Scientist Stories: Adrian Veres, Engineering Therapeutics

Welcome, Adrian Vares. He's an MD-PhD coming out of Harvard Medical School and the Melton Laboratory, and I, I thank him for joining us at the Sugar Science. Our mission is to connect scientists in type 1 diabetes and related research fields, enable their collaboration through our website, internet, and events, and accelerate research. In type 1 diabetes. Welcome, Adrian.

Thank you, it's a pleasure to be here.

I wondered if you could give us a, you know, quick thumbnail sketch of your career path, what excites you most about the research you're working on right now.

Yeah, great. So my quick career path is I trained initially mostly in physics and chemistry with a focus on kind of applying computational methods to biology, and at the time that really meant primarily working in microbiology where you could run sort of big ambitious experiments where you could measure a lot of different components. And then essentially, luckily for me, as I was starting the PhD phase of my MD-PhD training, realized that a lot of new technologies that were coming on board in human cell biology allowed us to kind of apply this thinking now to development. And then kind of this notion of applied developmental biology, working in vitro, having a lot of control over the system, and going after this really challenging question of transforming the cells into specific cell types that kind of we want and we desire became really interesting to me and has been the focus of my work since. I think as we're looking at this field in sort of the decades to come, I think it's going to be an absolutely fascinating field. Beta cells are clearly the front runners in terms of a challenging question and also pretty interesting answers, but I think it's setting up the foundation for how do we think about constructing tissues constructing beta cells initially, but also tissues and organs in vitro and architecting them throughout. And I think both learning from natural development but also not being constrained by it and kind of understanding that in vitro all bets are off and it's important to keep an open mind, but also we have much more control and being able to be creative about the means of controlling that I find is really exciting.

Yeah, that's well said, and it will be really interesting also to see how this, you know, this system will be able to be rolled out to, you know, sort of through the QC system and, you know, rolled out in a very orchestrated manner so that it's available and in the clinic.

Absolutely.

Yeah. So I just wanted to ask you, you know, if you could do a little bit, you know, what you're going to talk us a little bit about what you're sharing today, and then we'll go through some of your slides and then we'll have some questions, Q&A.

Yeah, so my— the main things I'll be sharing today is really focused on what I would call core Melton Lab work related to creating beta cells in vitro, and really the culmination of several years of work where we've initially set off to kind of comprehensively map what is it that we make during the differentiation protocol alongside and on the way to making beta cells, and then more recently work where we focused on now that we have that map, how can we think about controlling that process further and going beyond protocol control. What I really like about this work is also the beta cell world is a perfect prototype for it in the sense that, as you said, we're so close to the clinic with these products, with what I would call Gen 1 products. Over the years, we really have this opportunity to go in and provide improvements and technical refinements to the way we create the beta cells that should have a pretty transformative impact, both in terms of the the purity that they have, which is going to be the focus of what I'll be talking about, but then also paving the road to enhanced beta cells that, produced in vitro, behave differently than any beta cell that we're making today in the lab.

That sounds really interesting. Let's go through the slides and then we'll talk a little bit more.

Okay, so can you see my screen?

Perfectly.

