You are what you eat, but your microbes may influence what you eat. At least, that’s what seems to be true in fruit flies. A new report in Current Biology found that fruit fly dietary choices are affected by the bacteria that they inherit early in life.
Most, if not all, complex animals have microbes on and in them. Fruit flies are no exception, and their eggs have a high abundance of two types of bacteria on their surface; Acetobacter and Lactobacillus, both of which are members of ubiquitous families of bacteria that also play roles in fermented food products, thanks to their ability to generate flavorful and fragrant acetic and lactic acids. These bacteria persist throughout the development of the egg into an adult fly, but can vary depending on what a fly eats.
To test the role that the microbes play on diet, the researchers, lead by Dr. Adam CN Wong, at the University of Florida, generated four groups of flies with different populations of gut bacteria; normal flies with a mix of various bacteria, flies raised with no bacteria, flies with a single Acetobacter species, and flies with a single Lactobacillus species. Next, they tested the dietary preference of each of the four groups by placing flies into a feeding area with seven meal options - three meals each with a separate Acetobacter species, two with a separate Lactobacillus species, one with a Staphylococcus species (which is not found to be associated with flies), and one meal with no bacteria.
The normal flies had a preference for food treated with Acetobacter strains, as well as some of the Lactobacillus strains, but flies inoculated with either Acetobacter or Lactobacillus had a preference for food treated with the corresponding bacterium. That is to say, flies inoculated with Acetobacter preferred food with that Acetobacter species. Moreover, flies raised with no bacteria had less of a preference for Acetobacter-treated food, and had more interest in the meal without any bacteria.
Furthermore, these dietary preferences were observable from newly emerged fly larvae, suggesting that early exposure to particular bacteria impacts diet choices in flies later on.
In addition, the researchers established that there exists an interplay between nutritional choices and the presence of certain bacteria. When the groups of flies were given another food choice test, this time with meals that had different protein-to-carbohydrate ratios, flies with no bacteria had the highest preference of all for a “balanced” meal, while Acetobacter-inoculated flies preferred a high-protein diets, whereas Lactobacillus-inoculated flies went for the higher carb diet.
This study further explores the role of gut microbiota in modulating - or even manipulating - the behavior and choices of a more complex organism, through chemical stimulus and responses. Similar findings have been observed in other organisms, and this study also establishes the early impact that bacteria inherited from the mother has on behavior in flies.
A regular round-up of recent notable articles and science pieces.
Tracing the spread of HIV in America and exonerating “Patient O”: This week, researchers report an analysis of the viral genomes from HIV/AID patients samples taken in the 1970s, tracing the emergence of the pathogen back to Haiti. This study exonerates Gaétan Dugas, a French Canadian airline steward, who contributed to an early CDC study on the epidemiology of Kaposi’s sarcoma (the development of which is now known to be a complication associated with HIV infection). Dugas was labeled as “Patient O” in the study, and was mistakenly referred to as “patient zero,” implying that he initiated and purposefully spread the disease.
(Nature, $)
further coverage:
HIV’s Patient Zero exonerated - Nature
Researchers Clear 'Patient Zero' From AIDS Origin Story - NPR
The battle for oxygen by gut pathogens: Last month, the Bäumler lab at UC Davis showed that Citrobacter rodentium, a model bacterium for studying gut pathogens, manipulates cells in the intestine, altering the microenvironment and affording a metabolic growth advantage over other microbes in the intestine. The system of host-pathogen interactions that the group has uncovered is ridiculously complex, and they did some really cool experiments to show it. I missed this article when it came out last month, but heard it recently on the This Week in Microbiology podcast the other day.
‘"The finding is important because it explains how some enteric pathogens can manipulate mammalian cells to get the oxygen they need to breathe," said Andreas Bäumler, a professor of medical microbiology and immunology at UC Davis School of Medicine and lead author of the study. "It also offers new insight into developing strategies targeting the metabolism of the intestinal lining to prevent the expansion of harmful bacteria in the gut, a situation that is exacerbated by the overuse of antibiotics.”'
(Science, $)
further coverage:
Gut pathogens thrive on body's tissue-repair mechanism - MedicalXpress
This Week in Microbiology 137: The battle for oxygen
Mitochondria inaction: A group of researchers from Canada, the US and Japan presented this week a story about the adaptation of a single-celled eukaryote to a low oxygen environment, particularly with respect to what happens to its mitochondria, an organelle which is often thought of for its ready use of oxygen, and found that key components of its electron transport chain (crack open those Bio 101 textbooks!) seem to be losing the ability to function.
Salient note from the news and views:
“Most textbooks tell us that mitochondria are ‘kidney bean’ shaped organelles with an extensively folded inner membrane and this is how most biologists would draw one if asked. This is rather surprising considering the enormous morphological diversity of the bearers of these organelles. Humans look nothing like birds or fungi (all members of the Opisthokonta supergroup), let alone like plants, diatoms or intestinal parasites such as Giardia. So, why would we assume, or even think, that all mitochondria of these organisms look the same?”
(Current Biology, $)
Blogs & Mainstream Science Journalism
Keith Robison’s article on How Genomes Enabled Proteomics
Ed Yong at the Atlantic on the use of a bacterium to curb the spread of viral diseases
“That they are so ready is a triumph of basic science—research for knowledge’s sake that isn't geared towards any practical application. Wolbachia was discovered in 1924 by scientists who were peering into the cells of another species of mosquito. They had no idea what they discovered. Other entomologists later showed that the microbe was found in some 40 percent of insect species; they knew it was incredibly common and biologically fascinating, but, again, they had no applications in mind. Even when O’Neill began his work, he couldn’t have known where it would lead.”
