‘Holy grail’ of naked mole-rat research reveals how queens rule

0:00A snarl in the woods. Ancient forest spirits nip at his heels. He can outrun a force of nature on foot. He needs an instant power machine built for the wild. The only escape is in the wild. The Toyota BZWoodland. Off-road handling. All-electric power. Leave evil in the dust. Jon Cho stars in Camping. Face to wild. Test drive Toyota's new all-electric family at your local dealer. Toyota. Let's go places.

0:45Not only refreshing, but at some level astounding. Nature.

0:55Welcome back to The Nature Podcast. This time, how a naked mole rat queen maintains dominance. And the science of how people approach board games. I'm Nick Poucher-Chow. And I'm Katrina Clark.

1:17First up on the show this week, reporter Sharmini Bundel has been finding out about how a special odour helps keep a naked mole rat queen on her throne. Naked mole rats are very unusual creatures. They're neither moles nor rats, but these burrowing rodents do have a number of features that make them stand out among mammals. They're hairless, wrinkly and almost blind. They're extremely resistant to getting cancer. They live for an extraordinarily long time. They're essentially cold-blooded. The list goes on. But for Melissa Holmes, a neuroscientist who's worked with naked mole rats for 20 years, there's one feature that's really fascinating.

1:57For me, it's always been about their social behaviour. Their social structure is one of the naked mole rats strangest features, and it's unique amongst mammals. Within the group, there's a reproductive female and we call her the queen. The queen is the only female who's reproductively active. She has her breeding male and all other individuals within the colony don't reproduce. While the queen produces offspring, the workers do things like forage for food, protect the colony and care for the queen's many, many babies. This social structure is often compared to that of social insects like honeybees.

2:33However, queen bees are raised to be queens from before they even hatch. In naked mole rat society, any female could theoretically one day become a queen. Naked mole rat queens aren't born. They're made or it's something that is controlled by the social environment. It tells us that there's something happening within the colony that is allowing one female to be a queen and preventing all of the other females from being that. It's not a genetic trait. And this has been the great mystery of naked mole rat research. What determines which female becomes queen? And what prevents the rest of the colony from doing the same?

3:10In bees, other females are prevented from breeding by pheromones produced by the queens. These volatile compounds waft through the air and stop the ovaries developing in the other bees. In rodents, pheromones are usually found in urine, so researchers looked at queen mole rat urine. They found nothing. It looked like there was no naked mole rat equivalent of bee pheromones to suppress worker reproduction. And that is where the story ended, until now. We keep more than 450 animals. They live in 16 different colonies. They live in a tunnel system where we give them chambers connected with tunnels to make it similar to the wild.

3:52This is Mohamed Khalaf from the Max Delbrook Centre in Berlin, Germany. He's part of a team that have a paper out in Nature this week that might have cracked the mole rat mystery. To do so, he had to collect as many odours as possible associated with mole rats, including the queens, which was easier said than done. It's hard to find the queen, especially in large colonies, because they sleep on top of each other. Once I open the cage, they will all run away because they smell me. However, Mohamed did manage to find them, and he collected samples of all the various smells coming from queens and workers from different colonies.

4:33It turned out there was one compound that was exclusively made by mole-rack queens, a chemical called isopropyl miristate. Here's Gary Lewin, the head of the lab, who worked with Mohamed and others on this new research to explain what it is. When you looked at the structure, it didn't look like what we think of as pheromones, right? It's a relatively simple molecule. It's a volatile. The first thing you do is you go in literature and you look up where can you find isopropyl mirror state. And it turns out that it's a relatively common molecule that's found in many plastics. But what was even better is that we can just buy this stuff for basically$5.

5:09And so that's exactly what we did. And then we can start to do experiments. And it was easy peasy. Isopropyl mirastate is a well-known chemical. It's found in skin creams and even head lice treatments. But what was it doing in naked mole rat colonies? Well, it turned out that this molecule was being produced by the queen and detected by the workers. Brain scans showed that it activated the areas of worker brains associated with smell. When the workers could smell the molecule, it was acting as a form of contraception, preventing females from becoming sexually active and reproducing. If a colony loses a queen, ordinarily other females fight amongst themselves to determine who will take up the mantle and begin breeding.

5:52However, when researchers added isopropyl mirastate to a colony that had recently become queenless, everything stayed exceptionally peaceful. There was no fighting and the workers acted as if their queen was still there. That is, until the researchers stopped adding the chemical. That's when the fighting broke out, with high-ranking females competing to become the new queen. Amid all the naked mole rat infighting, it might be that whoever gets pregnant first is then able to suppress reproduction in everyone else. So it's actually quite a nice mechanism to make sure that once someone's pregnant and they're making isopropyl merosate, they're going to be the queen and nobody else can be the queen.

