Exposed, indeed.

The article “Exposed: The Great GM Crops Myth” published by The Independent is surprising, and at first glance looks like it could be about significant research indicating that genetic engineering has unforeseen yield decreases. However, I’m not very good at first glances. Instead, I read deeply and find out more. This post is going to be long, but I think the length is justified since the article has been diligently repeated on Common Dreams and Grist, and of course picked up by Digg and used as part of an anti-GM riff at the DailyKos and other political blogs. Frankly, I’m tired of seeing science get twisted to suit an agenda, and I’m going to report some facts.

The first sentence of Exposed is clearly sensationalist: “Genetic modification actually cuts the productivity of crops, an authoritative new study shows, undermining repeated claims that a switch to the controversial technology is needed to solve the growing world food crisis.”

Nevermind that scientists never state findings in such definite terms. Any result is simply a hypothesis that hasn’t been rejected. It isn’t fact until it has been corroborated by multiple studies by other researchers, and until it has been published in a peer reviewed journal of consequence. That’s simply the way science works. I suppose the enthusiasm can be chalked up to journalistic license.

The results of this study were published in the quarterly Better Crops (the full name of the publication is Better Crops with Plant Food). Better Crops is published by the International Plant Nutrition Institute. This is the first time I’ve heard of IPNI, but admittedly that doesn’t necessarily mean anything. The website states:

The International Plant Nutrition Institute (IPNI) is a new, not-for-profit, scientific organization dedicated to responsible management of plant nutrients — N, P, K, secondary nutrients, and micronutrients — for the benefit of the human family. With established programs in Latin America, North America, China, India, Southeast Asia, and planned expansion in other areas of the world, IPNI is a global organization ready to respond to the world’s demand for food, fuel, feed, and fiber.
IPNI provides a unified, scientific voice for the world’s fertilizer industry; independent of the industry, but scientifically credible and recognized by governments, academia, NGOs, the public, and the industry. Its scientists are working to help define the basis for appropriate use and management of plant nutrients, especially focusing on the environmental and economic issues related to their use and to provide comprehensive and regional information and research results to help farmers, and the industry, deal with environmental and agronomic problems.

So, IPNI is controlled by the fertilizer industry, which is one of the agricultural input industries that anti-big-ag advocates fight against. Looking over the website, this NGO seems to have a lot of information on fertilizer. Not genetic engineering, biotech, plant breeding, or any similar topics. Just fertilizer.

Of Better Crops, IPNI has this to say:

It’s not easy to describe this unique magazine. With an identity somewhere between an agronomic research journal and a marketing information series, BC provides a steady vehicle for reporting news from research related to nutrient management. While constantly evolving to serve its target audiences, the magazine also serves as a mirror of the agronomic research and education programs of the Institute.

In other words, Better Crops is an industry newsletter, similar to pamphlets on topics like “Beef, it’s what’s for dinner” or “The incredible edible egg.” I may be a bit skeptical, but I don’t trust information that comes directly from individual companies or from large industry groups unless similar findings are reported elsewhere.

Because Exposed came out on 20 April, I assumed that the article of interest would be in the current issue of Better Crops. Instead, the article by Barney Gordon was in the fourth issue of 2007 (way to keep on top of things, Independent). The abstract:

This study was conducted to determine if glyphosate-resistant (GR) soybeans respond differently to Mn fertilizer than conventional soybean varieties in an irrigated high-yield environment, and if so to develop fertilization strategies that will prevent or correct deficiencies. Yield of the GR variety was less than the conventional variety without Mn fertilizer. However, Mn application (banded at planting) to the GR variety closed the yield gap. The conventional soybean variety was not responsive to Mn fertilization. Conversely, yield was reduced at the highest rate of Mn. A second phase of the study showed that a combination of Mn applied as starter and foliar application provided maximum yield response.