A small molecule specifically stalling the protein synthesis of a single gene? Turns out that’s a thing. Derek Lowe has more
Epigenetics’ growing pains, reproducibility in the age of big data and ENCODE as a WPA program for geneticists: a conversation with Dr. John Greally
Dr. John Greally is a physician and faculty member in the Department of Genetics, at the Albert Einstein College of Medicine in the Bronx, NY. His lab studies large swaths of genomic data, and tries to make sense of the complex patterns that influence how genetic information is expressed in the body, and inherited from generation to generation.
I sat down with John recently and spoke to him about his background, growing up in Ireland and becoming a physician, as well as his opinions on scientific reproducibility, the contentious field of epigenetics, and whether consortium science projects, such as the ENCODE project (which he worked on), are worthwhile ventures.
———
How did you first get into science?
It was genetic. My parents are both doctors, so we were brought up with an atmosphere of experimentation in the house, but I think that my father in particular became an early influence on the idea that you should question the status quo. When I was very young - six months old - we moved to Boston, and my parents had another child, my little brother, and he developed leukemia and died at the age of two. So my father was in Boston training as a pathology fellow, as a physician, and he was faced with the situation where he was looking down the microscope at these cells, but he didn’t know what they do, these cells that killed my brother, and he wanted to understand more about those cells. So after my brother died, my dad packed us all into the car and we drove across the country to live in Saint Paul, Minnesota, where in the University of Minnesota there was this phenomenal researcher named Bob Good, and Bob was kind of like the founder of immunology in the United States. So they were figuring out what all of these cells of the immune system are actually doing. It was an incredible era of discovery.
And so when I was about five, we moved back to Ireland, and dad became Ireland’s first immunologist. So we were brought up with dad saying, “It’s so funny about these eosinophils, they look just like the neutrophils, but they do something completely different! And why do they have acid granules?” We were just being brought up with this idea that you should and could question everything.
But I think the other interesting thing is that my mother is a clinical geneticist. And her mother, I think, was the original scientist in the family, because after World War Two, in the west of Ireland, in County Roscommon, there wasn’t a huge amount of income for farmers. So she decided she was going to get pedigree hens and breed them. She built a chicken coop and basically did all of the crosses that were needed to keep her colony alive and would sell pedigree hens to the farmers around the west of Ireland. So she was a geneticist and entrepreneur. And a very successful farmer as well.
This completely informs some of my other questions. You’re from Galway originally—
Yes, and you pronounce it correct as well which is unusual. Well done.
[We began a brief digression into the Irish language itself, some of its intricacies and idiosyncrasies, as well as John’s experience spending time in Galway. The following passage condenses some of the more salient points of that digression.]
With the Irish language, you have to use your tonsils. It’s like Russian, except less civilized. You start with the verb just to get it out of the way.
And if you’re educated in the east of Ireland, you might as well be in Britain. The amount of teaching of Irish we got was appallingly poor. So when I moved back to Galway in my teens, after having been living in the east, everyone was able to speak very beautiful, fluent Connemara dialect.
So Galway was the hometown. It’s where I was born originally. And I’m now, having graduated from NUI Galway [the National University of Ireland in Galway], serving on their governing body, called the Údarás, which is the Irish word for it. So I go home regularly. I’ve been doing it for a couple of years and I’ve just been reappointed to it for a few more years. So I get to go and see how much the university has progressed since I was there. It’s really lovely to see that. I’m also trying to help them set up a genomic infrastructure in Galway, and if that were to happen, it could be very useful transfer of knowledge from here to there.
So is the idea that they would they be looking at genetics questions related to Ireland?
Ireland is very underexplored as regards genetics. There are some great initiatives going on in the UK as regards the 100,000 Genomes Project, but while Ireland has very good basic science research in genetics - there are people doing evolution in yeast, and genetics that has to do with animal husbandry - there is not so much as regards human genetics.
People have asked how much of the West coast of Ireland would have, say, Atlantic influences from North Africa to Spain, and so on. Obviously Dublin would have had traffic back and forth with the UK.
But the other thing about Ireland is you’d say, “the Murphys are from Cork, the Sullivans are from Kerry, the O’Neills are from Tyrone or up in the north, and the O’Flahertys are from Galway.” By last names, you’re able to tell generally where people are from in the country, and it reflects the fact that up until a couple of generations ago, there just wasn’t that much mobility within Ireland. People who moved basically left the country entirely. Because why would you move five miles down the road, right? Those people are all crazy.
It’s interesting because with the 100,000 Genomes Project, that was a big consideration, right? To make sure that the people whose genealogies you were studying were people who hadn’t moved around much within the UK.
Yes, so they did a very clever thing there. In the UK, they went to rural areas and asked individuals, “were all four of your grandparents from an area within 80 km radius of where you live right now?” So they were able to find the genetic history from around the turn of the 20th century, and that definitely shows patterns which would probably be quite blurred with subsequent movement within the country.
So Ireland has massive potential as regards understanding these distributions of people within it, and understanding our history a little bit more. But we’re also quite susceptible to certain diseases. Celiac is very high rate, phenylketonuria, cystic fibrosis, and I myself am a carrier for a hemochromatosis variant. Which is almost like getting a genetic test and hearing that you’re Irish. But the other thing about Ireland that I think has to be taken into account, is the importance of understanding the genome and the variants of it that we find there, we’ve exported our alleles all over the world. There’s 40 million people within the United States who claim Irish ancestry. So if we can understand disease risks and susceptibilities, and variants and stuff like that within the Irish population, we could immediately transfer that knowledge to tens of millions of people here.
Inscription on the glasses:
“Ever tried. Ever failed. No matter.
Try again. Fail again. Fail better.”
“You could use it as an inspirational quote or slogan, but it’s Samuel Beckett, so that’s so much cooler,” he said. “But that’s lab life, you know. You try. You fail. You then try and fail better next time. That’s how we do what we do.”
