Worrying directions in plant breeding

Here is an article about strawberry breeding in California.  It seems that there is a controversy arising because the leading lights in strawberry breeding at UC-Davis, California's flagship land grant university, are leaving to work in the private sector.  I don't know the details of the story, but it's already troubling to me that UC-Davis is already firmly inserted in the mercantile intellectual-property game of protecting and selling the varieties it develops.  This is common these days in public and private universities, but it runs counter to the very spirit of free exchange of knowledge in benefit of the "industrial classes", which is the mandate of UC-Davis and other land grant universities according to the Morrill Act that founded them in 1863.  It is also not how the CGIAR centers midwived the Green Revolution in the world, which was based on open-access plant varieties.  At any rate, this is the current reality, and it seems that this mercantile logic has come to a crossroads; the breeders who have been the prime actors in developing many of UC-Davis's strawberry varieties are now looking to move to a private company, and this threatens the very continued existence of the UC-Davis strawberry program.

Here is an article by Calestous Juma, a tireless cheerleader for genetic engineering.  Ostensibly he is arguing the rather uncontroversial point that genetic engineering may have a useful, though by no means exclusive or predominant, place in ensuring food security in the future.  However, his assertion at the article's end that future food security may be most threatened by the non-adoption of genetic engineering betrays the more inflated, central role he personally sees for the technology.  Note also his conflation of biotechnology (which is a broad field including a wide range of techniques, one of the least important of which is genetic engineering) with the narrow technical procedure of genetic engineering.  This confusion of biotechnology writ large with genetic engineering writ small is common among most people.  When genetic engineering critics make the mistake of condemning all biotechnology, advocates rightly point out that that would rule out in vitro tissue culture, marker-assisted breeding, and a number of other uncontroversial technologies.  When a genetic engineering lobbyist like Juma conflates the two, it seems like a smokescreen so he can defend the larger, uncontroversial category of biotechnology in case of criticism, even though what he's really talking about in the article is exclusively genetic engineering. 

In any case, Juma discusses out what are perhaps the most promising potential applications of genetic engineering for peasant farmers right now:  wilt-resistant banana, Bt cotton, and weevil-resistant cowpea (he doesn't mention brown-streak-virus-resistant cassava, but that's another promising one out there).  These technologies are promising in that they would allow farmers to control major pests that were either uncontrollable in the prevailing farming system (banana bacterial wilt) or controllable only through the use of lots of toxic pesticides (cotton and cowpea weevils and other insects).  They allow a simple "patch" plugged into the current system that would overcome a major limitation, which obviously can be very useful. 

The problem is that a healthy agroecosystem is complex, and doesn't allow for quick patches.  This isn't to say that the genetic engineering patch won't work in the cases cited above, but rather that the farming systems being "fixed" with the patch are not healthy to begin with.  Growing bananas year after year on the same soil, and using disease-ridden planting material to establish new banana plantations, is not healthy or resilient.  Growing vast expanses of cowpea monoculture on stripped soils is not healthy or resilient.  Growing cotton monocultures with little or no rotation is not healthy or resilient.  Accordingly, many of the problems that genetic engineering patches attempt to fix would not even be major problems in a healthier farming system.  Granted, we need to work with the prevailing farming system, not some fantastic ideal of what it should be, so perhaps in the short time we need the quick patches to the system.  But the long-term goal, and hence the focus of any serious ag research and development program, should be gradual change of farming systems toward more diversity and more reliance on free natural dynamics as opposed to purchased inputs.

A clear example of this is Bt corn in the US.  I would consider Bt corn here to be a major success in reducing use of some of the most toxic pesticides.  Prior to Bt corn, farmers tried (often fruitlessly) to control corn rootworm through lots of soil-applied insecticides.  Because the rootworm larvae lived in the soil, you had to really drench the whole field with chemicals, and because the soil naturally has organisms that tend to break down most pesticides over time, you had to use particularly persistent, toxic substances to reach the rootworm.  Bt corn has totally changed that, eliminating this major source of toxins from the environment. 

However, corn rootworm had a much simpler remedy, for those who were willing to consider the system as a whole and not just the patch fix of the day.  Corn rootworm lays its eggs in corn stubble, so you could eliminate it by rotating your field out of corn.  Any eggs laid in the corn stubble would hatch into larvae that would starve on whatever other crop you'd planted.  The only reason it was a problem was that, in our collective idiocy, we'd been planting corn after corn in the same fields, year after year.  Even when we employed our pitiful excuse of a rotation, one year of soybean alternated with one year of corn, it wasn't a robust solution, because the rootworm evolved to lay its eggs in soy fields to await the next year's corn crop.  If the US corn belt had, on the other hand, employed the complex rotations of the past, alternating corn, soy, wheat, alfalfa, oats, pasture, field peas, or any other number of possible crops, we would have had a complex, robust system in which Bt corn wouldn't have been necessary in the first place.  And it appears that, once established, such complex rotations can end up being just as profitable for the farmer. 

