Robin D. Graham, Ross M. Welch and Howarth E. Bouis

Department of Plant Science, University of Adelaide, Glen Osmond, 5064, Australia; U.S. Plant, Soil and Nutrition Laboratory, USDA-ARS, Tower Road, Ithaca, NY, 14853, U.S.A.; International food Policy Research Institute, 2033 K Street, N.W., Washington, D.C. 20006, U.S.A.

 

I. Introduction

II.. The need: The global perspective on human nutrition

A. The speed of change

B. The paradox of iron

C. Plant foods as sources of micronutrients for humans

III. New directions for world agriculture

A. The old paradigm

B. The new paradigm

C. A new paradigm

IV. Agricultural strategies for improving micronutrient concentrations in plant foods

A. Food systems

B. Fertilizers and organic manures

C. Varietal selection and plant breeding

D. Molecular-genetic crop transformation

E. Farm management

V. The case for plant breeding

VI. Germplasm resources, genetics, G*E, high yield, seedling vigor

A. Germplasm resources

B. Exploiting genetic variation in nutrient composition

C. Results of the CGIAR micronutrient project

D. Genotype environment interactions

E. Genetics of nutritional traits

F. Plant breeding for nutritional balance

G. Seed nutrient content

VII. The importance of bioavailability

A. The complexities of bioavailability

B. Bioavailability techniques

C. Intrinsic verses extrinsic labeling of plant foods for bioavailability determinations

D. Antinutrients in staple food crops

E. Promoter substances

F. Bioavailability issues concerning human studies

VIII Marketing strategies, economic and social constraints, cost-benefit analysis

A. Food Demand Patterns

B. Cost-Benefit Analysis

IX. Conclusions

A. Biological

B. Socio-economic

Abstract

Five years ago, with international funding, several international agricultural research centers undertook to explore the potential to improve the micronutrient quality of some staple food crops. Five objectives were identified; all needed a favorable result if breeding for higher micronutrient density in the staples were to be deemed feasible. Useful genetic variation to exploit was needed, the traits needed to be manageable in a breeding program (simple screening and high heritability), and stable across a wide range of environments if impact was to be large. Above all, the traits needed to be combinable with traits for high yield to ensure farmers chose the improved lines. Finally, it was necessary to show that the new types actually improved the health of humans of low nutrient status representing the target populations: that is, the extra nutrients are bioavailable to the gut. Today, only this last essential criterion remains to be fully satisfied. All other criteria are met to levels that lead us to claim that breeding for nutritional quality is a viable, practicable and cost-effective strategy to complement existing interventionist strategies. Subject to satisfying the last criterion, and results are encouraging, we call for a major funding initiative, and the installation of a new paradigm for 21st century agriculture: one espousing food systems that are highly productive, sustainable and nutritious. This paper reviews the case for and the rationale behind the project that is under way, gives an overview of the results to date and looks at the critical issues that still remain to be confronted.

What is agriculture to human existence? Agriculture provides the nutrients essential for human life. The reality of this is hidden when we use the less definitive term, ìfoodî; food may or may not provide all the necessary nutrients. If agriculture fails to produce adequate amounts of foods containing enough nutrients in balance to meet human needs, peopleís health will deteriorate, livelihoods will diminish, national morbidity and mortality rates will rise, national development will stagnate or decline, discontent and civil unrest will swell, political upheaval will ensue and human suffering will dramatically increase. Insufficient output of even one essential nutrient over a long time will produce these dire consequences. Therefore, it is imperative that the worldís agricultural institutions understand that the nutritional health of humans globally is largely dependent on the nutrient outputs that agricultural systems produce. In the twentieth century, agricultural institutions have not viewed themselves as suppliers of nutrients with an explicit goal to improving human nutrition and health. Such a view must be reached if we are to reduce malnutrition around the world and prevent much human suffering resulting from the ever increasing demand on our food systems for nutrient resources brought on by the increasing population pressure.

The worldís population has grown rapidly since 1950 swelling from 2.52 billion to over 6 billion today - an increase of nearly two and one-half times - because of dramatic reductions in mortality in populous developing countries (Figure 1). While the population growth rate has diminished from its maximum of 2.0 % per year during 1965 to 1970, population is still growing at a rate of 1.3 % per year, adding annually another 78 million mouths to feed. This growth in population is not expected to subside until well into the twenty-first century. Most of this increase in population will occur in developing nations, especially countries in South Asia, Africa, and Latin America. By 2050, the worldís agricultural systems will need to generate enough food to support the nutrient requirements of over 8.9 billion people (United Nations, 1998). To avoid widespread famine, this will require that agriculture, once again, match its earlier success in increasing food production demonstrated during the ìgreen revolutionî. By 2020, meeting the energy demand of an expanding population will require that farmers produce one and a half times as much food as they did in the early 1990s. This must all be accomplished with about the same land under cultivation but with dwindling water resources and declining soil fertility (Brown and Flavin, 1999; Pinstrup-Andersen, 1999). As we have said, growing sufficient food will not in itself assure adequate nutrition and healthy, productive lives for all.

Today, most agricultural systems in the developing world do not provide enough nutrients. Many fall short of supplying enough micronutrients (nutrients that are needed in only small amounts including 14 trace elements and 13 vitamins) to meet human needs, even though the production of energy and protein via cereal crops appears to be adequate to feed the world (Figure 2, Welch et al., 1997). However, the problems of poor food distribution means that over 800 million people do not receive enough energy (calories) and protein to meet their daily requirements (Table I) (Uvin, 1994). In 1995, of the estimated 10.4 million deaths among children less then five-years-old, protein-energy malnutrition was a causative factor in 5.1 million of these deaths (World Health Organization, 1999).

These figures, disturbing as they are, do not tell the full story: over half of the worldís population does not consune enough of each micronutrient in their food to support good health. Tables II, III, and IV show the extent of the micronutrient deficiency problems worldwide for only three micronutrients - iron, iodine and vitamin A. Other micronutrients, including zinc, selenium, folic acid, vitamin C, vitamin E, vitamin D, thiamine, vitamin B12 and niacin, are almost certainly impairing the health and productivity of large numbers of people (especially resource-poor women, infants and children in the developing world), but there are no data available to quantify the extent of deficiencies of these nutrients globally because of the lack of reliable and affordable clinical tests (Maberly et al., 1994; World Health Organization,1999).

Even though micronutrients are needed in minute quantities (i.e., micrograms to milligrams per day), they have tremendous impact on human health and well being. Insufficient dietary intakes of these nutrients impair the functions of the brain, the immune and reproductive systems and energy metabolism. These deficiencies result in learning disabilities, reduced work capacity and serious illnesses and death. Thus, micronutrient malnutrition is a serious global affliction that limits the work capacity of people and seriously hinders economic development (Anonymous, 1994).

Economic theory predicts that with increased income, individuals should be able to purchase more food and diversify their diets, especially with animal products, thereby improving their micronutrient status. This does not appear to be the case: for example, in Asia and Latin America, the availability of iron in food has declined even though income (Anonymous, 1994) and the availability and intake of foods containing high amounts of energy (i.e., cereals) have risen significantly (Figures 2 and 3). At the same time, iron deficiency in women, infants and children in resource-poor families has risen dramatically. Indeed, in South East Asia, iron deficiency now afflicts 98.2 % (over 1.4 billion) of the people in that region (see Table II)! Within the developing world, serious vitamin and trace element deficiencies persist and are not necessarily corrected by increased income within an acceptable period of time (Anonymous, 1994).

Dysfunction of the food system from low micronutrient output is affecting more people every day (see Figure 4 for examples of global trends in iron deficiency anemia) (Combs, Jr. et al., 1997). Agricultural systems must increase micronutrient outputs as a primary tool to eliminate micronutrient malnutrition (Combs et al., 1996; Welch et al., 1997). Without the cooperation of agriculture, finding sustainable solutions to this developing global nutrition crisis will not be possible.

In the past, supplementation and fortification programs have treated the symptoms of micronutrient malnutrition rather than the underlying causes. While many of these interventions have been successful in the short term, and especially so for the individuals reached by them, they have proved to be unsustainable and incapable of reaching all the people affected; indeed, they are least likely to reach those most at risk, namely resource-poor women, infants and children that live in remote areas either far from a clinic or that do not have ready access to processed and fortified foods. In spite of these interventions, the problem continues to increase.

In developing countries, the rise in micronutrient deficiencies is linked to the shift in cultivation towards dominance by cereals. Pressure on a fixed land base to produce more food has driven a shift in production toward cereals. High cereal productivity, the result of extensive research, has ensured that cereal production is relatively profitable with a relatively low risk of failure through disease, drought or post-harvest spoilage. Cereals are generally low in micronutrients, compared to many other food crops; consequently, food systems dominated by cereals are low in micronutrients. Moreover, we do not see that this trend can be reversed while the global population growth rate remains high.

To address micronutrient deficiencies in the comprehensive way that the figures above demand, several approaches are needed simultaneously. The requisite agricultural research to correct these deficiencies will take some time to come on line even if funded in proportion to the magnitude of the problem. Therefore, all current interventions, where cost-effective, should be continued to treat as many people currently at risk as possible.

The development of new food systems to deliver the required nutrients sustainably will take much effort and research. Allocation of funds for agricultural research must take into account the balance of food items that can optimally satisfy nutrient and energy requirements. Research must be devoted to the yield-improvement of nutrient-rich crops (e.g. legumes) that may have declined in production as a consequence of their being out competed by improved cereal cultivars. Finally, attention must be given to increasing the micronutrient density of the major staple food crops in order to help redress the decline in mineral and vitamin intakes. Results outlined in this paper show it is possible to shift the nutrient balance of cereals, and so of diets dominated by cereals, in the direction of better balance. Therefore, even when socio-economic factors make it difficult to change the diet (Gopalan, 1998), the nutrient balance of cropping systems where cereals figure prominently can be improved. Our paper addresses this last question, the issue the authors consider the most promising of the sustainable agricultural options that might be delivered in the shortest time.