So you did PhD work as well as medical work through NUI Galway.
Yes, it was a little bit ad hoc. I did an internship in Ireland and a pediatric residency out here, then went to Yale University and did my clinical genetics training. After that, I realized I wanted to dive deeply into basic science.
Francis Collins was out there talking about the human genome project, and I was passing where he used to work in Sherman Weissman’s lab, everyday to get to my lab. And Francis is banging on about how if we find the sequence, we’ll know all about human disease. Meanwhile, I was seeing patients and realizing, there’s just no way. You know, it’s much more complex than that. And I was reading this work by Hal Weintraub, in University of Washington. He was writing about how DNA was actually packaged up by proteins, and he had this theory that these proteins could be semi-conservatively associating into daughter cells and this could be a source of information. And I thought, “wow, that’s really cool, I need to study something like that.”
So I started working on genomic imprinting. And since I was working on that, I thought I really should try and get some formal degree out of it. So I approached my university back in Ireland and asked if they did PhD programs for people in my situation, and they said “we could, because you got honors in microbiology and immunology during medicine, which gives you the ability to go for the PhD. You just need to have a relationship with a supervisor in Galway and work out what he or she would need you to do.” So the supervisor in Galway gave me a number of parameters for how I would need to update him on my work, including coming home every six months, which was not very onerous.
So I did the PhD through Galway while working as a postdoc at Yale. I was my own graduate student. It was kind of an odd situation.
But Yale was interesting because I was given lab space by Dave Ward who was a pioneer of FISH, but as regards the mentorship for the work that I was doing, it really ended up being Frank Ruddle, a Hox gene superstar, technology innovator in his time, and just an incredibly inspiring and thoughtful guy. Just some wonderful people up there.
It sounds like a very interesting trajectory.
Weird trajectory. I always tell people to look at the way I’ve done things and try to do it better, because I did things really awkwardly, and took more time than I should have. But you end up where you end up and it’s all for the best, really.
What would you consider to be the biggest discovery of the past 20 years in biology?
I think you’d tend to gravitate towards technological things, so Yamanaka and reprogramming would be hugely enabling, and then obviously CRISPR in more recent era.
As regards insights, I don’t know. I think we’ve been very technology driven for a while. I think we’re sort of struggling to stay on top of all the information that’s flowing in and we’re not going as deeply as we could.
Why do you think that is?
Just volume. I had set up this literature search which would give me a couple of topics that I wanted to get papers on every month. And I used to get 30. It’s now up to about 180 in each one. So 30 you can manage easily enough. You’re sitting there over breakfast, there’s three here you’d like to read. 180? I’m not going to have time to go through 180 now, so a week goes by and you haven’t kept up with the literature, then another week then another week.
And then you kind of end up relying then on Twitter for people to tell you about interesting stuff. But often that stuff is contentious. So yeah, I don’t know.
I think that the technologies are going to lead to answers. As a geneticist, I’m very interested in the insight that we had from GWAS [Genome Wide Association Studies] that said that the variants that are associated with a lot of the common diseases are in regulatory regions [quick reminder, the majority of genetic information relates to controlling how and when a gene is expressed. —SB].
Now, that’s great because it leads us to the next thing that we can hit with the technology to kind of understand what these things do. But it doesn’t tell us what happens to the cells. If your study found a hit in the CDKN2A tumor suppressor gene, which is a very positive hit for Type II diabetes, that’s great, you’ve got an association, and it’s very robust, and you know what that gene does, but you don’t know what that mutation does to cells. You don’t know what happens during embryology, or during the response to challenges. So I think there’s an awful lot to be learned.
Of course, the CAR-T stuff that’s happening in cancers is the massive big elephant in the room now, when it comes to personalized cancer treatments. Maybe the genomics stuff really doesn’t matter, and we just need to do CAR-T. But the way it’ll shake down probably is that the personalized medicine stuff will help with targets and to help us understand why some people respond to CAR-T and others not, genomics may be informative for that.
We wave our hands about big data. We have a couple of choices there. One of them is to try to train every person in medicine and science to be good at everything that they should have been good at to start with, as well as big data, so they can assess the results themselves. The other is to tell the specialists to turn the big data into something that is consumable, which is the topic of a talk that I gave recently, and with visualization, that’s the goal.
My wife is a radiologist, so I always think that the people who figured out big data in terms of intuitive representations are the radiologists. When they do a CT scan, it’s not blowing an image onto a piece of film. It’s getting 1’s and 0’s. It’s digital information that they then reconstruct and can say “I know this is the front of the abdomen.” It’s an anatomical representation.
We kind of need that for a lot of areas of big data, especially genomics in order to make some sense of all this stuff. But we’re not investing in it to any great extent. We’re investing in generating the data without really getting the artists, the data visualization people, the computational people and telling them that they are actually the people who need to bear the brunt of this work.
What’s the genetic counterpart to representing anatomical information in a CT scan?
Right now we’ve written a few grant proposals and in thinking about this, I like the idea of having a series of windows into the data. So you have some sort of screen that you’re working on. You can’t understand the effect of a DNA sequence variant for example in terms of just one property of the genome. So if you said it’s a DNA sequence variant in a coding sequence of a gene, then you need to know a lot of stuff about the gene itself; what it is likely to do, where is it expressed in the body, and what stage of development it plays a role in. You’ll ask all of those questions, and there are ways of visualizing each of those things. You’ll want to know if this variant is present across populations, and what the phenotypic characteristics of those people are. You probably want to have the ability to look at multiple facets of the information that we have collected about that individual gene, and vectors of information in terms of gene characteristics, the characteristics of sequence across population, the protein characteristics, embryology, etc. You have to sort of address each of those vectors separately, in order to try to make sense of it. Hugely challenging. I don’t think there is just one image that would allow you to dig into this.