6:34Finding this chemical cue is a big deal in the naked mole rat community. Melissa, who you heard from at the start and who wasn't involved in this research, is excited about the new finding. This is like the holy grail for naked mole hat research. For those of us interested in the social structure, we figure out how this reproductive suppression is maintained. So identifying the cue that does that is amazing. There are, of course, many questions that remain to be answered. Melissa says that working out how this mechanism is controlled at a genetic level and how it evolved are yet to be understood.

7:08But she says that this research opens a window into understanding other social species, including ourselves. It provides really, really important insight into animal communication, animal social organisation, and all of these things, and potentially allows people to consider all of these other ways that animals are sharing information with each other that we've previously overlooked, or at least underappreciated. Isopropyl mirror state isn't unique to naked mole rats, It's been found in other species as well. But now that it's been shown to have such a dramatic effect in these rodents, researchers might want to see what impact it has, if any, on other mammals.

7:46Here's Gary again. There's one study where people have measured this substance coming off the breasts of breastfeeding women. And they measured several different molecules, and isopropyl meristate was one of them. And in this study, they even took this mixture of volatiles and put it on like a Q-tip and showed that newborns would start sniffing or trying to suckle this Q-tip. But that paper was like a one-off paper, I think quite a number of years ago, and nobody really followed it up. Whether it's looking for the effects of isopropyl mirastate in other species, or digging further into the unusual lives and biology of naked mole rats, it's clear that this paper opens up a lot of pathways for future research.

8:31But at least one naked mole rat mystery is now solved, and a new line can be added to future textbooks covering the unusual social structure of these unique mammals. Sharmini Bundel there. You also heard from Mohamed Khalaf and Gary Lewin from the Max Delbruck Centre in Germany, and from Melissa Holmes from the University of Toronto in Canada. For more on that story, check out the show notes for some links. Coming up, researchers have been uncovering what goes through your head when you're presented with a new game. Right now, though, it's time for the Research Highlights with Dan Fox.

9:10Two new exoplanets have been spotted in a finding that is surprisingly insubstantial. As at around the size of Jupiter but more than 20 times lighter, the pair are some of the lowest-density planets ever discovered. Researchers used telescopes around the globe to investigate the planets, which circle a star named TOI-791. One of the planets orbits the star every 232 days and the other every 139 days. Small variations in their orbits enabled scientists to measure the planet's masses and calculate their densities, which are lower than that of candy floss. Only a few dozen of these super-puff planets are known to exist.

9:54Finding more examples of them, and especially two in the same system, can help astronomers to work out how these bizarre worlds fall. Weigh up that research in the monthly notices of the Royal Astronomical Society.

10:13In 218 BC, Roman armies were famously caught by surprise when Hannibal led his force of 46 ,000 men and 37 war elephants on a trek across the Alps. No records by contemporary historians documenting the exact route they took have survived, but now researchers have calculated the metabolic cost that the army would have incurred over several possible mountain passes from present-day France, Hannibal's starting point, to what is now northern Italy. They found that the most favourable option was a pass called Col de la Traversette and calculated that the 15-day crossing would have required 230 tons of food and supplies.

10:56And even with abundant food, maintaining body weight would have been impossible. Crossing the Alps probably caused the elephants, each weighing around 3 tons, to lose 4 % of their weight. Under normal conditions on flat terrain, African elephants have to forage for 14 hours per day just to maintain body weight, an impossibility on the rapid forced march. You don't need to cross the Alps to find that research. It's in the proceedings of the National Academy of Sciences of the United States of America.

11:32Let's play a game. You're sat down with some friends. You've just been poured your favourite drink. There are delicious snacks everywhere. And the conversation is flowing. Hey, did you hear about the guy? No, no, you know, the guy. And then someone gets a glint in their eye and says, how about a board game? Let's imagine two scenarios. In the first, they open a lengthy rulebook and start listing off how the game works. OK, so there's 8 to 12 players. Two wizards, a maverick, the arbiter, two warriors, a corporal and a ledgerman. OK, now let's go back and try the second scenario. This time, they open a much smaller-looking rulebook and explain how it works.

