Ugh. I did put “pureed bee heads” in a title, didn’t I? How did I not see I was titillating to get attention, not to inform? A friend called me out on my last post. I was aiming to get across the awe I’d felt reading that paper and learning the very amazing ways scientists learn new things — such as testing the effects of chemicals on neurons by isolating them from the animal and measuring electrical activity. Amazing! My friend, however, saw a crass post encouraging a simplistic view of the complexities of research (especially the last line). By focusing on the gross-out aspect, I was encouraging a kind of science conversation that can’t talk about how and why we do research in favor of “shiny!” or “gross!”

So I screwed up (a bit). But the bonus is that now I have to write about research ethics. This post looks at animal research ethics by example: here’s a study and what they did and whether or not it seems ethical to me. For now, I’m going to avoid well-known historical examples: there are numerous cases of inappropriate methods on humans and other animals as well as good discussions for what went wrong. I’m also going to avoid (in this post) looking at ethics of research on humans in favor of focusing on how we treat animals.

Case: Does killing a small number of bees to study imidacloprid seem ethical?

First up, let’s take the research in the last post. One guiding principle in research ethics is that more intrusive or harmful methods need to be justified with strong societal benefits. In this case, the researchers wanted to better understand how the insecticide imidacloprid affects bees. By the time they were doing their study, the insecticide was already being used widely and there were concerns that it might be more dangerous to bees than other insects. Our agriculture is dependent on insecticides to maintain yields in the face of pest species so we (in general) need to allow their use. However, we must test them: instead of intentionally harming a small number of research animals and any target pests, we would be exposing all animals (including ourselves) to unknown risks. Research on insects and other animals to determine the effects of pesticides thus seems generally ethical.

However, just because we have a strong societal benefit to doing this research, still doesn’t mean it’s okay to kill these bees. Is this the only way we can figure out this question? In this case, the researchers are trying to understand how the chemical (and some other closely related chemicals) affect bee nervous systems which could help us mitigate harm to bees in the field. While I’m not an expert, it seems fairly hard to figure out the exact mode of action in the bee nervous system without isolating parts liked neurons and testing them. I can imagine that someday we might be able to grow bee neurons that never came from actual living bees and use those for research, we don’t have that today (so far as I know).

Finally, while bees are not exactly like us in their capacity to feel pain, it fits my sense of ethics that we should still minimize the harm we impose. In this study, the researchers used bees that were raised under proper conditions. The bees that were to be killed for my gruesome headline were frozen using dry ice which is, I’m given to understand, likely a low pain way for a bee to die. However, we really don’t know that much about invertebrates and pain. Moreover, while vertebrate research subjects (mammals, birds, etc.) have long had some legal protections on what is permitted to be done to them for research (and those protections have grown stronger over time), invertebrates have not historically been protected though it’s starting to change. That should help encourage more research to find ways to do research more ethically.

Overall, this research seems fairly useful to society, there don’t seem to be practical alternative methods and the bees were cared for in a humane way as far as we know. I’m comfortable saying these research methods were ethical.

Case: Was raising cancer-prone rats to test the effects of glyphosate-tolerant corn ethical?

You may remember a kerfluffle last fall when a study by a group of French researchers claimed that transgenic, glyphosate-tolerant corn (and glyphosate itself) caused cancer in a long-term feeding study. The study was widely criticized for poor experimental methods that likely made the results useless, as well as breathtaking media manipulation. Was it ethical to raise those rats and (ultimately) euthanize them for this study? Some were questioning whether the study was conducted in an ethical fashion right after it came out. Several of the many responses by other scientists specifically questioned the ethics of the study.

The first point to bring up is related to the criticized study design. In this study, the researchers raised animals that are prone to tumors well into their old age, then euthanized them for various samples. If we’re going to raise animals in captivity and then kill them, we should be sure we’re at least producing strong results. A widely criticized aspect of their study design was that they only used one control group for nine experimental groups per sex with variation in what the animals were fed (different amounts of GMO corn and amounts of glyphosate). Since each group had ten rats, that means there are 90 animals getting experimental treatments and only ten control animals to compare. Simply rolling dice will show you how easy it is for one or more of the nine groups of cancer-prone rats to show problems with an apparently more healthy control group. Nothing in the paper (or subsequent responses) have changed my mind that the study simply had poor design of experimental groups. If they couldn’t use more animals (fewer animals killed is better), they could have reduced the number of different treatment groups. Instead of having groups of rats each getting 11%, 22% and 33% of the GMO corn, they could have just done one percentage. That would have allowed more animals to be raised as controls, improving the study quality and ethics.