I freely admit that I don’t know anything beyond the basics of fertilizers, so feel free to take my analysis of the article with a grain of salt. On the other hand, I am knowledgeable enough in plant physiology to make reasonable conclusions about the article. Geoffrey Lean, environment editor of The Independent, is presumably not, since he completely twisted the facts to make his story.

First, I’d like to mention that Dr. Gordon of Kansas State (not University of Kansas as Lean reported) studies fertilizer and farming methods. Not plant breeding or plant physiology. He makes no claims to the contrary.

Next, I’d like to call attention to the title of the article: Manganese Nutrition of Glyphosate-Resistant and Conventional Soybeans. Titles of scholarly articles are always about the topic at hand, or they will be rejected by editors of the journal. If this article really was about differences in yield between GM and non-GM crops, the title would say so. Occasionally an experiment on one topic will result in exciting data on a topic other than the one that was intended, but the title would certainly reflect that. Scholarly articles typically start with an introduction that indicates past research about the topic. This article starts with:

Glyphosate-resistant soybean variety planting dwarfs that of conventional varieties in the U.S. by a factor of about 9 to 1. Nevertheless, GR soybean yield may still lag behind that of conventional soybeans, as many farmers have noticed that yields are not as high as expected, even under optimal conditions. In Kansas, average yield seldom exceeds 60 to 65 bu/A even when soybeans are grown with adequate rainfall and/or supplemental irrigation water.

There is evidence to suggest that glyphosate may interfere with Mn metabolism and also adversely affect populations of soil micro-organisms responsible for reduction of Mn to a plant-available form. Manganese availability is also strongly influenced by soil pH. As soil pH increases, plant-available Mn decreases. It is unlikely that Mn deficiencies will occur on acid soils. It stands to reason that the addition of supplemental Mn at the proper time may correct deficiencies and result in greater GR soybean yields.

Dr. Gordon hypothesizes that the herbicide glyphosate interferes with Mn uptake (although this is not what this study tested). All plants could have this response to glyphosate application, but we can’t test this hypothesis on plants that are not engineered or evolved to resist glyphosate, for hopefully obvious reasons. The evidence that glyphosate might interfere with crop mineral uptake is serious and must be further investigated, because crops won’t be able to reach their full yield potential without proper mineral uptake. I’m also concerned that glyphosate might affect soil micro-organisms, because research in organic farming methods shows that soil microbes are crucial to soil and plant health. These results might mean that we should discontinue or decrease glyphosate application, but more experiments to investigate these preliminary results must be conducted.

This experiment was designed to test the response to Mn fertilizer of one particular line of soy that has one particular transgene. The results showed that this particular line did respond to Mn fertilizer, indicating that it might not be as good at Mn uptake as the non-transgenic control. Since only one event was tested, though, no conclusions about the gene itself can be made.

An experiment to compare the overall yield of GM crops to non-GM crops (regardless of fertilizer) must include multiple lines of multiple species and include multiple transgenic traits. It must also include comparable non-GM crops for each GM crop in the study. The reason for the repetition is to avoid choosing particular lines of plants that naturally have higher or lower than average yields. Multiple transgenic traits must be used in order to prove one way or another whether “all GMOs” have lower yields than non-GMOs. Each transgene or cisgene is different, and we can’t assume that drought resistant GM rice is the same as beta carotene enhanced GM rice, for example. Despite peoples’ claims to the contrary, making general statements about all GM crops is impossible due to the wide diversity of traits that are available.

The experiment must treat the GM and non-GM plants exactly the same (same planting time and method, same fertilizer, same irrigation, same pest control, and so on) so that the results will actually tell us about the difference between the GM and non-GM plants, not about the differences in farming methods. The experiment must also take place in multiple climates, to ensure that the crops in the experiment will act the same whether it is warm or cool, dry or wet during the growing season. Analyzing the results from a comparison of farming methods would be a lot more complex because there would be multiple differences between experiment and control plots.

Note: In Dr. Gordon’s rebuttal to Exposed, he says that the third year of the experiment showed no difference between the GM and non-GM lines, probably due to environmental variation from year to year.