So the lesson from Bt corn, and I believe most other genetic engineering "quick fixes", is that it may be a useful patch for a broken system, but the long-term goal should not be a patched broken system but rather a healthy, robust system.  The problem is that, in devoting lots of effort to the quick patches, we often neglect the research, extension, and advocacy that would be necessary in the long term to bring about such a system.  A similar line of reasoning applies to attempts to provide missing vitamins in monotonous, starchy diets, as profiled in this and this article on vitamin A-enriched genetically engineering bananas (and similar to the long-standing Golden Rice story).  The root problem is a broken system for accessing land, water, and the other inputs (not to mention well-paying jobs) that would make it possible for peasants to consume a more diverse diet, as opposed to just bananas or just rice.  If people were eating a bit of squash, or sweet potato, or meat, or milk, or leafy greens, or mango, or any number of other foods in addition to their prime staple, then vitamin A deficiency wouldn't be an issue.  While inserting a bit more vitamin A into bananas may be worthwhile in the short term, the real questions we should be asking and researching relate to how to improve the overall farming and social system in a place like Uganda such that everyone can enjoy a more abundant and diversified diet.  If not, we might as well just mount an effort to have people eat grass and garbage to make up for their dietary shortcomings.

One last comment on the Juma article is his citation of a PLOS ONE study that claims to show benefits in income and dietary quality for farmers who planted Bt cotton in India.  While my anecdotal, common-sense impression is that Bt cotton would indeed be an effective way of cutting costs and decreasing really toxic chemicals in peasant households, there are quite a few problems I see with this article's statistical analysis and the conclusions it draws.  Namely, the study follows some 533 households over a number of years, but then at least in the characterization of the households it acts as if each observation is a separate household.  Furthermore, it purports to measure differential outcomes between Bt-adoption and non-adopting households, and to ascribe these outcomes to the adoption of Bt cotton, but to me the differences seen are not easily attributable to the adoption of the technology.

I am not well-versed in the details of what makes or doesn't make a solid statistical treatment in social science, but my instinct is that it's problematic to take multiple observations of the same household and then use the overall number of observations as your denominator when describing household characteristics.  You end up double- or even quadruple-counting the same households, so the characteristics of these households gain undue weight in your description of the overall group.  This might not be as much of a problem if the household characteristics you're measuring are relatively static over the study period (for instance, age of farmer, number of children, etc. shouldn't change too much in 3 years), but if your study is set up precisely to measure major changes in income and well-being, then this is a fundamental flaw in the experimental design (or at least the statistical treatment).  I believe they avoid this problem in some of their regression analysis, but it still seems like a problem to me.  Beyond this, and perhaps related, the way the study is set up seems to measure more the type of household that could afford to be an early adopter of the new technology, as opposed to the real gains in wellbeing deriving from use of the technology.  Since it seems that almost all households had adopted Bt by the last year of the study, the study becomes less a comparison of adopters vs. non-adopters, and more a before and after snapshot.  This being the case, if there are real gains in farmer income and diet over the study period, the most logical control group would not be the dwindling number of non-Bt adopters, but rather the overall farm populace in the region, both cotton-growers and non-cotton-growers.  If prevailing socioeconomic conditions meant that everyone was earning more and eating more, then you can't attribute this change to the adoption of Bt cotton.

I want to stress that I think Bt cotton is probably a good thing for peasant farmers in India and Africa (though I worry about genetic contamination of local populations of wild relatives and local landraces in both areas).  It allows them to reduce a major cost, and more importantly a major source of toxins, from their farming system by eliminating the use of the harsher pesticides.  But this study seems a bit shaky to me (which makes me wonder how it got into PLOS ONE), and certainly not a sound foundation for Juma and his ilk to make broad statements about the crucial role of genetic engineering in ensuring food security for the world.

In case you didn't feel like plant breeding was all going to hell, I hope these articles have convinced you otherwise.  Just kidding; I actually don't have a clear idea of "where plant breeding is going" on a worldwide scale, nor if it's heading in the right or the wrong direction.  But I will say that generally plant breeding (and almost any other measure taken to intensify agriculture by increasing yields) tends to augment productivity but at the cost of stability and resilience.  Namely, by selecting for just a few traits (usually yield or resistance to a specific pest), you attain uniformity in those traits (uniformly high yields, etc.), but also in other traits that may not be desireable (vulnerability to certain diseases, for example).  But beyond the desireability or not of a given trait, the very spread of uniformity makes the system more vulnerable.  Diversity is always more resilient, whether you're talking about a cornfield, a rainforest, or a stock portfolio.  So by opting for higher yield but less diversity, you create a system that is at once optimized for present conditions but also more precarious if the present conditions may change.  In a world of increasing climate fluxes and generalized unpredictability, this can lead to big problems.