Hopefully something more than just a linear genome browser.
The genome browser was a great tool for its time. But we’re ready for another big jump there. The moment that you start adding a lot of information to it, you’re dealing with very cumbersome representations of the data. They are doing great work as UCSC, but that’s about as good as you can get with that approach.
Why is epigenetics so often a lightning rod of controversy? Does it have something to do with the field? Something to do with disagreements between researchers? Or is it a concept that is difficult to articulate to laypeople?
If you have a very loose definition of epigenetics, it’s far too easy to communicate it to a layperson in all of its vagueness. To encompass anything you want it to encompass. So if you have a situation where you say to somebody, “I believe that there is a way that the genes you are born with can modified by some sort of external environment influence, so that your genes are not your destiny,” that is a hugely attractive concept to a lot of people. And I think that’s a starting point for a lot of people in the world of epigenetics. There’s a second constituency which is “I’m going to define epigenetics in terms of molecular events in a cell. I can study molecular events in a cell, so if I find differences in terms of these events in a cell, I can then overinterpret those differences to say whatever it is that I like the sound of.” The third constituency is the idea that there are ways of transmitting information from previous generation to offspring and subsequent generations that are over and above what you transmit through DNA itself. So people have been very attracted by the idea of stressors, unusual exposures and so on, continuing to propagate their effects over generations. In funny ways, the epidemiological evidence for some of this stuff is actually quite good. In mouse models, you can see certain things happening. One example is of mice exposed to vinclozolin, which is an endocrine disruptor. It’s certainly true that after several generations they continue to have an oligospermia phenotype. The problem is, you take one of these stories where the phenotype is actually quite solid, or you just do some sort investigation where there was some grandparental trauma or whatever, and you do a study, which usually focuses on DNA methylation, and it’s usually on an accessible cell type, like blood. You will see changes in DNA methylation. You will always see changes in DNA methylation. That’s the rule. There will be changes.
The problem is that DNA methylation is the readout for so many different things that you can’t tell what it is that’s causing that difference. So in the case of vinclozolin, I heard a very interesting talk at a transgenerational inheritance workshop at the NIH, by Mike Skinner, who did the original study, and what Mike basically was saying is that he has now seen evidence for pretty major DNA damage associated with vinclozolin in sperm. So if that’s the case and you do a genome wide assay for something like DNA methylation and you see differences, of course those differences could reflect the DNA itself having been rearranged, deleted, amplified or whatever, and if you don’t test for those influences, you’re not going to be able to dissect out why the methylation pattern changes. So we have to be kind of open minded about the possibility that yes, there could be some real cellular reprogramming going on, and these things could be real. But the problem is that pretty much every study that’s been done, to date, has not shown any of this.
In fact, there is a study I’m very excited to see published, from Stephan Beck in UCL, and he did what I think is the perfect epigenome wide association study, and because it’s unpublished work I won’t talk about it great detail, but he took into account the genetic influences in DNA methylation, by looking at purified cell types, performing comprehensive genome wide assays as well as monozygotic twin discordant studies. Very careful analysis. One of his trainees presented this at a meeting at the end of January, and they found basically no changes in DNA methylation, which of course you have to be open to the idea of it, because DNA methylation changes may not be something happening in every disease. But what it makes me wonder is whether when you do study right, you’re just not going to find noise, but if you do it poorly, you’ll find noise and you’ll report noise. So I’m really looking forward to Stephan coming out with study because I think it’s going to be a very informative negative study and will really set the benchmark for everyone in the field.
It’s an emotive topic because the joke about scientists, we prefer to use each others toothbrushes rather than each others theories. Once you’ve latched onto this really attractive idea of epigenetic influence modifying phenotype and heritability, it’s tough to countenance the idea that we may be wrong, and I’ve had to go through that myself. You know, that a lot of the early stuff that I did was based on uninformed decisions, but you know when you come to revise your opinion it’s disturbing.
But the GWAS field went through this in the early 2000s. If you wanted to do a GWAS of schizophrenia, you’d get your hundred patients that you see in clinic and your hundred controls, which might be you, people in your lab, people in your family, you know, it wasn’t particularly well designed. And a hundred patients is nowhere near enough statistical power. But you’d publish something and say you’ve got some interesting hit here. It was all non-replicable. You had problems with stratifications of populations. So they realized, and went through this and said, “hang on, we can’t do a schizophrenia study on a hundred patients, we have to do it on ten thousand.”
So they went probably went through massive trauma at that time, trying to think of how to get ten thousand patients. But they did it. They were ballsy enough that they basically figured it out, self-corrected, and have been producing great stuff in that area since. So epigenetics is going to go through a difficult growth spurt, but if we come out the other side and mature in terms of how rigorously we do these studies, we can rescue some of the damage that the brand has taken over the last few years.
We need to be much more careful about how we design, execute and interpret the studies. In order to do the studies right we have to be very careful about the cell types we study, we have to be comprehensive with the assays we do, we have to do genotyping, transcription work, probably chromatin studies, as well as DNA methylation. We can’t look at just one in isolation because you can’t interpret that. So the studies are going to become much more expensive and onerous, but they’ll be beautiful rich data sets of really informative findings we should be able to exploit in a lot of different ways. So I wouldn’t call them epigenetics studies at that point. I’d call them functional genomic association studies or something like that.
Big integrated gorgeous mess of data basically.
Hopefully not that much of a mess.
Genomewide data are always a mess, that’s just the name of the game. Which is why you have to go in and be very careful about the findings that you’ve identified within the genome. But it’ll be big sloppy hot messes of data.
Do you feel that scientists have an onus to be advocates of their own work?