Okay, great. So it's a pleasure to be here. Really briefly, the— this is, I think, the most important intro slide in my mind is the context to the work of creating beta cells in vitro really begins with the challenge tied to the fact that islet— that to the knowledge that islet transplantation is known as a possible cure of diabetes. Typically this means, though, taking a deceased donor pancreas, isolating islets, then transplanting them and beginning immunosuppression, and in some cases, although not in all, we can see good glucose control without need for exogenous insulins years out after transplantation. Now, this is really limited by the lack of tissues that we see for diseased organ donors, as well as this need for immunosuppression. So that's been a challenge in terms of rolling this out on a large scale. It's important to note that because of this, in many cases, we do islet transplantation when it goes alongside another transplanted organ, most often perhaps the kidneys, which can fail as a result of diabetes itself, and then you have to do immunosuppression anyway, ways so you can transplant islets. The challenge— and this really gets at the heart of what's the promise of regenerative biology to this question— and that challenge is essentially to try to address both of those questions. First of all, it's address the supply question by simply being able to produce as many cells as we need in vitro to be able to do the therapeutic approaches that we need to do. The second layer, and this is kind of still further out but a very active part of the field now, is since we have the cells in vitro and we have control over the cells, we have the opportunity to immunoprotect them in various ways. For now, it's really micro or macro encapsulation techniques, but soon there's a lot of activity in the field surrounding genetic modification to cloak these cells from the immune system in a number of ways. And then the focus there is once we create these islets, we can transplant them and restore this homeostasis. The focus in the Melton lab really over now the past 2 decades has been in creating these— this differentiation protocol. And in the last few years, we've been able to start with embryonic pluripotent stem cells. Now the protocol also works with induced pluripotent stem cells, differentiate them over the course of several weeks where we— all we do is add various factors and then create cells that are functional in vitro they can secrete insulin in a glucose-responsive manner, they have various other secretory aspects that match, kind of, islet to bona fide beta cells. Then more importantly, they're functional in vivo, meaning once we implant these in various diabetes models, we can see restoration of glucose homeostasis. Now the challenge of the work is kind of what is actually in this mixture that we make when we do the differentiation protocol. And if you looked about 5 years ago, so kind of when our most current version of the protocol was first coming out, the landscape, the picture that we had was really one based on essentially immunostaining or inspecting various genes one at a time. And you could create a pretty compelling map of what's going on from this, but it's actually far from comprehensive. And at that time, the characterization really focused on, well, let's understand the beta cells. You know, they're functional. How functional are they? What gene expression, what markers do they express, and how can we compare them to beta cells throughout normal development. Luckily, essentially for the field, in this past kind of 5 years, when this protocol was coming out was also roughly when the advent of single-cell RNA sequencing and other single-cell techniques was really starting to come to the front, to the fore, and we started applying these techniques to the differentiation protocol. And I won't go into kind of all the details of the experiments we did, but we essentially very systematically marched along the protocol at different time points across different versions of the protocol, different starting cell lines, and looked for what are the cell types we make in a kind of completely unbiased way. Now the summary of this, if we look at the end of the differentiation, which is the material that we might transplant, for example, we can essentially group the cells into 4 big picture categories. Some of these, first of all, we have the SC beta cells. This is the thing we're trying to make. They're typically between 20 to 40% of the differentiation product. We also have their neighbors, which are what we call SC-alpha cells. They produce glucagon. They really match beta-alpha cells from, from the islets in terms of their transcription profile. Interestingly, they expressed briefly— they expressed briefly insulin at part during part of the development. That has been a bit of a confusing question in the field. We also saw a new endocrine cell type that had previously not been known about in this differentiation because we'd approach the question with this unbiased perspective. And we understood these cells eventually to resemble enterochromaffin cells, which are a common endocrine cell type in the guts that produce and release serotonin.

