Plant breeding thus set the stage for sequential cycles of pest resistance and pest susceptibility of crop plants. We have no direct record of the consequences of this ancient ecological meddling, but myth and historical accounts tell of disastrous disease epidemics and insect outbreaks, so one can assume that from time to time large plantings of crops that were uniformly susceptible to a new kind of insect or disease fostered increases of that pest to epidemic proportions. Resistance genes were essential for crop domestication and monoculture but they did not guarantee perfect safety.

We have no record at all and little or no speculation about how the newly domesticated crops might have affected their wild relatives, which no doubt were growing in close proximity to the domesticates.

As years went by, breeders found that some kinds of resistance did not fail, and that such resistance often was less than complete; the plants suffered some damage but gave satisfactory performance overall. This longer lasting resistance was dubbed "durable" resistance. Further, the breeders discovered that durable resistance usually (but not always) was governed by several genes rather than by one major gene. The multifactorial kind of resistance has been called "horizontal resistance." The major-gene resistance has been called "vertical resistance."

The good news, then, was that breeders could identify and breed for durable resistance. The bad news was that the breeding was more difficult because several genes had to be transferred at one time, thus requiring larger populations for selection, as well as multiplying the usual problems with "linkage drag" (undesirable genes that are tightly linked to the desired ones). To this day, breeders use both kinds of resistance in varying proportions, according to the crop and where it is grown.

At first, breeders found and used resistance genes from the adapted, local landrace populations that also were the initial gene pool as a source of resistance genes for their new varieties. As years went by, these gene pools began to dry up and breeders looked further afield, turning to exotic (unadapted) landraces, and even to wild relatives of their crop. Sometimes they made extraordinary efforts to hybridize the domestic crop with a very distant wild relative—making a cross that could not succeed under natural conditions. Embryo rescue and even x-ray treatments were used to make "unnatural" crosses and derive breeding progeny from them. The breeders fooled around with Mother Nature; they moved genes farther than natural processes would allow.

But the breeders as a whole preferred to not breed from exotic varieties or distant and often wild relatives. They used exotic material only when there was no other choice. This preference was due not only to the difficulty of wide hybridization, but also to the fact that exotic germplasm exacerbates the problem of undesirable linkages. Few or none of the foreign genes—except the desired resistance genes—were suitable for the needs of high yielding, locally adapted varieties. But often the breeders had no choice; either they got the needed resistance genes from a distant relative, or they got nothing at all.

At about this time, breeders realized that it would be important to conserve remnant seed of landraces from all around the world, but especially from the centers of diversity of their crop. As farming worldwide grew more commercial, farmers turned more and more to professionally bred varieties that were better suited to commercial production, and in so doing they abandoned their landraces. If remnant seed of those landraces was not collected and saved in special storage facilities, the genetic base for crop breeding in the future would be drastically narrowed. Seed "banks" were needed. Through the efforts (especially in the 1960s and 1970s) of a few far-sighted plant breeders, seed banks were established in several countries and in international research centers.

So at the end of the 20^th century, plant breeding for pest resistance had laid out the genetic framework of vertical and horizontal resistance, and identified important sources of new resistance genes, i.e., plant germplasm from anywhere in the world. Sources were limited, however, to the crop species itself or its relatives, either wild or cultivated. All of the introduced genes therefore came from plants.

Plant breeders selected not only for tolerance or resistance to disease and insect pests, they also selected for tolerance to abiotic stresses such as heat and drought, cool temperatures, or nutrient imbalance. Much of this selection was involuntary; in selecting varieties with top performance over many seasons and many locations the breeders necessarily selected varieties with tolerance to the prevailing abiotic stresses of the diverse seasons and localities. In selecting for tolerance to environmental stresses, breeders necessarily changed the genetic makeup of the crop species, altering it still further from that of the original wild species, which had been restricted to certain environmental niches. Witness teosinte (the probable parent of maize), restricted to certain habitats in Mexico as compared to maize that now is grown in nearly every country of the world except Iceland.

Global distribution of crop plants often means that they are grown with no proximity to wild relatives that might intercross with them. Teosinte is not found in Germany or China, nor for that matter in the US Corn Belt. In other cases, however, wild species with hybridization potential coexist with their cultivated crop relatives, often as weeds. Canola, sunflower, and grain sorghum are examples of crops with hybridization potential with either a related species (canola with wild mustards) or with a weedy form of the same species (sorghum with shattercane, cultivated sunflower with wild sunflower).

The above discussion shows that plant breeders have changed the genetic composition of crop species to a large degree as they selected for pest resistance and also for resistance to environmental stresses. Such changes are in addition to the major phenotypic changes (e.g., non-shattering, uniform and fast germination) that were a consequence of domestication. What have been the consequences of such alterations, either on the crop species and its near relatives or on the ecosystems in which those species are grown? Twenty experienced plant breeders addressed this question as they responded to four queries I sent to them. My questions were:

1. Have the resistance traits been stable over time?
2. Have they led to undesirable consequences with respect to weediness of the crop or its relatives?
3. What have been the major sources of pest resistance genes as used in classical breeding (e.g., same species, related species, mutation)?
4. Are there relevant differences between the resistance genes currently being engineered into plants and those that have been transferred by conventional breeding?

In the following sections I summarize the responses from the breeders, and add commentary of my own.

