, now you get the full complement. So, having to do both of those steps is far more challenging, and represents an additional hurdle that you have when creating these iPS or stem cell-derived oocytes. 

Lauren: So, let’s talk about how you convert one cell type into another. What are, kind of, the — how do we think about how cell fate is controlled?

Judy: So, the way that we think about cell types is often which transcription factors are present. And those transcription factors are proteins that will bind to the DNA and basically cause the production of RNA from a given genomic locus, so from a given gene in the DNA. So, transcription factors are used as the sort of master regulators of a cell type, where one transcription factor might be relevant. 

Just to make it super simple, let’s say a certain transcription factor is relevant for neurons, and another one is relevant for hepatocytes. You would expect that all of the genes related to hepatocyte function and development are going to be controlled by that hepatocyte transcription factor. And in the neuron cell, that transcription factor is not present, so none of those relevant genes are made. That’s an oversimplification, but that’s kind of the idea of how these genetic regulatory networks work.

Lauren: Right. And so, when we’re thinking about in the lab, in the clinic, if we want to convert one cell type into another, we can express these transcription factors, which then leads to the expression of all their downstream genes. And then that, kind of, reprograms the cell and says, “We’re a hepatocyte now, we’re a neuron now, we have this transcription factor that’s driving this — let’s call [it] the gene regulatory network.” And that guides the cell to a specific identity. 

So, in this paper, what they’re doing is they’re trying to identify the transcription factors that govern this development of the oocyte so that they can take those transcription factors, express them in a stem cell, and then encourage that stem cell to become an oocyte. So, with that in mind, let’s start with how the authors identified these key genes — these transcription factors that are involved in oocyte development.

The development of the oocyte

Vineeta: My understanding is that they did whole transcriptome profiling of cells at different stages in mouse oocyte development. So, basically, compared the differential expression of lots and lots of genes at each of those different time points, and constructed a network analysis to nominate specifically not just genes, but regulatory genes. And they used a bioinformatic GO search to get to the subset of transcription factors that they believed were driving the evolution of the transcriptome through the differentiation process.

Judy: Yeah. I think a lot of those genes were already known from studies that had done in vivo work, so I think it’s important to make that distinction. You can do this work in vivo, which means taking oocytes at different stages of development in a mouse and actually measuring the transcriptome there. Or, they created a sort of — an organoid model, for lack of a better word — where they put cells into an environment that is similar to what they would experience during oogenesis, and measured the transcriptome at different points in that process. 

And so, I think the purpose of this was partially to find the genes that are relevant, but also partially to identify the period of time within this model system that should map onto the period of time in in vivo oogenesis. To be able to, basically, run their experiments, and know the right time period to look at the cells.

Lauren: I think that’s a very good point. It was both a — can we identify the correct transcription factors? But can we also validate this very handy in vitro system that we can then use to do further study?

Vineeta: And another way to think about it is — in biology, we talk a lot about necessity and sufficiency, and I actually think they did a better job with necessity than sufficiency. They proved that if you knock out any of these top-nominated transcription factors, that you can’t get differentiation past a certain stage. And you really need this set of factors to be expressed in order to get to a primary oocyte. Sufficiency is a much higher bar, right? You have to prove that the thing you got, the primary oocyte you got, can then go on to do all of the things that you expect it to do. And I would say it’s almost impossible to prove until you’ve generated the end state of, like, a mouse baby. So, they make some progress towards sufficiency, but less.

Judy: There’s also a sort of curious result that I would like to, I don’t know — talk to the authors about and maybe understand a little bit more. Which is that they find these eight genes that they collectively called PPT 8, and they do all of their experiments with these eight genes. But there’s a paragraph in the paper that talks about how they found a subset of four that is sufficient to get the same phenotype you see. But if you have some subset of five, six, or seven that contains that subset of four, it doesn’t necessarily mean that that’s going to work. 

So, there’s sort of this — they are sufficient, but then some other part of those eight — that set of eight — has some interaction with that set of four, such that they didn’t trust that the set of four was truly sufficient. So, there’s some complexity going on there that they’ve definitely moved on from by just using the set of eight.

Lauren: Yeah. There does seem to be maybe a higher level of regulation that hasn’t been elucidated yet.

Justin: A couple other things I thought were interesting, too, is that they had to also expose them to just some somatic cells from the ovary to get the development to take place. And so, I think it, again, reiterates the point that while these transcription factors are likely necessary, there’s also some dynamic signaling that’s happening from somatic cells that surround, as opposed to just being purely driven by those transcription factors. 

One other thing that also caught my attention is the timeline that this all took place in. If you look at what’s happening in vivo, this is usually happening over seven or eight days inside the mouse model. But within the in vitro model, they saw this happen over a couple of days. And so, this is likely a necessary set of transcription factors and regulators, but there are likely other regulators — potentially checkpoints or others — that are absent in the system that’s allowing it to run through this process on a much more accelerated timeline. And potentially could explain some of the other issues that they see downstream with having the morphological appearance of an oocyte, but a lot of the functional aspects of it didn’t quite get that.

Limitations of the current research

Lauren: So, what elements of a functional oocyte were they able to recapitulate with these oocyte-like cells, and what’s missing?

Judy: So, what they did successfully show was growth, the morphology. They also showed a couple of expressions of certain factors that we associate with oocytes, but what they didn’t show is that there’s the right dynamics surrounding the DNA. For example, meiosis didn’t occur, which — it’s also not something that they were aiming for, so I don’t think that’s the bar that we should hold them to, but there was no meiosis. So, what you end up with in the end is a cell with a lot of extra DNA in it. And then the other really important piece is that the methylation pattern is incorrect. It’s completely different from what you would expect in an oocyte.

Lauren: The methylation — that’s one of the key epigenetic modifications. So, that governs how the chromosomes are packaged, and that leads to how accessible certain genes are to be turned off and on at particular rates.

Vineeta: The maternal and paternal imprinting, the mechanism by which is most commonly methylation of the DNA, is really important for health and disease. We know now of many different disease states that are actually attributable to incorrect maternal or paternal imprinting. And so, it’s not a minor issue that methylation wasn’t solved, and one that we’d have to pay a lot more attention to as this research advances.

Lauren: Yes, that’s a really good point. So, the real test — the final test of whether you got an oocyte or not — would be to fertilize it with a sperm and to grow a new being — in this case, a mouse pup — up from that. In this paper, they tried that, but it didn’t quite work. What was the result of this experiment?

Justin: So, what they saw — and somewhat not surprisingly — is that when they did fertilize it, a very small percentage of the cells actually went on to cleave at all, and even of those that did, very few made it beyond the two or four-cell division. Which, a lot of the early cleavage in cells is dependent on having that haploid one single set of chromosomes. And so, in this scenario where they weren’t able to achieve meiosis, which means that they weren’t able to go in with a haploid cell, it’s unsurprising that when the cell was fertilized, it resulted ultimately in non-viable embryos — given that the chromosomal count mix is not consistent with traditional fertilization.

The future of oocyte biology

Lauren: Yeah. So, the paper makes some really great advances in our understanding of how an oocyte develops, what the gene networks are, what the transcription factors are that are regulating these — but there are still a lot of mysteries, and still a lot left to study in this process. When you think about these unanswered questions at the end of the paper, what are the questions that interest you? Where are you interested in seeing this work go next?

Justin: I think for me, when I look at this paper, there are two, kind of, key avenues you could take. One is to look at this as, “Can we use the oocyte as a structural scaffold for other scientific applications?” And we see this happening already today, where enucleated oocytes, or rather the oocyte without the nucleus of the DNA material, are used in applications for mitochondrial disease to other potential therapeutic applications. Those are relatively limited. 

And so, for me, the biggest question that this tees up is, “What are the next steps that need to be taken to really understand how we get the nuclear — how we get the meiosis portion of this correct? Because if we’re eventually going to reach, kind of, the vision that we outlined at the beginning — of having this be a critical asset and enabling greater access to fertility treatments — it’s really an absolute necessity and table stake in order for this to progress.

Judy: Yeah, I totally agree with Justin. At the end, the authors say that this is a potential — a potential use case here is somatic cell nuclear transfer, which is another way of saying cloning. I think that there’s not that much need for this kind of a solution for those applications. I think we should really see this as a stepping stone toward [an] entirely ex vivo generation of an oocyte for the purpose of in vitro fertilization.

Vineeta: Yeah. I think one of the things that’s hardest about this particular — the fertility use case of stem cell research is that, presumably, the parents of a prospective child want to see their genome represented in the progeny. A lot of other applications of stem cell research don’t actually have that requirement, especially if you can design creative ways to avoid immunosuppression and to create immune cloaking of a stem cell-derived therapy, and so on. You might actually envision in a lot of other fields an off-the-shelf stem cell-derived or iPSC-derived cell therapy that can be really therapeutic for a lot of patients with different diseases. 

Here, we can’t have that, or at least it doesn’t solve some part of the core fertility challenge. And so, because you’re so dependent on actually running this process on a case-by-case basis with each couple, I think we just have an even further way to go on this. It’s not like you could create a bank at some point, and then differentiate them from that every time you need to spin up a new embryo. You really have to get the whole process right end-to-end from the point of a patient-specific iPSC cell line. And that’s really hard.

Lauren: There’s not to be an allogenic option.

Vineeta: Exactly. There’s no allo-embryo to be had here.

Justin: And the bottom line for me is — in thinking of the ultimate translation of this, we’re obviously in the earliest of early innings in terms of this actually translating to being, kind of, that holy grail for fertility and the fertility treatments that we talked about earlier. I think this does clarify a lot of our understanding about what it takes to create a structurally similar cell to an oocyte. But ultimately, to reach the broader vision that we want, there’s still a lot of work that needs to be done.

Judy: This is actually a really important step for developmental biology as well, because I think it’s one thing to do descriptive research, where you understand an existing system and you characterize all the pieces of it, and it’s an entirely different level of understanding when you can actually rebuild it. So, I think there’s the phrase Feynman says: “You don’t really understand something until you can build it,” or some variation on that. And so, this is a perfect example of a paper that’s using building to get to an understanding that is deeper than what we had before.

Lauren: I think that’s a perfect note to end on. Justin, Judy, Vineeta, thank you for joining me on “Journal Club” today.

Vineeta: Thank you, Lauren.

Judy: Thanks, Lauren.

Justin: Thanks, Lauren. Thanks, Judy. Thanks, Vineeta. This was fun.

Lauren: And that’s it for “Journal Club” this week. If you enjoyed this episode, please subscribe, rate, and review wherever you listen to the podcast. And to learn more about how biology is technology, subscribe to our newsletter at

  • Vineeta Agarwala

    Vineeta Agarwala MD, PhD is a general partner at a16z investing in bio and healthcare technology. She is also a practicing physician and adjunct clinical faculty member at Stanford.

  • Justin Larkin

    Justin Larkin is a deal partner at a16z where he focuses on healthcare technology companies. Prior to joining the firm, he led strategy and operations at Verily and cofounded Wellsheet.

  • Judy Savitskaya

    Judy Savitskaya is a deal partner at a16z where she focuses on bio companies. Previously, she worked on synthetic biology research at UC Berkeley and was a computational modeling and neuronal networks researcher.

  • Lauren Richardson

Value Versus Volume (in Healthcare)

Todd Park, Vijay Pande, and Hanne Winarsky

The way we pay for healthcare in the US has long been by fee-for-service: per doctor visit, per test, per surgery, per hospital stay. But that system has led to rapidly escalating volumes of services and cost to the system—without actually improving outcomes. What if we shifted everything towards paying for value—and outcomes—instead? In this episode, Todd Park, co-founder and executive chairman of Devoted Health, and formerly Chief Technology Officer and technology advisor for President Barack Obama; a16z General Partner Vijay Pande; and Bio Eats World host Hanne Winarsky—talk all about the megatrend of value-based care, and how it is redefining healthcare itself. Why is now the moment for this massive shift? How do we implement it? What does it mean for doctors and patients, insurers and policymakers? What is tech’s role in making it possible, and what’s the business model and incentive for creating value?

Show Notes

  • Discussion of what value-based care is [1:20], and how it emerged as an alternative to the current system [4:35]
  • The importance of targeting healthcare [7:11], and how to scale the value-based system [13:11]
  • Why there is no “silver bullet” to fix the current system [16:00], but how we might shift toward a new paradigm [24:13]
  • Discussion of first steps that may be taken [31:06] and what it would look like to rebuild the healthcare system from scratch [34:00]


Lauren: Hi, I’m Lauren.

Hanne: And I’m Hanne. And this is “Bio Eats World,” our show where we talk about all the ways that our ability to engineer biology and reengineer healthcare is transforming the future. And when it comes to reengineering healthcare, there’s one concept that gets a whole lot of airtime — the concept of value-based care.

Lauren: Value-based care is a term that we’ve thrown around on many different episodes about how the healthcare system is evolving, but we’ve never really gone straight to the heart of the matter.

Hanne: So, that’s what we do in this episode with Todd Park, co-founder and executive chairman of Devoted Health, and formerly chief technology officer and tech advisor for President Barack Obama — along with a16z general partner, Vijay Pandey, and me, Hanne. So, what exactly is this big mega trend of value-based care all about? And how is it redefining what we think of as medicine, treatments and healthcare? What does it mean for doctors and patients, insurers, and policymakers? Why is now the moment for this big shift, and what exactly is tech’s role in it?

Defining value-based care

Hanne: We hear the term value-based care thrown around an awful lot, but we’ve never really talked about what that means. So, this conversation is really about what is value-based care? How do we implement it? Why is it better? And what is technology’s role in that? So, maybe we could just start with — how is it different from the healthcare system today…

Todd: Oh, it’s really different from how the healthcare system generally works today, famously, or infamously.

Vijay: Even aspirationally. I mean, when you think about how the plumbing works.

Todd: That’s right. The root cause, honestly, is how the healthcare system is paid for. So, historically, U.S. healthcare has been paid for in what’s called, “a fee-for-service.” Frank, right. So, doctors, hospitals, and healthcare providers are paid per thing they do — like, per doctor visit, per test, per surgery, per hospital stay. That has led to a situation where we have rapidly escalating volumes of services being delivered. But, unfortunately, we don’t actually have commensurately improving outcomes. We spend the most per capita of any country in the world. We rank at the bottom of the developed world on metrics like avoidable death, preventable death, adverse events, healthy life expectancy — and that’s because we fundamentally have a pay-for-volume payment system. 

If you want to change the situation, then change how healthcare is paid for. Basically, [we should] move away from pay for pure volume to engage in value-based payment — which, in a nutshell, is a payment system that actually financially supports, aids, and abets right care, right place, right time. Value-based care, to me, is the right care, including non-clinical support, delivered in a consistent, coordinated, and proactive way that both improves outcomes and lowers costs, [thus] saving money.

Vijay: It’s interesting to think about how this came about historically. Basically, [it was] a perk given to workers to help them stay with a given employer instead of somebody else. And so, that perk was, like, “We’ll get together, and we’ll pay for your medical bills or pay for things.” It’s kind of akin to, like, I don’t know, like if we gave a perk to say, “We’re going to pay for your plumber bills,” or, “If you have some major catastrophic problem, we’ll pay for the plumbing.” But that doesn’t mean we’re going to keep track of your house, or try to see if your plumbing is in good shape to avoid the problems. We’re not paying for that. That was never part of the plan.

Hanne: We’re not paying for copper pipes. We’re paying for when something really goes wrong.

Vijay: We’re paying for when the pipes burst. But we’re not paying for maintaining your house. That’s something that was always assumed to be on the patient side, so to speak. The thing is, we’ve gotten very good at trying to come up with therapies for cancer, for stroke, for massive heart disease. But, what we’ve come to realize is that that’s actually more expensive, because if we wait that long, those therapies can work. But [these therapies are] painful in many different ways, emotionally and physically for the patient. We want to get there before the pipes burst. Think about the house and the overall health of it.

Hanne: So, can we talk about when this concept started to emerge? Why was there this gradual dawning of realization that [value-based care] was a better north star to orient towards, or would everyone say this is a better north star?

Todd: I think in the last five years, the move to value-based payment and care has gone from, “Is it going to happen?” to, “It’s going to happen. It’s just a question of how fast.” It’s an idea that’s been around for decades. I mean, it goes in the category of a super obvious idea.

Hanne: Right. Right.

Todd: For example, if you have diabetes, or hypertension, or congestive heart failure, there’s an incredibly well-known best practice pathway in the form of medication regimes that get adjusted, based on each patient’s evolving situation. Along with very basic coaching on diet and activity. It’s very, very straightforward to execute. If I’m a primary care doctor being paid fee-for-service, and I’ve got a patient in front of me who has diabetes and hypertension, and I want to spend an hour with that person — to really help educate them about their condition, and really get into it, in terms of coaching and what to do, in terms of coordinating their care and providing them with the right support — I literally can’t afford to do it. Because I have to, in that same hour, see another three patients and get paid the fee per those services to stay alive. It is a bankrupting action for me to actually spend the extra hour; it is far more difficult to actually financially support and execute those pathways than it should be.

Poorly controlled chronic illness is the greatest single driver of more serious events like heart attacks, strokes, kidney disease, eye and nerve damage, and vascular disease. And it’s just nuts. We as a country don’t do these incredibly basic things to keep people healthy during these chronic conditions. In addition to these patients being cared for the way they should be, you will save so much money, because I’m going to actually save you one, two, three hospitalizations that cost $20,000 each. We’ve invested so much in developing incredibly advanced therapies for acute situations, but we aren’t doing the basics. Because systemic execution of the basics, with process control and improvement that every other industry would find routine — it really does require value-based payment and value-based care as the paradigm to make that happen.

Vijay: Yeah, otherwise, there’s no incentive.

Todd. Right.

Getting patients the care they need

Hanne: Todd, when you say the right care at the right place and the right time, can you talk about what exactly that means? And what would that mean on the entire country model?

Todd: Maybe the simplest way to explain it is, as opposed to me as a primary care doctor being paid X dollars to see people for 15 minutes, I am given what’s, in effect, a global budget for all medical spending — physician, drugs, tests, surgeries, and hospital stays — for my patients. It’s risk adjusted. If I’ve got a patient panel that has significantly sicker patients, then my budget’s higher, because essentially, the budget’s set to be equal to what the healthcare system has generally been spending to care for folks in a situation where patients have not been getting — in a systematic, universal way — right care, right place, right time. Highly prevention-oriented care. So, I have this budget that I’m working off of, and my goal as a primary doctor is to proactively get patients the right care in the right place at the right time. If you do that, then it’s been shown, for example, that you can cut hospitalizations versus the status quo by 40% or more.

Hanne: Incredible.

Todd: It’s a significantly positive financial ROI transaction for them, because the way these arrangements work is that they get a share of the savings from keeping folks out of the hospital. That’s why primary doctors are so much happier when you put them in value-based paying arrangements, because under those arrangements, they can afford to spend that extra time with the patient. They can afford to hire personnel on their teams to help care for that patient, and give them the really straightforward care and support that those patients need to actually stay healthy [and] out of the hospital. 