And finally, we—

so all three of these are endocrine cell types, meaning they're, you know, they're very similar to neurons. They're electrically active and release various hormones in response to input signals. And we also have some fraction of non-endocrine cells, which really match various components of the pancreas, as well as a couple much more rare endocrine cell types. So then once we had this map and we understood the balance of these different cell types, the challenge became to How can we do better? How can we make more of the things we want, less of the things we don't want? And at a very abstract question, the place where this kind of the answer typically begins is, well, let's go change the protocol. If we spend 20 years optimizing the various factors that we add and how, surely we can do better. That said, we're also in a place of diminishing returns there. We've put in a lot of effort in the protocol, and now as we make changes, we're just as likely to break it than we are to improve it. One place that we really hadn't put much effort in is in optimizing the actual cell line that is going through the protocol, and that's really the focus of my work, has been to take the starting cell line, and if the wild-type cell produces some deterministic mixture of cell types when we run it through the protocol, to identify changes that we can make to the starting pluripotent stem cells such that when the differentiation proceeds, we make more of the cells we want, for example, more beta cells, the same number of alpha cells, and then fewer fewer non-endocrine or fewer EC cells. The approach that we decided to take on for this was also in part due to the ability to start doing this, to do these types of experiments in human cells in the last few years, was what we call a pooled perturbation screen. So essentially we set off to systematically knock out every gene in the genomes of these cells, essentially treating each individual cell in the differentiation as a replicate of an experiment and then being able to separate out the populations that we care about and measure how frequently a specific perturbation made it into the final population that we make. We essentially do this experiment where we change the stem cells. Each individual cell gets a different change, a gene that's knocked out, as well as a different label that labels what gene we did knock out. They undergo the full differentiation protocol. At the end, we separate them out based on their identity and then measure how many labels do we see in each one of those categories. And then we essentially did this screen in two rounds. Once we targeted the whole genome, but with relatively little effort on every one gene just because there are so many. But then once we had our list of hits that were interesting, we went back, took all of them, and did a secondary targeted library that gave us really precise and sort of crisp statistical understanding of those perturbations. It's actually quite an interesting challenge to think about the data that we get out of this kind of screen because we essentially have this data that's the frequency of labels of given perturbations in each of our 4 populations of interest. We can then ask various questions. We can compare, for example, all 3 endocrine populations— beta, alpha, and EC— to non-endocrine cells. We can also ask questions that are focused to the balance of endocrine cells across them. And since we only have a few minutes, I won't go through how we do the math and statistics to arrive to this analysis, but we can really just look at one example. So here's a gene that's known to matter to beta cell differentiation, to beta cell development. Called PAX6. It's a transcription factor, and when we knock it out, we see that the rate of formation of beta cells is really low. The rate of formation of alpha cells is also very low. The rate of formation of EC cells, if anything, went up, so it was perhaps unaffected. And when we ask— summarize this information in a balance of different— of the endocrine cell types, we would qualify this knockout as being very pro-EC, in a sense the opposite of what we're looking for, although it gives us a clue that maybe we should go overexpress PAX6 and similar genes. But allows us to kind of start thinking about this question. Once we actually do the whole screen, we have many replicates of the same legibles that target the same genes. We can build statistical models to understand how confident we are that any one gene actually changes the process of differentiation. And at the end of the day, we arrive with a list of just over 200 genes that we believe have an effect once we knock them out individually. Most of these genes just tend to break the differentiation process, which perhaps is not surprising. It's a very precise orchestration that should be broken by many changes. But then we also have a whole set of genes that are actually having an effect that changes endocrine balance in one direction or another, and those really give us the opportunity to go and exploit them in some way. Really encouraging for us, since we're developmental biologists, we often think about transcription factors, and in this screen we sort of immediately see the effect of about 20— about 67 transcription factors that really have effects in every direction of pro-beta, pro-EC, or pro-alpha, essentially cluing us into those developmental regulatory signals that are actually driving the process of differentiation. It was really satisfying in one experiment essentially to recapitulate many, many years of careful work from developmental— from the developmental biology community applied to the pancreas. But this is really a picture of what's possible today when we combine these types of experiments with in vitro differentiation. We've also taken some of these hits and essentially— and started now early— doing early validation work to see what happens when we focus in on one gene and perturb it and measure the signal. And very encouragingly, we found genes that can now push the process towards beta cells or reduce the incidence of EC cells, and we can start to playing around with these combinations, which was really, in a sense, the reason we set off on this whole experiment in the first place, was identify perturbations that allow us to make more beta cells. So kind of just wrapping up, I know this is pretty quick, but we have this very complex developmental process in vitro by which we arrive to beta cells. They go through a number of progenitor states that have a number of branching points, but we can really map out what that looks like using single-cell methods, essentially. And then this gives us the opportunity of, once we know what are the main types of cells that we make, to start manipulating things explicitly in the starting material, in the starting pluripotent stem cell, to create new tailored cell lines that should be better for differentiation-based therapeutics. This is really a project that's been going on for some time now, and we're also really excited by the kinds of questions that it opens up. You know, one is really looking at combinations. So everything I've talked about was knocking out one gene at a time. There's a lot of information that is embedded in how genes interact and how perturbations interact. Combinatorial overexpression, combinatorial knockouts of various key components, TFs and otherwise, are really, really interesting to us here. And then also, this also kind of serves as a backbone to approaching other complex differentiation systems that create more cell types that also have anatomical considerations that need to be taken into account. So yeah, so that's my quick run-through. I'd love to to discuss anything.

This is great.