It is important to remember that the phrase "stability of resistance" refers to whether or not a previously resistant variety is overcome by a particular species of disease or insect. It does not infer that individual resistance genes lose their power to hold individual pest biotypes in check. The resistance genes are stable, but new (or previously undetected) pest biotypes appear, with types of virulence that are not curbed by the now-outdated resistance genes. The variety succumbs to the disease or insect pest once again, albeit to a new race of the pest, and breeders say that the variety’s resistance was unstable.

Genes for herbicide resistance (the archetype example of potentially dangerous genetic transformation) are not necessarily imparted by means of genetic transformation. Such genes are found within crop species or their relatives, or have been created by means of mutation. These genes, bred into a specific crop variety, theoretically could move from the crop to cross-compatible weed species and impart unwanted herbicide resistance to the weeds. But in order to cause a new problem, resistance genes would have to introgress into weeds that had not contributed the resistance genes in the first place. This example shows how difficult it can be to decide whether or not a given resistance gene in a crop plant will increase competitiveness in weeds or make crop plants into weeds. Presence or absence of genetic engineering is not the major determining factor.

The breeders look to a future generation of engineered plant genes that will provide greater diversity and utility than genes presently available in any one crop. Genes from related taxa, from very distant taxa, or from within the crop species may be altered to provide improved resistance, but they will be plant genes rather than genes from extremely different organisms. It may be difficult to identify the point at which such new genes should be called "unnatural."

Until recently, plant breeders did not worry about how their breeding affected weeds, or whether their crops could become weeds. Weeds were looked on as potential sources of genes for pest resistance if they could hybridize with crop species, but almost no one thought about whether or not the population genetics of weeds could be altered by introgression from crop species. A very few students of crop evolution studied the weeds that may have been ancestors of cultivated plants. Plant taxonomists and ecologists usually ignored weeds because they weren’t considered as parts of natural ecosystems.

Genetic engineering has changed all of that. If genes from far afield can be added to crop plants, giving them marvelous gains in pest resistance, tolerance of environmental stress, or enhanced seed production, one can imagine that those transgenes could enhance the power of weeds in the same ways.

The analogy may not be as simple as it sounds, however. Two concepts must be clarified and data need to be assembled before one can make firm predictions.

• Will a gene that greatly enhances survival chances for a crop plant perform the same service for a weed? (Crops grow in crowded monocultures; weeds usually grow in dispersed "polycultures.")
• Will the presence or absence of genetic diversity within a crop or weed population, or among crop or weed species in a site, affect the utility of a given resistance gene? (Crop varieties usually are genetically uniform, weed populations are not.)
• Should one distinguish between dangers of imparting genes for resistance to natural restraints, such as disease or insect attack, and resistance to man-made restraints, such as herbicides?
• Do we have any reason to believe that selection for new (or previously undetected) kinds of herbicide resistance in weed species operates on different principles than selection for new (or previously undetected) kinds of virulence in disease or insect species?

The breeders, in answering my four questions, were considering these two main points and the subsequent questions that they raise. My sense is that they did not want to classify resistance genes into only two categories—natural or engineered. Further, the breeders said we know so little about the molecular nature of resistance genes that we cannot yet categorize them in any meaningful way. I think they do not believe that mode of transfer or kingdom of origin is a meaningful classification. But I did not get any hints as to what would characterize a meaningful classification.

Despite their reluctance to sort genes into "engineered-bad" and "natural-good," the breeders acknowledged that whenever we fool around with Mother Nature we get surprises, some of them bad. Therefore we need to look with caution at any novel breeding technology, predicting possible consequences as well as we can, with the modicum of data we may have in hand.

We need to know more about the effects of genetic background on gene action. Location within a genome seems important, and the entire genetic background seems important. We have little or no understanding of these interactions.

We need to know more about the consequences of hybridization of crop species with related weeds and the potential for introgression in both directions. Jointed goatgrass hybridizes with common wheat and viable backcross offspring can be produced. Have resistance genes from wheat moved into jointed goatgrass and changed its survival potential? A similar question can be asked for sorghum and shattercane, sunflower and wild sunflower, canola and mustards, or maize and teosinte.

So we must ask ourselves, do we have data to answer either of these key questions—effect of genetic background, or consequences of hybridization—or at the least do we have enough data to let us speculate from a firmer foundation than we have at present?

In my opinion, we have fragments of data for some crops and/or their weed relatives, but rarely do we have enough for firm predictions about gene introgression or about gene action in the genome or the population.

What are the consequences of adding new pest resistance genes to a wild species, either a weed or otherwise? How plentiful and how powerful must the genes be to change the genetic balance of the wild species, make it a stronger weed, transform a non-weed into a weed, or, conversely, reduce the weed’s viability as a competing population?

How about the "function" of related weeds as a reservoir of new biotypes of pest species, disease, or insects? Are the weeds more dangerous to crop plants when they lack resistance and so are a constant source of pest infection and infestation? Or are they more threatening when they contain many of the same resistance genes as carried in the crop species and therefore encourage the multiplication of new pest biotypes (biotypes that are not bothered by the weeds’ resistance genes)?

The recommendation arising from these questions seems obvious. Whenever a worrisome outcome seems likely but data are too sparse for firm conclusions, scientists need to work hard to fill the void. They need to plan the right experiments, gather the needed data, and publicize the results in both public and specialist media. And the public needs to provide the funds—the tax dollars—to support this work, since most of it will need to be done by scientists in public institutions.

A final consideration: sometimes the odds of a bad outcome from not doing a particular action may be much higher than the odds of a bad outcome from performing that action. Sometimes it may be better to take action with uncertain outcome than to stand still. Life always works on probabilities.