The primary care doctor, basically, generally speaking, doubles their pay under value-based payment by delivering care the way they always thought they were going to [when] they graduated from med school. As a bonus, then you significantly increase your pay, because you’re saving the healthcare system so much money.

Vijay: Yeah. The proof of this is where you have a cohort at risk, and it’s the same provider, and the outcomes are fundamentally different.

Todd: Oh, yeah. And the evidence is conclusive, as well as it being commonsensical.

Hanne: It strikes me that — in the category of super, super obvious, but it’s a different way of measuring. It takes a different amount of time, a different perspective. So, how do you overcome that challenge, to measure how long somebody stays healthy for?

Todd: One interesting challenge that primary care doctors face is the so-called “foot in two boats challenge.” It’s part of the exercise of transmogrifying from a fee-for-service paid operation to a value-based paid operation. Fundamentally, they’re two completely different kinds of operations. So, in a fee-for-service paid operation, if you’re a primary care doctor, you have to maximize throughput in order to actually stay alive. In a value-based payment paid operation, you’re focused on how do I take the best possible proactive care of my patients and keep them out of harm’s way.

Hanne: Prevention.

Todd: Exactly. It’s much more proactive, personalized care of people, and following up with them out of the office to ensure that they have their medications. Any changes in their circumstances actually get reflected in a change of their treatment. That things are going well, and they have the right support, and the right iterations are made to their care.

Hanne: So, it’s a different muscle.

Todd: Yeah, it’s a completely different mode of operation. And so, the “foot in two boats” is a famous articulation of this problem. What if part of my patients are being paid for fee-for-service, and part of my patient panel is being paid in a value-based payment mode? Then, I’m trying to do max throughput and also proactive, systematic care.

Hanne: Yeah, and you’re stretched super thin.

Todd: And you’re trying to be, like, two different modes at once. That’s why the most successful provider organizations under value-based payment have gone all in on one boat. And that’s helped by the fact that, for example, Kaiser and CareMore have a built-in health insurance plan, inside themselves, that then actually pays — in a value-based way — the providers that they employ, that operate in a value-based way.

Scaling value-based care

Hanne: So, is that part of the reason why we haven’t seen it happen faster? Because you have to create something from the ground up that is a total systemic shift?

Todd: I think that’s a really, really good way of encapsulating why there hasn’t been this kind of shift nationally — because you have to go all in. As you think about the Kaisers, the CareMores, and even the ChenMeds — they take global risk payments from health plans, meaning, they go to health plans and say, “Pay me a global capitation payment, in effect, for all care expenses.” So they, in fact, are their own kind of mini-payer, if you will. Their existence proves that if you actually do that, and you get people the right care in the right place at the right time, it both leads to significant improvement outcomes and lower costs.

The common denominator across the successful early American experiments is that they’re a full payer/payer provider stack who can therefore actually act with the right incentives. And, by the way, have the right information at the fingertips to then take the right care of people in the right place at the right time. That is a really tall order to replicate. The rest of America is neck deep in fee-for-service, and the payers pay a fee-for-service. The providers operate fee-for-service, and all of their business systems and operations are optimized for fee-for-service.

Hanne: Yeah, it sounds like trying to renovate the foundation of a house, almost.

Todd: Right. You’ve just got to honestly roll in with a new house.

Hanne: Right. Build one.

Todd: That’s right. I mean, that’s effectively the magnitude of the challenge. That’s a good way to think about it. And so, the question has been, “How do we, as a country, take those archetypes — take those results — and scale them to much more of the country — to the whole country?” As opposed to it being available and accessible to limited populations of people in certain pockets of the country.

Vijay: The thing Todd really pointed directly to is that we have a sense for what to do, but how do you scale it? You have all this data and all this logistics. It’s just having to make tons of different decisions in complicated ways. This, ironically, seems like something that’s very well suited for tech, something where if you could build the infrastructure to do that, you could take all the little things that have to be done and do them that much more efficiently. Think about something like Amazon. Amazon and Sears, they both sell things. And actually, they both sold things over the internet. But, by taking a tech-first approach all the way through, whether we’re talking about the website, or the back end, or delivery, just everything — that’s the best bet to try to save and improve every little part. Because there’s not going to be a silver bullet that, like — with this one idea or killer algorithm, that suddenly healthcare is solved, or healthcare is easy or cheap. It’s going to be lots, and lots, and lots, and lots, of little things.

Hanne: When you say tech from A to Z, and there’s no one silver bullet, but it’s all little aspects and incremental accelerations or improvements, what exactly do you mean? Where does it look different?

Todd: You have to think about it as honestly reinventing each layer of the stack of American healthcare. You want to have a health insurance plan layer that’s explicitly optimized, from birth, to support value-based payment of care. So, the historical American health insurance company was born in a world, as Vijay said, where they’re paying fee-for-service bills. And so, that’s what they did. As those bills began piling up, escalating speed, their response was, “Okay, I’m going to erect administrative infrastructure that micromanages doctors and patients, and polices what they’re able to do, by making them ask me and give permission.” That’s probably not how they pitched it to people, but that’s effectively what it was.

Hanne: Yeah.

Vijay: Because the one thing they can do is say no.

Todd: Right. So utilization management and pre-authorization, right? And, things that, for all the physicians listening to this, are epithets. That, then, led to doctors erecting their own administrative infrastructure to interact with the payers, to make arguments about what should actually be paid for, which has massively escalated administrative spending on healthcare in America. It also led to escalating mistrust, distrust, between patients and insurers, and doctors and insurers. It has generally not solved the problem. You want a payer that says, “Look, we’re not going to pay a fee-for-service, we’re going to actually pay in a value-based way,” which is a totally different mode of payment.

Hanne: But that feels like it’s about, again, a kind of framework and mindset shift and payment shift. I don’t understand tech’s element in that.

Todd: You cannot do anything I just said without software, and you certainly can’t do it scalably without software. So, first of all, the notion of paying a primary care doctor in a value-based way — you’ve got to be able to, for that patient panel, for that doctor, set the right risk adjusted global budget. You’ve got to actually be able to actually track everything that actually happens. So you’ve given the doctor visibility into what’s happening to their patient base. And you’ve got to provide the doctor with data from all points of the compass — pharmacy data, lab data, electronic medical record data, claims data about what kind of care that patient’s getting from across the system. To put all that at the doctor’s fingertips, to actually make sure this person is healthy. Beyond that, you need to establish a relationship with a member that is incredibly supportive, where you’re also getting the member information about where they stand and what needs to happen, before they even know it.

Hanne: So, there’s like a whole other information flow happening.

Todd: That’s right. Mark Smith, who’s a visionary healthcare leader, has this great analogy. He says, “I go to Harrah’s casino. And before I even know I’m thirsty, there’s someone there with a drink. Just when I’m about to leave the blackjack table, someone says, ‘Can we comp you some free food? Take some chips while you’re at it.’”

Hanne: As if by magic, yeah.

Todd: Exactly, right. And so, he says, in all seriousness, healthcare needs to be like that. So, before you even realize that you’re about to have a problem, someone calls you and says, “Hey, I think you might want to actually get this med. I think I want to see you and check something out.” If the American health system operated like Harrah’s casino, it would save many more lives and cost a lot less. 

This is a classic data and tech problem. I have friends who work in AI in healthcare, and they said, “Look, at this point, the holes in American healthcare are so big you can see them from space.” So, if someone hasn’t refilled their med — they have a chronic illness, they don’t take their meds, they’re going to the hospital. And so, the low hanging fruit in U.S. healthcare is so plentiful — it’s fruit pies on the ground, right, with a cold glass of milk next to it. There’s so much progress we can make if we apply tech-enabled process control and improvement, as has been routine in virtually every other industry, to healthcare. But to do that, you need to actually have a full stack, payer provider ecosystem, where there is a business case for the use of those approaches.

Vijay: All of the innovation we’ve been thinking about has been — let’s say, in therapeutics or how medicine is done — what we’re really talking about is basically how to get better at the existing game. If the game is fee-for-service — how to do lots of services, better services, more services. Almost like if you’re building a machine to do chess, you can do chess really fast, but this is about getting rid of the chessboard entirely and playing a different game. The real point is in giving us the whole stack. That’s really, really hard to do. 

But to your point, I think it is particularly intriguing to ask now, if we’re going to start with a new game, how can we set up the game to have the best chance to win — the best chance to benefit the patients, and to reduce costs? How do you construct that game? How do we make sure that we’re constructing the right game? Because whatever game we do, someone’s going to try to win, and will win at those rules. But that may not be what’s best for patients and may not be what’s best for cost.

Hanne: It’s interesting, because when you’re describing this, I’m thinking about all the ways that healthcare as an industry is particularly challenging to introduce innovation into, because of how unique it is, and certain regulatory hurdles. And I’m also thinking about the move towards more consumer facing, and the market-driven forces that are pushing in that direction. So, I’m going to ask either a really hard or really dumb question. I don’t know which it is, but — how does this value-based shift work with that, kind of, market-driven shift? You know, you see these two big forces. Are they opposing tides? Do they come together? What’s the fit between those two?

Todd: So, in a nutshell, look — if you lower costs and improve outcomes, you can actually put a better health plan product in front of people in a market like the Medicare Advantage market. And more people will buy it.

Hanne: So it’s quality really?

Todd: It’s quality, and it’s cost savings. It’s then the ability to use those cost savings to fund better benefits in your health plan than the competition. And this intersects with consumerization, right. So, in a market where you can offer a health plan that is better for consumers, there’s also a significant positive ROI in investing in and delivering on a world class consumer experience. Why is this? Because if you actually deliver a world class consumer experience, that’s both helpful to you as you seek to win more customers, but it’s also directly helpful to your ability to deliver better outcomes and lower cost. Why? Because if you’re actually working with a member, and they do not trust you, then when you ping them and say, “I think you…”

Hanne: It all breaks down.

Todd: Yeah, the member will say, “Pound sand. I don’t trust you further than I can throw you.”

Hanne: Yeah. So, it’s about strengthening the relationship so that when you need to direct them, it works.

Todd: Exactly. Both when you need to direct them, they will actually listen to you, and B, when they have a problem, they’ll call you.

Hanne: Oh, wow.

Todd: As Atul Gawande says, “The true superpower of a primary care doctor is that people will tell them things that they don’t tell anyone else.”

Hanne: Oh, my gosh.

Todd: So, basically, a patient will call a primary care doctor and say, “I am not feeling so great about X,” right? And that’s an early warning signal to jump on. So, if you are a health plan, and you are genuinely trusted by your members, not only will they listen to you when you call them and suggest something — or ping them and suggest something…

Hanne: They tell you more important stuff.

Todd: They will call and say, “You know, I’m having this issue. I’m not sure what’s going on. Can you help me?”

How to shift towards value-based care

Hanne: Yeah. So, if this shift towards this model depends on gathering new types of data and knitting them into a more holistic picture of the patient for the provider and for the whole healthcare system, what are the types of information that we’re using now that we haven’t been using before? How are we thinking about that whole picture of the patient in a different way? And what is it that we need from the provider lens for this whole model to work?

Todd: Yes. So, building on what we were talking about earlier about the operation of a primary care physician practice, you’re really moving from a paradigm where you’re maximizing throughput to  a paradigm where you are maximizing the outcomes you’re delivering to patients. And so, armed with the right comprehensive data, you want to actually make sure that you are getting members the right care in the right place at the right time in a highly coordinated, proactive way. The role that technology plays in that is not just the assembly of data and the catalyzing of the right action, but also actually enabling virtual care and home care on an epic scale. Because it’s increasingly obvious that the right place for the right care is the home.

Hanne: Right. Not getting you into the ER, not in a hospital.

Todd: Or forcing you to come to the medical office, right. So, going to the early examples of someone who has diabetes, or hypertension, or congestive heart failure, as opposed to you going into the medical office, which is a big logistical exercise. Getting a measurement taken, and then the doctor deciding to put you on this particular med and saying, “Come back in three months.” You come back in three months, they make another adjustment, and then so on and so forth. That takes months or years to get your chronic condition under control. 

Instead, a patient has a continuous glucose monitor that streams data in the software. It has a wireless scale — streaming data into software. It has a wireless blood pressure cuff — streaming data in the software, which then analyzes the data in combination with humans, if necessary — then triggers a set of actions where the care provider can, basically, through a televisit, say, “Okay, I’m going to adjust your med by X,” and then see in 24 hours what the result was. And then adjust it again, then adjust it again, and adjust it again. Within a matter of one to two weeks, literally get that chronic condition under control — vis-a-vis a pattern in the old world, where it would take years, and in a way that’s dramatically more convenient for the patient. You’re basically dramatically — through remote monitoring and virtual visits aided by software, you’re both detecting issues a lot earlier, and you are increasing, by orders of magnitude, the speed of “intervene, see what happens, and adjust.”

Hanne: The whole feedback loop becomes much faster.

Todd: Exactly. That’s one of the many possible examples. But overall, that’s what’s happening. The accessibility of care dramatically improves, the richness and the timeliness of information dramatically improves, the frequency of touch points dramatically improves. And you experience a significant acceleration of improvement of outcomes, and associated with that, lowering of cost.

Hanne: So, I’m thinking about providers listening to this and thinking, “Well, the information itself sounds like a dream, to have all that at your fingertips, to understand all that about your patients. But actually parsing all of that?” How do we make sure that we deliver that to providers in a way that it’s not this giant, hairball mess of more data and information that they then need another administrative layer just to figure out?

Todd: Well, this is a classic problem where software could help — is helping today. What I would say is that the optimal approach to acting on this opportunity is where you’re fusing software and humans together in an optimal combination, such that you could actually deliver the actual result.

Hanne: Right. So, it’s precisely where those two meet that you have to make sure it’s joining well.

Todd: Yeah. You want to design a tech-enabled service, which leverages both software and humans in the right combination with software doing a ton of the work to be able to efficiently, effectively and scalably deliver the actual outcome to the patient.

Hanne: So, if we’re talking about this massive shift from treating sick people — treating [the] chronically ill — to keeping people healthy, healthier. Preventing illness, catching it earlier — how do we begin to shift the whole system conceptually around that? What other types of information do we need to be thinking about? Transportation, food security? Or new types of treatments that we haven’t been thinking about?

Todd: If you have a full “payvider” stack — and by that I mean a payer and provider healthcare stack that is optimized for value-based payment and care. One inevitable additional layer of action you take on is non-clinical drivers of terrible health outcomes. And again, you talk to every primary care doctor in America, they say, “A huge portion of what actually drives terrible health outcomes for my patients are not clinical.” So, social determinants like transportation, food, how your house is equipped, social and emotional support, etc. I’ve talked to primary doctors in Florida who say, you know what the number one thing that you could do to help people with COPD in Florida, and reduce adverse clinical events?

Hanne: No. What?

Todd: Get them air conditioning.

Hanne: No.

Todd: One of the things — it makes a lot of sense for payvider stacks to provide aid to members that includes fixing up your bathroom, with bars and mats that stop you from slipping and break your hip.

Hanne: Oh, my gosh, yeah.

Todd: To be able to actually get you healthy food that is tailored for your chronic condition, and other non-clinical artifacts and services that have a huge impact on ultimate clinical status and health status.

Hanne: It’s so interesting, because while you’re telling that story, I’m thinking about the unseen caregivers of those patients that are doing those things now, and how if the healthcare system can take over that role of the son, or the daughter, whoever it is taking care of that person that needs that extra air conditioner, or needs the handrails set up. It feels like the healthcare system is doing more, but it’s actually making it more cost efficient. It feels like everybody wins then.

Todd: I think you hit upon something very important there. In that, for the average patient, the American healthcare system can seem very forbidding, and confusing, and fragmented. And it’s very hard to understand. It’s the healthcare system that only works when it works, because the individual doctors and nurses in it are by far the best in human history. The problem with our healthcare system isn’t the doctors and the nurses. Far from it. It’s that the system in which these professionals are operating is so disorganized, fragmented, confusing, incentive misaligned, information poor, reactive. The notion of having a professional daughter and son to help you navigate the system as a patient, and be able to get the right care, right place, right time — is, I think, a crucial, crucial role. That’s, again, another role where technology can be extraordinarily helpful in helping that happen. So, a major barrier for a lot of patients is literally transportation. This ability to literally get to your doctor’s appointment, or get to where you’ve got to go as you travel through the healthcare system. And so, a very significant benefit to patients getting the right care, right place, right time is to actually provide them with medical transportation.

Hanne: Right. It’s not just the doctor’s visit, it’s getting to and from the doctor’s visit.

Todd: It’s literally getting to and from the doctor. Key. Really important, completely solvable problem in a tech-enabled way.

Vijay: You know, the thing about social determinants is that I think there’s so many more low lying fruit. It’s not something where people have put much effort. It’s not something that payers have really thought about. 

I think there’s a huge opportunity for just actually applying analytics, data science, tech, to figuring out what these low lying fruit areas are, and what could be done about it. It’s probably not even a hell of a lot of dollars, it’s probably just figuring out what’s the best place and what’s the biggest need, and when is that need. And that combination of who, what, where, when — that’s particularly hard to figure out. If we could provide a coach or a parent, or, like, a doctor that was with you 24/7, that person would probably know, but we can’t know that. So, that is where tech can sort of fill in the gaps.

A new future for healthcare

Hanne: So, just to go back to where we began, and think a little bit about this major new orientation towards, like — it’s a very obvious shift, but it’s a massive shift to make for the system, right? When the entire architecture of the system and the way it’s been developed over the last decades, and the mindsets and the education and the processes — all those things have to be shifted. If you were able to full stack the entire system — if you were able to start from scratch — what would that look like today? What does a real revolution American healthcare look like?

Vijay: I would love to just go right at the question, which is, what would it take to make the American healthcare system the envy of the world?

Todd: I think a way that the U.S. can leapfrog, going from the bottom of the rankings in the developed world to the top, is to make an increasingly strong move toward value-based payment care. See the rise of more and more tech-enabled payvider stacks across the country. Have those tech-enabled payvider stacks, in all their different forms, compete with increasing energy on the basis of outcomes and cost and consumer experience.

Vijay: And fueling innovation that way.

Todd: And fueling massive innovation, right, versus if the entire U.S. health system were actually being run centrally. We just have to get to a place where we create the right magnetic field from an incentive standpoint, by continuing to move strongly toward value-based payment. So that value-based care innovation and value delivery innovation can really blossom through these tech-enabled payvider stacks and all their configurations competing with each other.

Vijay: Well, how can that come to be? I think that makes a lot of sense from, sort of, a bottom up, but how do we get to it from the top down?

Todd: I actually think that in the Medicare space, all the conditions already exist to enable this to happen. It is entirely possible to build tech-enabled payvider stacks that are wired for value-based payment and care with no incremental policy change. One key to that is that in the Medicare space — say the average Medicare Advantage plan keeps a member for eight years. The cycle time between intervening and helping and saving money is short, given the population. And so, that means that if you’re a tech-enabled payvider, or tech-enabled payvider stack in the Medicare space, you have a very strong business case to make investments in better care. And you’ll realize the payoff. One challenge in value-based care oriented payvider stacks in the under-65 segment of the population is that people tend to have their insurance plan for far shorter than eight years. And so…

Hanne: And that’s because you change jobs, you change…

Todd: People change jobs.