The second point was raised by several animal research ethicists. The Sprague-Dawley rat line used in this study are very prone to tumors: up to 80% or more will develop tumors over the period these rats lived, regardless of treatments. This raises an important ethical consideration: large tumors can be painful, cause skin lesions and impair movement of the subject animals, requiring that we do something to make them more comfortable. Sadly, the main option is often euthanasia and animal welfare guidelines for research purpose require that animals that are in significant pain be euthanized if they cannot be cured. The final study as published included photographs of some of the rats in the study, showing significant tumors. Many scientists who commented on the ethical issues noted that the photographs showed that the rats were not euthanized when they should have been. For example, one group of research pathologists wrote:

As most members of the ESTP are veterinarians, we were shocked by the photographs of whole body animals bearing very large tumors. When looking at the lesions, we believe those animals should have been euthanized much earlier as imposed by the European legislation on laboratory animal protection (http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri = OJ:L:2010:276:0033:0079:EN:PDF).

A similar response from a French group:

Last but not least, we were shocked at reading the ethical rules followed for euthanasia (“25% body weight loss, tumors over 25% body weight…” leading to euthanasia; Anatomopathology, §2.5) and at looking at Fig. 3J–L: the size of the tumors, with skin erosions and ulcerations, having certainly an impact on movement, feeding and pain, is unacceptable under well-known guidelines (Workman et al. 1998). This should have led to a much earlier euthanasia with respect to ethical humane concerns and casts doubts about the “careful monitoring” (Anatomopathological observations, §3.2) of animals. No argument, apparently to leave tumors develop as much as possible, should have prevailed. Again this demonstrates a lack of understanding of animal physiology and ethics, and a lack of supervision by the Ethical Committee and by a site veterinarian (“vétérinaire sanitaire”, a function mandatory under French law, see Article R203-1 5°). We are surprised that these major ethical issues were not clarified during the review that the paper underwent before approval for publication.

I don’t think this study was ethical. While there is public value in doing long-term studies on the effects of new foods, such studies must be done in an ethical manner. If this study had used proper experimental groups and the animals cared for appropriately (including appropriate euthanasia), I would probably have considered it ethical. The research team responded to some of the ethical concerns published in response to their original paper. I find their responses unconvincing and only increase my belief that this study was not run in an ethical manner.

Case: Is it right to capture and kill around 4% of the bats living in a cave to study a rare, but deadly, human disease?

Finally, I want to tell you about a borderline case where I’m just not sure. A major focus of the book Spillover by David Quammen are the behavior of possible reservoir species. Reservoir species are non-human animals who naturally carry an infectious organism (it may or may not cause significant disease in the animal) that can be transmitted to humans and ultimately causes disease. The rabies virus has a natural reservoir in bats which turn out to be very common reservoirs for diseases we get from animals.

In the book, Quammen describes cases where European or American tourists visited a Ugandan cave (called Python Cave — you can guess the attraction), went back home and subsequently became very ill or died. The disease they contracted is called Marburg virus (which is related to Ebola) and it causes significant, life-threatening illnesses. The likely reservoir for Marburg virus are bats but little is known about how Marburg virus infects bat populations Spillover is well-referenced, so I was able to look up one study on the bats in Python Cave.

The methods in this study are straightforward: go to the cave, collect a certain number of bats, and run tests on samples to try to detect Marburg virus in different tissues. However, these tests included taking samples (liver and spleen) that require killing the collected bats. Unlike the previous study, the authors are very forthright about the protocols they are conforming to and I have no reason to think the animals were mistreated thru improper techniques or inattention. All due care seems to have been taken.

However, the researchers collected 1,622 bats out of an estimated 40,000 bats living in this cave. That’s about 4% of them. That’s a lot of bats to kill. Was it absolutely necessary to collect so many bats? In order to get their exact results, it seems they did. The overall rates of infection they found were pretty low: out of those 1,622 bats, they only found 40 that seemed actively infected. The researchers were also able to identify different strains of the virus. This allowed them to show that the bats (or the virus at least) travel further than expected, including populations in Gabon. Further, the wide variety of strains in this single population of bats supports the idea that bats are the long-term reservoir of the virus. If they had collected fewer bats they might not have gotten a large enough sample to do this analysis.

