Beyond Red Queen Breeding

“It takes all the running you can do, to keep in the same place.”

— The Red Queen, in Lewis Carroll’s Through the Looking-Glass

Norman Borlaug won the Nobel Prize for leading an international effort that doubled wheat yield in the early 1960s. A few years later, Peter Jennings and colleagues at the International Rice Research Institute released the yield-doubling rice variety IR8. This period of rapid progress is usually called the Green Revolution.

Crop yields have continued to rise since the Green Revolution, but not fast enough to keep pace with expected increases in global food demand, driven by population growth, increased meat consumption in growing economies, and industrial uses for farm products. New crop varieties are released every year, but mostly to keep up with evolving pests and pathogens. Whereas the Green Revolution doubled the yield potential, or yield under pest-free conditions, of wheat and rice, much of today’s crop genetic improvement could be called “Red Queen breeding” — running in place rather than increasing potential yields.

Side-by-side comparisons of old and new varieties often appear to show major progress, but this can be misleading. For example, in a 1998 comparison, the newer rice variety IR72 out-yielded the original Green Revolution rice, IR8. However, while over the intervening 30 years evolving pests had reduced IR8’s yield, IR8’s yield back in 1968 was greater than IR72’s yield in 1998.¹ Ongoing Red Queen breeding is certainly critical to maintaining the progress we’ve made so far, but running in place won’t give us the yield increases we need for future food security.

Instead, new approaches will be required to take yield potential to new heights. A closer look at the evolutionary history of our crops can help identify where the most promising opportunities lie — especially when it comes to recent developments in biotechnology — and which areas represent expensive distractions that are likely to preclude meaningful growth.

What natural selection left behind

Today, overall yield trends are worrisome. A recent survey of rice, wheat, and corn yields found linear increases in most countries.² A linear trend in food supply cannot keep up with exponential increases in food demand from increasing populations, as Malthus noted,³ or from billions of people eating more meat as their incomes rise.

Nor can linear yield increases continue forever. At some point, each crop will reach the biophysical limit of what is possible. Some major crops seem to have reached their limits in high-yield countries; rice yields in California, Korea, and China and wheat yields in the Netherlands, the United Kingdom, and France have not shown any clear yield increase since about 1995.²

These “yield plateaus” may not reflect unbeatable biophysical limits, but they do suggest the need for new approaches. Transferring genes to crops from other species has made crops more resistant to pests or to herbicides, but with no significant increase in yield potential. Will new DNA-editing methods do better? Evolutionary history offers some insight.

For millions of years, the wild ancestors of our crops were improved by what Darwin called “natural selection.”⁴ They were often exposed to unfavorable conditions, like drought, and attacked by a wide range of pests. When mutants arose with greater resistance to drought or pests, they tended to displace their less resistant parents, except when resistance came with significant tradeoffs. Similarly, mutants with more efficient photosynthetic enzymes displaced those with less efficient enzymes. So the wild ancestors of our crops already had very efficient enzymes and a wide range of sophisticated adaptations to drought, pests, and other forms of stress even before humans started improving them.

Then, for thousands of years, humans further improved crops by selecting which seeds to plant and tend. This “artificial selection” operated similarly to natural selection, although humans sometimes selected different traits than nature did. For example, selection imposed by insect pests tends to preserve plants with defensive toxins, whereas humans have selected less bitter cucumbers. Similarly, natural selection favored plants that scattered their seeds, whereas humans only harvest and replant seeds retained on the plant.

So after millions of years of natural selection and thousands of years of selection by humans, what improvements are left for biotechnology? New techniques to edit a crop’s existing genes may be considered safer than past gene-transfer methods because the resulting genotypes could have arisen as natural mutations, but that also means that the “new” crop traits we get by editing existing genes have probably arisen repeatedly in the past, only to be rejected by natural selection.

Repeated rejection by natural selection means that a trait decreased survival or reproduction in past environments. Often, natural selection was constrained by tradeoffs. For example, conservation of matter limited the ability of wild plants to simultaneously increase root growth and seed production. Similar tradeoffs constrain plant breeding today.

On the other hand, as I argued in my book Darwinian Agriculture,⁵ some of natural selection’s rejects represent “low-hanging fruit,” or relatively simple opportunities to increase yield potential. These opportunities fall into two main categories.