Hanne: …situations. Yeah.

Todd: They go from Medicaid, to the exchange, to an employee and back. And the cycle times between intervention and payoff, on average, tend to be longer, because you are dealing with a population that doesn’t have the level of illness burden of seniors. So, I think a really interesting area of policy innovation could be how would you actually solve that problem? Because if you did, then I think that it would then create the right kind of business case support to do the tech-enabled payvider play outside the Medicare space. I think it’s still possible to do, it’s just a lot harder for this reason.

Hanne: And to your point about the importance of the relationship with the patient being such a powerful tool when you only have a short-term relationship.

Todd: That’s right.

Vijay: You know, what we’ve been talking about trying to innovate a complicated system of analytics, decision making, logistics. These are all things that are, in many other areas, well approached by tech, well improved by tech. If you think about even just the A/B testing for websites, or for services, or for anything — that just constantly, sort of, trying new things, experimenting, having the analytics, seeing if it improves, and carry on. Ironically, that’s also not alien to medicine, that’s also an RCT, in a sense. I think we need to innovate, we need to try new things, see what works, see how it changes outcomes, and incent people to do it. Once you realize that you can set up a system where we are incented for innovation, that will automatically bring tech in.

Hanne: A system that’s incented for value, that’s incented for innovation, and that’s incented for actual health before we even get sick.

Thanks so much for joining us on “Bio Eats World.” If you’d like to hear more about all the ways biology is technology please go subscribe to the a16z bio newsletter at, and of course, subscribe to “Bio Eats World” anywhere you listen to podcasts.

  • Todd Park

  • Vijay Pande

    Vijay Pande is a general partner at a16z where he invests in biopharma and healthcare. Prior, he was a distinguished professor at Stanford. He is also the founder of [email protected] Distributed Computing Project.

  • Hanne Winarsky

The Machine that Made the Vaccine

Stephane Bancel, Jorge Conde, and Hanne Winarsky

In this special episode, which originally aired the day the FDA authorized the world’s first mRNA vaccine for emergency use, Moderna CEO Stephane Bancel tells the story of the machine that made the vaccine: the platform, the technology, and the moves behind the vaccine’s development.

This episode of Bio Eats World takes us from a world of pipette and lab benches to a world of industrial robots making medicines: We used to grow our vaccines, now we can “print” them, getting them to patients faster and more efficiently than ever before. In conversation with a16z general partner Jorge Conde and Bio Eats World host Hanne Winarsky, Bancel describes the exact moment he realized they might actually be able to make a vaccine for COVID-19; what happened next to go from pathogen to design; how this new technology that uses mRNA works (in a chocolate mousse metaphor!), and what makes it different from “old” vaccines; and how to think about managing both innovation and speed in this world. Why is this such a fundamental shift in the world of drug development? And where will this technology go next?

Show Notes

  • What happened when the SARS-CoV-2 virus was discovered, and Moderna’s response [1:44]
  • How Moderna’s vaccine was developed digitally [4:02]
  • Description of how mRNA drugs work [5:43] and the history of this technology [7:30]
  • The advantages of mRNA drugs [11:18]
  • How Moderna learned about the positive results from Phase 3 trials [14:41]
  • Why mRNA drugs can be produced more rapidly than other processes [19:06]
  • How mRNA technology might be used for other diseases, and where it is limited [22:18]
  • Details around Moderna’s manufacturing process [27:44]
  • How Moderna was founded, their original goals, and challenges faced [32:01]


Hi, I’m Lauren.

Hanne: And I’m Hanne, and this is “Bio Eats World”, our show where we talk about all the ways biology is technology.

Lauren: This week, in place of “Journal club,” we have a very special episode featuring Stéphane Bancel, the CEO of Moderna, in conversation with you, Hanne, and a16z general partner Jorge Conde. And we’re talking about the COVID-19 vaccine, right?

Hanne: Yep, that’s exactly right. The conversation is a really incredible dive into how they developed one of the world’s most awaited vaccines. Bancel describes everything from the moment he first realized they could make a COVID-19 vaccine with their technology to the day he heard the first data on how effective it was in humans. In this episode, which is airing just after the Vaccines and Related Biological Products Advisory Committee meeting makes its recommendation to the FDA, Stéphane tells the story of not just how the vaccine got made, but everything about the machine behind this vaccine — the fundamentally new platform and mRNA technology behind the vaccine’s development.

Lauren: This vaccine is really one of the first medicines that is part of a bigger transformation from a world of pipettes and lab benches to a world of industrialized machines making medicines. We used to grow our vaccines, but now we can print them — getting them to patients faster and more efficiently than ever.

Hanne: Bancel describes what it took to go from pathogen to design to clinical grade product; how mRNA works, in a chocolate mousse metaphor; and what makes it different from old vaccine technology — why exactly this is such a transformative shift in the world of drug development, and where this technology will go next.

The discovery of SARS-CoV-2

Jorge: Stéphane, sitting here now, in December of 2020, could you imagine a year ago that mRNA as a concept would be a household name?

Stéphane: No, and we have a lot of things going on in vaccines, in cancer, in autoimmune disease, in cardiology, in viral genetic disease. But I had no idea that 2020 was going to look that way.

Jorge: So, if we flash back to January of 2020, can you talk a little bit about the process that you went through to realize that you potentially had the technology that could be a solution for this emerging pandemic?

Stéphane: Yes, so I’ve been working in infectious diseases all my career, and I’ve developed an eye for outbreaks. So, one of the things I do is I read the Wall Street Journal and the Financial Times every morning as I get up. And between Christmas and New Year of last year, I noticed an article saying that there is a new pathogen agent in China giving pneumonia-like symptoms. That’s all it says. And so, I sent an email to somebody working for Tony Fauci — Barney Graham, who we’ve been collaborating with for years designing several vaccines together. 

And I said, “Barney, have you seen the new pathogen in China? What is it? Is it a bacteria or is it a virus?” And he replied to me a few hours later, and he says, “It’s not a bacteria. It seems to be a virus, but we don’t know which one yet.” And a day or two after, Barney sent me an email and said, “Hey, we learned from our contacts in China, it’s not flu, it’s not RSV. We don’t know what it is yet.” And then another day goes by and he says, “It’s a coronavirus, but it’s not SARS and it’s not MERS. It’s a new coronavirus. Within a day or two the sequence should be put online by the Chinese.”

And so on January 11, the Chinese put the sequence online. And our team at Moderna used the sequence to design a vaccine. In parallel, Barney’s team does the same thing. And when they shared notes after around 48 hours, they had designed exactly the same vaccine.

Creating Moderna’s vaccine digitally

Jorge: A couple of things that are fascinating about this — number one, the fact that the digital copy of this virus came from China before the biological version reached our shores. That’s remarkable in and of itself, that we knew what we were dealing with, at least digitally, in a matter of days thanks to all of the advances with genomic sequencing technology. But the other remarkable advancement in technology here is what you just described — you were able to design a vaccine based on the digital version of the virus, also in a matter of days, is it?

Stéphane: So, you’re right, Jorge. And this is the piece that I think most people in pharma don’t appreciate yet — the power of modern technology is, in 48 hours we designed and locked down the entire chemical structure of a vaccine.

Hanne: Unbelievable.

Stéphane: And we click “order” on the computer — so it all happened in silico, we never had access to a physical virus. And we designed the vaccine. And with, again, the two teams at NIH and Moderna, because we were so worried — to make a mistake in the vaccine design, as you can imagine.

Jorge: Of course.

Stéphane: So, we were super happy when the team literally compared notes after two days and they had exactly the same design for a vaccine, because it was an outbreak and we knew every day mattered. At the same time, we started to make clinical grade products to go into a Phase 1. And that is really remarkable — is the vaccine that is reviewed by the FDA on December 17, it’s exactly the same vaccine that our guys designed in January in silico. We never changed one atom. It’s exactly the same molecule.

Hanne: So, it’s the same vaccine that took 48 hours to design.

Stéphane: That’s going to help hundreds of millions of people next year, yeah.

How mRNA vaccines work

Hanne: Can we take a moment to just get really simple and talk about how you would define this messenger RNA technology, and what you wish the public understood about how mRNA works?

Stéphane: Right. Yeah, so it’s a molecule that exists in every one of your cells that is basically the xerox copy of an instruction of your genome for one gene at a time to make protein in your cells. So the way I would describe it to my two young daughters is — think about DNA like the hard drive of life, where all the instructions of all your 22,000-ish genes are stored. And think about it a bit like this is a recipe book that your grandma gave to you before she passed away. All your favorite recipes, that’s the hard drive, that’s DNA. And when you want to make, let’s say, a chocolate mousse, if you go with your grandma’s precious book into the kitchen, you’re going to damage your book a lot. There’s going to be flour, and eggs, and sugar. And after a few times, you might not be able to read the recipe anymore. So what evolution has done, which is beautiful, is to protect the integrity of the instruction in the hard drive, in the book. When the cells want to make one protein, like let’s say insulin, what it does — it makes a copy of the instruction only of insulin in the book, like my example of chocolate mousse — a xerox copy — and takes it into the kitchen (i.e., the cell) to make for a little machine called the ribosome, that I describe [to] my kids as a little 3D printer that reads the message with the instruction of mRNA, and makes the protein by adding one amino acid at a time. So it’s a natural molecule that basically carries genetic information to make proteins.

History of mRNA technology

Hanne: But using mRNA as a tool in the way you’ve been doing it, that was not always an obvious approach. So can you talk a little bit about where that began, that idea, and what it first looked like?

Stéphane: Yeah, so it’s actually very interesting. When mRNA and DNA were discovered, actually people in a lot of universities tried to make medicines out of mRNA because it was a very logical use of mRNA. Just copy nature — make a synthetic mRNA, inject it into animals before humans, and it should make a protein. Because of what was known about science at the time, including immunology — all the analytical tools that did not exist as part of manufacturing purity and so on — when they would inject mRNA in animals, animals would have flu-like symptoms — a fever, vomiting, diarrhea — because mRNA, if you remember, most viruses in life including COVID-19 is made of mRNA. And so through evolution, mammals have developed mechanisms to recognize foreign mRNAs. And of course, when you inject mRNA as an ID for a drug, that’ll be a foreign mRNA. And so actually people abandoned and just quit on trying to make mRNA as a drug. What happened in the 2010, 2011 timeframe, here in Boston, is you had a set of academics at Harvard and MIT who started to play with mRNA again because there [had] been some new discoveries made in the immune system — that they believed at the time that if you modified uridine, which is one of the four letters of mRNA, you can make an mRNA that’s immuno-silent.

Jorge: In some ways, when you think about it, Moderna doesn’t make therapies, right? You make instructions that the cell uses to make its own therapies.

Stéphane: Yeah, correct. We don’t give you the vaccine. We give you an instruction for the cells of healthy people, in that case — to read the instruction, to make one protein of a virus, to make it as well as if they had been infected by the real virus, to show it to the immune system so the immune system can make a neutralizing antibody and mature it. So that if, later, they get infected by the real virus, the immune system is ready to prevent the virus from [replicating] in their body and getting them sick. But what gets people sick in infectious disease is you have too many copies of a virus.

Jorge: Yeah, and so I think this is a remarkable thing for a couple of reasons. What you’re essentially doing is you’ve looked at the virus’s, you know, genome, and you’ve said, “Okay, if I take certain pieces of code from this virus and encode them in mRNA and deliver them to human cells, I am basically giving the human cells the instructions to make pieces of the virus that the immune system will train itself on, recognize, and eventually neutralize.” And in this particular case, that target was the spike protein in the SARS-CoV-2 virus. Is that accurate?

Stéphane: It sounds correct, Jorge. And the reason that mRNA, in my opinion, is so powerful is that you totally mimic to a human cell the natural biology of an infection without giving the virus at any time. We never give a virus to people — we give, as you said, a piece of the virus. In the case of corona, because it’s a pretty simple virus, we believed — and the clinical data have shown in the Phase 3 that we were correct — that one protein of a virus — a very important one, the spike protein — if you were able to get a high quality of a high quantity of neutralizing antibodies, you should be protecting people if they become infected with the real virus.

Advantages of mRNA vaccines

Hanne: Why is it better to have the cell mimic this natural process than in the old technology?

Stéphane: And that’s a piece that is really unique, because when you think about it, when you get an infection by an mRNA virus in your body, what happens? The virus of mRNA gets in your cell, you use your own cell machinery to make the protein — to basically self replicate inside your cell — and then it escapes your cell. And this is what your immune system sees. And so if you think about it, the old technology of vaccines, where you’re making an E. coli cell or CHO cell — a protein that then you inject in a human, and that just circulates in your blood. That is not mimicking the natural biology. 

In our case, the spike protein — we designed the vaccine, so it’s made inside the cell. So in a human cell, not an E. coli cell, and then we design it to be transmembraning — to stay attached to the cell, and to be presented to the immune system that basically backfalls, you know, in your blood, your body. And we see that thing sticking out of a cell — that is not [itself]. If you think about the 3D configuration of a B cell coming onto that protein, it is exactly like if it was a natural infection — which is why if you look at the data across the nine vaccines we put in the clinic, the antibody level is so high because it’s perfectly mimicking nature.

Hanne: How did you know which protein and that one was enough? How did that process work?

Stéphane: That’s a very good question. And, as Mr. Pasteur would say — and, of course, he has a big role in vaccinology — “only with a prepared mind.” So, one of the things we were doing with Dr. Fauci’s team for the last couple of years, is we [were] collaborating on studying viruses that could become outbreaks. None of us thought we [would] see over our lifetime a global pandemic. The last one we were all aware of as students of infectious disease is, of course, the Spanish flu. And so, one of the things where we got lucky is, we had been working for a few years together with Dr. Fauci’s team as part of that project for outbreak readiness on the MERS vaccine — the Middle East Respiratory Syndrome. 

Which, if I had used those words a year ago, nobody would have known what I was talking about, but today everybody knows it’s another coronavirus. We wanted to provide to them mRNA for research grade — so, animal testing, antigen design, picking the protein that makes sense. Because mRNA is so easy to make once you industrialize it. We were able to send to NIH, to the team working on MERS, all the different vaccine designs they wanted to try in animals. They would vaccinate the animals and then they would challenge them by giving them a high-dose of a virus. The one that was most protective was always the spike protein. They tried a lot of combinations, but spike by itself was always the best.

Jorge: And the theory, I assume, is because you’re essentially putting neutralizing antibodies around the spike and the spike is what the virus uses to get into cells in the first place.

Stéphane: Correct. The full-length spike protein was always the best. Some companies went into a clinic with three, four, five candidates. And there were different hypotheses they were testing. We did not have to do that because we had tried it for a couple of years. We knew that with our mRNA, our best guess was going with the full-length spike protein.

Success of Moderna’s vaccine in trial

Jorge: At a very high-level, you are essentially printing these vaccines versus growing versions of a virus or a denatured virus. So you can design it, you can print it, and then you can, you know, obviously get this into people very quickly as a result. That is a remarkable part of this entire story that is probably somewhat underappreciated, that allowed you, and collectively us, to move so quickly. When did you know, Stéphane, that, all right, this is going to work, this is going to work for COVID?

Stéphane: I had a very high belief that this should work since the beginning, so since January. Because this was the 10th vaccine we were working on. So it’s in the human data of a previous one. And in infectious disease — unlike in oncology, where the animal model tells you nothing — the infectious disease, if you look at a lot of data, there is extremely high translation from animals into humans. I saw MERS data before we started, of course, dosing in humans. So I knew the data in MERS looks great. So because we had done nine vaccines before, I knew it was going to look great in humans, which we learned all of this in May.

Jorge: Can you describe, Stéphane, when you first saw the interim Phase 3 data and what your reaction was?

Stéphane: So, it was a Sunday in November. I knew the independent NIH-led Safety Data Monitoring Board was going to meet at 10 a.m. on Sunday. And so I told my wife and my kids, I’m going to be a wreck the whole morning. I tried to pretend to work, but I was so distracted, I would check my email every two minutes, my phone every two minutes for a text message and so on. Maybe a bit before 1, I got a text from my team saying, “Hey, get on WebEx, we’re going to get the data.” There was not even a slide made. It was just somebody talking, literally reading to us the data.

And so I learned about the close to 95% efficacy. It was already a big N and the p-value was very, very low. Very, very low. So this was real. And the piece that was almost the most exciting to me and my team was the severe case of disease, which there were, I think, eight or nine on the interim data. We have now 30 on the final analysis. And there were zero on the vaccine — they were all on placebo. And you think about what this means, when you connect those two data sets together, it means if you get our vaccine, you have a 95% chance of having zero symptoms if you get infected by the virus. You will not even know you are sick, you’ll just go live your normal life, zero symptoms. And in the 5% case, where you will get disease, it will be mild disease. You will get no severe disease. 

And when you think about what has happened to our society — the elderly, people with high comorbidity, from hospitalization, when it gets bad [it] leads to death, and the total impact on the economy, the loss of jobs in so many industries, and so on — that whole cascade. If you could have a vaccine where most people, 95% get no symptoms, and the 5% who do get mild symptoms — never go and walk into a hospital — that will be a total game changer. So I listened to the data, then we talked to my team [for] a few minutes. No, I don’t think we were processing — and then I left my home office and I called my wife, she was in the house. And I told her, and I just started crying in the house.

Hanne: I think that’s what it felt like for all of us hearing it too. It felt like, you know, normal life could return. It was the promise of something like that.

Stéphane: We are losing, right now, 3,000 people in this country — I think it’s more than 10,000 people a day around the world. And it’s going to be a very tough winter. And that’s only the human toll, which is gigantic, but the piece I don’t think is talked about enough is the mental health toll happening to, you know, people at every age. All the young in, especially, you know, more disfavored communities where, you know, people are living in the small apartment, where Mom is trying to work. And kids trying — without a computer, without a good internet line, to learn remotely — the impact this will have in terms of equality. 

And then, of course, so many industries have been totally destroyed. I mean, look, they are closing indoor dining again, which I think is the right thing to do. Because I think the most dangerous thing right now is to have dinner indoors. I have not walked into a restaurant indoors since March, and I won’t go until I’m vaccinated.

Rapid manufacturing process

Jorge: So, as amazing as I think the COVID vaccine story is, I think it’s also worth talking about the machine that made the vaccine — the technology platform that you have built over the course of 10 years that allowed you, in January of 2020, to say like, “Hey, we need to develop a COVID vaccine.” I remember coming to visit Moderna on Kendall Square, that first facility you had. And what was interesting about it is you walked in, it didn’t look like your typical biotech company. It was a row of machines, a row of printers, a row of robots. And that’s very different than what your traditional biotech company looks like. And it looked a lot more like an assembly line, in some ways. Where you can order something up and out the other end would come the mRNA medicine that you had ordered.

