Tag Archives: agriculture

Who’s Afraid of the Big, Bad ORF?

A recent paper in the journal GM Crops and Food has generated an outsized splash in the press, particularly in biotechnology-averse Europe. I won’t reward a muckraking tabloid with a link, but here’s a screenshot that shows the basic theme:

Daily Mail hype.

Oh No, Toxic Genes!

Apparently the genetically modified food crops that hundreds of millions of people around the world have been eating without incident for more than a decade are in fact horribly toxic. But it turns out that the research that triggered this alarm proves no such thing. How did an arcane scientific finding get turned into a completely incorrect, apocalyptic headline? Let’s dig into it like scientifically educated journalists.

If we start by going to the source, we immediately hit an obstacle: there’s the abstract, but if we want to read the paper itself we’re expected to pony up $29. It would probably help a lot if journals made papers about important public policy issues freely accessible by default, but we don’t live in that world yet. Fortunately, journalists have an easy way to get around this: contact the authors directly. The Daily Mail appears to have failed at this, as all of the quotes in their article are from other sources. Other articles on the new work similarly lack any representation by the folks who actually did it.

It’s rare for scientists to blow off reporters completely, but sometimes they can be hard to reach, out of the office until after the deadline, or just uninterested in helping. Perhaps that was the case here. Let’s see. The first author is Nancy Podevin of the European Food Safety Authority in Parma, Italy. When I sent a note to her identifying myself as a journalist and asking for a reprint, she replied minutes later: “Please find the article attached. Please be aware that the content of the article has been incorrectly reflected in recent press articles.”

Not exactly hard to reach. Or reticent.

Alright, let’s dig into the work. Here’s the basic plan from the introduction:

Bioinformatic tools are increasingly being used in the evaluation of transgenic crops. Guidelines, proposed by WHO/FAO19 and EFSA, include the use of bioinformatics screening to assess the risk of potential allergenicity and toxicity. With this aim, the EFSA GMO Panel has updated its guidance for the risk assessment of GM plants and proposed to identify all new ORFs due to the transformation event. New ORFs are defined as strings of codons uninterrupted by the presence of a stop codon at the insert genomic DNA junction and within the insert. The putative translation products of these ORFs are then screened for similarities with known toxins and allergens.

This is a study done entirely on computer databases, in which the scientists looked for novel open reading frames (ORFs) in the transgenes of modified crops, then checked to see if any of those ORFs match any known allergens or toxins. The existence of an ORF doesn’t prove that it gets transcribed and translated into a stable protein, so we’re still several steps short of reality here, but it’s a useful exercise to define what might be possible. In this case, the investigators are looking specifically at a sequence called P35S, a gene promoter borrowed from cauliflower mosaic virus (CaMV). P35S promotes constitutive (constant) expression of the gene in front of it, so it’s been a popular choice for driving introduced transgenes in genetically modified crops. 54 of the transgenic crop strains currently approved in the US use this promoter.

In its original context, the P35S sequence overlaps with a CaMV sequence called gene VI. That means that the P35S sequence could potentially encode a piece of gene VI. Podevin and her colleague Patrick du Jardin searched the various P35S sequences used in transgenic crops, and identified a couple of ORFs. Remember, this is all on a computer. The paper contains no wet lab experiments showing that these ORFs are actually producing stable proteins in any cell. But let’s assume they do for now.

Translating those ORFs on the computer and searching against databases of known allergens and toxins, the researchers found … wait for it …


That’s right, these hypothetical proteins that might not even exist don’t match any known allergens or toxins anyway. They did an additional test that sets the bar lower, and found that by this standard, one of the putative proteins might be allergenic. But it’s a stretch:

The vector support machines (SVM) in AlgPred indicated on the basis of the dipeptide composition that the ORF that encoded part of P6 might have some allergenic properties. The sensitivity and specificity of this method is 88.87% and 81.86% respectively and should therefore always be used in combination with other tools.

All the other tools, though, found no allergenicity. Having established that there’s essentially no human risk, the authors speculated that there could still be effects on the plants themselves, such as plant stunting and late flowering. Considering that the entire point of most crop biotechnology is to increase yields, it seems unlikely that this applies to any of the current commercial strains, but product developers should probably keep an eye out for it in future strains. Either that, or they could simply follow the authors’ final advice:

The -343 variant [of P35S], identified by Odell and colleagues, contains all of the necessary elements for full promoter activity and does not appear to result in the presence of an ORF with functional domains, rendering it and its related variants the most appropriate promoter variants for avoiding unintended effects.