But the question sticks in my head: could they have done it another way? Did they have to kill so many bats, who definitely feel pain and know when they are being hurt? Even more, Marburg is a relatively rare disease in humans. While it has a very high fatality rate (up to 88%), the number of known human fatalities is less than 400 cases since the 1960s. The earliest described cases are in Europeans handling infected primates with inappropriate safety protocols: those problems are largely fixed. Most of the ongoing cases are in people who work in caves (in generally awful conditions) in sub-Saharan Africa. Before this study, it was already known that the primary way humans are getting infected is by close contact with bats in caves. The solution would be for people either to not work in bat caves or to wear safety equipment. Knowing that Marburg virus is transmitted further than expected or having greater assurance that bats are a reservoir doesn’t help solve the social and economic problems that result in humans working without safety equipment in bat caves. That makes it harder for me to support killing so many animals for research.

But, I come back to what they are finding out. It’s really pretty interesting (and that paper is surprisingly readable). Maybe knowing more about how Marburg virus is transmitted in bats, including the different varieties of it, will eventually help us figure out effective treatments. But I’m conflicted. 1,622 bats dead is a lot of dead animals.

claimtoken-5132a95215847

---

Updated 2013/03/08: The link to critical responses to the GMO corn feeding study was fixed.

Updated: I’ve written a follow-up on research ethics to make up for some problems in this post.

Yes, that title does in fact include the phrase “pureed bee heads”. While on vacation, I was doing some reading on basic stuff related to neonicotinoids, my favorite pesticide class¹. One paper that came up was a 2001 toxicity study done by researchers at Bayer AG on the effects of imidacloprid on honey bees. The results are not terribly surprising to anyone who knows anything about the (acute) effects of neonicotinoids on bees (neonics kill bees at high doses, make them behave funny at lower doses). Even though the results are unsurprising, the methods were interesting to me. Partly this is because I just don’t know that much about the field. But bonus for you is that you get to learn about interesting (but gruesome!) methods for find stuff out.

When I was reading that book on pesticide metabolism, there were a lot of passing methods to interesting methods. One method involved taking rabbit or other animal skin and spreading it thin to test skin absorption of a pesticide. A bit “ewww”, right? But, neat that they can figure things out using it, right? In this study², two of the things they wanted to study were if proteins in the brain bound to imidacloprid or chemicals that are by-products of organism reactions with imidacloprid (called metabolites) and if neurons in the presence of those chemicals reacted differently to electrical signals. To study this, they cut off bee heads and pureed them.

No, really:

For biochemical studies, worker honeybees were carefully collected from hive combs (collection site: Burscheid, Germany) and directly frozen using dry ice. Bee heads were then separated from other body parts by vigorous shaking and recovered by sieving. The heads were then frozen at -40°C (usually not longer than 6 weeks) until use.

…

Bee heads weighing 10g were homogenized in 200 ml ice-cold 0.1 M [molar] potassium phosphate buffer, pH 7.4 containing 95mM sucrose using a motor-driven Ultra Turrax blender. The homogenate was then centrifuged for 10 min at 1200*g* and the resulting supernatant was filtered through five layers of cheesecloth and then used without prior purification.

In short, they froze bees, separated their heads by shaking and tossing thru (probably) a small mesh, then later pureed them using a fancy science blender device and mixed it with sugar solution and other chemicals. Then they did some tests to see if anything in there reacted with imidacloprid or with common metabolites of it. In a related experiment they separated neurons out of those bee heads and ran electrical currents thru them when they were floating in solutions containing different amounts of chemicals that might interfere with electrical activity.

This is gruesome. While doing more population based studies is useful — for example, feed bees the chemicals of interest and see what happens over time — it’s also necessary to find out exactly what is going on at the physical level rather than just correlations. In this case, they are trying to see if and how imidacloprid affects bees in the brain by directly extracting the relevant parts and doing experiments on them. One result they got which would be much harder to tease out using what I would think as an obvious way to test toxicity (feed the chemical to bees, see which ones die), is that there are different forms of the same brain receptors that are reacting to imidacloprid and to different degrees. To do that, they had to puree bee heads. No doubt these are obvious methods to someone in the field, but to me they are gruesome (but neat!) science methods³.