Stéphane: Yes, and this goes back to this incredible property of mRNA, which I’m surprised that so many have missed — is that this is a disease and information-carrying molecule that you can industrialize. When you are in an analog business — which is what I think all pharma and all biotech is, in my book — it’s because every molecule is a different chemical entity, you cannot industrialize the making of a lot of it at the research grade. You have to literally have chemists, and pipettes, and so on. You know, doing like we all did in chemistry class, writing the synthetic route to get to a molecule that they want to do the biological effect they want. 

And then they have to design that chemical equation, and then all the pipettes and test tubes to do that. And when it’s another molecule, they have to invent another synthetic route. So it’s very — an analog world where you invent everything once at a time for one product. Because if every product is different, you have to re-optimize every time, and sometimes it’s very complicated because of very complex biological systems. So sometimes it will take you 6 months, 12 months, 18 months to get ready from preclinical data to be making clinical grade product that you need to file to FDA so that they give you the green light to go into testing this in humans. It’s highly regulated — as it should be processed to protect people’s safety. But in our case, it’s always the same thing, because mRNA is always made of [the] four same letters — the four letters of life, like zeros and ones in software. It’s the same manufacturing process. 

This is like software or LEGO, this is an engineering problem. It’s an engineering technology, it’s a platform. The only difference between all Zika vaccines, or all CMV vaccines, and the COVID vaccine — it’s only the order of the letter; the zeroes and ones  of life. The manufacturing process is the same, the equipment is the same, with the same operators. It’s the same thing. And so this is why we could go so fast. It took us 60 days to go from a sequence of a virus presented by the Chinese to dosing a human. The first SARS — SARS-CoV-, or as it was known before SARS — it took the NIH 20 months <Mmhmm.> to go from sequence to starting the Phase 1 study. So, you went from 20 months to 2 months.

Possible future uses for mRNA drugs

Jorge: Which is remarkable. Are we in the plug-and-play future for vaccines?

Stéphane: Oh, 100%. We’re going after making a seasonal flu vaccine — because, as we all know, still 10,000 Americans die every year, on average, of seasonal flu. We believe that we should be able to make a big dent [on] flu. And today we have six vaccines in development. We’re going to have many more soon, because for 10 years, you know, Jorge, we hoped that mRNA vaccines were going to work. We believed scientifically they were going to work — but until you have a Phase 3, randomized, placebo-controlled study where you test for the prevention of disease, you don’t know. Now we know.

Hanne: Are there limits right now to how sophisticated these instructions can get, or can we essentially give them as sophisticated instructions as the human body is capable of?

Stéphane: It’s — when the mechanism of a disease is not well understood. So we spoke about vaccines, and we said, “Look, coronavirus,” as I said, “is actually a simple virus.” We, as a society, got lucky. Think about HIV. HIV [was] discovered 40 years ago. There is still to this day no approved vaccine against HIV. Think about the awful world we would be in right now if Dr. Fauci had been standing on the presidential podium back in spring, and told them, “Folks, I’m sorry to tell you, but this is an awfully complex virus. We have no idea when we might have a vaccine.” Think about the state of mind we would all be in now. The biology of viral genetic disease is very well understood. Why? Because kids got two biogenetic information from their parents that they cannot make a correct protein, and that is what causes their disease. They have a wrong instruction in their DNA. 

You can give them an mRNA from our technology, coming in their cells with the right instructions — then they will have the right protein and they won’t get sick. If you think about cancer, on the other hand of the spectrum, or Alzheimer’s now, if the disease mechanism is not understood, we cannot drug it easily. We can try things, of course. We could make an mRNA behind that hypothesis, go try it in a clinic — but a lot of things will fail because we are guessing. And so the piece where I think we have an incredible tailwind — basically overlaps doing academic biology work around the world are helping us. Because if tomorrow there is a paper published by our lab in the U.S., or in China, or anywhere in the world that says protein X, Y, Z is the root cause of that disease, or those five proteins in this ratio are the root cause of that disease, then we can literally turn on the computer and, you know, design a drug to go test that hypothesis in an animal.

Jorge: Basically, the power of this approach works when you know what you want to make and then you just need to deliver the instructions to make that. Where it doesn’t work as well is when you’re not quite sure what it is that you need to make.

Stéphane: This is basically biology complexity or biology risk. The other dimension for us is the ability to deliver the mRNA in the right cell. We actually have become a “delivery of nucleic acid” company. We realized that what would allow us to maximize the impact we could have on disease, or helping as many people as we can over the next 5, 10, 20 years, is the ability to bring up mRNA to different cell types. So a good example today is, if you say, look, there is this university that published the mechanism of Alzheimer’s disease. If it happens in the brain and we don’t know how to bring mRNA [to] the brain safely, we cannot drug it. So the biology will be understood, but the delivery technology will not be there.

An example where we’re making a lot of progress right now is the lung. <Mmhmm.> We have been working with Vertex around how to deliver mRNA via an aerosol via your mouth into your lung, because they know the biology very well. And we work together to develop a delivery system to bring mRNA safely into your lungs, and to bring enough mRNA at a safe dose to get the biological effect. And we’re getting very close now. Once we can prove in the clinic that that delivery system works, then the next morning you can make any other drug you want that you need to get into the lung, because it’s getting another set of zeros and ones coded differently, with the same delivery system into the lung. And that’s the power of the technology — which is why with vaccines we’re able to go so fast.

Jorge: Yeah, the instructions have gotten so sophisticated over time that now the next sort of horizon is, you’ve got to get the vehicle for delivery equally sophisticated.

Stéphane: We’re adding vertical, after vertical, after vertical — then we bring mRNA into a new cell type. So the vaccine is one vertical. Getting mRNA into a tumor is another vertical. We have a very cool drug, where we inject mRNA in people’s hearts after a heart attack — and here we code for a protein called VEGF, for the biology geeks on the podcast, V-E-G-F. That is a protein that we all have the instruction in our DNA, which basically tells your body to make a new blood vessel.

Hanne: Amazing.

Stéphane: You use that protein every time you cut yourself.

Using robotics to manufacture drugs

Hanne: Stéphane, you’ve mentioned, you know, kind of the fast design of the vaccine, and then you mentioned even robots printing medicines. Can we get your version of what that machine assembly line looks like?

Stéphane: So, the robotics farm we have in our factory is basically just an assembly of robots that get instruction coming directly from computers. There’s no human interaction. And basically, you start from a piece of DNA. That is basically your template. You put that in a reactor with water. There is no cell — it’s a cell-free manufacturing process, which is why it’s so fast. And you put enzymes. And basically, what the enzymes do, they attach to the DNA, and they read the DNA template. And they quickly tell pieces of nucleic acid — i.e, the zeros and ones, the four letters of life — they bind them next to each other to make an mRNA molecule. Then the robot goes to the next step, which is you add a cap. 

Think about it like the nose of a molecule, that you add again with another enzyme. Then what you do, you purify the mRNA. So basically, you pick the mRNA from all that water, enzyme, and nucleotide, nucleic acid, and so on. And then when you have a pure mRNA molecule, after purification, you mix it with a lipid, i.e, fat. And that fat basically goes around and packages, like in a little bowl, the mRNA to protect the mRNA in your blood, and to get the mRNA inside your cells. When it’s inside your cells, the lipid — the fat — falls apart, the mRNA is released inside the cell, and the little ribosome — the little 3D printer of your cell — is going to read that message, make the protein on demand, and here you go — the patient, the human is making his or her own medicine.

Jorge: I remember from the earliest days you were obsessed with the operations. You were obsessed with turnaround time, with throughput, with, you know, cost per output. And the benefit of that approach is that it obviously just compounded over time. The benefit of the technology, as you’re describing it, is that you have a machine that prints the instructions that go into the cell — that uses the cell’s machine to make the medicine, or to make the vaccine. And that’s this incredibly powerful paradigm, you know, to taking therapeutics or vaccines from being very bespoke efforts to being truly industrialized, designed efforts.

Stéphane: That’s what is really so powerful is that the whole drug process is all about information. The piece that is remarkable is you have this very modular technology, because what happens in our cells is actually extremely logical. We start from the sequence information of a virus, like in the case of a COVID vaccine, or we use the human genome. We put [it] into a technology genetic-based cassette, and then you click “order” on the computer and you go again. And that’s the vision I always had since day one. And a lot even of my scientists thought I was crazy, because this industrialized, engineer-driven approach to drug discovery has never happened [before].

Hanne: So, Stéphane, you’ve described this process which is, you know, much more efficient, industrialized in nature, incredibly fast compared to the old process. Is there a world in which that gets even faster? Are there other things — you know, other increases in technology that would speed this up even more?

Stéphane: Yes, so it took us 42 days to go from sequence to shipping the human grade vials to Dr. Fauci’s team. The big bottleneck is sterility testing — a very important quality control test that is done for any injectable pharmaceutical to make sure that there’s no bacteria in the product. That test takes two weeks, because what basically you do, you take a sample of your vaccine and you wait enough time. If there’s only one copy of a bacteria, by that time you have enough multiplication of bacteria, through the detection of the assay of a test that you will see it, you will not miss it. It’s very important for people’s safety. Well, if there was a technology developed where you could do sterility testing in one day with high sensitivity, then you could take our process down to two weeks.

The history of Moderna and its approach

Jorge: So, we’ve talked about the vaccine. We’ve talked about the machine that made the vaccine. I’d love to take a second to talk about the company that built the machine. So from the moment that you started this company, you took a very different approach. And you’ve described it as having an engineer’s mindset. Can you talk a little bit about what you did, and how you thought about the early company build?

Stéphane: I had never built a company in hypergrowth. You know, I worked at Eli Lilly, I ran bioMérieux, which is a big diagnostic company. But I have never built myself a company building very, very quickly. We decided to do something very atypical, because most biotech companies are one-drug company at a time. What was very clear to us, because mRNA is an information molecule, is it made no scientific sense that this will be a one-drug company. It will be either zero, because we run out of money before we can safely get the drug approved, or it’ll be a company with thousands and thousands and thousands of drugs because of the platform.

And so, once we realized that, in the first hours of talking about Moderna, we started to become very worried and paranoid about, “Geez, we don’t know what we don’t know about this technology because it’s new. It has never been approved.” And, “Geez, if we pick one drug, if we are wrong and it doesn’t work in the clinic, everybody will believe what people have believed for 50-plus years” — which is, mRNA will never be a drug. And we most probably are going to go bankrupt. But if mRNA could’ve worked, we will have failed society. Because if we find a way to make this work, this will [mean] thousands and thousands of drugs that are undoable using existing technology — like the VEGF in hearts — and we will shortchange societies, shortchange patients. And that was just unbearable.

And so we spent a lot of time thinking about, okay, what are all the things that could make us fail? We ended up zooming [in] on four risks that we say — if we can manage and reduce those risks, we will have [the]  best chance to be the best version of Moderna. Those risks — we’ve talked about very publicly, especially when we went public. It’s technology risk around the mRNA technology. So, of course, if you do a new technology you don’t know what you don’t know. There’s going to be a lot of risk there of things not working as you expect. Two is the biology risk. You can have incredible risk that your scientific hypothesis on the biology is incorrect and the drug will fail — not because the technology wasn’t working, but because the scientific hypothesis on the biology is incorrect. Then there was going to be a lot of execution risk. And then, of course, financing risk. Because we said, like, you know, asset managers build a portfolio — we said “Picking one drug is crazy, it’s like buying only one stock.” And so we said, “Let’s build a full portfolio of drugs.”

And after many, many months of discussion, we designed basically a pipeline of 20 drugs that we said we’re going to take all those drugs in parallel to the clinic, so there would not be a binary event that the company makes it or not on one drug. So we diversified the technology risk around six different technology applications, from vaccines, to [a] drug in the heart, to a drug in the liver for a genetic disease. And then for every application, we took several drugs to diversify the biology risk. And we launched that crazy experiment with, you know, 17 drugs in the clinic so far — which was going to create incredible execution risk because it’s harder to do 17 at the same time than 1. And incredible financing risk because we needed a lot of capital. But we traded those risks with our eyes wide open, because the other risk could kill us with much higher probability — the technology and the biology risk.

Jorge: It’s very difficult in this industry to take that balance, platform versus programs. And, you know, what tends to be the case very quickly is most companies when they have to choose where to put an incremental dollar, or an incremental head, they put it on the programs because those are the golden eggs and they want to move those forwards to create value inflection. And as a result, the platform ends up getting starved. <Yes.> You started the other way around. You actually fed the platform, and you fed the goose, and then let the goose lay its eggs.

Stéphane: Yeah, exactly. The goose is more valuable than any egg. If you really believe you have a goose that’s going to be making thousands and thousands and thousands of eggs, you don’t want to kill the goose on the first or second egg.

Jorge: Although most geese are not that fertile in biotech. Yours… <crosstalk, laughter>

Stéphane: And that’s why I told you both that I was not interested [in going] public early because the capital markets were going to force me to not invest in the goose. Because biotech firms like to bet on eggs, not on [the] goose, because there has not been a lot of geese before in this industry. So we’re not used to it.

Jorge: I mean the record will show that you did a lot of things right. As you built the company over the last 10 years, can you talk a little bit about the things you did wrong, that if you could get them back you would do it over?

Stéphane: Well, [the] easiest one, given the COVID situation is — it took us three years to start working on vaccines. So think about how the world would be different and Moderna would be different if we started working on vaccines from day one. We might have been able to go even faster for COVID. So that’s a thing I regret, and that’s on me. I made quite a lot of mistakes hiring people, because I underestimated how intense our company is because I live it every day. I thought initially that it was obvious that this is a small company fighting for its life, so people are going to work hard. It’s brand-new, cutting-edge science, so it’s going to be complicated because every other thing is not going to work. 

So, being able to manage uncertainty — people having a lot of grit. Collaboration, because making a drug is a team sport. A drug is a system of so many capabilities — the biologists, the [toxicology] people, the chemists, the engineers to make the drug. And a lot of times, people coming from big pharma are used to working in silos. And people who come from academia don’t know how to develop drugs. It’s a system. And like any system, you get the best outcome if you really optimize the system working together.

Jorge: So, last question I would ask you — what advice would you give to the engineer that wants to get into biotech?

Stéphane: So, first he needs to learn a bit about biology. I mean, I had a chance, as I spent my entire career in biology, so I’ve learned a lot on the go — I’ve learned a lot by reading. I’m a curious guy, so I read a lot. You can get biology books and learn. And I think it’s understanding enough of biology so that you can be part of a conversation, so that you can have an impact on decisions and scientific choices that happen. And then you can go from there.

Hanne: That’s wonderful. Thank you so much for joining us on “Bio Eats World,” Stéphane. We’re so grateful for your time.

Thanks so much for joining us on “Bio Eats World.” If you’d like to hear more about all the ways biology is technology, please go subscribe to the a16z bio newsletter at, and of course, subscribe to “Bio Eats World” anywhere you listen to podcasts.

  • Stephane Bancel

  • Jorge Conde

    Jorge Conde is a general partner at Andreessen Horowitz where he invests in companies at the cross-section of biology, computer science, engineering. Before a16z bio, he was CSO at Syros, cofounded Knome, & more.

  • Hanne Winarsky

The Cost Disease in Healthcare

Marc Andreessen and Vijay Pande

How come things like healthcare, education, and housing get more and more expensive, but things like socks, shoes, and electronics all get cheaper and cheaper? In this episode of Bio Eats World, a16z founder and internet pioneer Marc Andreesen, and general partner Vijay Pande, discuss the lesser known law of economics that explains why healthcare, education and housing is so damn expensive, and getting worse.

What’s really at heart is tech’s ability to transform (expensive) services into (affordable) goods: think of the cost of a live string quartet, versus a streamed recorded track; or the cost of a custom-made shoe, versus a factory-made one. Until now, using tech to similarly transform services into goods in healthcare has seemed like an impossible dream — how would you do this for, say, the service of doctors providing care? But in this wide ranging conversation all about technology and society across all industries, Andreessen and Pande talk about the massive new gains recent technologies have begun to make this seem within reach, from eye surgery in malls to using AI in processing medical claims. Is there a future in which what doctors are doing today feels analogous to farmers hand plowing fields 300 years ago? And what would the role of that doctor of the future be?

Show Notes

  • How Baumol’s cost disease is distorting pricing in healthcare and education [1:56]
  • How technology could reduce healthcare costs [5:52], just as it has in the past with other goods and services [10:41], and LASIK as an example of market-driven healthcare [13:33]
  • The role of individual behavior in chronic health conditions [17:20], and ideas for how this can be managed [19:56]
  • Using apps and wearable devices to drive positive behavior change [25:15]


Lauren: Hi, I’m Lauren.

Hanne: And I’m Hanne. And this is our show, “Bio Eats World,” where we talk about all the ways our ability to engineer biology and re-engineer healthcare are transforming our future.

Lauren: So, Hanne, this episode is called “The Cost Disease in Healthcare.” What disease are we talking about?

Hanne: It’s actually a reference to what’s called Baumol’s cost disease, or the Baumol effect, which is a phenomenon first described by an economist named William Baumol in the 1960s. In short, the Baumol effect is when there’s a rise in wages and jobs and industries that then haven’t had the same rise in productivity.

Lauren: Okay. But what does that really mean, and what does it have to do with healthcare?

Hanne: That’s exactly what this episode is about. a16z founder and internet pioneer Marc Andreessen and General Partner Vijay Pande discuss the economic forces that make some things like healthcare, education, and housing get more and more expensive but things like socks, shoes, and electronics all get cheaper and cheaper.

In this wide-ranging conversation about how society and different industries work and what that means for consumers, Marc and Vijay talk about tech’s ability to transform expensive services into affordable goods. Think of the cost of a live string quartet versus a streamed recorded track, or the cost of a custom-made shoe versus a factory-made one. But until now, using tech to similarly transform services into goods in healthcare has seemed like an impossible dream.

How would you do this for, say, the service of doctors providing care? Marc and Vijay talk about the massive gains in new recent technologies that have begun to make this seem within reach, from laser eye surgery in storefronts and malls to using AI in processing medical claims.

Is there a future in which what doctors are doing today feels analogous to farmers hand plowing fields 300 years ago, and what would the role of that doctor of the future look like? Take it away, Marc and Vijay.

Baulmol’s cost disease

Vijay: So maybe, you know, the place to kick this off would be to talk about what is Baumol’s cost disease and why it’s so important. I think maybe the Twitter version of it is, how come things like healthcare and education and construction exponentially increase in cost, whereas socks from Walmart or many other goods, especially anything electronic, decreases exponentially. How could that be? I mean, we’re living in this world where things magically get cheaper, but that college education or healthcare just is becoming this massive challenge for us as a nation.

Marc: Yeah. So, the way that you measure the impact of technological change in society is through what economists refer to as productivity growth. It’s the process of figuring out how to make more output with less input, right. And so, normally, we kind of expect the world to work this way, which is, every year over the last 300 years with a couple of exceptions, most industries got a little bit better at making things and costing a little bit less, and that led to this huge rise in living standards.