What a—

I mean, that's a huge body of work, Adrienne. Thanks for sharing all you've done and all the contributions you've made. It sort of strikes me when you're at first glance that it's a bit like a safe cracker, right? I mean, listening, you know, tuning, turning the wheel or turning the dial, listening, you know, for the click and then turning back. And that's the kind of work you're doing, very precise, precise and very much, you know, data-driven, I guess, too, at the same time. So it's very carefully done.

I do—

I have a couple of questions.

One, I know that José— I think it was José Rivera Feliciano— he had talked about some data that suggests an inherent clock or circadian rhythm in the development of the beta cells from the stem cells. I saw like HES1 was one of the genes that was on there, but, you know, recently there was this Nature article talking about HES7. And I wondered, you know, if you guys have looked closely at the clock mechanism, developmental clock, I mean, sort of driving the, you know, the beta development.

Yeah, yeah. So a postdoc in our lab, Han, has actually put in quite a bit of effort on understanding, is there a circadian rhythm component to the way the cells behave? And if so, what is it regulated by? It's actually— and the interesting answer was that yes, it turns out that there is an oscillatory behavior to both the expression of the cells. And now we're really talking about the final product of the differentiation that's completed and not changing anymore. Are, as well as to their ability and sort of capacity to sense glucose metabolites and secrete insulin. It's actually very fascinating if you think about it, because these cells have last been in a body essentially decades ago. They've been in the freezer, they've been growing in our incubator since, but it shows that these mechanisms, you know, circadian mechanisms, are really present throughout the body, both kind of centrally regulated and synchronized, but then also essentially embedded in each individual cell. Now, in terms of how these mechanisms are present in our differentiation process itself, we don't know how that's driving the signaling and the fate commitment yet. It's likely connected to cell cycle regulation, which is inherently oscillatory and can be connected there, but the sort of precise detail of that mechanism in the system we don't know yet, although it's an exciting frontier.

Yeah, it is really.

And then I guess I wanted—

I loved how you showed, you know, you've got like the 20% of the beta cells and around that same amount of the alpha cells. What about the delta cells?

Um, who's making— I mean, those seem like they should be in the mix. Um, who's, who's making the best delta cells? Do you, do you know, uh, which lab is really sort of progressing there? I mean, are you guys making delta cells?

Yeah, so there's been several labs now that have been working on the kind of creation of delta cells. I think I've seen various projects that have all shown kind of different percentages in our differentiation protocol. So we should mention we see about 1 to 2% of somatostatin-expressing cells. And this really starts getting into how we define cell types, which is a very interesting and important question. We feel that the cells, that the 1-2% that we see in our differentiation, are actually relatively good stand-ins for the bona fide delta cells in the sense that they express markers that are quite unique to the delta cells in normal development and in the adult pancreas. That said, in terms of who's making the best ones, that's a tough question.

Maybe that's too far, not the best, but I mean, Who has some robust Delta cells where they show data for expression of somatostatin? Yeah, they're fully differentiated from the—

so our own lab has had projects that have shown success in that regard. We also have— I'm mostly familiar to the work actually on Delta cells coming from people from our lab who'd previously been in our labs, such as Quinn Peterson, now at Mayo. But I think the jury kind of remains out on, you know, clearly they're present in the islet. What function do they have in terms of regulating these synthetic islets, and do we need them, and how many do we need, is going to be a pretty interesting question. So yeah.

Yeah, no, it's definitely an interesting question, and I think it's, um, I think some people are talking about it. This morning we had a, we were on a talk with, um, INSPIRO.

It was kind of an interesting discussion, you know, Matthias von Hereth and, um, Matthias Hebrock was on there, and they were talking about, you know, can we establish libraries, global libraries of some of these iPS cells, and, or, you know, uh, what are your thoughts on that? I mean, it seems everyone has their proprietary way of creating these cells. Could there ultimately be repositories, or, you know, uh, I guess sort of, you know, like we said, libraries of these, of these cells? That would exist. Do you think that would— or is this going to be more like a personalized approach, personalized medicine?