Another complication is environmental variation from year to year. One of my research projects includes the hypothesis that a certain type of selection will improve seed protein over a period of years. To show this change over time, I have to save seed from multiple years and plant them side by side. I can’t just compare the data from 2006 to the data from 2007 because of all the tiny details that I couldn’t control in the field. Maybe the control plots in 2006 had a worse aphid infestation. Maybe the experimental plots in 2007 had slightly better soil…

Not only does the experiment to compare all GM crops to non-GM have to meet all of the above requirements, it also has to include multiple years worth of seed for each tested line.

It is sometimes possible to use multiple separate experiments to support a hypothesis, in a type of scholarly article called a meta analysis. These articles are often used in medicine in cases where larger human studies are not possible due to cost and other factors. A meta analysis can be used to collect information, but must be very extensive to allow conclusions to be drawn. Often, differences in the way each experiment was done prevent strong conclusions from being made.

I’d like to expand upon something that was mentioned briefly, but not explained, in Exposed. Companies such as Monsanto take years to develop and test the GM seed that is available for sale. The regulatory process is so stringent that, once approval is applied for, the trait can not be improved upon (to be more clear, once approval is applied for with one event, another event can not be substituted), or face a whole new round of application for approval. Regulatory hurdles and other issues make GM seed very expensive for the company to develop, so they must develop new seed to recoup their losses. For these and other reasons, GM seeds are often “one hit wonders” that excel in one specific trait, but not particularly for increased yield. Non-GM lines, on the other hand, are improved every year, with the best yielding plants being used to produce the next year’s seed. I recently attended a seminar presented by a scientist from Pioneer where he said that they were working to develop better yielding lines that would work in conjunction with their primary transgenic traits. The companies are aware that this is a problem with their products, and are of course working to solve it, to avoid losing sales.

The second study mentioned in Exposed also investigates “yield drag” in commercial soy that has been engineered for resistance to glyphosate. In the 2002 Yield Suppressions of Glyphosate-Resistant (Roundup Ready) Soybeans, Roger Elmore (now of Iowa State, previously of University of Nebraska) performs a similar experiment to the one in Manganese Nutrition, albeit without the manganese. The abstract:

Herbicide-resistant crops like glyphosate resistant (GR) soybean [Glycine max (L.) Merr.] are gaining acceptance in U.S. cropping systems. Comparisons from cultivar performance trials suggest a yield suppression may exist with GR soybean. Yield suppressions may result either cultivar genetic differentials, the GR gene/gene insertion process, or glyphosate. Grain yield of GR is probably not affected by glyphosate. Yield suppression due to the GR gene or its insertion process (GR effect) has not been reported. We conducted a field experiment at four Nebraska locations in 2 yr to evaluate the GR effect on soybean yield. Five backcross-derived pairs of GR and non-GR soybean sister lines were compared along with three high-yield, nonherbicide-resistant cultivars and five other herbicide-resistant cultivars. Glyphosate resistant sister lines yielded 5% (200 kg ha21) less than the non-GR sisters (GR effect). Seed weight of the non-GR sisters was greater than that of the GR sisters (in 1999) and the non-GR sister lines were 20 mm shorter than the GR sisters. Other variables monitored were similar between the two cultivar groups. The high-yield, nonherbicide-resistant cultivars included for comparison yielded 5% more than the non-GR sisters and 10% more than the GR sisters.

In other words, this is more of the same: plants bred for high yield perform better than plants that were bred for something else. More information about the difference in seed weight and height between the glyphosate resistant and non-GM soy can be found in the discussion section of the paper:

On average, non-GR sister lines yielded 5% more than the GR sisters when averaged over all locations and both years (Table 5). Non-GR sister grain yields were greater than those of their associated GR sisters in two of the five pairs… Grain yields of sister-line pairs are shown in Fig. 1. The greater number of data points to the right of the 1:1 ratio line indicates that the non-GR sisters yielded more on the average than their GR sister counterparts.

