Short Wave - Thanks, Neanderthals: How our ancient relatives could help find new antibiotics

Episode Date: October 30, 2023

Antibiotics have changed the world. They've made it possible to treat diseases that used to mean anything from discomfort to death. But no new classes of antibiotics have made it to the market since the 1980s. What if humans' closest, ancient relatives held the answer to antibiotic resistance? Some scientists want to discover new antibiotics using machine learning ... and some very, very old relatives of humans. Host Aaron Scott talks to César de la Fuente about using computers to discover the first therapeutic molecules in extinct organisms. Have a question? Email us at [e-mail] pcm.adswizz.com for information about our collection and use of personal data for sponsorship and to manage your podcast sponsorship preferences.NPR Privacy Policy

You're listening to Shortwave from NPR. Antibiotics have changed the world. They've made it possible to treat so many diseases that used to mean anything from discomfort to miserable death. But there's a problem. We're facing a silent pandemic where more and more bacteria are becoming resistant to available antibiotics. Cesar de la Fuente is a professor of bioengineering at the University of Pennsylvania School of Engineering. Today, over one of one. 1 million people die every single year as a consequence of untreatable infections.

And that's projected to worsen and to actually lead to the death of 10 million people by 2050, unless we do something about it. No new classes of antibiotics have made it to the market since the 1980s, just variations on existing antibiotics. That's in part because finding and testing new drugs is hugely expensive. But as a postdoc at MIT, Sasser had an idea. What if they could use machine learning to get a computer to do it? The first challenge was how to teach a computer to innovate at the molecular level. And after much thinking with our collaborators,

we thought that the best way to go about it was to actually mimic the greatest engine that we have for diversity and generation and to innovate at any level, which is evolution itself. So we decided to train a machine to execute Darwin's algorithm of evolution, where it took initial antibiotics that were not very effective and it was capable of evolving them to become much more effective. The computer came up with a number of molecules and then Sessar and his colleagues took those, synthesized them,

and tested them against bacteria and petri dishes. And we saw that a number of the computer-created antibiotics were capable of killing clinical isolates quite effectively. Then they tested the best one in mice and found that it killed bacteria there too. In 2018, they published the first. study that, to their knowledge, used AI to find a new antibiotic. The next thing they explored was using their computer models to dig through all the

proteins in the human body, what's called the proteome, in search of tiny proteins called peptides that might play a role in the immune system. Pepties are actually part of that innate immune system. So they're the first responders to any infection that we face. And so they're actually critical to allow us to live. They discovered more than 2,500 peptides with anti-infective traits. And I got them thinking, why stop there? When every species has its own proteome, its own treasure chest of potentially antibiotic molecules.

Every species, that is, living and dead. In one of the brainstorming sessions, one of the concepts that came about was the concept of Jurassic Park and of the extinction. Jurassic Park scientists came along. Using sophisticated techniques, they extract the preserved blood from the mosquito and, Bingo, DinoD DNA. The idea in the movie, of course, was to bring back to life dinosaurs, entire organism, so it's organismal the extinction. And of course, that has a lot of limitations, right?

There's the fundamental limitation. DNA is fragile. It can't survive for millions of years. There is no such thing as workable dino DNA. But even for more recently extinct animals where we do have some DNA, like the woolly mammoth, there are all sorts of ecological and ethical questions, too. But thinking about how overcome all those limitations, we came up with the concept of molecular de-extinction. You know, instead of bringing back entire organisms, why not just bring back

molecules from the past to solve present-day problems? And we decided to look at our closest relatives, which are Neanderthals and Denisovans. Today on the show, in celebration of all the ghosts and skeletons out there haunting the Halloween Graveyards. We're going to talk about resurrecting molecules from our extinct relatives to fight deadly modern-day infections. I'm Aaron Scott. You're listening to Shortwave, the science gravecasts from NPR. And so, Cesar, you said you were inspired by Jurassic Park, but I'm guessing there was no mosquito preserved in amber with Neanderthal DNA in it. So can you tell us how you accessed the DNA from our long-extinct ancestors? So just like it usually happens in science, there have been many, many people, many amazing men and women scientists over decades that have been developing painstaking methods to sequence mitochondrial and genomic DNA from ancient humans amongst other organisms. And the really great thing is that now we have those data available and we can access genomic and proteomic data from Neanderthals and the nisovans. and all we had to do was develop the right computational tools,

in this case a machine learning model, to really mine all that information that is out there and try to find interesting things. Can you tell us a little bit about how that worked? I mean, what is the machine learning model looking for? Yeah, so what it does essentially in broad terms is it takes the whole proteome. Again, these are all the proteins encoded by the genome

of both Neanderthals and Denisovans and also modern humans as a reference. It chose all these proteins up into little fragments, and then we ran a number of additional machine learning models and also human expert filters. And after those filters, we have a number of sequences from archaic humans that are predicted to be good antibiotics. And then we reached a point,

which was probably the most exhilarating moment of the project, which was we were about to resurrect molecules from extinct organisms that as far as we know are not present anywhere in the world, anywhere in nature. And as you can imagine scientifically, that was extremely exciting. But then, of course, that also brought about other thoughts of, you know, what does it mean to resurrect a molecule? There is no longer present around us. And from a bioethical perspective, so we started consulting with bioethesis, you know, at the time. And we continue to do so because we believe in.