12:18Okay, I shall introduce you to the mad world of Snap. Okay, we each deal a card in turn, and if it matches, you yell Snap and slap your hand on top and take the cards. And whoever gets all the cards wins. Which of these games sounded more fun, or more like a game you'd want to play? I know which one I would like to play, but you may have a different idea. And it turns out, when we think about brand new games we've never seen before, our brains are doing quite a lot of complicated simulations of how they might go and whether we'd enjoyed them. This has been revealed in detail in a new Nature paper, a finding that the team hopes could help them unlock more efficient AIs, as well as helping understand our human decisions better.

13:02I called up one of the authors, Katie Collins, to ask about the new study, and I started by quizzing her on her own interest in games. Ever since I've grown up, I've loved playing games. My family would host board game nights the day before Thanksgiving. We had our friends come and play games in different rooms. So I've always been very interested in games. My family are athletes across different sports and I've played sports in college specifically cross-country and tracks. I've been very interested in games in kind of the more classic sense for a long time. But as I've kind of grown in cognitive science and AI and worked very closely with Josh Teneman in particular and Tom Griffiths, started to realize how much of individual and social life could also be cast as a game, not always as fun, but also very serious games.

13:45Games are a really nice framework to think about so many aspects of what makes life so interesting. So I'm very interested in games on the philosophical and kind of mathematical framework as well. And do you have a favorite? Oh, I did love playing Sims a lot in any kind of tycoon game when I was growing up. So those are my favorite kinds of games. Anything that's open-ended that kind of forces creativity and inflexibility. I'm with you on that one. But getting back to your actual research. So games have been well studied in psychology and also computer science. Researchers tend to love to get tools to master games.

14:16They like them to train them to play Starcraft or Go or that sort of thing. So it seems like well trodden ground. What were you hoping to try and understand that maybe people haven't looked at before? There's two kind of elements that we wanted to look at that had been understudied. one, how people reason about games for the first time when they're novices, that a lot of the work on games has excitingly and awesomely been done on experts, but much less is known about how people think about a game for the first time. And that brings to the second prong, which is how people think about games. People don't just play in games or decide what actions to take, but they also decide whether a game is even worth playing at all.

14:52And that's another angle that we wanted to take in this paper, which is kind of understanding what's going on in people's heads when they're thinking about a game really for the first time, even before they play a single move in the game. And that's kind of what we started to look at in the paper. And how do you explore that? How do you understand what's going on inside someone's head? This requires very careful experimental design. We put a lot of thought into having people read a bunch of descriptions of the different games and then make judgments about what they thought the expected outcomes are going to be.

15:18Do you think player one's going to win? Do you think player two's going to win? Do you think the game's going to end in a draw? There's one group of people made those judgments and another group of people made judgments about how fun they thought the game would be. That's on the experimental design side. And then a core part of computational cognitive science then tries to understand what is going on in people's heads by building computational models that can make testable predictions that we can then compare to the human data that we end up collecting. Tell me about the games in the first place.

15:44What games are we looking at here? We're looking at variants off of classic games like Tic-Tac-Toe or Connect Four, where we start to make a few modifications off of these to turn them into new games. Because if we're trying to understand how people think about games for the first time, we need new games that they probably haven't thought about before. So we take something like Tic-Tac-Toe, which is played on a three-by-three board or knots and crosses in the UK, where you need three in a row to win, and you can win either horizontally, vertically, or diagonally. So then we would say, take three-by-three, three-in-a-row wins, and then say, how would you think about a game where it's three-by-three, two-in-a-row wins?

16:17Do you think player one would win? Do you think player two would win? Do you think it would end in a draw? What about five by five, three in a row wins? What do you think the outcomes would be there? So we made all of these kind of modifications to kind of push the boundaries of what people might have thought about before, so that might change the dynamics and strategic interactions of the game. And then how does this fit in with the computer model you've mentioned? You've got the people, they're playing the games, they're giving you their feedback on it. Is this helping build the model or is the model sort of representing what it is that they're thinking or is it sort of the two together in tandem.

16:48Josh Tenenbaum likes to call this model stimuli co-design, where usually we will kind of do three things in parallel, creating stimuli, building out our model, because our model is probably what we're using to kind of encode the hypotheses that we're making, run the model to assess what we'd expect if we ran people on this and they actually were like our model. That's kind of one angle that we do. Then we'd be collecting various pilot data to make sure that people are understanding our task, that it's sensible, and then kind of collect the full data. So it's very much kind of back and forth, doing little parts of the scientific process on your model, stimuli and pilot.

17:23So then ultimately, are you coming up with this model that sort of guesses how people are going to approach a new game? And then you see if this model matches how people do it in real life? Yes, yes. And so by doing this, what did you find? Were you able to create a model that accurately, you know, showed how people would approach a game for the first time? Yes, this is one of the things that we're extremely excited about. We have a model that we call the intuitive gamer that models people as engaging in a couple of properties when they're thinking about a new game, engaging in fast, flat, probabilistic and goal-directed mental simulation, which broken down into a few parts.