Gruesome Science Methods might become a series. We’ll see. Pesticide studies definitely lend themselves to it.

---

A recent column in Nature argues for evidence-based policy on the subject of neonicotinoid pesticides and their effects on bees and other pollinators. Huzzah! I agree. More evidence in policy! But just because we want politicians to make decisions based on evidence, that doesn’t mean public opinion and values — which are partially driven by their knowledge of the issues — doesn’t also matter. So I was dismayed to see the piece end with:

You can’t switch off the lies and exaggeration. But don’t worry about them. When I saw the exaggerated pollinator-decline claim attributed to me in The Guardian I did not seek to correct it, because the correct information, with references, will go into a forthcoming parliamentary-committee report. Unlike stories in the press, that report will definitely be read by officials who advise the politicians who, for the United Kingdom at least, make the final decision. And because of such reports, and a recent risk assessment from the European Food Safety Authority, we can be fairly sure that the decision on whether to restrict neonicotinoid use in Europe will not be made on the basis of avoiding 20% yield losses in crops, or saving the world’s bees from extinction.

Let me reword this: “Don’t worry about the public being incredibly misinformed about the real risks of neonicotinoids and their effects on bees. The politicians will do the right thing because I’ve made sure our scientific reports are accurate.”

My reaction? Your ivory tower is showing.

Let’s assume the politicians listen to your science and not the public. Likely they won’t outright ban all neonicotinoids because there are clearly some reasonably safe uses of them. Let’s next assume a majority of citizens take the Guardian’s fear-mongering exaggerations at face value (and various activist groups continue to exploit those fears). To the public, it appears they have just been disenfranchised because they don’t understand why neonicotinoids aren’t banned. After all, the media is telling them that “the scientists” think neonics are causing huge bee declines! Why aren’t the politicians listening?

Let’s instead assume various politicians listen to different lobbying groups, including some scientists and their reports. Some lobbying groups are arguing that banning or restricting neonicotinoid pesticide use will cause huge crop yield declines. Others are arguing that neonics will starve us all by killing the pollinators. Some compromise will probably be reached, but it’s not going to use scientific evidence as much as you would want.

Public opinion and knowledge matter. Consider climate change. No one would say it doesn’t matter that the media (depending on outlet) either exaggerates or denies climate change. It does matter because it shapes public opinion which in turn shapes what is politically possible. If you want policy made based on scientific evidence, then you’d better step out of your ivory tower.

---

Thanks to bug_girl and cotesia1 for retweeting this link to me today.

Without pesticides¹, farmers risk not growing enough to recoup their costs or growing crops that are too blemished to sell at market². But by using pesticides, farmers take on other risks: many are hazardous to those applying them, they can kill beneficial “non-target” bugs or plants³, runoff can harm ecosystems⁴ and long-term over-use of pesticides can make convenient and safe methods of control less effective.

So how do farmers balance these competing risks? One method is to predict whether or when a pest will be a problem. Farmers already do this for many crops (and various practices under the name “integrated pest management” usually require trying to predict pressure before using pesticides). But several common ways farmers control pests in crops like corn require up-front decisions because the pesticide is in the plant. This post was intended to jot down a sci-fi idea I’ve had stewing with no story to go with it. It turns out that idea isn’t as science fiction as I thought!

How do farmers predict pests?

If a field is not going to have enough pest activity to cause economic losses, there’s usually no reason to risk using a pesticide. But a farmer is going to want to be sure before making that decision. As an example, strawberries are often affected by molds that (as you can imagine) make them unsellable. Farmers usually control this by applying fungicides (pesticides that kill fungi usually yeasts and molds) repeatedly during the growing season. It’s a tricky balance. Apply too often and a farmer is wasting money. Don’t apply at the right time and she might lose a lot of berries (also, most shouldn’t be applied too close to harvest). Kevin Folta posted an example in Florida of using prediction to decrease application of fungicides. The mold is only likely to cause problems under certain environmental conditions, so they’ve built a network that keeps track of weather and other data and predicts when the fungicides are actually needed. The result is less pesticide being applied — good for workers, the environment and consumers.