A correlation of 0.75 is very strong. This correlation (as shown in Fig. 1 from Dr. Elmore’s paper) means that, if a GM-soy was low yielding, there is a strong probability that its non-GM sister would also be low yielding. The 5% average difference is undeniable, but the GM plant in some sister pairs out preformed the non-GM plant. In short, I wouldn’t say that this is conclusive, and there would still have to be additional studies on other types of genetically engineered plants to show a difference between all GM and non-GM.

Dr. Elmore concludes: “Cultivar choices are best based on (i) previous weed pressure and success of control measures in specific fields, (ii) the availability and cost of herbicides, (iii) availability and cost of herbicide-resistant cultivars, and (iv) yield.” I read that as: if your fields have stubborn weeds and glyphosate is easy to use, then a slight decrease in yield may be preferable to having to either using a more dangerous herbicide or having your yield decrease anyway when your field is overgrown with weeds.

He also says something more concerning: “Based on the results of this study and those of Elmore et al., 2001, the yield suppression appears associated with the GR gene or its insertion process rather than glyphosate itself.” I would like to see further studies on this possibility, and I would be very surprised if Monsanto isn’t already frantically working to solve any related problems. The best way to test the second hypothesis would still be the laborious experiment with a wide variety of GM traits in different crops. The first hypothesis (that the glyphosate resistance gene itself is causing a yield decrease) could be tested in a few ways, including a study of markers for low yield in populations that include the gene, and testing the yields of plants that have naturally evolved glyphosate resistance compared to their “less evolved” relatives.

This type of study is important to help farmers choose the best seed each year from thousands of choices. They also help farmers to choose the best pest management strategy for their particular situation. This is the whole point of ag extension, and is Dr. Elmore’s job. The purpose of Dr. Gordon’s study is similar: to help soybean farmers achieve the highest possible yields, even if it means applying additional Mn fertilizer. I’m fairly confident in saying that these scientists don’t appreciate having their research misinterpreted to make over reaching conclusions, even if they appreciate the attention.

While researching the background of Exposed, I came across a semi-meta analysis of GM crop yield studies compiled by Clio Turton of the Soil Association (a non-profit promoting organic ag) that was posted on Check Biotech. Every study is listed with the goal of saying that GM crops yield less than non-GM. However, even Mr. Turton even says, “First generation genetic modifications address production conditions (insect and weed control), and are in no way intended to increase the intrinsic yield capacity of the plant.” Instead, they decrease competition from weeds and decrease insect damage which increases yield by corollary. I’d be happy to discuss the articles he presents, but this post is probably long enough.

Bear with me while I use an analogy. Saying that we should stop using all GM crops because they don’t yield as high as crops that have been specially bred for yield is like saying that you are going to throw away a business laptop because the processor can’t handle graphics intensive games. We all know that business laptops were designed with other functions in mind, and don’t necessarily need whizbang graphics cards. If graphics cards were less expensive, they would be in all laptops. If getting a GM crop to market wasn’t so expensive, we would see a better selection of seeds on the market. Right now, the only ones who can afford to make them are Monsanto and Pioneer. Public researchers at universities and small companies can’t even hope to get seed to market, so research in GM crops has been slowed to a trickle in the US. This is not the fault of the scientists or of the companies. Instead, we can blame the anti-GM hysteria that caused regulators to make things so difficult.

Thanks, Mike for pointing out the Common Dreams article, and thus for making me stay up all night and spend half of today researching this post!

Barney Gordon (2007). Manganese Nutrition of Glyphosate-Resistant and Conventional Soybeans Better Crops, 91 (4), 12-14
Roger W. Elmore, Fred W. Roeth, Lenis A. Nelson, Charles A. Shapiro, Robert N. Klein, Stevan Z. Knezevic, & Alex Martin (2001). Glyphosate-Resistant Soybean Cultivar Yields Compared with Sister Lines Agron. J., 93, 408-412