Much of our experience has come from involvement in a feasibility study conducted within the last five years with colleagues at several CGIAR centers, including the International Center for Tropical Agriculture (CIAT), the International Center for Wheat and Maize Improvement (CIMMYT) and the International Rice Research Institute (IRRI). The philosophy of the approach developed for the study, significant results obtained and the major challenges still to be met will be reviewed. The overall aim is to make clear the way that all agricultural scientists need to adjust their thinking in order to help meet these challenges in the early part of the new millennium.

Underwood (1998) has presented a view from data of the World Health Organization (WHO) and the Food and Agriculture Organization (FAO) on the extent of micronutrient deficiencies in the human population, and how perceptions of these debilitating deficiencies within the nutrition community over the past few years have been extended to the economic consequences of them. More than three billion people are deficient in iron, 1.5 billion are deficient in iodine or live at risk, and some 250 million preschool children are deficient in vitamin A (a nutrient normally delivered in plant foods in the precursor form, b-carotene). While these international organizations do not have any figures on the extent of zinc deficiency (owing to the lack of a simple clinical screening procedure), specialists in zinc nutrition consider zinc and iron deficiencies to be of similar extent and distribution. These three elements and vitamin A are of first priority in primary health care. Other micronutrient deficiencies of growing concern to the human nutrition community include selenium deficiency and boron deficiency, though their roles in human nutrition are only emerging. Also, concerns for the vitamins, a-tocopherol, ascorbate, cobalamin and folate are increasing. Calcium, having some characteristics of a macronutrient and some of a micronutrient, is of much concern in early childhood and old age.

The four micronutrients, iron, zinc, and vitamin A are all involved in brain development in the critical sense that clinical effects can be seen at levels of deficiency commonly found in practice. When present in the pregnant female, or for up to 2 years post partum, permanent damage to brain development in the offspring is possible. Thereafter, further loss of cognitive ability is found, but this component of loss is reversible by better nutrition. These deficiencies also compromise immune competence, greatly increasing sensitivity to infectious disease-causing organisms, thereby increasing morbidity and mortality rates among those populations affected. Deficiencies of these nutrients curtail worker productivity; moreover, they interact in their function to aggravate the morbidity. For example, intake of vitamins A and C may increase the absorption of iron. All together, micronutrient deficiencies greatly contribute to the degenerative cycle of poverty as they limit the capacity of parents to earn an adequate income that limits the level of nutrition they can provide to their children, in turn limiting their childrenís work and cognitive potential (WHO, 1996).

These micronutrient deficiencies appear to be increasing in prevalence as the WHO data show. At the same time, FAO data show that diet diversity is declining as population pressure impacts on patterns of land use. Under high population density, cereals are often the most productive, most reliable and most profitable crops and their dominance of the landscape has increased (Welch and Graham, 1999). Cereals, when they dominate the diet are inadequate nutritionally, especially in micronutrients (both trace elements and certain vitamins) and certain amino acids. To us, this restriction in diet diversity appears the most likely reason for the emergence of micronutrient deficiencies over the last 15-20 years in such extent and severity. Cereals provide by far the most calories to humans and so dominate the diets of most resource-poor people in the world. One significant consequence of this premise is that the situation is unlikely to improve in terms of diet diversity until population declines, which in turn is unlikely for nearly 50 years, barring calamity. The only conclusion we can reach, then, is that a major effort must be made to improve the micronutrient content of cereals, and other staples, and there are ways that this may be done by both agricultural and food-processing industries. In many developing countries, agricultural solutions are expected to be the dominant answer as infrastructural limitations and poverty will curtail the impact of food processing approaches.

The time it takes for changes in the nutritional status of a population to emerge as clinical deficiency has recently been shown in the district of Chakaria, Bangladesh. Here a few years ago, a French medical team (Fischer et al., 1999) reported severe rickets in children; the diagnosis was calcium deficiency in the diet rather than vitamin D deficiency, the common cause. A surprising finding about this case is that there is no word in their local language for rickets, indicating recent etiology of this disease; indeed, there is hardly anyone in that population today over 25 displaying residual signs of rickets that normally afflicts children during growth spurts. Thus, the problem has become significant only in the last 20 years or so. An obvious cause has not emerged from the studies to date but a likely scenario is in the changing cropping systems as increasing population has dictated changes in land use. These people describe themselves culturally as rice and pulse (grain legume) eaters, but today their fields are dominated by rice; and rice as eaten (milled and polished) has much lower calcium contents than the pulses, as the analyses of foods from a local market and household showed. While orthopedic surgery can help the existing children in some moderately wealthy families, this is not a sustainable or desirable long-term solution. Food systems must be changed to deliver the required nutrients (in this case, calcium) in adequate amounts. Better agronomic practices could increase rice yields so that some land can be spared for production of foods that help to balance the diet nutritionally. Aditionally, rice cultivars vary in their grain calcium and high-calcium lines could be introduced to contribute to the supply (in the case of calcium, it can be easily calculated that high-calcium rice will only contribute perhaps 20% of the required shortfall).

In addition to the gross visible deformities of the calcium-rickets problem, all evidence suggests that several other micronutrient deficiencies (e.g., riboflavin, vitamin A, zinc, and iron) afflict most sectors of this population and all these nutrients must be part of a satisfactory, sustainable food-systems approach for this community. This study demonstrates how fast the nutritional status of a population can change, and the need to view agriculture as a source of all nutrients, not just calories.

One of the paradoxes of iron is that the earthís crust itself has an abundance of iron, but aerobic life on earth struggles to get enough because of the insoluble nature of iron in an oxidizing environment. All life forms together only require a mere one trillionth of the total in the earthís soils. On exposure to water and air during weathering of rocks, iron becomes oxidized and extremely insoluble; it is not leached out of soils into the oceans by rain, except catastrophically by erosion. The extreme insolubility and immobility of iron in aerobic environments ensures its presence in all soils providing universal stores of iron for soil microbes, plants, animals and humans. Yet, its very insolubility in the presence of air denies ready availability to living aerobic organisms. Sophisticated biological mechanisms have evolved to allow either dissolution and absorption of iron from the environment or acquisition of iron by parasitism and predation upon other living organisms. Many higher plants (excluding cereals) rely on root-cell membrane-bound ferric reductases to acquire iron from soil solution in the form of the ferrous ion. In the case of the cereals that are low in these reductases, the chelates excreted by their roots are highly sophisticated phytometallophores that are absorbed intact by root cells via a specific ferric-phytometallophore transport system in the cell plasma membrane. These mechanisms allow plants to acquire iron that is otherwise tightly bound in soil, and this iron is then fed into the food chain.

Another paradox of iron is that it can be a highly toxic element when absorbed by aerobic organisms in excessive amounts because it can undergo a series of chemical reactions with oxygen, ultimately producing highly reactive and damaging oxygen free radicals (e.g., the hydroxyl free radical). Thus, iron uptake by all aerobic life is highly regulated to prevent excessive levels of iron from being accumulated in cells. For this reason, trying to make food crops absorb significantly more iron than that required to meet metabolic demands is difficult and must be done with care.

Iron deficiency occurs mostly in infants, children and pre-menopausal women, that are dependent on plant foods as their major source of iron. Monogastric animals, particularly young pigs, are also commonly deficient in iron if housed indoors during winter and fed feeds primarily composed of cereal grain. Grazing animals appear to pick up significant iron by soil ingestion and this iron is made available either by rumen microflora or by the acid conditions in the true stomach. Because of their iron absorption efficiency, cereals provide an important entry point for iron into the food chain, even though the concentrations of iron in the grain are quite low compared to meat-iron sources. The bioavailability of iron in plant foods is important as not all iron can be absorbed and utilized from the monogastric gut. Some iron is required by the gut microflora and fauna, and some is simply unavailable for absorption because it remains in forms that are insoluble.

Humans require the following trace elements and vitamins for growth and health: arsenic, boron, chromium, copper, fluorine, iodine, iron, manganese, molybdenum, nickel, selenium, silicon, vanadium and zinc; water soluble vitamins, ascorbic acid, biotin, cobalamin, folic acid, niacin, pantothenic acid, pyridoxine, riboflavin, thiamin; fat-soluble vitamins, retinoic acid, calciferol, tocopherol, phylloquinone and menoquinone. The bioavailability to humans of the fat-soluble vitamins requires adequate fat in the diet (Combs, 1998). Undoubtedly, more essential trace elements and vitamins will be added to the list in future.

Widespread deficiency of these nutrients in developing countries curtails economic development and stability. Food diversity is the traditional means of ensuring a balanced diet containing all these micronutrients, but this is becoming increasingly difficult as the global population continues to increase. In announcing the urgency of addressing dietary deficiencies in human populations, WHO/FAO (United Nations, 1992) called on governments to fund food-based solutions. Plant foods are able to supply all these micronutrients in adequate amounts with the exception of cobalamin that comes mostly from animal products and bacterial contaminants on plant food.

The path of development set down for modern agriculture to follow into the 21st century changed only a decade ago, but we need further change for the new millennium. The necessary changes have much to do with the impact of the green revolution and its perceived failings; so far, the changes in agricultural thinking have addressed only one of the major shortcomings of modern, technological agriculture, namely, the environmental concerns. Evidence, increasingly alarming, is showing us that in spite of our technology, modern agriculture is failing to deliver nutritious food.

Progress in agriculture in the last 100 years has been science-driven, producing an increasingly technological operation which has shown itself capable of achieving each lift in productivity needed to feed the world and even to provide more calories per person. The technology included new varieties, chemicals ranging from mineral fertilizers to pesticides to synthetic plant hormones, and machines to supplement and replace the labor force. This technological revolution in agriculture we call the production paradigm and it culminated in the ìgreen revolutionî, a series of highly orchestrated, global strategies developed under the threat of starvation, to expand the global production of food, ensuring we could feed the increasing human family. This tremendous effort began in the 1960s and achieved adequacy in world food production in just two decades, an effort for which one of the leaders, Dr NE Borlaug, received the Nobel Peace Prize in 1980.