A similar logic operates not just in agriculture but in many (most?) aspects of industrialized society.  We optimize processes for one or a few factors (usually profitability), but in doing so we create precarious systems that are dependent on all the inputs and surrounding infrastructure that allowed them to attain these optimized output levels.  The Toyota model of just-in-time parts delivery offers more efficient factory operations, but it is also more vulnerable to delays or any other unforeseen setbacks, since its very efficiency is a result of cutting out the margins for error.  Even athletes today are more streamlined, and hence more fragile.  An NFL running back may fine-tune his body for the most efficient, explosive short bursts of exertion, but if he breaks through the line for a 60-yard run, afterwards he needs oxygen from a tank like an asthmatic in the Andes.  Baseball pitchers are pitching faster than ever, but they pitch fewer innings and get injured more.  In short, efficiency consists in "trimming off the fat" from any process, but sometimes that fat wasn't just dead weight; it served a role as a reserve for resilience, or even for other functions that we had taken for granted before (pitcher stamina, or the myriad of unrecognized values provided by an unmanaged ecosystem).

This problem of throwing out valuable things when you "trim the fat" applies even to efforts to improve fruit and vegetable flavor in modern plant breeding.  For a long time breeders have been criticized for favoring industrial factors (durability, suitability for packing and transport, etc.) over flavor.  The 20th century plant breeding complex gave us tomatoes, melons, and many other plants that could be harvested in Guatemala or Chile in December, shipped to Appleton, WI, and still look good.  But they were picked unripe, and never developed much flavor.  Now it seems that breeders are refocusing on flavor, and turning in part to precise techniques like marker-assisted breeding to do the job.  I applaud their new focus, but the same risks of fine-tuning and efficiency apply here. 

Flavor is a complex thing, presumably influenced by lots of genes in any given plant (what is known as a quantitative, not qualitative trait because it is the accumulation of many factors and not drastically changeable by just one gene).  If you focus on just a few easily-inherited traits like sweetness or lack of fiber, you are sure to miss out on some other valuable aspects of flavor.  This is especially the case if you use marker-assisted selection to hone in on just a few genes "of worth", to the exclusion or ignorance of other flavor-influencing genes.  The result is clear to me when I eat fruits or vegetables in Colombia.  Those that are imported, or grown from imported, improved varieties, are much less complex in their flavor than local produce and varieties.  Whether it's broccoli, or blackberries, or mandarins, or grapes, the Colombian, unimproved (or less-improved) varieties have much more complexity of flavor, while the US or imported version tends to be simple--sweet, with most fiber and bitterness removed.  Sometimes this makes for more agreeable eating--US sweet corn tastes more like my idea of what corn should be, without the pastiness of "unimproved" corn.  But it is usually a blander, more cloying experience.  US mandarins (or those imported to the US from Peru or Israel or wherever) are uniformly, uninterestingly sweet, and have less of the fiber that makes them so good for your digestive system.  US or Chilean grapes just taste like sugar, vs. the mix of tart, bitter, and sweet you find in Colombian Isabella seeded grapes.  As I've hinted at in the case of fiber in mandarin oranges, sometimes these "bad" factors that breeders try to get rid of are in fact good for us.  The bitterness and other off flavors in more primitive crops and varieties often indicate some slightly-toxic secondary compound that serves the human body as a health enhancer in low doses.

The biggest problem with this quest for short-term efficiency, whether in horticulture breeding or Major League pitchers, is that it may damage the long-term prospects for future improvements. Honeycrisp apples are a big improvement on Granny Smith or Red Delicious, but the flavor is still much less complex than the many varieties I encounter at the fall farmer market.  If all breeding were focused on the few gene markers that seem to make Honeycrisp so good, we'd lose all the other complexity and nuance of the other varieties.  The diverse apple varieties in the US today are a cornucopia of differing shades of sweetness, bitterness, overtones of pineapple, orange, or other flavors and scents, plus the texture and color of the fruit, which also weigh in on the eating experience.  While lots of breeding work has been done on apples, it's mainly based on the many varieties that arose randomly in the US frontier days; breeders just mix and build on that base.  If we'd been focused on marker-assisted selection and efficiency in those pioneer days, we would never have developed all these varieties to begin with.  The random mutations and the careful observation of cider orchards by their owners would have been regarded as inefficient.