To put this all in context, plant viruses commonly infect all sorts of crops. One survey (PDF here) found CaMV and its colleagues widespread in numerous types of produce. We’re already eating huge quantities of plant viral proteins – not hypothetical ones, real ones – all the time. If there is an ORF from CaMV gene VI being expressed as a protein in transgenic crops, it’s likely one you’ve digested before, even if you eat exclusively organic food.

So there you have it. This was a research paper that used bioinformatic methods to ask yet again if GM crops are any more dangerous than non-GM crops. It ended up adding to the large pile of established data showing that they are not. Through what can only be described as laziness and ideologically blinded reporting, it served as a handy news hook for stories claiming exactly the opposite.

Update 2013.1.22 12:49: After writing this post, I saw this discussion thread, in which several smart folks make essentially the same points.

Update 2013.1.23 7:07: After Dr. Podevin graciously sent the paper, I pinged her with a few additional questions about the work because, well, that’s what I do. I received her reply this morning:

I have been overloaded with requests for the paper and as I am no longer working at EFSA it is difficult for me to react.

To answer you[r] questions I am not planning to work on this topic further. It is difficult how headlines on toxic genes in GMOs can be seen to be linked to our paper as we concluded that there are no indications for toxicity of the encoded protein. This virus has been infecting Cauliflower and related plants with no recorded health effect.

It should also be noted that this promoter [has] an ORF overlaps with Gene VI but that no functional gene is present. So in most cases this gene fragment will not lead to the production of a protein.

Update 2013.1.24 15:06: I’ve now received a note from the journal publisher as well:

I am the publishing director at Landes Bioscience – and for GM Crops & Food. Thanks for your excellent piece which was just brought to my attention. Would also quickly like to note that we have now made this paper OA, ie, freely available to anyone who wants to download and read. [link]

Colony Collapse Disorder: Dead Bees and Sloppy Science

A flurry of recent scientific papers, and a blizzard of subsequent news hype, has led a lot of people to conclude that the mystery of colony collapse disorder (CCD), which causes beehives to die suddenly, has been solved. Indeed, a Reuters reporter recently proclaimed exactly that in an editorial published on the wire service’s site.

All of these reports have converged on a single culprit: neonicotinoid insecticides, a category that includes some of the most widely-used chemicals in agriculture. According to this story, the pesticides aren’t present in high enough levels to kill the bees right away, but low-level exposure over a period of weeks slowly poisons them.


A beehive. Image courtesy artethgray.

Of course the pesticide industry hasn’t been taking this lying down. Agrochemical giant Bayer, for one, has been issuing testy press releases faulting the new studies. Bayer is a leading supplier of imidacloprid, a very popular neonicotinoid compound that is used in both agricultural and home pesticides.

Imidacloprid was also the focus of the most recent scientific study to pin CCD on pesticides, and in this case, at least, Bayer may have a point.

I’ve found this new study, by Chensheng Lu of the Harvard School of Public Health and two collaborators from the Worcester County Beekeepers Association, particularly interesting – and not in a good way. The press release about the paper has been the source of most of the news coverage, so I suppose it made a better impression on other science journalists than it did on me. Here’s how it starts off:

The likely culprit in sharp worldwide declines in honeybee colonies since 2006 is imidacloprid, one of the most widely used pesticides, according to a new study from Harvard School of Public Health (HSPH). The authors, led by Alex Lu, associate professor of environmental exposure biology in the Department of Environmental Health, write that the new research provides “convincing evidence” of the link between imidacloprid and the phenomenon known as Colony Collapse Disorder (CCD), in which adult bees abandon their hives.

The study will appear in the June issue of the Bulletin of Insectology.

“The significance of bees to agriculture cannot be underestimated,” says Lu. “And it apparently doesn’t take much of the pesticide to affect the bees. Our experiment included pesticide amounts below what is normally present in the environment.”

Let’s take this a little bit at a time. First, we’re being told the “likely culprit” has been found in a condition that’s baffled researchers for several years. That’s an extraordinary claim, so I’m expecting extraordinary data to back it up. Apparently the new paper will contain just that, because it’s supposed to be “convincing evidence.” Anyone setting the bar that high is either sitting on rock-solid results, or full of shit. In my experience the latter is much more common, so my skeptic senses are already tingling.

Then things really start to go pear-shape. The Bulletin of Insectology? I try to avoid being a journal snob, but come on, insectology? The name of the field is entomology, and a quick Google search confirms that “insectology” appears nowhere else in science except for the title of this journal. Their web site doesn’t exactly scream “high publication standards,” either. If you’re a fan of impact factors, the B of I scores a whopping 0.371, so apparently it’s not going to be rivaling Nature for citations anytime soon, either.