Agriculture is kind of the classic case where food is really cheap as a consequence of a lot of technological leverage applied to the challenge of growing food, and we just generate a lot more calories of food for a lot less money now than we did 100 years ago or 300 years ago.

The problem or the challenge is that different sectors of the economy have different rates of productivity growth, basically, depending on their idiosyncrasies and then depending on the extent to which technology is empowered or allowed to actually have its effect on things.

And so you see these industries like consumer electronics and media and food and clothing in which you’ve got this spectacular productivity growth and then correspondingly these spectacular price declines over time. And then you have these other economic sectors — education and healthcare and housing as three in which the price curves are going in the wrong direction.

The cost of a college degree gets more expensive every year. The cost of heart surgery gets more expensive every year, which is going backward, basically neutral, and maybe even negative productivity growth. Like, they might be getting worse over time.

Then you basically got this problem where you’ve got certain industries that are racing ahead in productivity growth, and so, those workers are kind of becoming super technologically empowered to produce a lot more with less input. And so, those workers are actually getting paid a lot more because they just got so much technological leverage to the work that they do.

Think about, for example, the producer of a TV show or something like the level of kind of power that you have with a modern computer to like produce a TV show. It’s leaps and bounds beyond what you would’ve had if you were literally cutting, you know, splicing film, you know, by hand with the scissors and tape, which is how things used to work.

And so you’ve got these industries in which productivity is growing very fast, prices are declining, and wages are exploding. And then you have these other industries like education and housing and healthcare in which that’s not happening, but the problem is workers can actually migrate from sector to sector. If I’m not excited enough about having a job as, like, I don’t know, a film editor or something, like, I can go to nursing school and I can become a nurse. And now I’ve migrated out of the media industry, and I’ve migrated into the nursing sector.

And then the problem is wages get set across these industries, and so you basically have industries with neutral and negative productivity growth that are now setting wages as if they had positive productivity growth, which they don’t. And then the result of that is just this explosion of price in those kinds of negative productivity sectors. It’s just horrible for consumers of healthcare, education, and housing because the same stuff gets basically more expensive every year without getting any better and maybe getting worse.

Oh, and then, the other big problem is, there’s no reason why this would ever stop. The way I would crystalize this whole thing is — because of rapid productivity growth in consumer electronics, a 100-inch big-screen TV that goes on your wall, in your house, and you can watch every movie ever made for $10 a month. The price of that TV is going to drop to, like, $100. That is, like, quite literally what’s happening.

Correspondingly, the price of a high-end private four-year university degree has leaped up dramatically over the last 20 years. It’s now in the order of, you know, $75,000 a year. So it’s like $300,000 for a degree. It’s not going to be that long before a four-year college degree costs a million dollars.

And so you’re going to have a $100 television set that covers your whole wall, and you’re going to spend a million dollars getting a college degree. And that’s just crazy. It’s just, like, such a horribly bad outcome, and yet there’s something in the structure of how these markets work that prevents us from kind of speaking openly about the trends that result in this.

Technology and AI in healthcare

Vijay: Yeah. I think, Marc, one key point that you laid out there was that this is very much the cost of labor and that there’s a sort of specialized labor. And in many of these industries that we’re talking about — healthcare, education — this is an apprenticeship, where you have to spend many years to be able to develop skills that are handed down from one person to another. Very, kind of, pre-Industrial Revolution kind of behavior. Whereas when you talk about goods, goods are made in factories that are completely automated, and that technology can be applied there to make them 10% better a year, and that leads to exponential performance over time.

And one of the key ideas that I was curious to hash out with you is what we’re seeing in AI. What we’re starting to see is that AI is turning what used to be something that had to be done by a service into something that can be thought of as a good. That instead of a person training in an apprenticeship-like way to do something, the machine learns these things. You can make copies trivially. You can get now the advantage of Moore’s Law, and this almost alchemical magical transformation seems to be one part of a potential solution to addressing Baumol’s cost disease. I’m curious how you see, at least, that part of it?

Marc: Yeah. So this actually goes to an example that Baumol used when he wrote the book, kind of on this topic. And so he used the hypothetical example of, like, a string quartet, right. There’s two ways to experience a string quartet, right, in your house. One way is the old way, which you can actually hire four musicians to come and set up in your house and play Beethoven quartets, and it’s going to sound great. And by the way, people still do this. You do this for, like, weddings, right, this is still a thing.

The other way to experience a string quartet in your house is electronic playback, a recording. And what’s the cost of a string quartet recording to playback in your house today? You know, it’s basically zero. If you just chart the price of getting four musicians to come play at your house from, like, 200 years ago to 100 years ago, to 50 years ago, to 20 years ago, to now, that price has exploded. The in-person version has gotten, like, wildly more expensive, right, because of Baumol’s cost disease, because those musicians actually work for a living, and they have other career options.

And then, exactly to your point, in AI, what’s the simplest form of AI? It’s a computer literally listening to what’s happening and playing it back. And it turns out that costs nothing. An enormous amount of progress in the modern economy is that, right. You also benefit from that, by the way, every time you buy a loaf of bread. Our ancestors were not buying loaves of bread carefully pre-sliced off the shelf. Our ancestors, to the extent that they were able to get access to the core ingredients in the first place, were, you know, making bread by hand.

Vijay: Yeah. Baker as a service.

Marc: Yeah. Exactly. This is actually the big lever that increases living standards. Exactly to your point also, it has been hamstrung by the fact that historically computers have only been able to do so much. Machines have only been able to do so much. And now we have these sort of much more flexible technologies kind of gathered under this term of AI that at least in theory give us the opportunity to now revisit a lot of our assumptions about what should be a product and what should be a service.

Vijay: Yeah. One of the fun things that we’re seeing is AI is nibbling in with the easy, mundane things that are annoying for people to do that they have to be trained to do, but then, that training that goes to people can now be done to the AI, and the AI can be trained once and then scale and actually learn from everyone else’s mistakes.

And so what we’re seeing, as initial go-to-markets in healthcare are in areas of billing or simple types of diagnoses or triage — areas where this isn’t trying to make some superhuman genius, which may in time come, but I think the first go-to-market is just taking the things that are just boring and reproducible that are just expensive because of the human power involved, not even necessarily because you need a super genius. And that’s something that we’re seeing right now.

Marc: There’s a famous story of Alan Turing where he was working on inventing the computer in the early 40s, and he was hanging out at Bell Labs in New Jersey with his friend Claude Shannon who’s the inventor of information theory. The two of them were having this heated lunch discussion at the AT&T headquarters building in New Jersey with all the top researchers and AT&T executives kind of sitting around nearby about basically this concept of AI. Like, what would it mean for computers to actually be intelligent to actually have brains.

And they were debating back and forth, and finally, Turing got frustrated, and he stood up and yelled at Claude Shannon. He said, “Look. I’m not talking about turning a computer into a super genius. I’m just talking about building a mediocre mind, like the president of AT&T.” And this gets into the emotions and the politics of how we think about automation, because the technological progress and productivity growth changes jobs, but in the fullness of time, what we will realize is that a lot of what doctors are doing today, for example — a lot of that work is going to be analogous to literally when farmers used to hand plow fields 300 years ago.

Like, if you took a modern farmer who’s running a fully computerized operation with all these modern combines and tractors and GPS and all these amazing hybrid engineered seeds and all these miracle fertilizers and everything, and if you told them that they had to go back to hand plowing fields, we would have much worse food at far, far higher prices and a lot of people would go hungry.

I am quite convinced the doctors in, you know, whatever, 50, 100, 200 years are going to look back at what doctors do today, and they’re just going to be, “My God. I can’t believe those poor people ever had to do all that.” And in fact, they’re also going to say like there was so much more important work to do.

Vijay: Yep. You know, it’s interesting to think about what the arguments against this could be, and one would be that — well, you know, could the industrialized process be really comparable to what a human being can do? People can do so many things. I was just thinking about how shoes were made. You would have a cobbler who could make shoes that were perfectly suited for your feet, and they’d be doing — everything would be one-off and bespoke and probably would be better shoes, maybe. But instead, you just define a series of shoe sizes — you know, I’m either like a 9, 9.5 or a 10 or whatever, and I could just get the closest one, and it’s good enough.

And the fact that it’s 10 times cheaper or whatever, and now with non-material so much better, that pretty soon you forget about the other experience, and you just get used to the new way of doing things. And that’s kind of my suspicion, that in the beginning there will be trade-offs that you have to make, and that people will have to get used to, but that in time I think you wouldn’t think of doing it any other way. And at least this follows industrialization in other contexts.

Marc: Yeah. And in fact, back “in the glory days” when like all shoes were made by hand, they were sold, like, [so] crazily expensive that you would have one pair of shoes, right. This idea that you’d have like a shoe closet would’ve struck people as just absurd because you have a pair of shoes. And then, by the way, they’re so expensive because of all the manual labor involved, right, relative to your ability to make money, you know, as sort of a normal worker that like if your shoes start to wear through the sole, you’re out of luck. You’re probably going to be wearing those things for five years.

Kids wearing, like, newspapers stuffed in their shoes, right, to be able to basically compensate for the holes in the shoes because shoes are just a lot more expensive to replace. Just imagine that shoes cost 1/6 of all GDP, right, which is where we’re at with healthcare, right. And so imagine if it was like 1/6 of all economic output had to be used to pay for shoes, and then it turns out nobody wanted to pay for anybody else’s shoes — and how terrible that world would be. And how that would really screw up, you know — we’d have all these crazily intense, like, political debates. We would’ve had these political debates between Trump and Biden on the national shoe policy.

Vijay: Oh, yeah. Yeah. Obama shoe, Trump shoe.

Marc: Yeah. Exactly. And then, you know, government-made shoes, right, getting these things out of the realm where you have to have these debates because things are like gigantic, expensive and nobody wants to pay for them is itself just a massive [inaudible] function increase in human welfare that you don’t notice it until you don’t have it.

Vijay: Well, that’s why I think it’s maybe not as much of a surprise why it’s showing up in healthcare because healthcare will eventually become 100% of GDP.

Marc: Right.

Vijay: So it’s something that’s not sustainable, this exponential growth in costs. So I think entrepreneurs are seeing that potential. They’re creating this in both front office for doing scheduling, for doing diagnosis, for doing back-office, for billing — all the sort of routine and horrible things that people hate. But I’m curious, let’s just posit that the technology will continue to advance and that AI will get a foothold and will do something and then eventually more and eventually more. I’m curious, Marc, if you think, is that it? Is that enough where AI is doing some large fraction of the work to really shift this cost curve, or is there more than just a technology that’s required as part of the solution?

LASIK as a case study

Marc: We actually do have a clear example of this happening in the area of medical treatment, and that example is laser eye surgery, right. Basically, LASIK — laser eye surgery which basically literally will fix your eyes, so you don’t need glasses anymore — is the kind of medical procedure that if you described it to somebody from 1950, they’d think you’d lost your mind. It’s literally beaming lasers onto the surface of somebody’s eye to change the shape of the eye.

Vijay: In your mall.

Marc: Yeah. In a mall, right, quite possibly right there in the front window. Right?

Vijay: Yeah.

Marc: And so, it’s an amazingly technologically advanced procedure. It’s actually gotten even more technologically advanced over the last 20 or 25 years. There was a point when you had to, like, try to hold really still because the laser needs to hit the right part of the eye. And now, they’ve got all this advanced 3-D cartographic mapping where the laser follows your eye movement in real-time. And so it’s become this incredibly technically sophisticated kind of thing. And while that’s been happening, the price has been dropping.

And in fact, the reason why LASIK outlets are in the mall is because they can afford to be, right? It’s actually become quite inexpensive to set up a LASIK operation, and it’s actually quite inexpensive to get LASIK. This is the kind of thing where it’s like, this procedure as a technological feat is not more advanced than heart surgery. It’s not more advanced than certain forms of even, I would speculate, brain surgery. This is advanced stuff, and yet this thing is on a quality improvement and cost reduction curve completely unlike any other surgical procedure.

And then you kind of say, well why is that? And, of course, the reason is because it’s paid out-of-pocket, right. So it doesn’t run through the insurance system. It’s not something that other people pay for. It’s not something that has any politics around it. It’s an outpatient procedure. It’s voluntary. And if you don’t get it, by the way, then you get glasses. And if you do get it, then you don’t need glasses.

And so, as a consumer, you can actually make the trade-off of, like, is it worth to spend whatever — $1,500 for this surgery, as opposed to spending, you know, $200 for glasses every few years. What if we could basically re-engineer our whole approach to how we think about all this stuff? And, you know, we can’t literally do that, because consumers might be in a position to decide whether they need eye surgery and how that should work. Maybe they’re not quite in the same position to understand what it means to have a quadruple bypass. And then there’s also, like, it’s an outpatient procedure. Inpatient procedures are a lot more expensive, have, you know, lots of care requirements. But nevertheless, it’s like, “Okay. There’s a shining beacon for what’s possible.” So there’s that.

There’s also this big definitional question in my mind which is, like, what is healthcare? And we tend to think that healthcare is like a discrete thing and the politics are kind of all calibrated around that. And so the big political arguments are always about, do you have healthcare, or do you not have healthcare — as if you’re saying, like, I don’t know, do you have a shirt or do you not have a shirt, right? But that’s not really what it is. The definition of what it means to have healthcare keeps expanding, right, as sort of the number of things that people consider to be conditions that they want treated and the number of things that are actually treatable keep expanding.

And then there’s this whole other debate of inputs versus outputs, which is, how are we measuring healthcare? Are we measuring it by how much it costs and all of the things that go into it and all the procedures, or are we measuring it in terms of outcomes and literally things like health and longevity, right, and sort of physical vitality? And you really start to have different views on basically what it is we’re all paying for, what value we’re getting for it, and then, by extension, what shape and form healthcare will have in the future where it could end up being very radically different.

I’ll just give a thumbnail sketch for how healthcare can be radically different in the future. It may be that all the medical procedures, surgical procedures get basically automated and become very cheap, but it may be that we end up spending more healthcare than ever because healthcare basically turns into advanced therapy. And so instead, like I said, it may turn out to be the physical issues are the easy and cheap ones to deal with, and it may turn out that it’s the psychological and sociological issues are the complex and painful ones to deal with.

And so, maybe the job in the future of “doctor” and “therapist” merges, and we end up with this very different type of profession that’s really oriented around helping people optimize their entire life. And then it’s like, “Oh yeah. Every now and then, you need to go get a little laser surgery, but that’s, like, not the major part of the spending.” And then as a consequence, like, maybe doctors, you know, 20 or 50 or 100 years from now are paid a lot more, because they’re actually a lot more valued in our lives despite the fact that so much of the work that they do today has been automated.

The sheepskin effect

Vijay: Yeah. That they become the focal point for all that automation and keeping the human element. And your point about inputs and outputs, I think, is super important, because if you compare it to other areas where Baumol’s cost disease exists, like education — that also seems to be very much measured more by inputs and outputs. People ask, “Oh, did this school get as much money per student than that school?” Not, “How well did the students do?”

Marc: So, the crack in the matrix that makes me really wonder about education as a service, as a product, or whatever, is something called the sheepskin effect. And so, basically you assume that, you know, you go to school for eight semesters, you know, four years. You come out the other end, you get a job, and let’s say the job pays you whatever — you know, $80,000. So then, apply the following thought experiment, which is, what happens to that income once you’re out of college? What happens to your rate of income if you only complete seven out of the eight semesters?

Logically you would think, “Well, if the value of the education is all the stuff that they’re teaching me, then I’m going to get 7/8ths of the wage, and I’m going to be making $70,000 instead of $80,000,” right, or whatever that correction is. Of course, that’s not what happens. What actually happens is if you only complete seven semesters out of eight, you’re going to get paid $40,000, right, because you’re going to be a college dropout instead of a college graduate.

Vijay: And you get paid basically what you would’ve if you didn’t go to college.

Marc: If you didn’t go to college. Exactly right. And so that’s the sheepskin effect. There’s two possible explanations for the sheepskin effect. One is, all the actual skills are taught to you in that last semester. That’s one possibility, but we know that’s not true. And so the other explanation is, college actually does not have that much to do with the skills that are being taught. It’s something else. It’s basically a stamp of approval that says you can execute a task all the way to completion. The education may be somewhat beside the point. It may just simply be the fact you demonstrated you can get through a program.

The healthcare crack in the matrix to me is the fact that it used to be the medical conditions that mattered were things that just happened to us that we had no control over — or you’d be in a factory, and your arm would get cut off, or you would just die, and you had no control over it. So many of the medical conditions that we’re dealing with today as individuals and societally are as a consequence or downstream in behavior.

Vijay: Yes.

Marc: Obesity is the big one, right. It’s like obesity is cross-linked to all these issues, right, including heart disease and stroke and cancers and, like, everything.

Vijay: Massive comorbidities all over the place.

Marc: All over the place. And so the most effective form of healthcare is don’t eat bad foods and then exercise every day, right. And then, if you’re going to drink like only drink a little bit, and by the way, don’t smoke. “The healthcare system,” as we understand it, is that by the time you show up having had a heart attack or whatever, you had 30 years of basically bad behavioral characteristics leading up to that.

What does it say about us that we treat the healthcare system as basically the last-ditch attempt to keep us from dying after we’ve basically spent our life behaving very poorly. And that goes back to this idea of the doctor becoming the therapist. The answer to the actual health outcomes is upstream of what’s happening in the healthcare system.

Drivers of chronic health conditions

Vijay: And it goes to the bigger issue which you asked about, which was, what is healthcare? Because there’s this kind of mind-blowing article that came out that talked about the reasons for death, and what healthcare deals with is relatively a small sliver of the pie compared to genetics and social determinants. And social determinants being the biggest pie piece, 40%. If your spouse smokes, you’re probably going to smoke, or you’re going to get a lot of secondhand smoke. If your spouse is overweight, you’ll probably be overweight. If your friends drink heavily, you’re going to drink heavily — that the social determinants around you have a bigger impact on healthcare.

And actually, we’re starting to see now when finally the healthcare companies are going full-stack. You’re seeing payer/providers thinking about an air conditioner as therapy or as a therapeutic, because that actually has a greater chance of decreasing mortality or decreasing the chance of going to the hospital if you’re living in Florida, for example, than other things.

And so, I think that’s really kind of a key point that we need to sort of think about, and it goes to the market. And part of the challenge here is that healthcare itself is this kind of artificial market that’s created by the government where certain things are healthcare, certain things aren’t healthcare. We’re seeing Medicare Advantage and other things that allow you to go full-stack affecting this, but part of maybe now that after we get the technology in, it seems like there is no choice but to really revisit what is healthcare.