Even before this monumental international effort for food sufficiency got under way, Rachel Carson had published her famous book, Silent Spring, depicting the threat to our environment from the indiscriminate use of toxic chemicals, much of it in "modern" agriculture. Her lead gradually became a popular movement as more and more instances of the danger and damage were publicized. Not only was there concern for the environment generally, but there was concern that the existing orientation to agricultural production and "efficiency" was threatening the resource base of land, soil, air and water through processes such as loss of soil fertility by erosion, acidification, salinization and desertification. It was in the mid-1980s that the existing philosophy was largely overrun and a new paradigm installed: we must have high productivity while preserving or improving the resource base of agriculture and the environment - the so-called sustainability paradigm. The ìgreen revolutionî failed initially to place enough emphasis on the sustainability of the increased productivity they set out to achieve.

As soon as agricultural research within the new philosophy of sustainable agriculture had become accepted, with remarkable global consensus, than new concerns have arisen (though there is still much to be done to realize our objective of sustainability). In a way, like the concerns about the environment, these new concerns have been brewing for some time, but they have been brought into sharp focus by the data and statistics of the nutrition community and the WHO in the last few years. The food supply, while it has been sufficient (wars and the like excepted), is simply not nutritious (section II). No previous disease or deficiency has ever affected over half of the total population, in reality most of the women and children of the Third World, together with a surprisingly large number in developed countries. Some blame the ìgreen revolutionî in that the new highly productive cereals did not provide nutrient balance, but we believe it is more rational to attribute any blame to population increase. As the basis of the effort to increase food production in poor countries, highly productive cereals have displaced other crops that are higher in iron. For example, in India where cereal production has increased more than four times in the two decades since 1970 (while the population has less than doubled), pulse production actually declined. To put an economic perspective on the magnitude of this problem, the World Bank has estimated that iron deficiency in India costs that country about 5% of its GNP annually, and 11% in Bangladesh. During the ìgreen revolutionísî push towards food security, precious little thought was given to nutritional value, and certainly almost none to the iron content of the new cereal varieties being bred, let alone to the iron content of the changing diets. To be fair, it must be said that the contemplation of mass starvation, inevitable without the ìgreen revolutionî effort, is far worse than the problems we must now tackle. The challenge now is to support a new paradigm for agriculture - an agriculture which aims not only for productivity and sustainability, but also, for balanced nutrition, what we have called the productive, sustainable, nutritious food systems paradigm.

There are a number of ways in which the micronutrient density of crop plants can be increased, and the necessary research to introduce these strategies successfully will require considerable laboratory support. Not only will agronomists and analysts need to understand the global situation and the complexities of a balanced human diet, but government policies will be needed to develop consumer awareness. Food industries need to be aware of the issues, whether they major importers or exporters of food and food products.

Historically, balanced diets have been formulated by nutritionists who selected from the foodstuffs available a combination of food products that together deliver all known dietary requirements in reasonable proportions. We have seen above, however, that population pressure has changed the balance of foodstuffs produced. Sometimes this has been favorable to health (Gopalan, 1999), but it is now the likely cause of widespread and increasing micronutrient malnutrition in many countries. Given that these changes were driven by economics, food systems for the 21st century will need to consider diet diversity in an economic way rather than from the viewpoint of the dietary ideal. Whatís needed are nutritionists working with agriculturalists, economists, policy makers and sociologists to consider food systems that are economically viable yet can deliver more micronutrients in an acceptable food supply. This will require consideration not only of the demand for calories and protein, but also their suitability in the cropping system and the labor demand and timing of any proposed new crop that might shift the diet in the desired direction. The economics of production, dietary balance and consumer acceptability will be the final arbiters of change. Developing better diets in isolation from the economics and practicalities of the food systems themselves is unlikely to succeed until population pressure is no longer the primary driver.

Fertilizer technology and use is widely understood and appreciated in modern agriculture so it is a major vehicle for change in plant mineral content and food quality. The density of several micronutrients can be usefully enhanced by application of the appropriate mineral forms (Allaway, 1986; House and Welch, 1989): these are zinc, iodine, selenium, copper and nickel. The effect of zinc fertilizer is shown in Table V, which also presents genotypic differences in ability to load zinc into grain. However, in this example, Excalibur is the best adapted to the zinc-deficient soil and has the highest yield with and without fertilizer added but does not have the high concentration of zinc in grain as the new breedersí line, VL660. It is desirable to combine the traits of high yield (tolerance to the deficiency in the soil) and high nutrient density in grain (see section V). Manganese can be increased only by late foliar applications to the generative tissues (Ascher, 1994).

However, because of its rapid oxidation in soil and because of its low mobility in phloem, soluble ferrous fertilizer is ineffective in increasing the iron concentration in plants, especially in the grain that develops months after application. Foliar applications are not much better. Chromium, boron and vanadium are also ineffective fertilizers because of their low phloem mobility (Welch, 1986). All the vitamins of plant origin are synthesized de novo by the plant and are not a consideration as fertilizers. Thus, for many of the mineral nutrients of concern, fertilization is a useful strategy, while for iron and the vitamins, it is not (although adequate nutrition is a prerequisite for optimum vitamin biosynthesis in plants). Nitrogen fertilizer can increase vitamin content, but excessive application appears to be generally counterproductive (Salunkhe and Desai, 1988) and is to be avoided in any case for well known agronomic and environmental reasons. Potassium fertilizers often increase the concentration of vitamin C. Lime is used to increase the pH of acid soils and can enhance the calcium concentrations in plants, but its use is well known to decrease the concentrations of the micronutrient cations. On the other hand, organic amendments, especially farmyard manures, increase the concentration of many nutrients and can be seen to enhance the nutritional value and nutrient balance of plant foods.

Exploiting the genetic variation in crop plants for micronutrient density is one of the most powerful tools we have to change the nutrient balance of a given diet on a large scale. Whereas conventional supplementation and fortification programs have been shown not to work well where the infrastructure is inadequate, delivery of nutrients in staple foods is certain to reach even the most disadvantaged. New varieties are for farmers the most widely and quickly adopted, and therefore the most effective, of modern agricultural technologies. This subject is discussed in detail in sections V and VI of this review.

Recent achievements in genetic engineering of new crop plants holds the promise of dramatic improvements in nutritional balance from fewer dietary components where the latter may be dictated by the agricultural consequences of excessive population pressure on food producing land. Genetic engineering involves first introducing foreign DNA into a crop species or artificially modifying its own DNA to achieve desired results. The process is such that often only a single cell is transformed in the desired way and it is then necessary to regenerate a whole plant from that cell by potentiating its whole complement of DNA through certain tissue culture techniques. An exciting recent example is the transformation of rice in Japan to have higher iron content in the endosperm of the grain (Goto et al., 1999). In this work, the gene from soybean that controls the synthesis of phytoferritin, a large iron-containing protein, was inserted into the rice genome. This was accomplished using an endosperm promoter, the rice seed-storage protein glutelin promoter, GluB-1, that ensures the expression of phytoferritin only in the endosperm, where it can be most effective in human nutrition. On average, the iron concentration in the rice grains was doubled, but potentially this advantage could be even greater after milling as up to half of normal iron in rice is in the outer layers and subject to loss in milling and polishing. Presumably, much of the phytoferritin would be retained after milling because it is located in the inner tissues (endosperm). While there is considerable debate about the bioavailability to humans of ferritin-iron from plant sources, and while the stability of such transformants is still a problem for this emerging technology, this success is a most encouraging pointer to the future. Indeed, several groups are attempting to transform other crops in the same way, and yet more are trying to transform food crops with hemoglobin-like molecules (see section VII).

Crop rotation is an important tool in food production as well as soil management. Rotation of cereals and grain legumes allows for better management of difficult weeds and the input of nitrogen fixed by the legume is important to the overall nitrogen balance and cost. At the same time, these two crops provide some nutritional balance to the diet. For example, while cereals are low in iron, manganese, copper, cobalt, calcium and magnesium, grain legumes are relatively high, and while grain legumes are very low in sulfur amino acids, the cereals are relatively better. In south Asia, cereal production has increased four times faster than that of grain legumes in the last 30 years (FAO production statistics). More productive grain legumes could redress this imbalance. Both these groups of staples are low in many important vitamins that can be supplied by vegetables and fruits (Combs, 1995), but the introduction of more balanced cropping systems in near-subsistence circumstances must occur in a suitable economic environment if they are to survive and have an impact. This involves a complex balance of factors that requires participation from several disciplines.

Tillage systems are changing, but the impact on plant composition is not currently considered, especially in respect to nutritional quality. We know that the shift towards minimum tillage tends to decrease nutrient density as deficiencies are more common, especially of nitrogen and phosphorus, but also of micronutrients. The long-term effects of such changes in cropping systems need to be fully understood.

The breeding of crop plants for nutritional quality has been strongly questioned by many plant breeders because the history of past attempts is not encouraging. The primary nutritional objective in the last 50 years has been to improve the protein content and quality of staple crops, and as we shall see, this has been quite difficult and has achieved only limited success. Nutritionists also question this strategy as, traditionally, they have advocated attaining nutritional balance by dietary diversity. The collective wisdom has therefore been that in a world of expanding population plant breeders of staples should focus on delivering calories, leaving nutritionists to deal with the need for dietary diversity to achieve balanced nutrition.

The issues of this debate are being raised again in the International Agricultural Research Centers (IARCs) while means of dealing with the extent and severity of micronutrient deficiencies in human populations are considered from every angle. As the IARCs debate the value of breeding for micronutrient-dense cultivars of staple crops, it is worth remembering that in order to make a sound economic assessment of the potential contribution of breeding to controlling micronutrient malnutrition, more biological information is needed than currently exists. The feasibility and costs of breeding can only then be assessed in relation to the magnitude of benefits to malnourished consumers. These benefits must be assessed from feeding trials (bioavailability studies) on the best germplasm identifiable. This section is intended to be a contribution to the debate, but more, it sets out the deficiencies in our knowledge on which we must base an informed opinion.

It is something of a circular argument that an economic analysis should be done to justify research expenditure in this area but some research expenditure is needed in advance to generate the data on which a realistic economic analysis can be based. We argue that efforts to identify and collect the essential biological information on germplasm resources, heritabilities, efficacy of selection criteria or molecular markers, as well as bioavailability of the extra micronutrient contained in the best germplasm found is justified ahead of a wide-ranging economic analysis.

