Why do bats carry so many dread diseases?

Ebola, SARS, rabies... Bats are replete with lethal pathogens. But why?
03 April 2020
Presented by Chris Smith
Production by Chris Smith.


A bat in a tree


This month, why screening at airports for Covid19 is unlikely to work, how flight forced bat viruses to become virulent, MRI scans of throat singers reveals how they produce multiple sounds at the same time, and the role that DNA does and does not play in education... 

In this episode

Travellers at an airport

00:42 - Screening for disease spread

Why questionnaires and thermometers at airports miss most cases of travelers with diseases...

Screening for disease spread
Katie Gostic, University of Chicago

When disease outbreaks - like the one the world’s in the grip of at the moment - occur, we put in place screening measures to try to prevent people from spreading the problem. At airports, this can include temperature monitoring and questioning about risky symptoms. But how effective is this? As she explains to Chris Smith, Katie Gostic, at the University of Chicago, the answer is quite alarming...

Katie - We were trying to assess how effective traveler screening for this coronavirus might be. That is sort of what fraction of infected people would we expect to be caught if they were screened for fever or exposure risk as they pass through an airport. And we came at this knowing that this sort of traveler screening historically has not been particularly effective for other pathogens that have emerged in the last 20 to 50 years. But we were trying to get a better estimate for how it might play out for this coronavirus specifically.

Chris - What sorts of screening are - and were - people doing?

Katie - So the typical screening approach involves a symptom screen, which basically means that your temperature is taken, and if you have obvious respiratory difficulties or a cough that would be noted by whoever's doing the screening. And then there's usually also a questionnaire that asks you things about your risk factors. So, for example, "have you been in an area where we know that there's a coronavirus outbreak," would be one of the biggest risk factors for this pathogen.

Chris - And when you sought to look into this, actually what was your approach?

Katie - The vast majority of screening does consist of this sort of fever and symptoms screening, plus a questionnaire. We came at this using basic probabilistic assumptions. The first thing we did was we asked sort of what's the probability that any single individual passing through screening would be caught. To do that, you basically need to come up with a probability that an infected person is detected in the symptom screening unit and then another probability that they're detected in the questionnaire based screening. We can come up with those probabilities based on what we know about the biology of the virus. So the biggest question that you have to ask first is, is this hypothetical individual that's passing through screening showing symptoms yet or are they still early in the course of infection and haven't yet progressed through the incubation period? And then the second question is, does this person realize that they might have been exposed to the virus? And we basically estimate that people are unlikely to realize they have been exposed to coronavirus based on the fact that a lot of cases were showing up early who weren't able to report a clear source of exposure. And then in terms of the probability of having started to show symptoms, we can estimate that based on how long the incubation period of the virus is known to be.

Chris - And what data do you then work with? So those are your assumptions, that's how you're approaching it. So do you then take, what, real world travel data, and ask, applied to this, with these assumptions, how many would we find?

Katie - Mostly what we do is we make assumptions and we sort of predict using those assumptions what fraction of cases we would expect to be detected. But we can test how well it's working by comparing it to real world data. Basically to do that, you know at what time different countries had implemented airport screening and we assume that that if infected people passed through that screening and then later developed a severe case, they would eventually appear in the data, basically. Like once someone passes through screening, if they reached their destination and later end up in the hospital, then we know that screening has failed to detect that person.

Chris - Right. And on the basis of your model, in your assumptions, how effective is this screening?

Katie - So we estimate that even in sort of the best case scenario, that screening is probably missing 50% or more of infected travellers with coronavirus. And this is not necessarily an issue with the same machinery used to do screening like thermal scanners or human compliance and implementation, it's more just a biological reality of the way this infection progresses. So we estimate that the vast majority of people who would feel well enough to travel probably don't show symptoms at the time that they pass through screening. And we know that the incubation period for this pathogen is pretty long. And we know that there are some people who are completely asymptomatic. And so basically those people who either don't yet show symptoms or might never show clear symptoms are just undetectable. And we know also for this coronavirus that risk factors are pretty nonspecific. Like essentially anyone who's passed through an area that's known to have an epidemic would be considered at risk. And so the risk questionnaire isn't extraordinarily helpful for detecting cases either.