Then it gets even worse. The press release came out in early April, with no embargo, but the paper is scheduled to be published in June. Nor is this an “advanced online publication” situation – this paper really isn’t out yet in any format. This is truly science by press release. Maybe we should just move on, forget we ever saw this, and also ignore the absurdity of the author’s quote (he didn’t really say “cannot be underestimated” did he?).

The subsequent media storm was deafening, though, so I felt compelled to dig in. Emailing Dr. Lu, I got a prompt and courteous reply with an attached PDF of the paper – or at least a “corrected proof.” After confirming that it was okay to discuss it even though it wasn’t slated to be published for two more months (a question I gather he hadn’t been asked yet), I started reading.

It wasn’t as bad as I’d expected.

I realize that’s faint praise given the foregoing, but working my way through the paper a picture started to emerge. This project seems to have begun as an earnest effort to do good science. Then, somewhere along the line, someone decided to push the data out the door in a big hurry, bypassing the revisions that a competent peer reviewer would have demanded. Perhaps it was because two other publications about neonicotinoids and bees had just come out in Science (Henry et al. and Whitehorn et al.). Or perhaps the collaboration fell apart, or the team decided that some of the additional experiments they needed would take another year to do and they were sick of waiting. Whatever the reason, the final publication suffered.

Nonetheless, the experiment – there’s only one in the paper – had a lot of potential. Hypothesizing that imidacloprid sprayed on corn crops could contaminate the high-fructose corn syrup (HFCS) that’s fed to commercial beehives, the researchers decided to see what eating sub-lethal doses of the pesticide would really do to bees under field conditions.

They placed five newly constructed and stocked honey bee hives on each of four field sites, for a total of twenty hives. The field sites were more than 12km apart, so the bees from different sites would forage on independent territories. Following conventional apicultural practices for commercial hives, the team fed HFCS to all of the bees to supplement their honey stores during the winter. Four hives on each site ate HFCS spiked with various doses of imidacloprid, while the fifth hive was a control, receiving unadulterated HFCS. At the end of the winter, fifteen of the sixteen imidacloprid-fed hives – and one of the four control hives – had died.

The authors claim that the hives’ deaths resembled CCD, but that may be a bit of a stretch. For one thing, they report seeing dead bees on the ground near the hive entrances, which isn’t typical of colony collapse. They also didn’t see any of the pathogens that often correlate with CCD, such as varroa mites, iridoviruses, and the unicellular parasite Nosema ceranae. In addition, these experiments all took place in Worcester County, MA, just east of where I live, during 2010 and 2011. That was an absolutely horrific winter, breaking all kinds of records for snowfall, ice accumulation, and cold. It was hardly representative of the way traveling commercial hives spend their winters (they go to Florida). Of course the usual numerical objection also comes up; this was a very small experiment that clearly lacked the statistical power to extrapolate to an entire industry.

The biggest problem, though, is that the work is full of provocative but completely unsupported speculation. The authors discuss imidacloprid use on corn in some depth, and outline a plausible route by which it could end up in HFCS – but that’s entirely theoretical. Nowhere do we see data or a reference showing that the pesticide was ever actually in the sweetener that commercial bees ate, or measuring its levels.

Even if we assume, without a shred of evidence, that imidacloprid routinely contaminates HFCS, that would raise a whole new problem. Control bees also got HFCS. That means the controls also would have been eating some unknown amount of the chemical, and the experimental bees would have gotten a double dose, rendering the result meaningless. To do the experiment right, one would have to test the HFCS for the pesticide to confirm the levels, and also find some source for uncontaminated HFCS for the controls. If I were reviewing this paper for publication, I’d demand those data, and would also insist that claims of “convincing evidence” be edited to more cautious language – which, I suppose, might drive the authors to send the paper elsewhere.

We should see whether low levels of imidacloprid are contributing to CCD. It’s an entirely plausible hypothesis. Unfortunately, it remains untested.

Lu, C., Warchol, K., Callahan, R. (2012). In situ replication of honey bee colony collapse disorder, Bulletin of Insectology (in press).

Henry, M., Beguin, M., Requier, F., Rollin, O., Odoux, J., Aupinel, P., Aptel, J., Tchamitchian, S., & Decourtye, A. (2012). A Common Pesticide Decreases Foraging Success and Survival in Honey Bees Science DOI: 10.1126/science.1215039

Whitehorn, P., O’Connor, S., Wackers, F., & Goulson, D. (2012). Neonicotinoid Pesticide Reduces Bumble Bee Colony Growth and Queen Production Science DOI: 10.1126/science.1215025