Yes. We don’t have the luxury of doing ivory tower science. You might be extraordinarily good, and you might be one of the five experts in your niche field. That’s great, but ultimately we’re publicly funded. The work is meant to be for the public good. It’s great if you’re able to make the case to your peers in grant applications and publications, but ultimately if you understand it you need to be able to communicate it. I think it’s useful for us as well.
I’ve got the grandmother test; can you explain what you’re doing in such a way that your grandmother can understand it? Now, some of us have very smart grandmothers, so if you do, just think of somebody who is so far removed from what you’re doing that they shouldn’t have any specialized knowledge. If you’re unable to explain it to them, it’s probably wrong. I think you need to take that as the starting point.
You’re not necessarily going to be the most popular person at the cocktail party if you’re talking about science to everyone you meet. But you should at least be able to communicate the enthusiasm and excitement and the promise and the rigor and the responsibility that’s involved. You want people to feel like good things are happening. Especially in healthcare. We’re in a medical school here, the science we’re doing here is supposed to be helping people. When people ask questions, at least try to give them some sort of an answer.
It’s not exactly vocational work. It should be fun for you. If it’s not fun for you to communicate this stuff, you have to question why you’re doing this in the first place.
Although you go through periods where you don’t know what it is that’s bothering you, about some question that you’re studying. And you can’t articulate it, but if you can get past that then you’re in good shape.
How do you feel about the recent discussion of a “reproducibility crisis” in science? Does a crisis exist, and if so, how should we be solving it?
Reproducibility is a very thorny issue in the era of big data. Because it’s impossible to review a paper. People use analysis when they create a genomics paper of whatever type. The software is sometimes described. Sometimes it’s described with version numbers. Sometimes it’s described with the adjustment of parameters they used. Sometimes they make the software that they’ve generated available for people to reuse. It’s very rare that somebody writes a paper in such a way that you can use their software exactly the way they did. When we have tried to reproduce people’s chromatin or other work and we follow what they did in the paper, inevitably we ended up contacting the authors because we can’t quite follow what happened.
The other thing, of course, is that the data sets are so huge that if you’ve got your review request from Nature and they want it turned around in seven to ten days, it may take seven to ten days to download the data. The GEO link that you get for the data sitting in a public database doesn’t allow you to download any data, but rather to see that the data exists somewhere. So it’s not as if the data are actually accessible to you; you do not have the ability to check and see whether the way they claimed to have done the analysis was the way they came up with the figures or the conclusions they draw in the paper. So we can’t analyze data at the time of submission. We fundamentally can’t review these papers, but we do. So there’s a lot of trust. Then the question is how often do people try to go back and re-analyze the data.
There are these forensic statisticians out there who go out and try to dig into these things. It’s sort of shocking to see how often they find that you can’t reproduce this stuff.
Now there is a component like the pharmaceutical companies have talked about how few genomics studies they can reproduce. If you have a model of a transcriptional difference that characterizes an aggressive breast cancer from less aggressive one and your model is 20 genes, any statistician will tell you that those 20 genes just happen to poke their heads just slightly above a lot of competing models.
So when a pharmaceutical company says that they can’t reproduce this at all, it may be that the model they’ve chosen is actually pretty close to the one that was published, but with one or two little tweaks, you can end up reordering this thing so that two very comparable models in terms of statistical association can look quite different in terms of the genes they contain.
So the roots of reproducibility are very complex in terms of being able to replicate a study, but also in terms of the issues that have to do with there being no single outcome to a lot of these studies.
Do you see a solution to this?
Yeah, but it’s going to be unpopular. You have to complement publicly accessible database with publicly accessible computing. So if we were able to take publicly accessible grid computer, for example, and build some sort of interface that says, “Okay, this group has put up these various analyses on GEO and embedded the pipelines they’ve used to analyze this data, which is federally funded. By clicking this button you can re-run the analysis, but also because they’ve been very transparent about the software they’ve put onto the grid, you can play with the parameters they’ve used.” You could allow two mismatches when they only allowed one, and ask if that materially affect the conclusions they’ve drawn. You have to have something like that in order to review papers, in order to be able to interrogate the subtleties and ambiguities that reside in any sort of data like that.
To build such a system is expensive and would place an immense burden on scientists to upload their analytical software into that environment in order to get things published. But ultimately that would separate scientists into those performing highly rigorous research and those who are not. I think if you have snob appeal you may actually get people to adhere to it. What you won’t get is the funding agencies driving it. Federal funding agencies like to get the scientists to lead and then they follow, rather than to impose requirements on the scientists.
Private agencies are much more comfortable with the idea of saying “you’re going to do things our way.” In fact, that often works. Often, that’s a really good model.
Of course, the other players in all of this are the journals. If they require stuff, things happen. If funding agencies and journals were to get together and say, “we need things this way,” that would be perfectly appropriate.
But to get back to the question of reproducibility in epigenetics, if we can’t even define epigenetics how can we get consensus on how to do these studies?
Right. The devil’s in the details. It’s quite easy to write out a reproducibility strategy on paper, but there are so many details that so many researchers will disagree with.
Yeah, obviously other communities have done this. Like astrophysicists have figured out how to share data, but it’s a smaller community, and they are very savvy when it comes to the issues involved. Aaron Golden and I wrote a little review article called “Astrogenomics,” where we talked about how to take the lessons of astrophysicists into the world of genomics. It’s a good idea, but it’s difficult to reach consensus among a whole bunch of people who are very inclined to do things their own way, which is a trait that we kind of want to encourage — to keep science innovative and independent.
Having worked on some consortium funded scientific projects, do you feel that there are inherent problems with large consortium-based projects, compared to investigator-initiated research? Does consortium science lead to groupthink, or stifle innovation?