Chris - Gosh, 50% that's enormous.

Katie - Yes.

Chris - So what are the implications of that?

Katie - Well, we think that this is one of the big reasons that the virus was able to spread so easily at first and one of the reasons that there was such a big delay in realizing that there was community transmission going on in places like Italy or Iran or Seattle before the first cases were detected.

Chris - We are where we are now. We've got a quarter of the world's population as it stands at the moment in late March, currently experiencing some kind of lockdown or restriction on their movements to try to break the transmission chain. What can we take away from what you've found here that obviously it may be a bit late for this situation, but so that we don't get SARS coronavirus Mark three in the future?

Katie - I think that's a difficult question. One obvious extreme end of the potential solutions would just be to shut down air travel networks as soon as we detect the next emerging infectious disease. That obviously comes with serious economic implications. And to be honest with you, even when we detect evidence that a new virus is emerging in humans, if it has an animal origin, it's actually surprisingly uncommon for that virus to transmit well from human to humans. And so I think shutting down global air travel networks every time could lead to a lot of false alarms and could really substantially damage the economy. But on the other hand, what we've seen here is that it's really difficult to screen for emerging infectious diseases at airports and that air travel is a really important driver of global spread. And so usually the way that public health agencies respond to these scenarios is that they know from the outset that traveller screening is going to be leaky. And so there's a lot of investment in more on the ground contact tracing - which basically means that public health professionals, as soon as the case is detected, go out and try to round up anyone that the first imported case might've been in contact with and infected before those people can start new chains of transmission. People in my field often talk about new epidemics as being like fires and these first imported cases like throwing sparks on the ground, so just because you have a couple of sparks on the ground is not a guarantee that one of those sparks is going to start a larger blaze. And contact tracing is a really good way to deal with the fact that we often can't prevent every spark from coming in, but we can do our best to prevent a big fire from starting.

A bat in a tree

08:21 - Why do bats carry deadly viruses?

From Ebola to Marburg, and coronavirus to rabies, bats carry some of the world's worst viruses. Why?

Why do bats carry deadly viruses?
Cara Brook, University of California, Berkeley

Apart from the fear that they summon up, what links Ebola and Marburg viruses, rabies, SARS, and even the new pandemic coronavirus we currently have circulating? The answer is that they’re all linked to bats, which seem to tolerate carrying these agents remarkably well. When they get into us though, it’s a different matter. So why are bats such a rich source of zoonotic viruses, and why are their infectious payloads so aggressive? The answer, it turns out, may be because bats are much more resilient, owing to the adaptations they’ve had to make in becoming airborne, and this means their viruses have had to become correspondingly more agile to compensate as well. Speaking with Chris Smith, Cara Brook…

Cara - My lab group published a paper last year where we were able to identify that bats are actually reservoirs for the world's most virulent viruses. But virulence means damage to the host. And in this case we mean virulence to a zoonotic host, so a human. So a zoonotic virus is a virus that spills over from an animal to a human. And in this case we collected case fatality rates of mammalian viruses that have spilled over to humans. And we were able to determine that those that come from bats cause higher case fatality rates than those derived from any other mammal. And so we wanted to understand why.

Chris - We're talking about things like Ebola here, aren't we? Which if it's in a bat, the bat doesn't fair too badly. When that gets into a person, you can have up to an 80% mortality rate.

Cara - Exactly. So with virulence it's important to define the host that you're talking about and what we've observed is that bats host these viruses as natural reservoirs and don't experience ostensible morbidity or mortality themselves. And they're reservoirs for several of the world's most virulent viruses, including rabies and related lists of viruses, Ebola and Marburg filoviruses, hendra and nipah, SARS, MERS, and most recently SARS-CoV-2.

Chris - Couple of questions immediately spring to mind. First question is, why have the bats got these really nasty viruses in the first place? And B, why don't they succumb?