The fundamental assumption in proposing a plant breeding contribution to overcoming micronutrient malnutrition is that nutrient density traits must be delivered in cultivars of the highest yield. In order to have maximum impact, top yielding lines are needed to convince local farmers to grow them when the target consumer is in no position to pay a higher price for quality. To do this, a major increase in breeding costs will be necessary in order to maintain progress in yield and quality concurrently, and the results of exploratory research will need to justify that increase to the donors supporting the major breeding programs in the international centers.

For a plant breeding approach, as with fertilizer use, laboratory support is essential. Selecting among existing varieties is the simplest approach. Results of a study at CIMMYT show that, of the wheat varieties released by that organization over 40 years of breeding, the best were about 20% higher in iron and zinc concentration in grain than the lowest (I. Ortiz-Monasterio, 1997, in Graham et al., 1999). A 20% difference is likely to be important to deficient consumers of this staple, provided they are capable of absorbing the iron. Surprisingly, the overall iron concentrations in grain were not depressed by the increased yield achieved over 40 years of breeding. The same seems to be true for Phaseolus beans. Over the whole process of domestication and modern plant breeding, despite a huge increase in yield potential, the mean concentrations and the range across 1,100 varieties and 200 wild types of beans were about the same for both iron and zinc (S. Beebe, 1996, in Graham et al., 1999).

Predictably, the potential for nutritional enhancement by deliberate selection within a breeding program is much greater than by selection within currently available varieties (Gerloff and Gabelman, 1983; Graham and Welch, 1996). We have found four or five fold variation between the lowest and highest micronutrient cation concentrations in the grain of several hundreds of accessions from the germplasm banks of the major cereals (Graham et al., 1999); the highest concentrations are about twice those of popular modern cultivars. For b-carotene in maize and cassava, the range is much greater than that for iron (Brunson and Quackenbush, 1962; Iglesias et al., 1997). In all crops studied, it is possible to combine the high-density trait with high yield, unlike protein content and yield that are negatively correlated. The micronutrient traits are stable across environments and the genetic control is relatively simple. For example, high iron density in rice is linked closely with aromaticity, a single gene trait, making selection easy in early generations (Graham et al., 1997).

Screening for micronutrient cation density is easy using the inductively-coupled, argon-plasma optical emission (ICP-OES) after digestion in nitric-perchloric acid. The cost is considered high for breeding work and ways of lowering the cost for dedicated servicing of plant breeding programs is needed. The alternative is to use the ICP-OES capability to help develop molecular markers that are expected to cost around $US 2.00 per sample tested, somewhat less than the ICP-OES laboratory. Ultimately, total nutrient concentration is not the objective, but utilizable nutrient in the human gut. The bioavailability of the nutrient in high-density types must be demonstrated, and this requires special feeding trials by human nutritionists. These are so expensive that they are only possible for advanced lines scheduled for release.

The micronutrient density of seeds is important not only for human nutrition, but also for animal production and the nutrition of the seedling in the next generation. A vigorous crop is established by seeds with a high density of nutrients, including micronutrients (Welch, 1986; Crosbie et al., 1993; Rengel and Graham, 1995; Mousavi-Nik et al., 1997), that confer better resistance to stress and disease, leading eventually to higher yield. These effects are most pronounced in micronutrient-deficient soils (Weir and Hudson, 1966; Gurley and Giddens, 1969; Longnecker et al., 1988; Yilmaz et al., 1997).

The prospects for improving the provitamin A carotenoid density of staple plant foods are very good indeed. Yellow types are well known in all the usual white starchy plant foods except rice, although the genetics are complex and somewhat obscure. Rapid breeding progress should be possible since genetic gain in carotene content can be visually estimated with accuracy (Simon, 1992). Moreover, the carotenoids tend to be endosperm-stored and therefore less subject to the milling losses seen for mineral content in cereal grains. It is necessary, therefore, only to characterize the parental material to establish that the sources of pigmentation do have high provitamin A activity in humans.

VI. Germplasm resources, genetics, G*E, high yield, seedling vigor

Germplasm resources can be exploited to improve the nutrient output of food systems in several ways. The simplest is by designing new cropping systems with better nutrient output, for example, by re-introducing micronutrient-dense pulse species into cereal-dominated production systems. However, in many places the trend has been exactly the reverse, driven by the forces of population pressure and economics, and so cannot immediately be sustainably reversed. There is urgent need to reduce the risk to farmers associated with pulse production and to enhance the yield and reliability of most pulses while at least maintaining or further enhancing their nutrient density. Through increased yield, the caloric output of the pulse phase of the rotation contributes both to the overarching need for calories and protein, and to adequate returns for the farmer. Only then will these cereal-legume rotational systems be productive, sustainable and nutritionally enhanced.

Within any plant breeding program, selecting micronutrient-dense lines among existing varieties is the first approach. Results of a study at CIMMYT show that of the wheat varieties released by that organization over 40 years of breeding, the best were about 20% higher in iron and/or zinc concentration in grain than the lowest (I. Ortiz-Monasterio, 1997, in Graham et al., 1999). A 20% difference is likely to be important to deficient consumers of this staple, provided the % bioavailability is not the same. Implementation of this simple strategy needs the establishment and acceptance of the importance of nutritional quality as a breeding objective by the breeders concerned and/or by agronomists who influence the choice of crops and cultivars in farming systems. Most important is the adoption of the objective by the breeding program itself, as in choosing lines for registration and release, breeders often have several sibs to choose from on the basis of small differences in expression of many traits. In making the decision, differences in nutrient output could be a priority when the breeders and their stakeholders rate it highly enough.

The potential is much obviously greater for nutritional enhancement within the breeding program by deliberate use of nutrient-dense parents followed by selection in segregating populations - than by selection among currently available varieties (Gerloff and Gabelman, 1983; Graham and Welch, 1996). The data in Figure 5 are typical of the genetic variation in seed nutrient concentrations among cultivars and breedersí lines in field trials. In this field trial, entries were mostly released cultivars or advanced lines from local wheat breeders, though some reference lines of known tolerance to soils deficient in zinc were included.

INSERT FIG. 5 HERE

The best varieties were about 2.5 times higher in iron concentration than the poorest, and the most agronomically zinc-efficient lines were among the more iron-dense lines. Since the entries in this trial were not genetically as diverse as one would find in a germplasm bank, the range of concentrations is not as great as that described below. Moreover, the mean concentration is relatively low as soil type and other environmental factors do contribute to the expression of this phenotype, so the range of values in this high pH sand are lower than that seen in better soils. However, the best varieties tend to be the best in all environments. Since environmental factors affect seed nutrient concentration, the only true and useful comparison of the composition in seed of many genotypes comes from growing them together in the same field in the same season and analyzing the seed produced rather than the seed sown, as the latter may have come from diverse environments or seasons. This technique was followed throughout the study described below. Within experimental error, all differences could be described, prima facie, to the genotype effect, although effects of unusual genotype x environment interactions cannot be ruled out.

Given that such genetic variation was known to exist and has been repeatedly demonstrated, we might ask the question, what is the full range of variation in a major germplasm bank for the traits of interest? The CGIAR was funded by the Danish International Development Agency to explore the potential in its germplasm banks of the major staple crops for micronutrient density. As a major report was published recently (Graham, Senadhira, Beebe, Iglesias and Ortiz-Monasterio, 1999), a summary of our findings and interpretations is presented here. (Besides the authors named and the current authors, Drs G. Gregorio, W. Roca, H. Cabellos, and M. Banziger have contributed to this effort.)

Wheat: From a survey of wheat germplasm, Dr Ortiz-Monasterio found four or five-fold variation between the lowest and highest iron and zinc concentrations in the grain among several hundreds of accessions. The highest concentrations were about twice those of popular modern cultivars, so that a factor of two is the likely potential for improvement over the currently grown varieties. However, Ortiz-Monasterio (1998) has found iron and zinc density in wild relatives of modern bread wheats to be even greater, up to 50% more again. Since the seeds of these wild progenitors are often small, it remains to be shown whether their iron and zinc density will still be fully expressed when the trait is introgressed into high-yielding cultivars.

Generally, what has been found for iron and zinc, has also been shown for other minerals in grain. Moreover, there is a positive correlation of iron and zinc concentrations in grain. This significant, but not perfect, ìlinkage dragî means that screening for iron alone is likely to improve the concentrations of several minerals, but the correlation is significant for iron, zinc and calcium, three cations of special importance to human nutrition. Lines especially high in both iron and zinc have been found, and moreover, among high-yielding lines, it has been demonstrated that the high-density traits can be combined with high yield. This is a most important point in view of our strategy of getting these benefits to consumers via varieties that farmers will want to grow. In another study, Ortiz-Monasterio reported that, of all major varieties released during more than 40 years of breeding for improved yield without any consideration of iron or zinc density, there was no trend in iron or zinc concentrations over time (Graham et al., 1999). The fact that we may be able to raise the concentrations of all or several minerals together, and especially the more critical ones, will benefit consumers in terms of nutritional balance as well.

Rice: The findings in the rice project are particularly encouraging. Iron density in rice varied from 7-24 mg kg-1 (all concentrations reported are on a dry weight basis) and zinc density from 16-58 mg kg-1. A benchmark was established in that nearly all the widely grown ìgreen revolutionî varieties were similar, about 12 and 22 mg kg-1 for iron and zinc respectively. The best lines discovered in the survey of the germplasm collection were therefore twice as high in iron and 1.5 times as high in zinc as the most widely grown varieties today. High iron and to a lesser extent, high zinc concentration, were subsequently shown to be linked to the trait of aromaticity. Most aromatic rices such as jasmine and basmati types are high in both iron and zinc, and as before, generally in most minerals (Senadhira and Graham, 1999; Graham et al., 1997, 1999). The close linkage to aroma suggests iron density in rice expresses as a single gene trait since aroma is itself controlled at a single locus. As in other crops, these micronutrient density traits have been combined with high yield.