responsible innovation. We want to make sure that everything we do is done correctly. And so the computer basically finds these peptides that have potential. Take us through how you actually tested to see if they work? Yeah. So the next step was actually doing the resurrection part. And for that, we use chemistry. We use a technique called solid-faced chemical synthesis, which essentially is like little robots that allow us to make the peptides. And they make one amino acid at the time and then they linked them in a chain to essentially get your final peptide, which again is a tiny protein. And then we expose them to bacteria that we grow in the laboratory and we see whether they're able to kill clinically relevant bacteria or not. And what we found

is that several of the peptides were able to kill bacteria quite effectively in petri dishes in vitro. And then so those best ones that we found, we transitioned them to animal models after testing for toxicity and things like that. And in one of the mouse models, which was a skin infection model, one of the Neanderthal peptides was able to reduce the infection to levels comparable to a standard of care antibiotic called polymixin B. And that peptide we call it Neanderthal in 1, because we consider it as sort of the first antibiotic coming from a Neanderthal, and that is capable of reducing infections in an animal model. And I'm curious, I mean, are ancient relatives? lived in a different microbial world from today.

Did you discover that these peptides maybe went about attacking the bacteria in a way that is different from how our modern peptides work? Yeah, actually, one of the results that were quite intriguing was that the modern sequences tended to target the outer membrane of bacteria, whereas the ancient ones tend to target the cytoplasmic membrane. And just to recap a little bit, gram-negative bacteria, this is a type of bacteria. that are particularly pathogenic or problematic in our society. They have two membranes, the outer membrane and the cytoplasmic or inner membrane.

So what I'm saying here is that the modern sequences tended to target the outer membrane, whereas the ancient ones tended to go after the cytoplasmic or inner membrane. And just to make sure that I'm getting this right, so many modern antimicrobial peptides kill things like bacteria by disrupting their outer cell membranes. But these ancient peptides pass through that outer membrane, and disrupt the inner membrane to kill things. Now, the sample size that we have for this is very small,

so I don't want to make any sort of correlations here, but this is just early evidence that perhaps these sequences operate differently on targeting bacteria in different ways. What are the next steps? Is this something that you're going to pursue further? Like, can we hope to see a dandothal-based antibiotic on the market? one day in the future? I mean, yeah, I think the next steps is obviously use this information, these molecules as templates for further optimization, for anti-infective activity and so on.

I don't think they're ready yet to enter the market or even pre-IND enabling studies, which are the studies that need to happen before phase one clinical trials. But I think they represent excellent sort of templates so that we can learn more and further optimize them. Going beyond molecular de-extinction, what are some of the other ways that your lab and other researchers are using AI to develop new antibiotics and other therapies? Like, what are the prospects that have you really excited? This is a really emerging field. If, you know, it's really only half a decade old. And if I had to sort of summarize the progress that we've been able to make is that typically to use in traditional methods in antibiotic discovery in the traditional pipeline,

to discover one preclinical candidate, it takes between three and six years. What I would say AI has enabled is that instead of having to weigh three or six years, we can now discover thousands of preclinical candidates in a matter of hours. And I think the really fascinating thing, if you step back, is that prior to all this work that I'm mentioning, the total number of peptidic antibiotics or peptide antibiotics known to humans was around 6,000. And now we have over 1 million, just in a couple of years of worth of progress.

And of course, our ultimate dream is that some of the work that we're doing, some of these efforts will lead into something that will, you know, save lives and help people. That's really what drives us every single day. Thank you so much for exploring this research with us. I mean, it's just absolutely fascinating. Thank you for having me. It's been really a blast. If you want to hear more about the science behind the de-extinction of entire animal,

like, say, the Dodo and the Woli Mammoth, not to mention the questions about it. We did a few episodes a while back. We'll link to them in our episode notes. This episode was produced by Rachel Carlson and edited by our managing producer Rebecca Ramirez. Anil Oza Check the Facts. Our audio engineer is Patrick Murray.

Beth Donovan is our senior director and Anya Grundman is our senior vice president of programming. I'm Aaron Scott, dressing up this year. years a Neanderthal peptide. Okay, maybe not, but happy Halloween to you all from us here at Shortwave.