Cara - Bats are the only flying mammal and we think that their resilience to viral infection is related to the physiology of flight. Flight is actually more metabolically expensive than any form of terrestrial locomotion. So a rodent at full speed running will raise its baseline metabolic rate up to seven fold when at full speed, and a human will raise it two or three fold, but a bat will raise its baseline metabolic rate up to 15 fold. So double that of all other mammals. And we would predict based on this high metabolic activity that they would be extremely short lived because normal metabolism results in the accumulation of oxygen free radicals, which then cause oxidative stress. But in spite of these really high metabolic rates, bats are actually the longest lived for their body size of any mammalian taxon. The oldest known bat, the Brandt's bat, it has been documented living up to 40 years in the wild. And so really what we think is that for flight to have ever become physiologically possible to begin with, bats had to first evolve really efficient mechanisms of oxidative damage mitigation. And actually viruses in a cell will incur oxidative damage on that cell and also recruit highly inflammatory immune cells to the site of infection, which causes further oxidative damage. So we think that bats evolved these cellular mechanisms that then enabled flight but had cascading consequences on their longevity and then on their resilience to viral infection.

Chris - So neat hypothesis. What did you actually do to test it though?

Cara - We did viral infection experiments on a series of bat cell lines and then estimated rates of viral spread. So in particular it's been shown that a couple of bat species have constituently expressed interferon alpha. And what that means is that they have a perpetually primed antiviral signalling system. So normally when a virus invades a cell, a mammalian cell will release this protein interferon that signals to neighboring cells that there's the presence of an infection and they need to induce this suite of antiviral responses. But in bats, or some bats anyway, it appears that this defense is perpetually primed. And so basically we grew three different cell lines, bat cells that were either constituently or induced in their interferon response, infected them with a suite of different viruses, and then used a monkey cell line as a control that's deficient in its interferon pathway. And then we modeled the antiviral response of the cell. And then when we estimated the rate of viral spread, we found that the antiviral response was matched by a higher within-host spread rate in the case of the virus. And so we hypothesized that viruses would evolve evolutionary optima in bats, so if a virus coevolves in a bat host, it's going, its baseline within-host spread rate is going to be fine tuned to match that of a bat, but then when it spills over or emerges into a non-bat, we would expect that that organism would experience extensive pathology.

Chris - In other words, the virus has tooled up in order to combat the bat resilience, which means that when it finds itself in a different host, for instance, a human where there isn't that enormous amount of pushback, it's got far more virulence than it needs to have.

Cara - Yeah, it's spreading faster and that then causes virulence. Usually viruses evolve to optimize or balance transmission and virulence. It's often asked why are pathogens pathological? Why do they cause damage to their host? Because it doesn't advantage the virus in any way to kill its host because the virus wants the host to sustain itself as long as possible so that it can continue to infect new hosts. So a virus will always evolve to maximize its rate of between-host transmission. If the virus kills its host too quickly, that's not going to optimize that between-host infection rate.

Chris - And just to finish, how does your current set of findings pertain to what we're seeing with this new coronavirus from China?

Cara - Covid 19 as a pathogen itself is less virulent than many of the previously observed bat viruses, but balancing that trade off it appears to be more transmissible. This virus has adapted and evolved to be transmissible in a human population and typically we would make the prediction that becoming more transmissible means that it's going to be less virulent. Many of these other bat-born pathogens, Ebola as a great example, cause human fatality much too quickly in order to transmit globally in the way that we've seen with this virus that has caused this recent outbreak.

Tuvan throat singer performing

16:02 - Tuvan throat singer gets in an MRI

Imaging reveals how the human vocal tract can sing two notes at the same time...

Tuvan throat singer gets in an MRI
Chandan Narayan, York University, Canada

Speaking with Chris Smith, York University, Canada's Chandan Narayan explains how he persuaded a Tuvan throat singer to perform inside an MRI scanner, revealing how the vocal tract's in their performers can create two sounds at the same time...