In addition to genetic variation in nutrient density in rice grains, there are genotypic differences in the relative loss of iron and zinc in the milling process. Both because iron is inherently low in rice and because milling removes half or more of that, rice is the poorest in iron of all the cereals.

Maize: A cereal with enormous yield potential, maize offers equally great potential nutritionally, although this is not generally accepted. Generally, maize kernels have higher iron and zinc concentrations than grains of the other high yielding cereals, though as the project leader, Dr Banziger has shown, the potential to increase these concentrations by further deliberate selection within high-yielding germplasm appears to be less than in wheat or rice. However, multiple aleurone layer (MAL) types are known and have been shown to have higher concentrations of all minerals than normal single layer maize (R.M. Welch, unpublished). Modern maize also has much variation for protein and the modern high-lysine types offer better amino acid balance in this limiting nutrient. Tissue culture selection (Phillips and McLure, 1985) has made available high methionine germplasm that further enhances the quality of the protein and the absorption of iron and zinc in the gut (section VII). Most importantly, in maize there are high-carotene types (yellow maize) that are very high yielding and available in combination with high lysine and the other nutritional traits. Bearing in mind the synergistic interactions among nutrients, this represents the best-balanced nutritional package available among the most productive staple crops. While it is obviously deficient in vitamin C and tryptophan, a nutritionally enhanced maize has the potential to help balance a poor diet that is based on very few components.

Beans: In beans, Dr Beebe has shown considerable genetic variation in mineral concentrations among both wild beans and modern cultivars but domestication has not changed the mean concentrations of iron and zinc in the seeds, nor the range. This is most surprising, given the marked increase in yield between wild and modern beans; obviously, the loading of iron and zinc into the bean seed has increased in proportion to the yield over the time of domestication. The concentration of zinc in beans is one of the highest among vegetable sources, and is nearly equal to dairy products (Pennington and Young, 1990). Unlike wheat and rice, the concentrations of iron are generally even higher than those of zinc, but like the cereals, iron and zinc are positively correlated. However, unlike the cereals, there is no correlation with calcium. The concentrations of iron at the high end (100mg kg-1) are much higher than in the cereals, although relatively, the increase that is possible through deliberate selection is less. Again, like wheat and rice, the high-density traits are fairly stable across environments, although, as in all crops, the environmental effect is also highly significant.

Beans are high in phytate and tannins, both considered antinutrients that inhibit the absorption of dietary iron and zinc from the gut. However, in the most comprehensive studies done to date, the bioavailability of the iron and zinc to rats and to humans is not dependent on the concentrations of either anitnutrient, even though the amounts present varied over a wide range (see section VI for details). Genetic variation for these antinutrients exists but its exploitation depends on further bioavailability data to justify the effort and to define whether to increase or to decrease these constituents. As discussed in section VI, there are nutritional arguments for not decreasing their concentrations. On the other hand, promoters of absorption, such as the sulfur amino acids, also vary with genotype and have heritabilities that make breeding practical, if needed. Since sulfur amino acids are generally low in beans, as in many pulses, such a breeding effort would appear worthwhile, though the potential for increasing them may be only of the order of 20-30% of the current mean concentrations.

Cassava: Like maize, cassava has much more potential nutritionally than is generally accepted. Its poor reputation is probably because the root has low protein concentration, and dates from the 1960s when protein was considered the primary limiting factor for nutrition. Cassava leaves and roots generally contain cyanogenic glycosides and so require careful processing to eliminate the cyanide before consumption. The toxin is apparently useful in protecting the crop from insects, wild pigs and other animals, so the crop can be stored in the ground as a drought reserve. However, sweet types exist that require less in the way of special preparation to remove the cyanide. The CIAT team, Drs Iglesias, Roca, Bellotti and Cabellos, have advocated the use of the leaves as a green vegetable because some Brazilian populations depend on leaves from this crop for additional nourishment. The study of the leaves strongly indicates considerable nutritional value, since they are high in protein, iron, zinc and other minerals, as well as vitamins A, B and C. They are therefore a useful supplement to the roots that provide lots of energy but are low in protein. Even so, studies have shown that there is considerable genetic variation for root protein, iron, zinc, calcium, b-carotene and vitamin C. At the high end of the range, these concentrations can make an important contribution to nutrition. For example, the concentration of vitamin C in the roots can be as high as 40mg kg-1 (fresh weight basis), and the iron can be as high as 10 mg kg-1, higher than in most milled rice. In particular, there are in the Brazilian collection many accessions of cassava with yellow and even orange-colored storage roots. For cassava roots, b-carotene varies from nil to 2.5 mg kg-1 of fresh root, sufficient to supply the recommended intake of a child from a portion the size of a small potato (Iglesias et al., 1997). They have shown that the genetic control of b-carotene, in a cross of high x low pigmented types, is by two genes with epistatic effect controlling respectively transport and loading of precursors. Genetic variation has also been shown in the extent of losses in nutritional value resulting from cooking and processing, a potentially fruitful area of research in food systems for health.

Carotenoids: The prospects for improving the provitamin A carotenoid density of staple plant foods are very good indeed. Yellow types are well known in all the usual white starchy plant foods, but less known in rice. Although the genetics are complex and somewhat obscure, rapid breeding progress is nevertheless possible since genetic gain in carotene content can be visually estimated with accuracy (Simon, 1992). Recently, carotenoids other than b-carotene have been shown to be present in the eye where they function in preventing age-related macular degeneration (Khachik et al., 1999). Such carotenoids are contained in many vegetables and fruits, as well as pasta and bread wheats (Rosser et al., 1999). Strong carotenoid pigmentation was common in older bread wheat varieties, and yellow bread is still common in China where it is prized for its aroma and taste. This century, wheat breeding has been focused on wheats producing white flour, driven by market demand, and the traits for high grain carotenoid content have been eliminated. By re-educating consumers, these types could be brought back into breeding programs in all countries, but the benefits would be greater for resource-poor consumers. The importance of carotenes to iron deficiency anemia is indicated by the recent work of Garcia-Casal et al. (1998) who showed both vitamin A and b-carotene increased the gut absorption of iron from fortified cereal-based meals (see section VII). It remains to be shown whether other carotenoids can have the same effect. Carotenoids abound in cereal germplasm, and even b-carotene can almost certainly be increased in wheat through breeding. It is already high in yellow maize and has eliminated vitamin A deficiency in pigs housed in winter (Brunson and Quackenbush, 1962). The significance of carotenoids in cereals is the high intake of these staples and the chemical diversity of carotenoids present, so giving a spectrum of potential for quenching damaging free radical reactions within cells.

Expression of the micronutrient-density traits has been tested over a wide range of environments, and although the environmental effect itself is strong, the genotype effect is consistent across environments (implying the G*E interaction is not serious), sufficient to encourage a breeding effort. Environmental factors considered by one or more of the crop programs include acid soils, alkaline soils, saline soils, acid-sulfate soils, iron-deficient soils, time of planting, field site, season, nitrogen fertilization, phosphorus fertilization, potassium status, elevation and drought stress.

The first genetic study of a micronutrient efficiency factor appears to be the classical work of Weiss (1943) on iron efficiency in soybeans, in which he showed that efficiency was due to a single, major, dominant gene controlling the reducing power of the root membrane surface. Since Weiss' pioneering study, several minor additive genes have been discovered to contribute to iron efficiency in this crop (Fehr, 1982), and these are of practical significance in improving tolerance to iron-deficient soils as all successful cultivars carry the efficiency alleles at the major locus. A major and several minor genes are likely to be the case with other micronutrients. Reports to 1970 were reviewed by Epstein (1972) who noted apparently simple genetic control of boron efficiency in tomato and celery, iron efficiency in maize and tomato, and magnesium efficiency in celery. More recently another iron efficiency trait in tomato has been shown to be due to a major gene, coding for a ferrous-iron transporting non-protein amino acid, nicotianamine (Ripperger and Schreiber, 1982). This gene is expressed in the shoot and appears to facilitate iron transport in the phloem and so may play an important role in transport into fruits and seeds. Thus, two suites of genes may be important to loading micronutrients into seeds: those involved in uptake from soil, known as micronutrient efficiency traits (Graham, 1984), and those involved in transport within the plant and to the seed, nutrient transport and loading traits. Little is known of the genetics of transport and loading of micronutrients, other than the pioneering work on tomato mentioned above. Nicotianamine has since been detected in a wide range of higher plant species and may be a universal ferrous-iron ligand facilitating both intercellular and long-distance transport in the phloem. Indeed, it is likely to be able to facilitate the movement of all the micronutrient cations (Welch, 1995).

The genetic control of iron density in rice appears to be relatively simple. High iron (and zinc) density in rice is linked closely with aromaticity, itself a single gene trait, so is also likely to be a major gene. The link to aroma makes selection easy in early generations by smell (Graham et al., 1997). Iron and zinc-dense rice lines have been selected in the breeding program at the International Rice Research Institute that are also high in yield, aroma and cooking quality. They are superior in other minerals as well and have proved to have high bioavailability to rats (currently subject to long-term human dietary trials in the Philippines).

Copper efficiency in rye appears to be a dominant trait controlled at a single locus on the long arm of chromosome 5R (Graham, 1984). Of several translocations that exist, the 5RL/4A translocation appears to be the most satisfactory agronomic type and has been successfully incorporated into adapted cultivars for South Australia (Graham et al., 1987). Work with rye addition and translocation lines has shown that copper efficiency is not linked to zinc efficiency, nor either of them to manganese efficiency. Thus, independent and relatively specific genes are involved, and root system geometry or size does not appear to be critical. For example, although wheat and triticale root systems are similar (Graham et al., 1981) and the triticales do not inherit the fine and extensive root system of rye, yet they do inherit all three micronutrient efficiency traits from rye. Manganese efficiency is located on 2R, a conclusion supported by the poor performance on manganese-deficient soils of armadillo-type triticales lacking 2R. Manganese efficiency in barley also appears to be simply inherited (Graham, 1984; McCarthy et al., 1988). A high percentage of wheat and barley cultivars currently have exceptional sensitivity to manganese deficiency, the genetic basis of which is unclear in wheat but in barley, many of them have a parent introduced from Alexandria, Egypt, in their pedigree.