18:01The fast is that you're doing just a few simulations, more than one, less than 10, where simulation is kind of imagining how two players might actually play this game. But then the players themselves, rather than modeling them as really expert players or really silly random players just taking random actions, that those players themselves are goal directed, that they're taking action toward whatever the goal of the game is, that they're flat by being depth limited, so they're not thinking too far ahead, which is, again, in contrast to these expert models. and probabilistic that agents might assess what a good next action is to take, but not always take the exact best action.

18:36That's kind of the heart of our model. And what's exciting is that that same modeling paradigm, the intuitive gamer, captures how people make judgments about these new games and also plays in the game for the first time. And so what did this reveal to you about how people approach games? I think it reveals a few things that people are not entirely unsystematic in how they're thinking about a game. So there's systematicity in how people think about a game for the first time, which is really, really cool, because that's a really hard thing to do if you think about it from an AI perspective, that in AI, many models are trained on tons and tons and tons and tons of data to reason about just one game, or increasingly many games.

19:14There's still lots and lots of data. People can very, very flexibly make judgments and make pretty systematic judgments across each other when they're thinking about a problem or a game or system for the first time. But I think also what we see is that people are not making predictions that directly align with the game theoretic optimal prediction that folks say in economics might imagine entirely rational agents would make. So they're not well captured by classic game theory accounts when they're thinking about things for the first time. We also call this kind of intuitive theory of games that people might have, that people have some intuitions for how to reason about new games.

19:45It's not perfect, but it's also a way to make reasonable kind of guesses about how the outcomes of a game might be. And, you know, you've mentioned AI quite a bit there. In the paper, you sort of put forward that this model you've created, this intuitive gamer, could help design of future AI models. Can you unpack that a little bit for me? How might this understanding help with that? There's at least two ways that it could help. One we see is that you don't need many simulations or much compute to get close to the game theoretic optimal in many of these cases. You could imagine a system that if it has a limited token budget or otherwise limits on compute, that if you just want to make quick guesses about whether a particular path or a particular plan or particular problem is going to lead to some fruitful output, then you could imagine doing something that's kind of like the intuitive gamer, where you just do a little bit of thinking to make a quick judgment for whether that path is worth exploring.

20:39and that could be valuable for AI to also do more resource rational reasoning. That's one. Two is that if we think about systems as thought partners that might work with us, it also helps if they have a better computational understanding of people. You could imagine a system then could be a partner to people then say, oh, actually, you might think that this particular game or problem is going to have this particular output, but actually, it's not so you can kind of imagine a system that could simulate a person and that could use any of these computational cognitive models as kind of an inner loop to that process as well at the start we talked about like the sims and things like that obviously these are like tic-tac-toe sort of games or knots and crosses how much do you think needs to be done to sort of bridge that gap to see how someone approaches something much more complicated like the sims or something like that versus like a tic-tac-toe sort of game?

21:32Great question. That is going to be a very open area for research, I think, for the upcoming years, which will be really exciting. I think one bridge that we're actively working on is to also look at how people not just play in existing games, but create new games. And that then is one element that brings you just one baby step closer to something like The Sims, where Sims is interesting in that it's so open-ended and you can choose your own goals. When you're creating something new, that's also a place where you're starting to bring other ideas in being more creative, being more innovative. And I think that that's one bridge towards these much more open-ended, open-world game environments and understanding what people are doing for all the kinds of rich cognitive activity that's going on when someone's thinking about sims.

22:11And is that your hope for the future to sort of understand all of these cognitive processes that we have going on? That would be my hope for the future. I think that is a really, really big challenge that will take many people working together in many, many more years. But yes, that would be a big dream. That was Katie Collins from MIT in the US. For more on that story, check out the show notes for a link to the study. And hey, if you like games, nature has made its own. If you've ever wanted to know if you're the best person to terraform Mars, you can now give it a go in our online game. I'll put a link to that in the show notes as well.

22:45And that's all for this time. If you'd like to stay in touch with us, you can. We're on email podcast at nature.com or you can find us on social media at nature podcast. I'm Katrina Clark. And I'm Nick Patrick Howe. Thanks for listening.

23:14A colossal monster. A city in ruins. No way out. Except inside of the all-electric Toyota CHR. John Cho stars in Escape. Test drive Toyota's new all-electric family today. Toyota, let's go places.