Chandan - My name is Chandan Narayan. I am a associate professor of linguistics at York University in Toronto, Canada. What we're hearing is a Tuvan throat singer. He's performing a style of singing called Sigit, characterized by a really amazing phenomenon, I think. There is a deep low frequency rumble, which is followed by a very high frequency whistle-like tone. He seems to be able to control both of these sounds independently and it's unique to the Tuvan people, the Tuvan culture.

Chandan - So that's the whistle that you hear on top of this growling low-frequency pitch. He's doing something that we all do when we speak vowels or when we sing: we produce a set of organized frequencies from our mouth. Now those organized frequencies are called formants. About two or three of these determine what our vowel characteristic is. So if we're saying an ah, or an oo or an ee, he's doing the exact same thing, but he is merging two of these formants to produce what's called a super-formant or a focused state. That's the whistle that we hear. We don't hear this in a normal speech because those two formants are separate. So he's done something with his mouth that drives these two formants together to produce this tone, this whistle-like tone. And he makes further refinements with his tongue to shift the locus of this tone so as to produce a melody.

Chris - What was the outstanding question then?

Chandan - We didn't know what he was doing until we did this study. So the question initially was how are they doing it?

Chris - Presumably these people learn this from other people. They're not born with the innate ability to do it, they won't know exactly what they're doing to get it to sound right. They just respond to positive feedback both in themselves and from others. And that's how they know how to do it?

Chandan - Correct

Chris - But how did you then approach this to actually work out what they were doing?

Chandan - We started off with pretty high quality recordings that we made in a sound booth at the university. And then we followed that up with MRI imaging. We had a singer from this group that was touring in Canada. We were able to get them to come to the university and produce these sounds in the MRI scanner. And so we took those images and kind of reverse engineered a model of their throats and their mouths, computing cross-sectional areas of the throat. So if you think of the throat as a giant tube with pinches all throughout the tube, those, those particular pinches affect the nature of the resulting sound. And so the MRI gave us a good roadmap as to where those pinches might be occurring. And then we gave that data to Brad Story, who is an expert in vocal track modelling. And he was able to produce a model, a computational model of the vocal tract and predict the types of sounds that the Tuvans were producing based on the shape of their vocal tract.

Chris - Based on what you now know from these observations, if we return to that sound we heard at the beginning, can you now explain in terms of the anatomy, how they're doing each of those two different components?

Chandan - Yes. The singers raise the base of their tongue towards the uvula, which is in the back of the throat. And what that does is it produces a very focused bundle of energy at a specific frequency. Now that bundle of energy at that frequency is then shifted by making a constriction at the front of the mouth. So that energy is what we hear as a whistle. And that whistle's precise tune is controlled by making a constriction at the front of the mouth, near the teeth with the tip of the tongue. So there are two constrictions being made: one at the back of the mouth and one at the front of the mouth. The back constriction produces this whistle and the front constriction plays the melody of that whistle.

Chris - And is that because by constricting the front and the back, you actually are creating effectively a resonant chamber out of the mouth that's got two different sizes: one which is the vocal folds up to the back of the tongue, and then another one, which is the bit to the front of the mouth as well?

Chandan - Correct. So when you make a constriction anywhere along this tube, you're actually coupling two tubes, one tube behind the constriction and one tube in front of the construction. So in the case of the vocal tract there'll be a resonating chamber behind the base of the tongue, and a resonating chamber between the back of the tongue and the tip of the tongue , and a small resonating chamber between the tip of the tongue and the end of your lips. So they are multiple resonating chambers all acting together to produce this whistle.

Chris - And to what extent, given it's so dependent on the anatomy, being able to put your tongue into those positions, is this unique to these people? Could anyone do this or is it something that they have the right head and mouth and vocal tract shaped in order to do this most effectively?

Chandan - No, there's nothing special about the anatomy of the Tuvan singers. It's a learned phenomenon. I've tried, I failed, but others have been quite successful.

Chris - Will you give us a demo?

Chandan - I'm sorry. No, I haven't had enough coffee.

Chris - It takes coffee, does it? Most people say whiskey.

Chandan - Actually, you know, this group, we, we took them out for a lunch after we did the recording and they were quite adamant that they should never have cold water. You know in North America people are really into having water with ice. So I offered them some and they were taken aback. They said that cold water would ruin their vocal folds and they would never be able to make money again!