Yeah, so I think there's two questions in there. One, to this notion of if we're building therapeutics, yeah, what, what are we doing? Are we going the sort of global, the, the universal stem cell route, or are we going the personalized route? And I think right now there's a lot of research in both fields, and I think insofar as making personalized stem cells remains hard, that makes, that makes universal donor stem cells very attractive. Also perhaps easier to understand from a safety and like regulatory point of view because it's the same material over and over again. But I think if, you know, creating the stem cells and the differentiation process becomes more easier over time, and that itself, that process can be QC'd and controlled, there's not necessarily a direct reason to prefer one over the other. I think time will tell. In terms of which becomes easier or not. To that notion of a library, I think that's a great idea. And I actually feel myself spending a lot of time in the software world. It's very clear that, you know, if we look at where the world of software is today, it's gotten there immensely because of very open sharing of the resources and the code and the work. And I think this kind of open source principles really apply to to the field of stem cell biology. It's also really difficult, right? Like, we have a stem cell line, if it's thawed and handled differently, even though it has the same label on a tube, might behave completely differently. This is what we're trying to exploit in this, in this, in our work, is really change the stem cell lines in a productive manner. But most changes are bad, and I think getting through that is going to take a lot of attention. But I also think it's perhaps not just a good idea, but something necessary for our community to improve the shareability of when someone says, hey, try this protocol on this stem cell line, you don't have to hesitate. You can just try it and get predictable results. I think that would be an amazing step forward and something various members of the communities are working on.

Yeah, completely agree. And have you heard about a small company in San Francisco called Minutia? They're making nanosensors to detect whether or not beta cells are still functioning once they've been implanted. And—

sorry?

I'm not familiar. I'm not familiar with them now.

Very interesting. I'd love to hear more. Matthias Hebrock is one of the co-founders. We're watching it very carefully because that's very important, right? I mean, it's awesome to be able to implant these cells, but we also need to know— I'm not sure the FDA is very concerned about whether or not they're functioning. How can we tell they're still functioning and how can we remove them if they're not functioning, etc. So the idea of the nanosensors is really interesting.

I absolutely agree. I also think the field of biological measurement right now is really in an explosion. We also have the opportunity to— because we again can control the cells that we put in, we have the opportunity to bake in sensors into our cell types, for example, that allow us to know, are they alive? Are they dying? Are they you know, very, very, like, responding to various signals or even being functional. I think that's a really exciting direction. I also feel, broadly speaking, this is a new frontier for us in terms of transplanting components, right? We've implanted biological devices which are inert, essentially. We've implanted whole organs in a variety of contexts where most things that happen are really, like, we can essentially assess them with biopsies. How we're going to understand, you know, we don't have something we can just go biopsy in most of the plans I've seen to implant stem cell-derived beta cells, but how we can actually set up these queries and get that information back will be really important to be able to improve. I think measurement, you know, there's this saying of measure what matters, and if we don't measure something, there's no way we're going to be keeping track of it and being able to track our progress against that metric.

Yeah, no, that's absolutely well said.

I wonder, so what's next for you? What are, uh, what's, what's ahead for you?

Yeah, so, so what I, I'm very happy in, in terms of having found this niche of controlling development in vitro and using in vitro development as, um, really a sandbox in which we can understand the mechanisms of cell type specification, identity, develop technologies to control those mechanisms better, and then all of this in the service of creating those cell types that we care most about. I think to me there's, there's a feeling with beta cells now we're essentially, we're not far from the finish line. There's a lot more improvements to be made, but in terms of just creating the beta cell, we're actually quite close to the finish line. Now the question has become What's the organ that the beta cell fits in? How is it assembled? How is it connected to other things? How does that organ interact with, for example, the immune system? I think are really important questions, as well as starting to think about organs where we can create the cell types in vitro. But it's not just about making one cell or a clump of one kind of cell, but it's about having an anatomically structured and organized organ, which is a you know, as daunting of a challenge today as I think making beta cells was about 20 years ago. And I think this— these types of approaches will be really key to it. So I'm trying to— so I'm looking forward to essentially moving in that direction.

Yeah, fantastic. I can't wait to see what you're going to do next. You've got a lot of— a wealth of sort of finely detailed science behind you, and you've got your MD-PhD almost in hand. And you've got—

Exactly.

And you've got some—

I mean, that, that intersection between clinic and the bench will be a really powerful place, I think, for you. And can't wait to see what you do next.

Thank you.

I was going to say thanks so much for taking the time to talk with us.

It's my pleasure.