I was at a meeting of the ENCODE consortium, way back, when they were trying to figure out which peak caller to use for ChIP-seq. There was a large number of people in the room, and no real merits to separate - MACS and PeakSeq. The guy who was moderating the discussion pulled out a coin, and we ended up with MACS.
In consortium efforts, you have to make decisions like that. You can call it groupthink, but it’s also the efficient way of herding cats. So if we get back to this question of the GWAS community and this question of how to scale up a 100 patient study to 10,000, you can’t use a mom-and-pop approach. So consortium science is absolutely bloody essential at times. You just can’t do science without scale, sometimes. But if everything became consortium science you would certainly lose the ability to have these kind of weird probing innovative studies which may take an entire field with them. There’s going to be failure in both.
But you can tolerate a lot of failure with individual investigator-initiated projects when you have a few projects that work out really well.
Whereas if consortium research fails, that’s a disproportionate loss.
I think consortium funding occurs when the risk is low. If it’s a matter of just grinding through a well-accepted set of steps to come up with what you need, you just do a lot of it. That gets funded. I couldn’t imagine a big consortium thing which was pure high risk technology development which was maybe 10% likely to come off in the first place.
What about HGP-Write? Where they’re attempting to synthesize a designer cell with a modified custom human genome.
I think it’s a waste a money. I haven’t heard a justification that says this will be an interesting intellectual exercise that’s going to tell us anything about human health or physiology of anything like this. It just seems like “if we can do it, we should try.” Now, if somebody wants to correct me on that and say, “these are the reasons why this will be a really great project and it’s going to open up a lot of understanding about physiology,” well then I’d love to be corrected on that. But right now, as far as I can see, it’s “we’re able to do this, so why not.”
Might makes right.
It’s like Team America: World Police, except in a science lab. You can fill in all the quotations.
What was your experience working with ENCODE?
It was great. You get to hang out with a lot of really interesting people who covered a big spectrum in very much an emerging field.
It was a time of poor funding, so I think while there’s been a lot of push-back about ENCODE, and how it wasted money, and overstated functional elements of the genome, and stuff like that-
We’ve all read Dan Graur’s opinions.
Exactly, but if you look at it almost like being the Works Program Administration during the Great Depression, it kept a whole bunch of people alive during that period. And in fact, a lot of the leaders of chromatin biology, bioinformatics, genomics, who are around today cut their teeth during that period. It trained a whole bunch of people in a rough funding environment.
I don’t think the ENCODE project was great value for money. It helped push a few technologies. It helped give us some insights into the genome, but ultimately we were working with crummy cell types. The technologies were kind of poor. You go back and look at some of those data sets and they’re not very reliable.
But the fact that we have so many really good researchers around the country who were grad students and postdocs around that time. They wouldn’t be here. We seeded a generation of scientists because of ENCODE, and there were connections made at ENCODE. I met a lot of people there who I am still collaborating with, or am in contact with, or know to call on them if I need a paper reviewed at PLoS Genetics.
Have you felt that scientific research has meaningfully changed since your career began?
It’s a really great era in which to do genetics. Because the tools and insights available to us, and the braintrust that’s out there of really, really smart people. That’s a massive meaningful change since I was trying to cobble together my PhD at Yale.
As a physician scientist, my first inclination is to talk about complexity, because you never see the same patient twice. If you ask a hematologist about sickle cell disease, it’s the same bloody mutation over and over again, and no two kids look the same in terms of presentation. So physician scientists will always talk about the inherent complexity to biology.
Whereas if you’re a bit more focused on an area of science, and you’ve come through PhD training, there’s a temptation to simplify and that’s great. That’s helpful. It’s very helpful. I think what is happening in genetics and genomics is that the complexity that was always evident from looking at the patient is now being evident from looking at the data. You’re beginning to understand how little effect a lot of these sequence variants have on the ultimate phenotype and how even when you have a strong rare mutation causing a coding sequence change how variable the manifestations of that can be in terms of the phenotype of the individual. So we’re catching up with the reality that we would have seen in the clinic all along. That to me is a really meaningful change, because it’s illuminating. It shows us the reality of what’s happening.
Pulling back the curtain.
Yeah, very much so. It’s frustrating when things are too reductionist. Unrealistically reductionist. You have to start there obviously, but if you know there’s a bigger picture it’s frustrating. So the meaningful change is predominantly in those terms. The practicality of day-to-day is that we have so much more sophisticated technology, assays, and computational approaches available to us now than we have previously. I think it’s more a philosophical thing of appreciating and beginning to embrace complexity.
You can’t start by trying to embrace all complexity. You have to be reductionist in the individual component. We do that in epigenetics. We look at DNA methylation. We do these association studies and we’re finding stuff.
We started off reductionist and now we know we have to broaden our scope, embrace this complexity and get a fuller picture. That’s going to be great, it’s going to be wonderful, but we all have to go there.
A new study in Cell Host & Microbe aims to zero in on how the Zika virus interacts with the molecules on the surface of cells in and associated with the developing fetus. The idea behind this is to understand why the virus has such a proclivity for infecting these cells, over others, and hopefully, to develop a therapeutic strategy that may protect newborns from infections, or to vaccinate the population against the virus.
pictured: Aedes aegyptii, the predominant vector of the Zika virus. Image source.
The researchers, from UC San Fransisco and UC Berkeley, found evidence from experiments with various placental cells, that the virus grows better in mid-gestation phase cells than in late-gestation phase cells, which aligns with clinical data suggesting a higher risk of Zika-related complications early in pregnancy.
Also of note was the observation that different strains of Zika virus have different levels of infectivity among cell types. This may hint at reasons why we have not seen birth defects associated with Zika in previous outbreaks.