Children working at school

23:06 - Does DNA contribute to school performance?

How well genetic scores predict school achievements...

Does DNA contribute to school performance?
Tim Morris, University of Bristol

To what extent does your DNA contribute to what you achieve in the classroom? Speaking with Chris Smith, Tim Morris explains how he has been trying to find out…

Tim - We have lots of information on why people perform the way they do in school. There's lots of background social reasons, parental reasons, there's intelligence and personality, and there's also an increasing argument that genes are really quite important for the way we learn amongst many other outcomes.

Chris - People obviously say, you know, bright kids tend to have bright parents, so that must be genetic. Of course, there's also the explanation that if you've got bright parents, you probably grow up in a household that's more educationally rich.

Tim - Yeah, that's right. There's probably some genetic components of it. But as you say, parents who are better educated are more likely to have more books in the household. They're more likely to take their children on museum trips or help out with their homework. So these things can operate kind of through both genetic and non-genetic pathways.

Chris - So how did you try and dissect apart the two?

Tim - So in our study we looked at the DNA of a sample of children from a UK cohort study called the Avon Longitudinal Study of Parents and Children, which has collected data on almost 15,000 children since the early 1990s. They've genotyped the children, and as a result of this, we know the parts of the DNA that they carry that may be related to things like educational attainment. And so we can build these genetic scores which provide a measure of kind of known genetic liability to a certain characteristic, in this case education. These genetic scores, they are called normally distributed. So some people score very highly, some people score very low or the bulk of people kind of are in the middle. And we're looking at how the differences on these genetic scores amongst the study children, how they relate to differences in exam performance throughout their schooling.

Chris - What did you find? Is there a really strong correspondence or not?

Tim - Well we find that on average, children with a higher genetic score tend to perform better than children with a lower genetic score. This is when we look at the population of children all together. But when we're looking at individual children, when we're interested in trying to predict how well a child will perform later in education, we see that at that individual level, these scores really aren't very useful.

Chris - Do you mean as in that there are no specific markers that you could say if you've got this particular makeup, you're going to be Einstein and if you haven't, you're going to do less well? It's not as predictive as that?

Tim - No, absolutely not. So one of the interesting things with things like educational attainment is that they are characterized by lots of tiny, tiny, tiny genetic effects. So there's no such thing as a gene for education or a gene for intelligence. We're seeing that there's combinations of thousands upon thousands of individual points of DNA that are very slightly related to educational performance.

Chris - So what's the take home message then?

Tim - I think the take home message is that that genetic data is really useful for asking and answering some questions on groups of people. But it's really not very useful, certainly with things like education, for predicting individual performance. So the idea of personalized education based upon genotype - we really don't see any evidence for that whatsoever.

Chris - So if it's not genes, what is it that determines a person's aptitude to learning and becoming very educated?

Tim - It's a whole complex range of things and this is the problem with the genetic prediction is that education is an incredibly complex characteristic. It's influenced by genes, by your intelligence, by your work ethic, by your parents, social class, by your parent's education, family structure, and there's so many different things feeding into education. It's completely unsurprising that using any one of those components will not give you a reliable prediction of how well a child will do. And you know what we've seen and what a number of studies have shown is that even when you're combining all of these different parts of information, it's still incredibly difficult to predict how well someone will do. There's too many unexplained factors and too much randomness, kind of inherent noise, in how well people perform.

Chris - Is that what you expected to find or were you hoping, when you launched into this study, that what was going to emerge was this really strong genetic predictor of education and you're going to go, there you go, this is the genetic hand you need to be dealt do well in school.

Tim - No, it wasn't that surprising to me. One of my motivations was to kind of stem the push that there's been from some scientists and nonscientists, including government advisors, that genetic data will be really useful for useful in schools for predicting how well children will do. So we weren't surprised to see that this wasn't the case. If anything, it was surprising to see just how little people's genes kind of, you know, their genotype, gave to the prediction of education. It really is, on top of everything else, almost nothing.


Add a comment