Less is known of the genetics of zinc efficiency (Graham et al., 1992). The study of addition lines of rye (Graham, 1984) suggests several loci on as many different chromosomes are involved in zinc efficiency in rye; likewise, a few genes are involved in zinc efficiency in rice. The largest single screening exercise was of 3703 lines of paddy rice (Ponnamperuma, 1976; IRRI, 1979) where 388 lines were judged to be tolerant. Following diallel analysis, a recent report suggested that the genetic effects responsible for the zinc efficiency trait in rice are mostly additive, and to a lesser extent dominant (Majumder et al., 1990). Soybean varieties differ in their response to zinc fertilizer (Rao et al., 1977; Rose et al., 1981; Saxena and Chandel, 1992). Such a result is suggested to be a consequence of differential efficiency of zinc absorption; the distribution of F3 lines from the cross between zinc-efficient and zinc-inefficient genotypes (330 F3 lines tested) suggested that only a few genes control the zinc efficiency trait (Hartwig et al., 1991). Recently, a recombinant inbred population of Phaseolus beans and molecular-genetic analysis of QTLs were used to map several loci controlling zinc concentration in the bean seeds (S. Beebe, pers. comm.).

The various mechanisms of zinc efficiency are likely to be additive as suggested by Majumder et al. (1990), putting great emphasis in a breeding program on step-wise pyramiding of genetic information. Combining traits for several micronutrient efficiencies into one locally adapted crop cultivar has been facilitated by the availability of rapid methods of producing doubled haploid populations that can be immediately phenotyped by conventional bioassays. Using the best local germplasm, cultivars with improved zinc efficiency may be expedited without severely disrupting the broad adaptation already achieved. A recent Australian study has indicated that it is possible to pyramid zinc efficiency genes in bread wheat to produce much more efficient types than currently exist in released cultivars (Grewal, et al., 1997).

Nutritionists stress that balancing diet is a complex issue. It not only depends on the diversity in the diet, but on what foods are eaten together as one influences the nutrients extracted from the other; for example, vitamin C from fruit can increase the absorption of iron from cereals. Food processing method also affects the outcome.

However, the CGIAR mission is empowering the poor through assisting in appropriate change to their agriculture. One of the poorest countries, Bangladesh, produces only 20% of the vegetables it requires by FAO standards. In all probability, this production is unevenly distributed to the wealthy, and millions of resource-poor citizens of Bangladesh eat virtually only cereal. This group is the target group: any genuine improvement in the micronutrient concentration of their staple will fully benefit them. This is the simplest scenario. However, consider people in progressively higher socio-economic strata, their diet will be diversified accordingly, and they will eat less staple and benefit proportionately less, but will also be less at risk.

This strategy for maximum effect through plant breeding should not, unwittingly, be projected as delivering a balanced diet, even though a cultivar might be greatly improved with respect to the most commonly deficient micronutrients, iron, zinc and pro-vitamin A. Improved intakes of target nutrients would be expected to have a major effect in improving cognitive and physical capacity of the generation born after the new cultivar goes into wide production. Additional increments in health and welfare, through supply of other less limiting nutrients will depend on improved work capacity and income, leading to better diversity of diet as this new generation grows up. Thus, the primary benefit will be breaking the cycle of poverty due to the perinatal effects of these major deficiencies on the mother and child.

Selenium is an essential element whose roles are only emerging at this time but three aspects of it suggest that it should be included in the initial screening work: it is intimately involved in the metabolism of iodine which cannot function without it, and deficiency of it is exceptionally widespread. Thirdly, it can be analysed in grain concurrently with iodine by ICP-OES or ICP-MS. In a recent paper Combs (1995) recognized bread as a useful source of selenium, and another source noted purple wheats as rich in selenium. There would appear to be good opportunities for finding useful genetic variation in selenium accumulation in wheat grain. Finally, the link of low-selenium intake and cancers of almost every kind is now stronger (Clark et al., 1996), and work on selenium will attract funding from wealthy nations.

The micronutrient density of seeds is important not only for human nutrition, but also for the nutrition of the seedling in the next generation. A more vigorous crop is established by seeds with a high density of nutrients, including micronutrients (Welch, 1986; Rengel and Graham, 1995; Mousavi-Nik et al., 1997), that confer better resistance to stress and disease, leading eventually to higher yield. These effects are most pronounced in micronutrient-deficient soils (Weir and Hudson, 1966; Yilmaz et al., 1997). Thus, micronutrient-dense seeds are desirable both for the farmer and the consumer, a ìwin-winî situation.

The seed is a store of nutrients that provides for the growth of the enclosed embryo to create the next generation. The embryo must draw on these nutrients to develop the leaves that can take over the role of providing energy to the seedling and to develop the roots that must ultimately deliver the mineral nutrients needed for further growth. It is necessary, therefore, that the seed contain enough nutrients, including micronutrients, to sustain root growth until there is enough root absorbing surface to supply all the young plantís needs. This is not a simple matter: the seed must have a bigger store of a particular nutrient when the soil is low in that nutrient, because it will take longer to create a root system of sufficient size to supply the plant at the pace required for maximal growth rate. Paradoxically but not surprisingly, it is under conditions of low nutrient availability in soil that the mother plant has most difficulty in storing adequate nutrient in the seed. This is a serious problem in agriculture as in subsistence farming systems where farmers keep their own seed, these seeds are likely to be low in the nutrient that is also low in the soil, compounding the problem of poor seedling vigor. Poor vigor, in turn, leads to poor establishment, more severe competition from weeds that are better adapted than the crop, more susceptibility to diseases and ultimately lower grain yield. The costs in terms of grain yield, on farm, from seeds of low nutrient content have been demonstrated for zinc in wheat (Marcar et al., 1986; Rengel and Graham, 1995), manganese in barley (Longnecker et al., 1991; Ascher, 1994) and lupin (Crosbie et al., 1993), molybdenum in maize (Weir and Hudson, 1966) and phosphorus in lupin (Thompson et al., 1992). Many more examples have been demonstrated under controlled experimental conditions; it is likely to be a general phenomenon. However, the case of zinc is of greatest interest here. Zinc is widely deficient in soils, crops, animals and humans, indeed the whole food chain. It follows, therefore, that breeding for zinc-dense seeds will benefit both producer and consumer. The agricultural imperative and the health imperative coincide. This makes the case for breeding for nutrient-dense seed compelling.

Seedlings under stress, such as from waterlogging, high pH, defoliation or herbicide damage often first show symptoms of iron deficiency, indicating that the in-built iron scavenging mechanisms (Welch, 1995) have been compromised. It can be deduced from this that adequate stores of iron in seeds is equally important as for the other nutrients above, though this is more difficult to demonstrate experimentally because of the inefficiency of iron fertilizers needed for the control treatment.

There are further implications in the seed-nutrient effect. In breeding programs, it has been customary to grow together in the same field trial breeding lines that have been collected from various places. This is true whether the field trial is of advanced breedersí lines, potential parents for the breeding program or segregating materials from within the breeding program. For example, parent materials might be gathered from around the world or around the country to identify types adapted to a particular type of soil that is among other things, deficient in zinc. Such a variety trial will be confounded because the seed of the entries in it will vary in zinc according to their source and not due solely to their genotype. Seed zinc content is controlled additively by both genotype and the environment of the mother plant. Thus, a line that has high seed zinc (because it was grown in a high-zinc soil) will have an advantage in the trial regardless of whether it has the genes for high zinc density or not. In the same way, a nutrient-dense genotype may be disadvantaged in this trial if it came from a zinc-deficient area. The implications of this are important for breeding in micronutrient-deficient areas and are equally so for breeding for high nutrient density. It is necessary to grow all seed together at the one site for at least one generation to level the playing field before looking for high-density phenotypes. Breeding programs in South Australia have recognized these constraints and adapted methods to deal with them. Of course, as we study the genetics of these traits, map the genes and determine molecular markers for them, it will be possible to characterize the genotype directly, avoiding these confounding effects on the phenotype, greatly speeding up the breeding process.

Crop and soil scientists are fully cognizant of the importance of the ìavailableî nutrient status of soils in determining fertilizer recommendations for optimizing crop productivity. They understand that the total concentration of a nutrient in a soil does not reflect the plant-available nutrient supply within the soilís rooting zone. This is especially true for micronutrient metals such as iron, zinc, copper and manganese which can be present in soil pools in forms that are unavailable for absorption by plant roots (Marschner, 1995; Welch and House, 1984). This concept of nutrient availability also holds for micronutrients in plant foods eaten by people consuming varied diets containing a myriad of other food ingredients. The total amount of a micronutrient in a plant food does not represent the actual micronutrient content of the food that is utilizable by the consumer. This quantity (i.e., the bioavailable amount) must be determined independently using methodologies especially developed for such purposes. In human nutrition terms, bioavailability is commonly defined as the amount of a nutrient in a meal that is absorbable and utilizable by the person eating the meal (Van Campen and Glahn, 1999). There is an immense body of research concerning micronutrient bioavailability in plant foods that cannot be covered in the brief review of micronutrient bioavailability presented here. For more detailed information concerning this topic the reader is referred to the following references: (Benito and Miller, 1998; Fairweather-Tait and Hurrell, 1996; House, 1999; Hunt, 1996; Hurrell, 1997; Matzke, 1998; Rao, 1994; Van Campen and Glahn, 1999; Wienk et al., 1999; World Health Organization, 1996).