Marc: It’s like, okay, then, how do we think about paying for this? What are we paying for, right? What are we getting for what we’re paying for, and then, of course, like, who’s paying for it? And, you know, I would just propose when you have a system that’s 1/6th of GDP in which, like, a gigantic amount of the adverse outcomes are being caused by people’s behavior or social context — and most people’s healthcare is being paid for at least in part or potentially entirely by other people, and where consumers have basically surrendered to the system and don’t feel like they have any choice whatsoever, and don’t exercise any choice. And then you have a system as a consequence that’s so heavily regulated and subsidized by the government, you can actually say it’s a minor miracle that it works as well as it does. We basically just designed the worst possible economic configuration for industry.

Vijay: Well, and it’s funny, because some of these things are hidden almost like the germ hypotheses where people didn’t realize there were germs. That was really the hidden danger that we weren’t doing, and really sanitation is the way to fix things. It could be that eating healthy and avoiding Type II diabetes is the new sanitation. The AMA feels that they have a new initiative that nobody in America should die from Type II diabetes. And if you think about it logically, that makes total sense. Just like no one in America hopefully died from issues from sanitation the way we would maybe 200 years ago. I think maybe now it suggests that even now when we have the technology, the question is, how can we go from where we are now to this direction that we’re talking about.

Marc: Yeah. So the positive view there was an economist named Herbert Stein who had this famous thing when he talked about these issues. He said, “If something can’t go on forever, it will stop.” And so maybe contrary to what I said earlier, if there are no limits on how far this can go, healthcare being 1/6 of GDP becoming 1/3, becoming 1/2 — at some point it becomes the most important issue in the world and people are just not going to be willing to put up with it. Maybe just the pain — like, the economic and political pain just gets, like, simply too intense and then you start to realize you have to kind of unwind yourself from some of these assumptions.

But I think honestly the other thing is just more things like LASIK. More things where we can actually demonstrate what happens when technology kind of hits in the positive way. Like, what technology does, right — dramatic boost to productivity growth — which means dramatic improvements of quality combined with, you know, dramatic decreases in costs. The optimistic view there would just be, like, as consumers we’re getting trained to basically be able to like comparison shop and evaluate and get, like, ratings or reviews on everything in our life. Literally everything, almost anything you choose to do, whatever restaurant, you know, you go on Yelp or even these days online dating. You’re used to a level of kind of consumer choice and selection and quality control and decision-making.

If you go, like, buy a new car or something is just, like, the wealth of data that’s available to you — it’s extraordinarily unlikely that you’re going to buy a new car these days and be disappointed, just because you’re going to know everything ahead of time and you’re going to figure out how to get the best possible deal for exactly the car that you want. I think the other part of it is supply-driven, which is just, like, we need to actually drive more technologically superior approaches into the market and like make them available and make them obvious. Like the payer/provider model you mentioned, of just — align with the people who actually want to improve outcomes and just kind of demonstrate the new way.

It’ll be a little bit as if we had only ever had state-controlled media or something, and then all of a sudden somebody kind of had the crazy idea to like maybe actually start making movies in the private sector, and then it just turned out that those movies were 1000 times better. At some point, we just may need to make the new movies.

Vijay: Yeah. And also, what you’re describing is full-stack healthcare on the enterprise side, which an employer will have, but also direct-to-consumer healthcare, and that we’re probably going to start to generalize that. We may start to view Peloton and Peloton-ish things as direct-to-consumer healthcare. I think part of the challenge is that — and this is true for diet and other aspects of healthcare — is that things are so tailored to the individual that it’s been so hard. And nutrition, we can do a whole podcast just on nutrition and the sort of mess that is, but I think now with all that you can measure, even to the point where you could have like a continuous glucose monitor on you, and measuring that every minute, and having that tell you what you should eat, how you should exercise.

As we move into that, sort of, something in between LASIK and Peloton, we’ll start to emerge where maybe it’s not surgery in your eye at your house, but things that are much more clinical and that are getting to these social determinants. Getting to exercise, getting to Type II diabetes and all of its morbidities, getting to diet. You could deal with several of the top killers. That at least would be such a fundamental transition and would be the type of thing that could bend the curve that we’re talking about.

Health apps and wearable devices

Marc: A lot of what you just described can actually be done today. It is actually fairly amazing what you can have, like, as a consumer today just to go through the list — these fitness trackers have gotten really good, whether it’s the Apple Watch or the Fitbit or whatever. They’re now doing, you know, pulse, they’re doing blood pressure, they’re doing kind of comprehensive health state tracking, and they’re doing sleep tracking.

So, you’ve got all the sensor platforms kind of in that thing. You’ve got the sensor platform on the phone. You can’t do laser eye surgery in the house, but people should be able to do eye exams, right, because you basically now have medical-grade visual sensors in the camera. And then you’ve got continuous glucose monitor kind of thing. And then, on your phone, you can have the fitness app that basically tells you what to do to stay fit. You can have the food app that basically helps you figure out what to eat. You can track every aspect of your behavior. You could track alcohol consumption or whatever other recreational whatever. You can aggregate all this data up, and there’s like tons and tons of apps on the phone now that will, like, basically do all this for you.

Now, you know, the people doing this today are like the hyper conscientious types that are super into optimizing every part of their existence. That’s only a small sliver of the population that will voluntarily do this. You can just imagine a mandate, right, for people to get “healthcare coverage” or healthcare insurance at some point, you know, they have to kind of sign up for a better kind of personal behavioral regime, and they might use these technologies to support that. Or, by the way, you could imagine the voluntary version of it.

One of the sort of consequences of healthcare being so expensive right now is this incredible rise in the individual deductible. It might be that the deductible for you just, like, laying around eating Cheetos and smoking pot is, you know, $1,500 or $4,000 or whatever, but the deductible for you with a healthy lifestyle is $200. And then you’ve got, like, the so-called good driver discount that they do for car insurance. And so then you have this sort of behavioral kind of push to be able to directly save money. And that’s an enticing idea, because that aligns the interests of the consumer, right, with the interest of the system and kind of maybe could throw things back into some kind of calibration.

Vijay: Yeah. You think about all the parts you just talked about, that you can get this to be more consumer-driven in a market-like way. Take your previous example of the string quartet that’s in your pocket with Spotify. Now you have your doctor quartet or orchestra in your pocket — with you all the time, giving you the cure that you need. We have the motivations. We have the technology, and actually, we have the startups building it. The optimist in me sounds like this is going to happen, and this is happening, but we just have to sort of get all the pieces together to make it happen.

Marc: So, I have forced myself to watch some cable television for the first time in a long time over this election. So, I’ve actually seen some TV commercials for the first time in, like, a year.

Vijay: Oh, wow.

Marc: By far, the best part of the election coverage was the meditation app And actually, their commercials are actually quite nice because it’s just literally, like, 30 seconds of rainforest sounds. There was also this company called which is an app to help you pray, like, if you’re religious. It’s got, you know, Bible study and guided prayer sessions and stuff like that. At first, I was like, “Okay. That’s a weird juxtaposition.” And then I was like, “Oh, no. I get it. Calm is basically selling secular prayer, right, or is selling religious meditation.”

Vijay: Exactly.

Marc: Which actually bears on health, right, because a huge driver of modern health conditions is basically stress and inflammation and, like, there are physical components to that, but there’s also a medical, psychological, sociological component to that. And so, if people are able to actually deal with stress in their lives, that could actually, like, you know, affect some of these things if it affects things like the rate of heart attacks. It can also affect things like stress-eating, which then leads to obesity.

Vijay: Absolutely.

Marc: It may be that the upstream apps that are, like, the key healthcare apps that we actually need on all of our phones are — take your pick — or You could hire a pastor or a preacher or a priest to come to your house and pray with you or whatever advanced meditation, Zen Buddhist meditation, but it’s going to be a lot cheaper if it’s an app in your pocket.

Vijay: Yeah. They’re just probably aren’t enough to go around.

Marc: The serious part of this is what technology should do is it should empower us, right. It should basically give us capabilities, and it should give us reinforcement and expansion of our capabilities, and help and assistance in ways that make our lives directly better. And I think there is a very big reason for optimism that there is sort of this complete set of ways that we can actually improve our lives that the technology can really help us with.

Vijay: Yeah. Absolutely. Amen.

Lauren: Thanks so much for joining us on “Bio Eats World.” If you’d like to hear more about all the ways biology is technology, please go to subscribe to the a16z Bio Newsletter at And, of course, subscribe to “Bio Eats World” anywhere you listen to podcasts.

  • Marc Andreessen

    Marc Andreessen is a cofounder and general partner at a16z. Marc co-created the highly influential Mosaic internet browser and cofounded Netscape.

  • Vijay Pande

    Vijay Pande is a general partner at a16z where he invests in biopharma and healthcare. Prior, he was a distinguished professor at Stanford. He is also the founder of [email protected] Distributed Computing Project.

Degrading Drugs for Problematic Proteins

Carolyn Bertozzi and Lauren Richardson

In Bio Eats World’s Journal Club episodes, we discuss groundbreaking research articles, why they matter, what new opportunities they present, and how to take these findings from paper to practice. In this episode, Stanford Professor Carolyn Bertozzi and host Lauren Richardson discuss the article “Lysosome-targeting chimaeras for degradation of extracellular proteins” by Steven M. Banik, Kayvon Pedram, Simon Wisnovsky, Green Ahn, Nicholas M. Riley & Carolyn R. Bertozzi, published in Nature 584, 291–297 (2020).

Many diseases are caused by proteins that have gone haywire in some fashion. There could be too much of the protein, it could be mutated, or it could be present in the wrong place or time. So how do you get rid of these problematic proteins? Dr. Bertozzi and  her lab developed a class of drugs — or modality — that in essence, tosses the disease-related proteins into the cellular trash can. While there are other drugs that work through targeted protein degradation, the drugs created by the Bertozzi team (called LYTACs) are able to attack a set of critical proteins, some of which have never been touched by any kind of drug before. Our conversation covers how they engineered these new drugs, their benefits, and how they can be further optimized and specialized in the future.

Show Notes

  • How conventional drugs work, and how PROTAC targets proteins differently [1:46]
  • Discussion of how PROTAC can only reach proteins within the cell [6:00], how LYTAC targets external proteins [9:56], and how LYTAC pinpoints specific proteins for degradation [12:54]
  • Implications for future use of LYTAC [16:37], and what work is needed to turn it into a therapeutic treatment [20:29]


Hanne: Hi, I’m Hanne, and welcome to Bio Eats World.

Lauren: I’m Lauren, and this is the first episode of “Journal Club.”

Hanne: So, tell me. What’s “Journal Club” all about?

Lauren: So, “Journal Club” is on Thursdays, and this is where we take a recent scientific article and discuss it, either with the authors of the paper, or with our own internal experts here at a16z. And we highlight what the paper shows, what new opportunities it presents, and how to take those research findings from paper to practice.

Hanne: Okay. So, tell me — what’s this first paper all about?

Lauren: The first paper is titled “Lysosome-targeting chimaeras for degradation of extracellular proteins,” and it was published in Nature.

Hanne: That’s a lot of words. What do they actually mean?

Lauren: The basic idea is that diseases are often caused by proteins that have gone haywire in some way. So, there’s either too much of them or they’re present in the wrong place or at the wrong time. And the idea here is to create a new kind of drug to degrade those proteins. So, if there’s too much of the protein, you’re reducing the levels. If it’s in the wrong place or the wrong time, you’re removing it from that area. And that’s a really exciting new type of drug molecule.

Hanne: Okay, cool. So, what’s important about this paper? How is this moving us forward?

Lauren: This paper is really exciting because it’s targeting a whole new class of proteins, some of which have never been able to be touched by any other kind of drug.

Hanne: And who is the guest joining you for today?

Lauren: Right. Today, we have the senior author on the paper, Carolyn Bertozzi, who is an amazing scientist. She’s a professor at Stanford and her work focuses on creating new methods to perform controlled chemical reactions within living systems. So, we’re going to lead off with Carolyn describing how conventional drugs work.

PROTAC vs. conventional approaches

Carolyn: So, most conventional medicines act by binding to a target — a pathogenic driver. That’s a protein in your body that’s contributing to a disease. And they act by what’s called occupancy-driven pharmacology. They bind to that target and block its function. So, ibuprofen binds to an enzyme and blocks its activity, which then blocks an inflammatory pathway.

Lauren: So, the normal typical drugs are working by binding proteins and blocking their activities. But in the last 10 years, we’ve seen some really exciting alternatives to drugs that rely on this specific model, with the most well-known and the most well-developed being what’s called a PROTAC, or a proteolysis-targeting chimaera. So, how do these new drugs differ from what we just described — the standard typical drugs?

Carolyn: So, that concept came out of academic labs in the early 2000s, and the two people who published the defining papers in this area were Craig Crews from Yale and Ray Deshaies, who at the time was at Caltech — now he leads research at Amgen. And they had this idea that another way to shut down a pathogenic protein would be to target it for degradation. And around that time, there had been some breakthroughs in our understanding of how nature normally degrades proteins, because she has to be able to do that — new proteins get made, old ones get degraded. And a central mechanism for degradation of proteins inside the cell is that they get marked with ubiquitin chains, and that’s a signal for the proteasome — which is like the meat grinder inside of the cell — to chop up these proteins and destroy them. And there are enzymes that put these ubiquitin chains onto proteins that are destined for degradation.

And so, what Crews and Deshaies realized is that you could build a molecule that artificially bridges the gap between a target protein and this ubiquitin machinery. And with that molecule, you could basically get a protein ubiquitinated intentionally, and therefore degraded. So, that was their conceptual idea.

Lauren: So, in the course of the normal function of the cell, you have proteins being produced, but you also have proteins being degraded. And so, one of the main mechanisms for degrading protein is by the ubiquitin-proteasome system. And that’s where the cell says, “Degrade this protein by adding ubiquitin molecules onto it,” and that pulls it to the proteasome where it gets chopped up. But what a PROTAC does is, it’s a molecule that can bind to target protein — so the one that you want to degrade — and brings the enzyme to it that adds the ubiquitin tag — adds the flag — and then that brings it to the proteasome to be degraded.

Carolyn: Right. And the reason that was so transformative is that not all proteins are easy to block, actually, with drugs. There are lots of proteins that are not enzymes, and they don’t even have a pocket, really, where you could put a drug and it would block the function. So, the cool thing about these PROTACs is that they don’t have to bind in a place that would block its activity, but instead bridges the gap to an enzyme that puts the ubiquitin on and drives the degradation. So, the promise, really, is that the PROTAC — or the targeted degradation approach — expands the druggable proteome because now more proteins can be drugged, because you have this other way of doing it — through degradation and not just blocking.

Lauren: So, you’re using the endogenous mechanism that the cell already has for flagging proteins that you want to be degraded, and using it now to target a much wider range of proteins than you could if you were only able to target those that have, you know, a really nice pocket that could be targeted with an activity inhibitor.

Carolyn: Right.

The benefits of LYTAC

Lauren: So, what are some of the limitations of these approaches?

Carolyn: Well, the targeted degradation field began with the PROTACs, but it has expanded over the last 20 years to include other types of protein degraders — but all of these processes function on proteins that are inside the cell — in the cytosol or in the nucleus. And meanwhile, there’s this whole other world of proteins which are outside the cell. So, these are proteins that are displayed on the cell surface — the membrane-associated proteins — many of which, the majority of the molecule is outside, presented on the surface, where it’s not accessible to the proteasome. And as well, there are many proteins that are just completely secreted by the cell and just released into the extracellular space, and those extracellular proteins are about 40% of the human proteome. So, that’s a pretty big chunk of the pie that is not available to the PROTAC strategy.

And many of these proteins — these extracellular and cell surface proteins — are important targets for drug development. And, you know, my lab had been working on a variety of different cell surface molecules and secreted molecules that contribute to things like cancer immune evasion, for example. And many of the molecules we wanted to drug were really not druggable using the conventional blockers. And that’s where the lysosome-targeting chimaera, LYTAC, research started.

Lauren: I see. So, the PROTACs that you described are a really exciting new modality, but they are limited in that they can only target the proteins that are within the cell. And there’s this huge world of proteins that just are not available to be targeted in that way. And they aren’t ones that rely on occupancy of, like, a particular binding site — they can’t be targeted by those types of drugs either. So, they’re really kind of an unmet need for drugs to target them.

Carolyn: I would go even further and say sometimes even targets that can be drugged with a blocker, you can get a more potent effect with a degrader at lower doses, right? So, even secreted and cell surface molecules that have been successfully drugged with monoclonal antibodies, you might actually do better if you convert over to a degradation strategy.

Lauren: Why do you think that is? Why do you think they’re better than activity modulators, or is that not known?

Carolyn: Well, I think with occupancy-driven pharmacology, you can’t ever get, like, 100% of the target protein blocked. There’s always an equilibrium, and you have to constantly pump the system with enough drug to keep the occupancy as saturated as possible. By contrast, the degrader can bind to a target and get rid of it, and then bind to another target, and get rid of it, and bind to another one, and get rid of it. So, you’re just reducing the level of the target protein, but because there’s the potential for one drug molecule to mediate the degradation of multiple targets, you could get a deeper inhibitory effect in principle. And that has now been borne out, even in some early-stage human clinical studies, with PROTACs. The same could very well be true with LYTACs. Of course, it’s a much earlier technology, so we don’t know that definitively, but there’s — I think — a rationale for thinking that way.

Lauren: Yeah. That’s kind of like pharmacodynamics 101, that you have a reversible inhibitor, and you’re going to have this equilibrium, but these degrading molecules, you know — they don’t get degraded when they tag the protein for degradation. They have a benefit of — one degrader molecule can target a huge number of target molecules. So, that’s really interesting, that even in, like, a head-to-head comparison on a known druggable target, that you can possibly get a better effect by degrading as opposed to inhibiting.

How LYTAC targets proteins

So, now that we have the background on why we need this new type of drug, why you decided to go after extracellular and membrane-associated proteins — let’s get into the details of how you develop these molecules. And, as we mentioned, the PROTACs co-opt this endogenous pathway, the ubiquitin-proteasome pathway — but they can’t reach these proteins outside the cell. So, what cellular pathway did you co-opt to degrade those proteins?

Carolyn: So, again, nature degrades these extracellular and circulating molecules, and she does this through what’s called the endosome-lysosome pathway. So, cells will basically internalize and engulf molecules from the extracellular space into endosomal vesicles that go through a maturation process to become the lysosomes. And the lysosomes — people from their cell biology classes might recall — that’s the organelle within the cell that has a lot of degradative enzymes. So, lysosomes can degrade proteins, polysaccharides, lipids. There’s a lot of hydrolases within the lysosome.

And so, we conceived of an idea where we would develop bifunctional molecules, where one part binds the protein that you want to degrade, and the other part binds a lysosomal trafficking receptor system. So, that’s the key — is that lysosomal trafficking system. And it turns out that in human biology, there are about a dozen known receptors whose job it is to grab stuff — either from the membrane or from the extracellular space — and pull it into this endosome-lysosome pathway for degradation. And so, what we have done is hijacked those pathways by basically building molecules that interact with those receptors, and then attaching them to a molecule that binds a target of interest. So, that’s the structure of the LYTAC — a binder on one side for the target, a binder on the other side for a lysosomal trafficking receptor.