The overarching mechanistic question which the study set out to answer is how Zika enters and infects these placental cells better than others. To answer this, the group looked at some factors for infection in Dengue virus, which is closely related to Zika (and indeed, previous Dengue infection may play a role in response to Zika infections), and asked what the levels of these factors are in cells infected with Zika virus. They found that high levels of one such factor, TIM1, correspond well with high Zika virus susceptibility, whereas other factors on their hit list were more variable from cell to cell.
There’s been some recent work showing that duramycin, a peptide antibiotic approved for use in animals, is effective at preventing Dengue virus infection, in a manner related to this host factor, TIM1. Promisingly, the UCSF and UCB researchers found that duramycin is also able to reduce Zika infectivity in the cells studied.
Of course, these results have the normal caveats associated with preliminary drug studies; the infection assays were in a cell culture model, rather than in a live animal (though these are human cells, and not a model organism), the pharmacokinetics and bioavailability of duramycin is not particularly well known, so it’s possible that the drug may never reach the placenta to prevent infection. On the plus side, there has been a clinical trial for using duramycin to treat cystic fibrosis, and while that treatment strategy doesn’t seem to have gone anywhere, there were no safety issues or terrible side effects.
Hopefully this strategy will bear some fruit, especially as duramycin treatment may be useful in preventing and treating other viruses.
A eukaryote without signs of mitochondria: In Brief
(Image: the oxymonad Monocercomonoides)
A forthcoming paper the journal Current Biology caught my eye the other day (as well the eyes of a bunch of people on twitter). A group of researchers at Charles University in Prague presented work suggesting that a unicellular eukaryote totally lacks a mitochondrion. Now, for those of you in the metazoan/cell culture/biomedical field who might be raising an eyebrow or two, this claim is not completely unbelievable; this microbe is part of a group of protozoa called oxymonads, who are all obligate symbionts, and as far as I can tell, they all live in oxygen-free environments. Furthermore, it’s known that many of these microbes have degenerate mitochondria and other related vestigial organelles (like hydrogenosomes). The amitochondrial nature of these things (or really, the difficulty in finding mitochondrial components) was a counterargument in the 1980s to the notion that all eukarya descended from a single endosymbiotic event.
The group generated a draft genome of this organism, an oxymonad called Monocercomonoides sp. PA203. It seems to be found in a wide range of guts of various mammals, insects and other organisms. So they started looking through the genome to see if they could find any traces of mitochondria (mtDNA, homologs of nuclear proteins associated with mitochondria, such as import machinery and chaperones, and components of iron-sulfur processing networks, a task relegated to mitochondria in eukarya) and they couldn’t find anything. Obviously, you can’t prove a negative, but it seems to me that the group did quite a lot of searching for hints of mitochondria.
So how does Monocercomonoides generate energy without a mitochondrion? From their sequencing based approach, the researchers propose that Monocercomonoides largely generates ATP through substrate-level phosphorylation, which is apparently how other protists get by. They also found some genes involved in arginine degradation, which is, again, thought to be an alternative source for ATP by protists.
So that’s about it for the paper in brief. It definitely looks like good work, especially for something so rooted in genome sequencing. However, I have some overhanging questions and issues with some of the data that just don’t make sense to me.
In the draft genome for Monocercomonoides, the authors found about 16,000 predicted protein coding genes, and about 32,000 introns. As you might expect, these other “amitochondrial” oxymonads have pretty pared down genomes - especially when it comes to introns.
Quick reminder; introns are interrupting sequences in genes that have to be removed after transcription, before the RNA can get out of the nucleus and do what it needs to do. In eukarya, they are removed by a large assemblage called the spliceosome. The important thing to take away here is that introns cost a lot of ATP. It costs a lot of ATP to transcribe RNA, and the spliceosome is dependent on a number of ATPases to work. Most eukarya are able to overcome this by generating a lot of ATP (with mitochondria), while oxymonads and other organisms with degenerate mitochondria have shed a lot of their introns, presumably because they are very costly.
So how does Monocercomonoides maintain so many introns with so little energy production (substrate-level phosphorylation and arginine catabolism don’t seem to be particularly comparable)? I don’t have an answer to this, and it doesn’t look like the massive energy cost was appreciated by this group.
Another thing which has always been striking to me about these “amitochondrial” organisms is how little we know about their cell cycles. I mention this because one of the less appreciated roles of mitochondria is that they are known to have a fairly strong connection to G1 to S phase cell cycle in many cell types — if there is a problem with mitochondrial fission and fusion processes, various cyclins (namely Cyclin E, I believe) build up and stop progression into S phase. Presumably without the mitochondrial machinery in Monocercomonoides, they have to regulate cell cycle differently.
This Species of Bee Uses Fungus to Protect its Larvae from Pathogens
Just about once a week, I pick up a little rectangular container of sushi for lunch. Sometimes, this is a bi- or tri-weekly occurance. I tend to drizzle one or two little packets of iodine-colored soy sauce onto the chunks of wasabi and ginger that come in the container, which I then dunk the sushi into, to the horror of any culinary purists. I’ve always wondered about soy sauce; particularly how it relates to the spongy white blocks of tofu that I tend to think of as “soy.”
As it happens, the difference is owed largely to microbes. Soy sauce is made from a complex fermentation of grains and soybeans. What’s more, it’s been made this way for a very long time; the earliest recorded mention of fermented soy beans dates back to 165 BC. And while the most famous microbes in the process of fermenting soy beans for sauce, vinegar and sake production are Aspergillus oryzae, and lactic acid bacteria, a number of other fungi are sometimes used. Red yeast rice is a rice inoculated with the mold Monascus purpureus, and this “red ferment” is used in the production of a number of familiar dishes and drinks; red rice vinegar, some Chinese, Korean and Japanese rice wines, and is the classic food coloring in char siu and other dishes.