Figure 6 depicts the dynamic factors and their interactions that affect the amount of a micronutrient bioavailable to a person eating a meal containing plant food. Micronutrients can occur in various chemical forms of differing proportions in plant foods and their amounts vary depending on numerous factors including the growth environment, plant species, genotype and cultural methods used to grow the plant. These forms have characteristically different solubilities and reactivities with other plant constituents and other meal components. There are multiple interactions occurring between micronutrients in plant foods and other plant substances once the food is consumed, such as with other interacting nutrients and chemical substances that can either inhibit (i.e., antinutrients) or enhance (i.e., promoters that can increase absorption and/or utilization) micronutrient bioavailability. Additionally, many other interacting factors, both genetic and environmental, affect micronutrient bioavailability to the consumer, such as food processing methods, meal preparation techniques, and an individualís personal characteristics (e.g., sex, age, genetic predisposition, ethnic background, economic status, physiological state, nutritional and disease status). Thus, determining micronutrient bioavailability in plant foods is beset with difficulties and uncertainties (House, 1999). Hence, micronutrient bioavailability is a confusing and complex issue for human nutritionists that remains a very active research venue for many researchers worldwide (Van Campen and Glahn, 1999). There is no one bioavailability method applicable for all micronutrients or for all circumstances and plant foods (Fairweather-Tait and Hurrell, 1996).

Various methods, have been developed to determine micronutrient bioavailability in plant foods to humans that encompass in vitro or in vivo models or combinations of both (Van Campen and Glahn, 1999; Wienk et al., 1999). Unfortunately, none of these models is ideal for all foods, nutrients, and circumstances. Ultimately, experiments with human subjects fed enriched staple foods grown locally and consumed in traditional diets under real world situations will determine the actual benefit of enhancing the concentration of micronutrients in staple plant foods to people at risk of developing micronutrient malnutrition. However, such human experimentation is costly and long-term in nature. Therefore, practically speaking, plant breeders must rely on other less costly bioavailability methods to screen relatively large numbers of promising experimental lines before selecting only the most promising ones to be used in conducting human feeding trials under free-living conditions in order to keep expenditures within the permissible limits of available funds.

In our international breeding effort, initial bioavailability screening of promising lines of micronutrient-enriched staple food crops were performed using a rat model (see reference Welch et al., 1974) for detailed information concerning the rat model used). Concernedly, some researchers consider the rat bioavailability model to be an obsolete model because it is only empirically based (Wienk et al., 1999). Although, quantitatively, rats are far more efficient at absorbing Fe and Zn than humans from plant food sources and sometimes they do not respond as humans do to certain inhibitors such as tea, rats have been used effectively to rank foods with respect to bioavailable iron and zinc. For example, the effects of some dietary factors, such as ascorbate, soy protein or bran, on iron bioavailability have been shown to be similar in both rats and humans fed the same dietary sources of iron although quantitatively, the actual bioavailability values were considerably higher in the rats (Reddy and Cook, 1991). Importantly, crop lines shown to have relatively low bioavailable amounts of iron and zinc in rats would, most certainly, have very low bioavailable amounts of these nutrients when fed to humans. However, some lines showing high levels of bioavailable iron and zinc in a rat model, may not contain highly bioavailable levels of iron and zinc when fed to humans; thus, the need ultimately to screen the top lines selected from rat models using human trials. While not ideal, for the purpose of initially screening promising lines of iron- and zinc-dense staple foods, the rat model is a useful tool for ranking promising lines for their bioavailable iron and zinc content. However, it must not be the ultimate test for determining bioavailable levels of iron and zinc in micronutrient-enriched staple foods to humans.

Interestingly, for comparative purposes, we screened some promising lines of rice and beans using both a rat model and an in vitro human intestinal cell culture technique. The bioavailability rankings obtained with rats were compared to those rankings obtained using a cell culture model (i.e., Caco-2 cell culture). Caco-2 cells were developed from an adenocarcinoma isolated from a human large intestine. Under appropriate culture conditions these cells differentiate into a polarized monolayer of cells having microvilli. Caco-2 cell monolayers are morphologically and physiologically very similar to the layer of mucosal epithelial cells that line the surface of the small intestine and are responsible for most micronutrient-metal absorption from the gut. Similar rankings of the lines tested were obtained for Fe bioavailability using either of these models, giving support for continued use of the rat model in supporting plant breeding efforts. Currently, the Caco-2 cell model does not require radio-labeling of plant foods with Fe radioisotopes which is a distinct advantage over using the rat model which does. However, the human cell culture model has not been developed for use in determining the bioavailability of non-radiolabeled Zn in plant foods. When such a technique is developed for Zn, the human Caco-2 cell model may become the technique of choice for use in screening large numbers of lines for bioavailable Fe and Zn because it is a human cell model and is relatively inexpensive, rapid and would not require the use of radioisotope or stable-isotope labeled plants (see Van Campen and Glahn, 1999 for more information on the Caco-2 cell culture model).

After selecting a technique to determine bioavailable levels of Fe and Zn in plant foods, one is faced with the decision to use either intrinsic or extrinsic labels of these nutrients in meals to be irond fed to experimental subjects. Commonly, bioavailability methods require that either radioactive or stable isotopes of Fe or Zn be used to label these nutrients in plant foods before their bioavailability can be determined in model systems or in humans. Intrinsic labels refer to the growing of plants in growth media to which the isotopes have been added allowing the plants to absorb them and to incorporate them ìnaturallyî into their metabolites. Extrinsic labels refer to adding inorganic solutions of these isotopes to meals prepared from plant foods after the edible plant portions have been harvested. Extrinsically labeled plant foods are relatively inexpensive to prepare and are easy to use, while obtaining intrinsically labeled plant foods requires that isotope labels be supplied in the growth media allowing incorporation of the isotopes naturally into the plants during growth is a more costly, time-consuming and difficult process. Therefore, many researchers have used extrinsic labels to perform bioavailability studies. While extrinsic labels have been shown to produce reliable bioavailability results under many if not most circumstances for several plant foods, they do not always reflect what is found using intrinsic labels. Therefore, the use of intrinsic labels is the only method that assures unequivocal Fe and Zn bioavailability results for all plant food sources when fed to humans in meals containing many interacting factors (House, 1999; Van Campen and Glahn, 1999).

Phytic acid. Plant foods contain various substances that can interfere with the absorption or utilization of Fe and Zn in staple plant foods. Table VI lists examples of antinutrients that are known to be present at high levels in many staple plant foods and their edible products. Among those listed, phytic acid [(myo-inositol hexakis(dihydrogen phosphate)], or its natural product, phytin (a K, Mg salt of phytic acid), has been studied the most because phytin is abundant in edible seeds and grains and forms insoluble precipitates with a number of polyvalent mineral cations (e.g., Ca2+, Fe3+, and Zn2+) in vitro, and when added to purified diets. Sodium phytate is known to decrease Fe and Zn absorption in a number of monogastric animal species and in humans (Erdman, Jr. and Poneros-Schneier, 1989; Pallauf and Rimbach, 1997). High dietary Ca accentuates the effects of phytate on Fe and Zn bioavailability, and using Zn-phytate/Zn molar ratios in diets has been reported to be a better predictor of Zn bioavailability than dietary phytate alone. For humans, ratios above 0.5 mol/kg dry diet or 200 mM/1000 Kcals may be a cause of concern for Zn nutriture.

There is no doubt that high levels of phytic acid can inhibit the bioavailability of Fe and Zn to people eating staple plant foods such as whole cereal grain products and legume seeds. However, plant foods high in phytin do not always inhibit Fe and Zn bioavailability, and the reasons for the discrepancies are not currently understood (House, 1999). For example, Morris and his cooperators (Morris, 1986; Morris and Ellis, 1982) identified the natural chemical form of Fe in wheat seeds to be monoferric phytate. They then reported that monoferric phytate was relatively highly bioavailable to rats, dogs and humans. Furthermore, in a mineral balance study with human subjects (see Table VII) they compared high phytate whole wheat bran muffins to dephytinized wheat bran muffins very low in phytate. Meals containing the muffins were eaten daily over the 15 days of the study. Initially, those men given high phytate whole wheat muffins were in negative Fe balance during the first 5 days of the study. After an additional 10 days, the men were in positive Fe balance eating the high phytate bran muffins. Those men eating the dephytinized muffins remained in positive Fe balance over the course of the study. Thus, there appeared to be some change in the men eating the high-phytate muffins that allowed them to become in positive Fe balance after 15 days of consuming the muffins. The reason for this change from negative to positive Fe balance in adult men while still consuming high-phytic acid muffins is still not known.

Some reports concerning the effects of naturally occurring phytin in seeds and grains on Zn bioavailability are confusing and contradictory (Hambidge et al., 1986; House, 1999; Morris, 1986; Welch, 1993). Certainly, soluble salts of phytate when added to purified diets used in animal models or fed to humans do reduce Zn bioavailability. However, other dietary factors, such as Ca and protein type, can interact with phytin causing large negative effects of phytin on the bioavailability of Zn. More research is required before all the mechanisms of phytin in reducing the bioavailability of Fe and Zn to humans consuming mixed diets are fully understood. The commonly accepted mechanism for phytateís antinutritive activity, metal-phytate precipitation in the gut, may not be the only way in which phytate changes Fe and Zn bioavailability because phytate and its hydrolysis products can be readily absorbed by mucosal cells in the gut (Sakamoto et al., 1993). Some of these compounds may produce hormone-like reactions within the cells that impact the regulation of nutrient absorption and utilization (Harland and Morris, 1995; Harland and Narula, 1999; Pallauf and Rimbach, 1997).

Some nutritionists are currently promoting the idea that the nutritional quality of staple plant foods could be dramatically improved if genetic means were found to reduce or eliminate antinutrients from these foods. This idea seems logical and sensible on the surface, but further thought leads to some questions that should be addressed before attempts are made to transform staple food crops in this way (Graham and Welch, 1996; Welch, 1993). Many of these antinutrients are major plant metabolites that play important roles in the life of the plant. Removing them from plant seeds may have unforseen implications for crop productivity, including reduced disease resistance, lower insect and herbivore resistance, lower seed-nutrient stores, and less stress tolerance. Logically, there must be important reasons why plants have evolved the complicated genetic mechanisms required to synthesize, regulate, store and degrade these types of substances in their reproductive organs.