In the case of crops like corn⁵, weather isn’t necessarily how they decide whether to use a pesticide. Western Corn Rootworm (Diabrotica virgifera virgifera) is one of the most economically important pests on corn crops in the United States and Europe. This insect (like many others) goes through several phases of life. It starts in the ground as an egg laid the previous summer which hatches in a larval form the following spring or early summer, feeding on roots of young corn plants (hence the name). It becomes an adult after it pupates in late summer at which point it mates and the females lay eggs. The adult beetles are usually not very damaging to corn crops. The larvae however feed on the roots. Smaller amounts of damage make it hard for the corn to take up nutrients and decreases yield. Greater amounts of root damage and the corn plant might fall over (or “lodge”) — and yields no corn at all. For most corn growers, they need a certain amount of yield to be economically viable so some pests are tolerable, but above a certain amount and large swaths of the field will be damaged to the point of economic loss. But the lifecycle of corn rootworms allows a farmer to predict when pest pressure will be high! Integrated pest management (IPM) practices recommend capturing the adult beetles the previous season to estimate the likely numbers of larvae the next season⁶.

Currently, the most popular way to control these rootworms are transgenic Bt varieties. There are several types of Bt toxins (called “Cry” proteins) that affect rootworms and different varieties of corn have genetic traits to express one or more of them. Under current EPA requirements, a farmer has to plant at least part of his field in non-Bt corn. The part of the field planted in non-Bt corn is called a refuge. The idea is that if any larvae survive eating the Bt toxins (and maybe have the ability to survive those toxins) they will breed with ones that lived off the non-Bt corn in the refuge. Hopefully this interbreeding makes sure that the pest insects as a group don’t have resistance. But there’s growing evidence that Bt resistance is developing⁷ and also that refuge compliance is lower than expected⁸. Moreover, some farmers are choosing to plant Bt varieties every season, regardless of whether or not they are likely to have a lot of rootworms that season⁸! The ag scientists are working on better ways to manage Bt crops including the idea of “refuge in a bag” which intermixes a percentage of non-Bt seed with Bt seed. The hope is farmers will find it easier to comply with refuge requirements (though it’s not clear how well it will hold off resistance — I have a post coming up on this). But it still doesn’t fix the problem of using a control method (in this case, Bt toxins created by the plant) when it’s not needed. Farmers rightly see this as a form of low cost insurance since if they chose not to use it and don’t really need it, it doesn’t cost much.

Triggered Expression: Science Fiction … or Not?

I had an idea a long time ago for a science fiction story that we could make seeds that have multiple useful traits. In the story idea (I never came up with a decent plot) a farmer would be able to “dial-up” his desired seed traits for the season. Punch some buttons and sometime later she has seed ready for planting, customized to her field conditions. Awesome, right? Of course, as I learned more about how seed breeding works, I realized that any particular combination of traits requires someone to have bred them into that seed at least the year before! Moreover, any one seed might require a complicated set of different parents⁹. So my story idea just wouldn’t work. Further, at least one of the traits I was imagining farmers “dialing up” were pest control ones like Bt toxins. Since Bt is still a pesticide, ideally farmers wouldn’t have to use it every year if there aren’t enough pests to bother, so obviously it shouldn’t be in the seed a farmer dials-up every year. Curses! There’s just no way to rescue this science fiction idea for increased farmer control of seed traits, is there?

The other day I realized there was an even cooler sci-fi idea (and for about five minutes I thought I was thinking of something new): what if a farmer bought seed with all the traits she might want to use but then used some harmless chemical to trigger the trait working? For Bt trait, the plants would always have the genes for Bt but they would only produce the toxins if triggered by a signal. If we could figure out different signals for different traits, a farmer could plant seeds with all kinds of traits, but decide later when to use them. Science fiction idea rescued! But this idea isn’t as science fiction as I thought: about five minutes after I mused about this on twitter, I was pointed to some research on using plants as chemical signals to detect chemical warfare attacks. The idea is we could grow genetically engineered plants around a city or other sensitive area that would react to common chemical warfare agents and change color or appearance as a warning. This seems awfully expensive as a chemical warfare defense method, but in crop plants it seems like it might be worth it. Give farmers the insurance that they can use a particular method of control, but allow them to make the decision very late. Unsurprisingly there’s already research on this idea¹⁰ but it’s no where near commercialization.

There’s a short post on science fiction ideas in agriculture that aren’t as sci-fi as I thought. The future is always coming sooner than I expect.

---