Lauren: Right. So, nature has already come up with a way to degrade the proteins that are membrane-associated and extracellular, and you just developed a mechanism that allowed you to say which protein you want to degrade and then extracting it from the extracellular space and degrading it inside the cell.

Carolyn: Yeah. And one of the best known lysosomal trafficking receptors is the so-called mannose 6-phosphate receptor. And mannose 6-phosphate is a sugar epitope that is found on lysosomal enzymes, and that allows them to be trafficked to the lysosome by this receptor — the mannose 6-phosphate receptor.

Lauren: So, you have this sugar molecule that if you attach it to a protein, that’s going to take it into the lysosome. So, how did you engineer the specificity to target the protein that you wanted to the lysosome?

Carolyn: So, you need a binding molecule that is very specific, and ideally also very high affinity, against your target of interest. And, you know, in our early proof of concept studies, we chose targets to degrade for which there already were high affinity, high specificity antibodies available — several of which are already approved human medicines. So, for example, we’re interested in the epidermal growth factor receptor as a target for degradation. This is an important cancer target. EGFR — it’s overexpressed or mutated in many cancer types, where it’s driving the proliferation of cells. And we made a LYTAC out of a human drug called cetuximab — it’s an antibody against EGFR that is used, you know, in the oncology setting. So, that process of taking an antibody against a target and just decorating the antibody with the mannose 6-phosphate groups — that converts it to a LYTAC.

Lauren: Great. So, antibodies are molecules that our immune system produces, and they are incredibly well-tuned to bind one specific protein — and there are many drugs that are actually antibodies — but their main function is to just block that protein. And what you did was you took that therapeutically active antibody and added the glycan molecules that you needed to turn it into a LYTAC. So, now not only is it blocking the protein, but it’s shuttling it into the lysosome to be degraded. It almost gives it, like, an extra function— like, making it even more effective at disrupting their target’s function.

Carolyn: That’s right. Also, the more we learn about biology, the more we are appreciating its complexity. And I think we also are now understanding that most proteins have functions that are not just binary, you know — like an enzyme is either on or off. Most proteins have multiple dimensions to their function. They interact with other proteins. So, when you block a protein through an antibody, or through a small molecule inhibitor, there are probably other interactions of that protein that you’re not affecting, which still contribute to the biology. And when you degrade the protein entirely, you take away all those dimensions of its function. And so, it’s not just that a degrader can be more potent than the inhibitor in an axis of biology — I think the degrader can have more axes of an effect.

Lauren: Do you think that that could lead to possibly off-target effects of disrupting, kind of, a bigger network than you anticipated?

Carolyn: That’s a good question, and I guess it depends on where you draw the line between on-target and off-target. Because, take a protein like EGFR. The biology of that receptor is driven by its interactions with other components of the signaling pathway. You know, EGFR binds its ligand — the epidermal growth factor — and the consequence of that is, that triggers a signaling cascade. And so, if you inhibit the activity of EGFR by just blocking, you don’t affect any of the downstream signaling biochemistry. 

However, if you drive the degradation through the LYTAC approach, and some components of that signaling machinery come down with it, that is actually a direct hit — I would say that’s on target, right? Because you’re hitting not just EGFR, you’re hitting the complex that drives its biology, right? So, again, the biology is never transacted by a protein in isolation — it’s by that protein and the network of its interactors. So, I would argue that if you can degrade some of its interactors, it’s a more profound influence that’s on-target.

Possible future applications

Lauren: So, now that we’ve talked about, you know, the details of your study, how you develop these bifunctional ligands that can bind to a specific target protein and shuttle it into the lysosome for degradation — let’s zoom out and put this research into the broader perspective. What are some of the new opportunities that this work provides?

Carolyn: We’re now exploring therapeutic applications of the LYTAC technology, and we’re interested in extracellular targets that have been very — either difficult, or really just impossible to drug, and there really is no option right now for patients for certain disease settings. So, for example, we’re very interested in diseases that involve aggregation of proteins in the extracellular environment. Proteins that in their misfolded or unfolded forms lead to toxic aggregates that can cause tissue damage. And so, these are diseases that are often called amyloid diseases. The ones that are most familiar to people would be neurodegenerative conditions like Alzheimer’s disease, Parkinson’s disease — it’s been very difficult to figure out, you know, how do you get rid of these protein aggregates that are pathogenic in the extracellular space? They’re not really amenable to inhibition — the process by which they form is often not well-understood. You really just want to get rid of them, right? You want to degrade them. And I think the LYTAC approach is perfectly situated to take on peripheral amyloid diseases. 

For example, there’s a condition called light chain amyloidosis. Antibodies have a heavy chain and a light chain. So, in patients with this condition, there’s too much light chain all by itself, and it’s not stable, and it’s forming amyloid aggregates — which deposit in organs throughout the body and they’re toxic. The standard of treatment for these patients is very poor. So, we think the LYTAC approach could be interesting in that setting.

Lauren: That’s a perfect example because those light chains don’t have an enzymatic function. They don’t have a nice pocket that you would be able to stick a drug in. So, the ability to pull those out of the extracellular space and degrade them with a LYTAC sounds like a perfect match between disease physiology and drug modality.

Carolyn: Yeah. So, that’s an example of a, sort of, secreted pathogenic molecule or system of molecules. There are other membrane-associated targets that we think the LYTAC is well-suited toward. And one class of molecules that my lab is really interested in are called mucins. These are transmembrane glycoproteins that are huge, and they’re kind of the giant redwood trees of the cell surface, so to speak. And they’re known to be associated with cancers. And cancers that overexpress these mucin molecules — they tend to be very aggressive and very difficult to treat. And we’ve done a lot of work to understand, like, what’s the function of these mucins that’s oncogenic. And the bad news, from the perspective of drug discovery, is that a lot of the biology of these mucins is a physical biology. So, they’re pathogenic because of their stiffness, and their rigidity, and their physical effects on the cell surface — not because they interact with a receptor, for example, which maybe you could block, right? 

And so, what do you do when the function of the molecule is a physical one, and not a biochemical one? And I think this is where you just want to get rid of them. I think you just want to degrade them. And fibrosis, right — that’s a disease setting where there’s pathogenic accumulation of collagen scarring. And you know, that’s hard to think about — how to drug that, you know, at least at the end point of the disease, where you have this material that you really just want to degrade. And so, again, I think a LYTAC strategy would be interesting to test in that setting.

Lauren: A lot of really important applications for this. So, what are some of the elements of the LYTAC design that still need to be optimized to turn them into therapeutics?

Carolyn: So, this was the “version 1.0” of the LYTAC technology, and the work that’s now going on is basically the second- and third-generation improvements. And those improvements have taken several forms. So, first of all, we’re interested in improving the structures. So, the second-generation LYTACs have a new chemistry, so that the conjugations are site-specific — that we can engineer the part of the antibody that actually gets coupled to the mannose 6-phosphate groups. And with our new chemistries, we can make different geometries of LYTACs and find what is the best geometry for a given target — and it probably will be target-dependent.

So, we’re kind of now writing the rulebooks, and in the publication, the LYTACs we made are built from these known antibodies. We are now developing LYTACs from other kinds of binders — including small molecules that might otherwise have been blockers — we’re now converting them to degraders through the LYTAC approach. Another dimension that we’re expanding upon is the lysosomal trafficking receptor that we hijack. So, the mannose 6-phosphate receptor was a great starting point — it’s expressed in virtually all cell types. But there are other systems that are more specific for different cell types or different tissues. So, our next LYTAC family are targeting a receptor called the asialoglycoprotein receptor, which is a liver-specific lysosomal trafficking receptor. And we have a preprint that we posted on ChemRxiv on this new generation of LYTACs.

Lauren: Yeah. Liver-specific makes a lot of sense because we were talking about fibrosis — liver fibrosis is a huge problem, and that’s caused by too much collagen in that area that you want to break down. But you don’t want to break down collagen everywhere in the body, you know — that’s really a critical molecule. You could get wrinkles, God forbid! <laughter> It’s really important in your skin, it’s really important in your joints. So to have that specificity of where you want to target the degradation is really important, and an additional, like, strength to this approach.

Carolyn: Yeah. And I think that then hints to a broader universe of LYTACs that target different receptors that are tissue-specific in different settings. That’s the tip of, hopefully, a big iceberg of interesting new degraders.

Lauren: And more to come. So, we’ll end with — what is the key take-home message from this article and from our discussion today?

Carolyn: I think the most important point is that this exciting, still relatively young field of targeted protein degradation has just been set free from the confines of the cell. So, extracellular proteins should now be added to the list of potential targets for a degradation strategy, and we hope with the LYTAC technology that we can bring added benefit to patients.

Lauren: Well, thank you so much for joining me today on “Journal Club.” I really enjoyed our discussion, and I’m so excited to see what comes out of this research.

Carolyn: Thank you.

Lauren: And that’s a wrap for the first episode of “Journal Club.” If you enjoyed this episode, please subscribe, rate, and review wherever you listen to podcasts. And to learn more about how biology is technology, subscribe to our newsletter at

  • Carolyn Bertozzi

  • Lauren Richardson

The Biology of Aging

Laura Deming, Kristen Fortney, Vijay Pande, and Hanne Winarsky

Welcome to the first episode of Bio Eats World, a podcast all about how biology is technology. Bio is breaking out of the lab and clinic and into our daily lives—on the verge of revolutionizing our world in ways we are only just beginning to imagine.

In this episode, we talk all about the science of aging. Once a fringe field, aging research is now entering a new phase with the first clinical trials of aging-related drugs. As the entire field shifts into this moment of translation, what have we learned? What are the basic approaches to developing aging-related drugs? How is studying aging helping us understand diseases like cancer and Alzheimer’s — and increasing the amount of time we are healthy — today?

In this conversation, Laura Deming, founder of The Longevity Fund; Kristen Fortney, co-founder of BioAge, a clinical-stage company focused on finding drugs to extend healthspan; Vijay Pande, general partner at a16z; and host Hanne Winarsky discuss the entire arc of aging science from one genetic tweak in a tiny worm to changing a whole paradigm of healthcare delivery.

Show Notes

  • Overview of the research on aging [1:54] and the current state of the science [4:16]
  • The three most common research approaches [6:19], why this field is expanding so rapidly [8:21], and possible applications for disease treatments [11:00]
  • Discussion of pure aging research vs. treatments for disease [14:46]
  • Getting this science into the healthcare system [18:42] and issues with research funding [22:16]
  • The types of entrepreneurs needed to expand the field [25:23]


Hi, I’m Lauren.

Hanne: And I’m Hanne, and this is our first episode in the new podcast “Bio Eats World,” where we talk all about how biology is breaking out of the lab and clinic and into our daily lives — and really on the verge of revolutionizing our entire world in ways we’re only just beginning to imagine.

Lauren: So, Hanne, the title of this first episode is “The Biology of Aging.” What aspects of aging are we gonna be discussing today?

Hanne: Well, really we’ve been trying to dream up ways of slowing down aging for as long as we’ve been aging, right? But the field of studying aging as a science is pretty new. So in this episode, we look at the entire, kind of, biology of aging — what we’ve learned; what’s reality; and what is translating into actually increasing our health span, and potentially — one day — possibly [slowing] down aging.

Lauren: What’s healthspan, and how’s that different from lifespan?

Hanne: Your first thought when you think about studying aging might be how we might slow it down, but really the way a lot of people in the field think about it is increasing our healthspan — which is the amount of time that we live healthy. What’s really interesting about this episode is, it’s about not just increasing healthspan and age span, but what we’re learning about disease — and particularly chronic age-related diseases — that might help us be healthier today.

Joining me for this conversation is Laura Deming, founder and partner of The Longevity Fund; Kristen Fortney, founder of BioAge, a clinical-stage company focused on finding drugs that extend healthspan using machine learning; and Vijay Pande, a16z general partner on the bio fund.

Lauren: Were there any insights from this episode that changed the way you think about aging yourself?

Hanne: Yeah, well, I definitely enjoyed hearing about the drug already widely available that really might increase our lifespan. And I also loved hearing about what the difference between Benadryl and Unisom is. So we start with a little bit of a history of the field, talk about where it’s come [from], and where we are today.

History of research on aging

So, where actually are we in the biology of aging today? There’s been a big surge of talk, even over the past few years, about what the science of longevity is — how it’s developed — but where are we actually today?

Vijay: Mortality is, like, this thing that philosophers opined [about] for millennia, but yet the biology of aging seems new. <laughter>

Kristen: Right. New, insofar as it’s new that anything actually works, I guess, right? One of the earliest discoveries in aging research that goes back decades is that if you could severely restrict food intake in animals — calorie restriction — they would live substantially longer. But it’s only been fairly recently that we’ve been able to actually intervene, and actually impact how long a mammal can live. And one of the interventions that was first shown to work in mammals is parabiosis — exposing old mice to young blood. And that really was first discovered 50 years ago.

The major acceleration came during the 1990s, the 2000s — and it’s mostly attributable to the first finding, you know — Cynthia Kenyon, Gary Ruvkun, Tom Hughes — that you could delete a single gene in the worm C. elegans and double its lifespan. Everyone thought aging — so complicated. You know, how are we going to have a dramatic impact on aging when it’s really all of these different systems and processes that are going wrong simultaneously?

And then, you know, wow — wait a minute. This one tweak, and then suddenly this massive difference in lifespan. So a lot of invertebrate geneticists went into the fields and mapped out all these longevity genes that impact worms, flies, and yeast, which is awesome. But now, you know, which of those translate to humans? Those are the ones that matter for translation.

Laura: Going back to, kind of, the history of the field — you, kind of, have these really — you know, sort of, highly-advanced intellectuals, going to the field and then, kind of, losing a lot of their momentum forward, practically — Nobel laureates like Elie Metchnikoff, claiming that gut bacteria, kind of, control aging. And maybe that’s coming back around now in some areas of current biology, but back then, it’s not as well supported. It’s only recently that you started to have the traction in the field to make specific discoveries.

That period of time was just so critical to the field’s birth. Cynthia Kenyon, when she was making these first studies, was told, “You’ll fall off the face of the earth, literally, if you pursue this research and you do the study.” And if you look at her first paper, she was the lead author because no grad student was willing in her lab to do the work. That was such a controversial first step to take as a, you know, young principal investigator. That was how unexpected it was <Mmhmm.> that people really thought that it would be the end of your career to, kind of, go into this field — and then she, kind of, you know, started it anew.

Hanne: They didn’t even want to touch it.

Kristen: Yeah, exactly. Worse than unexpected — like, bad science.

Current state of the science

Hanne: So, can we talk about what that traction actually is looking like right now? What is the most promising traction?

Laura: I think one thing that we feel really strongly is — this is the critical decade. Patients are for the first time receiving drugs that were developed in the context of aging biology. And it’s fascinating to watch these first clinical trials occur, where companies are actually developing drugs.

And when that first patient gets an actual clinical benefit, we’re gonna see — you know, people actually affected by these, kind of, ideas that have percolated in the field for decades.

One of the, kind of, examples of this that’s most, sort of, prominent in the field was a trial testing a drug called Metformin in the elderly. And so it’s actually looking at all-cause mortality, not just a specific disease as an endpoint. Metformin itself, you know, is this drug which retrospectively has been shown to be somewhat correlated to a decreased mortality in, for example, diabetic patients.

Kristen: Well, it was discovered just by analyzing health records, right? So, so…

Vijay: Which itself is kind of fun.

Kristen: Which itself is, like, yeah — that’s a great way to find, sort of, repurposed drugs.

Vijay: Yeah, and who’s living longer.

Kristen: Exactly. Yeah. So it’s one of these drugs that millions of people have been taking for decades. You can actually go back in time and ask the question — you know, are people who are on Metformin living longer? And they are, and it’s kind of amazing. So that’s, sort of, where the whole hypothesis for this compound came from that’s now being tested in the clinic, which is so exciting.

Hanne: I gotta go get me a prescription right now! <laughter> Are there key approaches that we haven’t touched on yet that we should be describing as this new field kind of evolves?

Laura: There’s also resTORbio, which, you know, was testing a molecule that’s similar to rapamycin. And that was for respiratory tract infections in the elderly. That trial did not work when trying to get into Phase 3, but if that had replicated, that would have been one of the more — big, sort of, examples.

There are some, sort of, drugs in the clinical, sort of, landscape today that are for metabolic disease. So things like NASH, or diabetes, or obesity — which when you overexpress these proteins in mice, make the mice live longer. So there’s this key link between things that we already are using to treat metabolic disease in the clinic, and, kind of — what might actually impact lifespan.

Vijay: So, that’s the connection with Metformin?

Kristen: Metformin impacts cancer deaths, too. So again, it’s like a broader aging-related mechanism.

Vijay: Okay. That’s interesting.

Common research approaches

Laura: One way that we try to classify these companies is in three generations. One is focusing on traditional pathways — so things that might affect, for example, insulin signaling in the body. And those are, kind of, known targets that people are drugging with existing modalities.

The second would be trying to screen for novel targets using platforms that are high throughput, and, kind of — either novel model organisms, or, kind of, novel in vitro or in vivo screens.

The third would be to actually target damage directly — where you’re not saying there’s an evolved pathway that we’re knocking up or down. You’re, rather, saying there’s a set of damage accumulated, and that’s what we’re, kind of, going after in a more engineered fashion. So, you know — for example, targeting what are called senescent cells — so, cells that get, kind of, old and decrepit with age.

Hanne: Mmhmm. The idea of zombie cells, right?

Laura: There’s damage that builds up in the lysosome of each cell, called lipofuscin. And that is an aging-related type of damage which, when targeted, you know, may be relevant to these neuro disorders that people are, kind of, starting to work on. So there’s three different, you know — just small examples of clinical, sort of, work being done, but for age-related diseases.

Hanne: That’s like three different frameworks, basically.

Kristen: Well, the question, right — for the first generation of companies — is what’s the low-hanging fruit? If something is very well conserved through invertebrates up to mammals, probably it’s gonna do something in humans too. So mTOR is a very interesting target. That said, the genes that are the most important for invertebrates are probably not the most important ones for humans, right? So I think a lot of those new pathways have yet to be discovered, and will have much higher impact on longevity — phenotypes as well.

And damage, I guess, also is sort of going directly to the major causes of disease. So I think those all make sense as approaches. I mean, it’s so unexplored now therapeutically, right? Even those drugs that have a very mild impact on longevity are, I think, going to be incredibly meaningful. I think that’s a really important consideration as well.

Vijay: And what do you call mild? Like, 10% increase in…

Kristen: Yeah, like a few percent increase in lifespan.

Laura: Rapamycin is probably the most well-validated drug for extending mouse lifespan, right? But, you know, the amount of compounds that were tested to that level of scientific rigor — it’s about 30 compounds. They put 30 drugs into mice, you know — did 30 random experiments. Right? <laughter> And one of them, you know, boosted lifespan by 14%. So, I think there’s going to be tons of things that have [a] much higher effect than rapamycin.