First, some traits that were harmful in the past may be useful today. For example, our crops are descended from wild plants that were exposed to less than 300 ppm CO₂ for most of their evolutionary history. Today atmospheric CO₂ is 400 ppm and rising. Mathematical modeling suggests that it may be possible to increase photosynthesis 3% by improving adaptation to elevated CO₂.⁶

Second, because natural selection sacrifices plant community performance when it conflicts with individual plant competitiveness, reversing past natural selection can sometimes increase yield potential. Green Revolution wheat and rice varieties provide classic examples. For these crops, redirecting resources from stems into grain doubled yield. Shorter stems could also support heavier heads of grain, but in the wild, short plants generally lost out to tall plants. By selecting for all short plants, the Green Revolution resolved the shading problem and vastly improved yields.

Darwinian Agriculture discusses many other examples of traits rejected by past natural selection that have been or could be beneficial in agriculture. For instance, tassels, the pollen-producing male flowers at the top of corn plants, cast shade on leaves below. Plants with smaller tassels have greater yield potential, so why did natural selection miss this improvement? For the same reason that natural selection failed to reduce the extreme size of antlers in Irish elk as they were going extinct. Natural selection favors traits that increase individual reproduction, whatever the consequences for the species.

Plant breeders took 60 years to reduce tassel size 50% as a side effect of breeding for yield.⁷ Jennings, in contrast, doubled rice yield potential in just a few years by focusing on specific traits linked to individual-versus-community tradeoffs.⁸

Steering clear of dead ends

Recent improvements in our ability to edit specific genes could lead to similarly rapid improvements in yield potential, but only if we make the right changes. Recognizing evolutionary tradeoffs will help us not only identify promising opportunities but also avoid costly dead ends.

When we find a gene key to drought tolerance or pest resistance, for example, we may assume that increasing the expression of that gene will also increase yield under drought. But before investing millions of dollars in that approach, we should ask why past natural selection missed that “improvement.” If the tradeoffs that constrained past natural selection are not relevant to modern agriculture, fine. But we may find that a gene that increases yield under drought comes with costs, such as poor flood tolerance or reduced competitiveness with weeds. It is essential that we attend to those tradeoffs.

Consider recent attempts to improve drought tolerance in corn. Corn recently transformed with a “cold shock protein” from bacteria⁹ reportedly averaged 6% higher yield under drought than corn without the bacterial protein.¹⁰ Why?

During the critical “silking” stage, the “drought-tolerant” corn was actually growing in wetter soil than the control, because the cold shock protein slowed growth of leaf area early in the season.¹⁰ With less leaf area, the corn used water more slowly, leaving more water in the soil for the critical stage.

Evolution missed this improvement because a plant that leaves water in the soil “for later” will usually lose that water to a thirsty neighbor. But if a whole field of plants leaves some water in the soil, as in this case, they can all benefit. Timing is critical, however, as slower leaf growth might decrease yield under other conditions.

The future of gene editing

Up until now, most improvements in crop yield potential have come from accepting tradeoffs rejected by natural selection.⁵ But what about the future? Can we do better than 6%? Here are two ideas.

One key to improving yield under drought is to get as much photosynthesis as possible from the water available. As CO₂ diffuses into leaves for photosynthesis, water vapor inevitably diffuses out. But the ratio of photosynthesis to water loss can be four times greater in the morning, when it’s cool and humid, than in hot, dry afternoons.¹¹ If we develop crops that mainly use water in the morning, when water-use efficiency is greatest, they would photosynthesize less per day but could continue photosynthesizing for more days by making a limited water supply last longer. Under some conditions, this could increase yield by 25% or more. Natural selection would have missed this improvement because plants that saved water in the afternoon would have been out-competed by wasteful neighbors that used water all day.

A more radical innovation comes from Adam Smith’s hypothesis that specialization increases efficiency.¹² Why, for instance, do we need both corn and soybean to produce both oil and protein? Corn is an efficient producer of starch and oil, but it depends on fertilizer nitrogen to make protein. Soybean plants get nitrogen from symbiotic bacteria in their roots, but their ability to produce oil is much less efficient than corn.

Replacing much of the protein in corn seeds with corn oil could greatly decrease corn’s fertilizer needs. We would then require more protein from other crops, like soybeans. Decreasing the oil content of soybean seeds would free up energy that could potentially power increased symbiotic nitrogen fixation, allowing for more protein production, again without nitrogen fertilizer.

Broadly speaking, many opportunities exist to intervene in the coevolution of species in ways that enhance overall efficiency.¹³ Seizing on such opportunities will require that we home in on the ways that new genetic technologies enable us to think past natural selection in our design and harvesting of crops. It will also require that we remain cognizant of the tradeoffs that directed natural selection in the past and that will continue to dictate performance in the future.

Some day, we may understand the inner workings of plants and their interactions with their environment well enough to design and create plants with complex traits that never arose naturally, so were never tested by natural selection. In the meantime, we should focus on identifying and exploiting those cases where tradeoffs that constrained past natural selection need not constrain modern agriculture today.