Red yeast rice has also been used pharmacologically since 800AD or so, to promote digestion, vitality and blood circulation. Which makes some sense, because this yeast produces compounds similar to statins (you may be familiar with some brand names: Lipitor, Crestor, Zocor) which inhibit the production of cholesterol.
So it may not surprise you to hear that humans may not be the only organism that ferment Monascus yeasts. Researchers at the University of San Paolo in Brazil reported that a species of bee makes use of a Monascus fungus, albeit for different reasons than we do. These bees appear to cultivate fungus to protect their larvae from pathogens.
In bee nests, larvae are reared in wax cells called brood cells. Nursery worker bees fill these cells with a semi-liquid food supply, a queen bee lays an egg on top of this mass of food, and the cell is sealed. Unlike other bee nests, the researchers noticed that the walls of brood cells of Scaptotrigona depilis, a stingless species of bee, are coated with a white fungus.
The growth of this fungus begins to spike three days after the egg is laid, when it’s nearly ready to hatch. The fungal growth continues and then abates, after the 5th day of larval development. With time-lapse microscopy, they observed the behavior of 3-day-old larvae. They saw that the larva made circular movements around its brood cell and consumed fungal outgrowths.
Just because the bee larvae like to chow down on this fungus, doesn’t mean it’s necessary for the bee’s survival. This fungus could just make its way into the bee nest and happen to grow in the larval food. In order to rule out any coincidences like this, and to establish whether or not the fungus plays some sort of role in the development or survival of the larva, the researchers created artificial brood cells. They placed larval food into this artifice. Then they sterilized the cell to kill any fungus or bacteria already in the larval food. In half of the fake cells, they re-introduced the fungus. Then they transferred newly hatched larvae into the artificial cells. In the fake brood cells that had fungus added, 76% of the larvae survived (150 total), while only 8% survived in cells without fungus, which pretty strongly indicates that the fungus is important for larval survival.
They noticed that the food in cells without fungus smelled bad and seemed to show signs of spoilage (such as stickiness). The larvae that were placed in these cells grew more slowly, and developed darkened guts and started to die in mass on day 7.
Okay, so how does this work? The fungus may be protecting the larval food from microbial contamination, by producing chemicals that kill pathogens. Monascus fungi are known produce a variety of chemicals that do this task -- it’s one of the reasons humans use it for fermentation, to preserve food that we’d like to eat later. The alternate hypothesis is that the fungus provides some sort of major nutritional benefit to the larvae.
In order to test the effect of the brood cell fungus against other microorganisms, they tested the growth of the bacteria Escherichia coli and Staphylococcus aureus in the presence of various larval food samples (new larval food, with no fungus, old larval food, and fungus from the borders of the brood cells). The results from this experiment were not conclusive; while old larval food inhibited bacterial growth, fungus from the borders of the cells did not. However, E. coli and S. aureus may not be the microbial contaminants that these bees are threatened by, and so lack of antimicrobial effect here does not indicate lack of relevant antimicrobial effect by the fungus. One thing I didn’t really understand about this study is why the researchers used these bacteria and not pathogens known to cause diseases in bees (other species suffer from contaminated brood cells, caused by a range of bacteria). Maybe it’s because they are hoping to find compounds with relevance to human diseases. Or maybe it’s dangerous to work with bee pathogens in a lab where you are working with bees.
This brings up another question; if these Monascus-fermented brood cells are so good at preventing diseases, why don’t other types of bees farm fungus? The answer may have to do with the unique behavior of nest-building by stingless bees.
When stingless bees start a new nest, they do so by swarming. Hundreds or thousands of workers and one or a few virgin queens gradually perform this task over several weeks, and they take building materials (particularly the waxy mixture of plant resin called cerumen) from the mother nest to start building the structure of the new colony. Then the workers bring food from the mother nest to the new colony. Once the new nest is ready, a few virgin queens migrate to the new colony. The researchers suggest that workers may also actively propagate the fungus from one nest to the next. The swarming behavior in which building materials are taken from one nest to the new one is not observed among other social insects.
Knowing that there is carry-over of building materials from the one generation of nest to the next, to try to figure out the source of the fungus in the nests, the researchers tested 11 types of material from the nests of three colonies, including larval food from provisioned cells, crop content of nurse bees, body structures of bees, building materials of the nest and honey and pollen reserves.
They used the larval food sterilized with UV light as a growth medium (since they know that the fungus can grow on that) and inoculated it with these samples of material. If a material from the nest was the source of the fungus, then fungus should grow on the food that it had been inoculated with.
The results? Fungus was only found on food that had been inoculated with cerumen, used for building the brood cells, food pots and all other structures in the nest. However, all of the structures in the nest are not covered in fungus, so why do only the brood cells have mycelia growing in them? The fungal mycelia does not grow out from the cerumen until the cerumen is used to construct brood cells, when it has contact with the semi-liquid larval food, and begins to proliferate rapidly.
Other insects grow fungus intentionally, but not in this way. Fungus-growing ants, ambrosia beetles and a few species of termites actively gather spores in special storage structures for fungal transport prior to founding a new colony. In previous instances of fungal-farming, the reproductive adults perform the task, not the workers.
This research evokes more questions than it answers (which is great!). Does this have wider implications for stingless bee nests? Or is this species unique in its fungal farming? Could we use this fungus to fight brood cell diseases in honey bees? Are these bees so reliant on this fungus that they are more susceptible to diseases? Which came first, the fungus or the pathogens it defends against?
The original paper can be found here:
A Brazilian Social Bee Must Cultivate Fungus to Survive
Cristiano Menezes, Ayrton Vollet-Neto, Anita Jocelyne Marsaioli, Davila Zampieri, Isabela Cardoso Fontoura, Augusto Ducati Luchessi, Vera Lucia Imperatriz-Fonseca