Vijay: Getting back to thinking about just the biology of it, it’s — is there any other trend for the “why now”? Is it just finally people like Cynthia Kenyon being brave enough to, sort of, help create the field? Are there any other, sort of, confluence of things coming in here?

Kristen: Mapping out every single molecule in a blood sample, in a human blood sample — proteins, metabolites, whatever we can get our hands on — and seeing which of those predict living a long, healthy lifespan — and going after those that are causal. Even five years ago, really, the technologies that we’re using didn’t exist.

Laura: Kristen, you really, kind of, changed my thinking here. When we first met, you were talking about biomarkers for longevity, and how important those were — and to be able to test our hypotheses in humans, and that’s where it all counts. And so, kind of, when you had pointed out to this was the key problem, I think that was such a big watershed for the field of — if we just make a fast, easy, cheap, reliable biomarker for aging, that’s really gonna change the whole field in a way that is more than just, kind of, getting one pathway to work it.

Vijay: The biomarker thing is actually very interesting, because — let’s make an analogy. We have cholesterol as a biomarker for heart disease. And because there’s such a causal relationship between cholesterol and heart disease, you don’t have to run a trial waiting for people to die of heart disease. And that’s huge.

And especially, also, you can measure it. You can see small changes go up and down. You have something that’s not binary — dead or alive. You have something that has a lot of nuance to it. And so, having biomarkers is both really useful, but — I actually think somewhat reflects just the maturation of the space, too.

Hanne: Is there another approach where we’re all aging differently, and we need to understand things on an individual level in terms of what our aging type is? That different systems age in different ways?

Kristen: It’s the same as with any biomarker, right? <Yeah.> Or with cancer. <Yeah.> You can, like, personalize the hell out of it, and say you’ve got these weird mutations — and therefore you’re part of this special subtype, right? And I kind of think that personalized medicine is where you go after you’ve, sort of, exhausted the things that are going to work for a broader population.

I mean, as we discussed earlier, there are already mechanisms of aging conserved across species — you know, from yeast to us. So certainly there are also really potent mechanisms of aging that are conserved across humans. We’re focused on targeting those first, looking at the commonalities first — but certainly, you know, for certain individuals, there will be particularities to how they age that you could also, you know, treat differently in different people.

Vijay: When we’re talking about changing paradigms, it’s not just a scientific paradigm, or even a clinical paradigm — but as a healthcare delivery paradigm as well.

Now there is this opportunity to say, “Given that knowledge, what can we do against existing therapeutic areas — existing disease?” We don’t have to talk about “fountain of youth” — we’re talking about learning new biology. Learning new targets that can directly go into a clinical trial for a new disease. And I suspect that could be a very interesting, sort of, initial area — initial application.

Hanne: So it’s, like, what can learning about aging actually do to make you healthier right now? In the age you’re actually in.

Vijay: Or it can actually help you cure a disease that you have.

Applications for treating disease

Hanne: How — what is that connection? Can we just spell that out?

Vijay: Yeah, well — and there’s a couple of variants of this. One variant would be an aging-related disease, like Werner’s disease — these diseases where you age rapidly. That’s kind of an obvious one, but maybe what’s less obvious is other diseases, like — could we be talking cancer? Could we be talking Alzheimer’s? What are the possibilities?

Kristen: It’s all of those, right? I mean, age is the single biggest risk factor for those diseases. Like, 20-year-olds do not get Alzheimer’s — and we cannot cure Alzheimer’s today, and therapeutically it’s been a disaster. Everything has failed in the clinic thus far, and part of that is probably because we’re studying it in the wrong way. I mean, when we’re testing drugs in animal models, mice don’t get Alzheimer’s, and young animals do not get Alzheimer’s at all.

Laura: Alzheimer’s disease, cancer, heart disease, and stroke — we have to study these diseases in the context of aging. And that, I think, is a new perspective.

Vijay: If you think about just the biology of Alzheimer’s, it’s not even clear what’s going on. Like, even which protein is it? A-beta? Is it tau? Is Alzheimer’s an A-beta aggregation problem? Is it a fibril problem? Is it a tauopathy? Like, even the field can’t even agree on the biology. Even targeting a fibril, or targeting tauopathies it’s not a traditional pocket that you get a small molecule to go into.

If you have something where the current drug design methods don’t work, it seems like applying the current drug design methods is not the right thing to do. This feels like the type of radical shift that could have an impact, and still keep us in small molecule land. When we think about this, then actually the translation part is pretty straightforward because I think the beauty of what we’re talking about here is, the current healthcare system won’t have to change.

Hanne: Interesting.

Vijay: That basically we have indications and, as Kristen mentioned like, not just any indications, but the biggest killers that we have to deal with.

Hanne: Huge amount of need.

Vijay: Huge amount of need.

Vijay: And Alzheimer’s, where there’s at least, to date, no drug at all. I’m curious, like you could have a patient with the early signs of Alzheimer’s like, you know, with MCI, mild cognitive inhibition. Could you reverse a phenotype, or could you just delay a phenotype?

Kristen: I think that is the whole promise and the practical approach as well. That really, if you have a drug in hand that treats aging fundamentally, it should treat several different diseases. And yes, we can work within that — the existing medical system. With the one caveat I don’t think an aging drug is going to be a great drug for metastatic cancer.

Vijay: Yes. So stage four is probably too far.

Kristen: Yeah. And, sort of, how far is too far? And really, these targets will probably have their most potential when they’re used in a preventative fashion. And, of course, that’s not something that the existing system can deal with. But I do think that early disease, like MCI you can at least halt progression, which would be massive, you know. And potentially reverse it with some of these mechanisms.

Vijay: Well, and the reversals would I think, gets everyone excited.

Kristen: Definitely.

Vijay: But even if you could just slow down — in Alzheimer’s, slowing down could still be very, very valuable.

Kristen: Yeah. It would still be disease-modifying. Yeah.

Vijay: And you could have <inaudible> point against that.

Hanne: So, it’s interesting you’re saying almost that, like, the biggest hurdle is getting the biology of aging in its approach of its own. And then once you can get the right targets, then you can, sort of, slot into the existing system and keep moving.

Vijay: I think there’s so much about the science the biology of aging that has been validated, that now has opened the door to now treating these as targets. And actually, you know, the <inaudible> is you could, like, just identify that target, toss it over the fence to your favorite pharma and it would slot into the same type of programs that they would be running right now. It doesn’t require a radical, sort of, reenvisioning of pharma to make this happen.

Moreover, I think you know, if you look at the history of pharma, it goes through waves of new technologies. And maybe it’s an interesting question when or if longevity becomes that hot new trend. And I suspect that in order for that to happen, you have to have one or two clinical trials that have shown this works, and then it probably just catches fire.

Aging treatments vs. disease treatments

I want to amplify one thing Kristen said that I think went by relatively quickly that is very, very important is that these compounds, if they are truly going after the biology of aging, will be useful in multiple indications. At first, that sounds magical, but there are actually precedents for existing compounds. So that alone is interesting that they’re already precedents.

Hanne: Can you compare an example there?

Vijay: I mean, some of my favorite stupid one is actually Benadryl and Unisom. So actually it’s the exact same drug. You go to the pharmacy. Often they just happen to be on opposite sides of the aisle and actually, when sold as a sleeping pill, it costs a lot more than as a… <laughter>

Hanne: Oh, I’ve never noticed that. Is that true?

Vijay: It’s the exact same compound, exact same dose. And if you ever take Benadryl for allergies, you get very sleepy. So that’s a simple example. There are better examples in other diseases.

Kristen: Humira, for example, is one of the ones.

Vijay: That’s a great example. Humira has like what, five or six indications?

Kristen: That’s right. I think even more and, like, the world’s most valuable drug as well, right? So…

Vijay: But this is a little different. I think in that one you just happen to like…

Hanne: They’re similar. They’re similar diseases they’re more similar.

Vijay: The Humira case, it’s similar diseases. In the Benadryl case, it happens to make you sleepy. <Right, right.> And it’s almost like taking advantage of the side effect. This is something fundamentally different. This is something where actually the, sort of, way to save all these diseases is to slow down aging and that’s why it has such a broad impact.

Hanne: So, is it oversimplifying it to say aging as a kind of root cause of all these diseases or is that…

Vijay: Or an amplifier of the diseases.

Kristen: Or a causal driver.

Vijay: Or a causal driver.

Kristen: Well, look at immune aging, right? I mean, your immune system declines horribly with age. You don’t respond as well to vaccines. You’re more likely to get incredibly sick when you do get the flu or a cold, and that affects everything in your whole body that makes everything worse.

Vijay: From a pure basing point of view, it is a causal driver.

Kristen: There you go.

Vijay: From just a mathematical-statistical point of view.

Kristen: By definition.

Vijay: And then that makes it a very natural, philosophical way to think about it.

Laura: One of the hypotheses about why we have genetic pathways that control aging is that we’ve evolved those for a reason. That there’s a benefit to living longer enough to have kids in a different environment. And it really wouldn’t do you well to live longer and be sick, right? You want to have ways to impact all your health that pushes back all diseases. Otherwise, kind of, you just get — you know, dead [from] a different thing earlier. And so that’s, kind of, perhaps why it’d be plausible to believe that there’d be, sort of, all-disease sort of efficacy for these kinds of anti-aging therapeutics.

Vijay: Actually, what is the evolutionary selection for aging or lack of aging? Because you could see that once you’ve given birth to children, or maybe gotten them to grandchildren — then you have no purpose, right? <laughter> I mean, you’re done from an evolutionary point of view, and you’ve — let’s say, diminished purpose from a purely, sort of, cold, evolutionary point of view — but you’re still taking up resources.

Laura: If you have a certain fixed mortality rate year over year — if that’s actually much higher than it is today in our developed society, your probability of being dead at any one point in time in your life is actually — it gets pretty high, even independent of aging over time.

And so, if there are any things that benefit you when you’re young that might be harmful to you older — or just, kind of, maybe things that accumulate randomly past the point at which you’re likely to be dead from other non-aging causes — they might just accumulate. And so, now that we have actually the ability to live long enough to potentially have benefited from the number of years, there’s been no selective pressure, potentially — to, kind of, live longer in that, sort of, period of life.

Vijay: One of the things that I’m always just curious about is — what don’t we know now that we need to know? Because the problem with biology is that it’s just so complicated. Longevity and aging biology seems to be amongst the most complicated. That’s the thing that I’m always wondering about, is — what is going to be the big surprise or the big curveball, and what can we learn from it?

Kristen: That’s a really good point, right? Because I think we’re all waiting for the first clinical trial to be successful, and that’s going to be so important for the field. So for pharma companies that traditionally don’t work in this area to really get confidence and excitement around it. But, yeah, there’s so much risk associated with bringing these first mechanisms forward and figuring out the indication path.

I mean, you can even have a good mechanism but have, you know — defining these indications for the first time. Of course, we’re gonna get it wrong the first few times. There’s so much to figure out because it’s really such a new field.

The current healthcare system

Hanne: Okay. So, we’ve talked about the explosion of the field — of the study of the science, the biology of aging. And then we’ve talked a little bit about what that brings us actually right now, in terms of understanding biology and disease — but where do we meet resistance again, where we try to get this into the health system that exists today, as a kind of preventative medicine? What does that look like in terms of the end goal being a healthier life, a longer life, a longer healthspan?

Kristen: I think that’s a great question, because you’ve got this therapy in hand — you think it’s actually slowing down aging, and, yes, you can work with the existing healthcare system and layer on indications one at a time. But really you’re not getting to the whole aging population as quickly as you can, right? And what could that path look like in the future?

So, biomarkers is one route. I mean, maybe people are still pre-disease, but they’re frail. There’s sort of functional and molecular biomarkers that predict they’re going to be sick soon.

Vijay: Like statins.

Kristen: Like statins. Exactly like statins.

Vijay: And satins will, you know — sort of, does handle a biomarker <Yep.> with the hope — when it’s done prophylactically — to avoid disease. People often say that people don’t want to pay for prevention, but we do pay for statins. There’s this old joke that plumbers have saved more lives than doctors. And that’s this point about sanitation — has just been this fundamental, sort of, floor just for human health.

And then I think the next level up, in my mind, is getting rid of the Fritos — and no disrespect to Frito-Lay or Pepsico, <laughter> or minimizing the Fritos, you know — as much as I do like them. That’s what comes to mind.

I mean, basically, no one should have type 2 diabetes. I mean, that’s another version of sanitation. And so now the question is — could you imagine, like, with longevity biology in hand, where you have these biomarkers, no one should have these aging-related diseases — or maybe nobody should have disease before the age of blank? And that blank goes from, like, 60 to 70 to 80 to 90 and onward?

Hanne: That’s right.

Vijay: Perhaps what we really just need is something to have this rock-solid biomarker that the clinicians are convinced is an issue — and then you have therapeutics that can help you manage to that biomarker. At least there’s a paradigm for that.

Hanne: Well, well, exactly — but any therapy that really delayed aging — that really delayed the onset of disease — would save a tremendous amount of money, you know. And you can put a number on that, and you can justify a certain cost. It shouldn’t be that hard.

Kristen: I think that’s where it comes back to “this is the decade” — because this is the first time that we’re going to see trials actually looking at all-cause mortality with therapies that are already on market today, and we’re going to see the impact of those readouts. That’s never been something that’s ever been done before. That’s truly different from any other time in history.

Laura: And that’s the proof we need to get the system to really start recognizing it that way.

Kristen: One would hope. <laughter> If that doesn’t move hearts and minds, what will?

Vijay: So, that’s a great point. I’m wondering, like, what would be the analogy? Like, are we at, like — first Lipitor, kind of thing. We’re looking — because then there’s been, what? Four generations of statins since then? Before then, actually, that model didn’t even exist.

Laura: It means you kind of form it responding to, like, the first watermark trial of the shift in paradigm — and that kind of occurring potentially as a result of these, but yeah.

Kristen: For the field too, right? I mean, we’re now at the point where several of these hypotheses are being tested clinically. We’re going to have to wait while we really get the human proof of concept for the idea, and then once that data comes in, I think that’s going to be huge.

Osteoporosis is a really good example, too, right? It didn’t used to be considered a disease, but there are, sort of, markers of — you know, your bones get weaker as you age, and that predisposes you to really severe outcomes and events. And now it’s recognized as one, and now there are drugs, and there’s a way forward, and payers were convinced, right? So there are case studies, I think, that we can follow

Hanne: Where it’s kind of flipped. The understanding has flipped.

Kristen: Exactly.

Vijay: The mentality towards it has flipped.

Issues around funding further research

Hanne: Where are we in the hype cycle, would you say?

Kristen: Yeah, aging and biotech generally, like — it’s shifted in the last few years to be a lot more accessible with, I would say, like, low upfront capital, right?

So, first of all, the data sets that my company relies on — you know, we were for the first couple of years a data company — you know, like, for people with laptops, vastly cheaper than biology. Even if we were doing biology, now there are incubator spaces. Now there are CROs like WuXi that can do all your chemistry outsourced.

So, I think the barrier to entry for biotech has gotten a lot lower, and really enabled a lot of these new and exciting ways to work on targets and therapies.

Laura: And in 2011, 2013 — like, there were so few companies that, like, just having enough money to finance those companies in the space was the limiting thing. Now I think there is actually enough money, just even from the past couple of years, to fund the good ideas and the good people.

And so when an entrepreneur comes to us and says, “Hey, I want to make,” this is a common thing, “make a lot of money and then put it back into biotech,” it’s like, no, no, no, if you’re actually a good entrepreneur, please start a company. That’s what we need more of — start a company if you want to impact the space. We lack people.

Kristen, one thing I’m just fascinated by — as you know, you were one of the first to really go out there and do a couple of things. One will say we need biomarkers for aging, but also just build an aging company at all. I mean, there were very, very few new companies when you started. What have been the, sort of, easier and harder things that you’ve encountered as a result of that focus?

Kristen: I mean, it’s new, right? So everyone, I think, understands that it’s riskier, I guess, than if you have, you know, another company for NASH, another company for cancer — where everybody knows exactly how that’s going to go, from discovery through validation, through your clinical trial design, through your reimbursement.

There’s a lot of uncertainties because the space is so new, but related to that, there’s also so much opportunity. I would say that there’s more awareness now that these drugs are in trials. That there’s more — I would say — also appetite for novel mechanisms now <Yeah.> that the usual approaches are not working. So I think the landscape has changed a lot — not just, you know, at the startup level, but in terms of, like, big biotech as well.

Vijay: Well, there’s — one, sort of, just common question for any founder serving the biopharma side — when you can do many things, what do you do first? How do you pick a therapeutic area? That’s probably one of the hardest questions an entrepreneur has to deal with.

Kristen: Yeah, so there’s no, sort of, clear, well-trodden path — but that means that we also have the opportunity to really optimize and build something new. We’re trying to design our first clinical trials. So should it be for an age-related disease? Should it be for something closer to aging? Again, uncertainty plus opportunity, right? And trading those two things off, and making a bet.

Laura: We’re really focused right now on just getting more people to be longevity founders. Early 2010s, it was lack of capital. Like, there was just no money in the space. Right now the big bottleneck is founders. And we’ve seen many amazing companies built by both grad students directly out of their, kind of, Ph.D — but also people coming from software engineering, managerial positions.

And a lot of these people self-select out of the population. They say, “I can’t start a longevity company because I don’t fit the profile of a brilliant scientist founder — or a, kind of, traditional, say, investment banker type.” But, you know, they make incredible founders, and there’s just a huge population of folks out there who, I think, should be starting companies. So just to double down the idea that, like, if you want to really impact longevity, start a company. That is, like, exactly what we need right now.

Hanne: What are the other types of founders that you tend to see coming into the field — you know, in this new field?

Vijay: The founders in this space typically combine a couple of things. They either are biologists who have embraced, you know, machine learning or other areas — or even people that are coming from the tech side that are learning the biology. It’s a really unusual time where you can actually learn both.

And maybe you’ve learned both from the beginning — but actually it almost feels like it’s never too late, because you can pick up both sides. But that if you can capture both sides, I think you’ll have a huge advantage.

A nontraditional founder for us would be someone that is coming, maybe, from the pure pharma side. And we haven’t seen that yet, but I suspect they’re coming on — and Kristen’s nodding her head. And I suspect they’re coming either to be founders or as, you know, CSOs — and that they may become some of the key employees for these companies.

Hanne: So the culture and the talent landscape [is] changing too, evolving and changing. Interesting.

That’s it for the biology of aging. And thanks so much for joining us on “Bio Eats World.” If you’d like to hear more about all the ways biology is technology, please subscribe to the a16z bio newsletter at And, of course, subscribe to “Bio Eats World” anywhere you listen to podcasts.

  • Laura Deming

  • Kristen Fortney

  • Vijay Pande

    Vijay Pande is a general partner at a16z where he invests in biopharma and healthcare. Prior, he was a distinguished professor at Stanford. He is also the founder of [email protected] Distributed Computing Project.

  • Hanne Winarsky