Recently, low-phytate mutants of staple plant foods have been identified (e.g., low-phytate maize kernels) (Ertl et al., 1998; Sugiura and Raboy, 1999). They were developed for the feed industry to reduce P inputs into monogastric animal and fish rations as well as to reduce P pollution of the environment from pig and chicken manures and fish feces produced when monogastric animals and fish are fed high-phytate rations. While phytate is not hydrolyzed appreciably in the gut of pigs, chickens and fish, the phytate in the manure of monogastric animals is easily hydrolyzed by soil microbes releasing inorganic P to the environment and causing P pollution of field run-off water and, ultimately, streams, rivers and lakes. Some have suggested that these types of low-phytate mutants be incorporated in to plant breeding programs to reduce phytate in these food crops thereby lowering the negative effects of phytate on Fe and Zn bioavailability in human populations.

A recent report supports this suggestion (Mendoza et al., 1998). These researchers studied the effects of genetically modified low-phytate maize (LPM) on Fe absorption by humans fed tortillas made from LPM. The LPM tortillas contained only 35% of the phytate found in tortillas made from wild-type maize (WTM). Iron absorption from the LPM was 8.2% of intake while absorption from WTM was significantly lower (P< 0.001) at 5.5% of intake. The authors concluded that genetically modified maize lines with the LPM-trait may improve Fe absorption in human populations that are dependent on maize as a staple food. However, this possibility should be critically scrutinized through further research before any attempts are made to incorporate the low-phytate traits into a plant breeding program for staple plant foods to feed people in many developing countries. The needs for further research are discussed below.

Phytin is the primary storage form of P in all higher plant seeds and grains of nutritional importance (Welch, 1986). Phytin is accumulated in globoid crystals in membrane-bound protein bodies of certain cell types within the developing seed, such as those occurring in aleurone cells of the aleurone layer and embryo of cereal grains (Lott, 1984; Mazzolini and Legge, 1981; Ogawa et al., 1979). The phytin deposited within protein bodies as globoid crystals, is associated with the accumulation not only of P, but also of other minerals including K, Mg, Fe, Zn, Cu, Mn, and, in some seed cell types, Ca. Phytin, therefore, plays an important role as a storage pool for mineral nutrient reserve required by the developing embryo within the seed or grain during germination, and during early seedling growth. It contributes to the viability and vigor of the seedling produced (Welch, 1986). Selecting for seed and grain crops with substantially lower phytin content to improve these crops as sources of bioavailable Fe and Zn could have undesirable effects on agricultural production especially in regions having soils of low P status and inadequate trace mineral fertility. Research should be conducted to assure that reducing phytate levels will not have adverse impact on crop production in developing countries which can ill afford to have reduced crop yields or increased production costs to feed their growing populations.

Interestingly, several researchers report that phytic acid (or certain hydrolysis products) may play important roles in lowering chronic disease rates in humans, including certain types of cancer and heart disease, and in performing other beneficial functions related to human health (see text box below) (Harland and Morris, 1995; Pallauf and Rimbach, 1997; Zhou and Erdman, Jr., 1995). Thus, reducing phytate levels in important staple food crops could have adverse effects on human health by increasing cancer risk and other chronic diseases. Therefore, caution should be used by plant scientists before they try and modify food crops with respect to their phytin content until further research is forthcoming on this important topic.

Another approach to eliminating the negative effect of phytate in staple plant foods on Fe and Zn bioavailability in human diets has been initiated by some researchers using transgenic techniques. For example, the gene for a heat stable phytase has been identified and isolated from a fungus (Aspergillus fumigatus). This fungal gene was incorporated into rice and was expressed in the rice-grain endosperm during grain development (unpublished results presented by I. Potrykus and co-workers in abstract form at the XVI International Botanical Congress, St. Louis, MO, 1999). According to information in the abstract, when the rice grain carrying the phytase was cooked, all of the phytic acid in the grain was degraded. While this type of genetic modification of rice would improve the nutritional quality of this important staple food with respect to Fe and Zn, it may also increase the risk of developing certain chronic diseases in people consuming rice as a staple food. Hence, more research is needed to determine the overall benefits of phytate to humans before attempting to reduce phytate levels in plant foods.

In contrast to antinutrients, certain dietary substances (i.e., promoter substances; see Table VIII) can enhance the bioavailability to humans of micronutrients, including Fe and Zn, in meals containing plant foods even in the presence of the antinutrients (Ashmead and Christy, 1985; Gordon and Godber, 1988; Michaelsen and Friis, 1998; Mulvihill and Morrissey, 1998a; Mulvihill and Morrissey, 1998b; Politz and Clydesdale, 1988; Welch and House, 1995). Many of these promoter-substances are normal plant metabolites and modest increases in their concentrations may not result in adverse effects on plant growth and development. Usually, these factors are rich in certain foods (animal meats and many types of fruits and vegetables), but are at relatively low levels in staple foods such as cereal grains and legume seeds. Increasing promoter concentrations in staple food crops by genetic manipulation is an attractive possibility if we know the chemical identity of the promoter substance and the genes responsible for its expression. Taking this approach would eliminate many of the concerns related to decreasing antinutrients in these crops discussed above (Graham and Welch, 1996). Thus, breeding for increased promoter substances in plant foods is an important opportunity for plant scientists to improve the nutritional quality of staple foods with respect to some micronutrients including Fe and Zn. Additionally, producing transgenic plants via modern genetic engineering techniques that express more of these substances in edible portions of food crops is an important area for additional research and development in plant nutritional genomics. Indeed, Potrykusís research group in Switzerland, that developed the transgenic rice line containing a heat stable phytase (see discussion above), also over expressed a rice gene for metallothionine in the same rice line increasing the level of this cysteine-rich protein in the grain-endosperm by about 25%. They believe that doing so would improve the bioavailability of Fe in the rice grain to humans although there is no animal or human data published to date to substantiate this claim.

Ascorbic acid. Among the promoter substances listed in Table VIII, ascorbic acid (vitamin C), the ìmeat factorî and b-carotene (a vitamin A precursor) have received the most interest scientifically. Ascorbic acid is thought to enhance the absorption of Fe from plant foods via forming Fe(III) complexes and by reducing Fe³⁺ to the more soluble and bioavailable Fe²⁺ valence state (House, 1999). While ascorbic acid markedly increases nonheme Fe absorption in humans, relatively log-term dietary supplementation with ascorbate had less effect on Fe bioavailability than expected based on experiments with single meals (Hunt et al., 1994). Consuming ascorbate-rich foods has been reported to counteract phytate inhibition of Fe absorption from wheat rolls (Hallberg et al., 1989). Another effect of ascorbate intake was to prevent the dose-dependent inhibition of Fe absorption by polyphenols and phytic acid in maize bran (Siegenberg et al., 1991) and to ameliorate the adverse affects of phytate in rice starch on Fe bioavailability to humans (Tuntawiroon et al., 1999). Therefore, increasing the ascorbate levels in plant foods genetically would help alleviate the negative effects of phytate and polyphenols in staple foods on Fe bioavailability and at the same time, enhance these foods as sources of the essential nutrient, vitamin C. Such efforts should be encouraged.

The ìmeat factorî. The chemical identity of the ìmeat factorî promoter has not been discovered, but it appears be related to cysteine-rich polypeptides in muscle proteins from animal sources (Gordon and Godber, 1988; Kapsokefalou and Miller, 1995; Mulvihill and Morrissey, 1998b; Mulvihill and Morrissey, 1998a; Politz and Clydesdale, 1988; Welch and House, 1995). Some studies have shown that cysteine and cysteine-rich polypeptides enhanced the bioavailability of Fe to monogastric animals and humans (Taylor et al., 1986); these findings support the contention that the ìmeat factorî is a polypeptide rich in this amino acid (Welch and House, 1995). Others have reported that methionine in addition to cysteine can promote Zn bioavailability (House et al., 1996; House et al., 1997; Welch and House, 1995). Greater efforts should be applied to identifying the chemical form and mechanisms of action of the ìmeat factorî. Once identified, ways should be found to incorporate this promoter into staple plant foods containing high levels of antinutrients, such as phytic acid. This would improve the bioavailability of Fe and Zn from these important plant foods, taking full advantage of any genetic enrichments made in Fe and Zn levels in these staples through plant breeding.

Phytoferritin. Ferritin is a major storage form of cellular Fe in animals, plants and certain bacteria that is also involved in controlling cellular Fe homeostasis (Andrews et al., 1992). In humans, serum Fe-ferritin stores are one factor used in determining Fe status of individuals (Dallman, 1990). Ferritin molecules are multimeric proteins that sequester up to 4500 atoms of Fe as hydrous ferric oxide-phosphate inside a protein coat having a molecular weight of about 450 kDa (Briat, 1996). The apparent bioavailability of Fe from ferritin to rats has been reported. Iron provided as ferritin, from horse spleen, in a Fe-deficient semi-purified diet was able to fully recover rats from iron deficiency-induced anemia demonstrating that the Fe in ferritin was bioavailable to rats. Additionally, rats fed soybean meal containing phytoferritin also were able to fully recover from Fe deficiency anemia although the recovery could have resulted not from phytoferritin but from other forms of Fe present in the soybean meal (Beard et al., 1996).

It has been proposed that significantly increasing phytoferritin levels in major food corps could contribute greatly to decreasing Fe deficiency globally (Theil et al., 1997). Interestingly, Goto et al. (1999) have enriched the grain-Fe content of a rice line by as much as 300% by transferring the entire genetic coding sequence for soybean phytoferritin into the rice line using Agrobacterium-mediated transformation. They used a grain-glutelin storage protein promoter (GluB-1) that allowed the expression of the gene in the rice grain endosperm. If the Fe in rice-grain phytoferritin is found to have a high bioavailability to humans and if it can be shown that phytoferritin has no negative effects on plant growth or harmful effects in humans, then this technique should be considered for use in developing other Fe-dense staple food crops.

Plant nonsymbiotic hemoglobin. The iron in hemoglobin from animal meat sources is relatively highly bioavailable to humans compared to non-heme forms from plant foods. Additionally, heme-Fe absorption by gut cells is not as affected by several antinutrients such as phytate and polyphenols as is non-heme Fe. Furthermore, the regulation and mechanisms of heme-iron absorption by gut cells are different from that of non-heme Fe (House, 1999). For these reasons, several laboratories are currently trying to express hemoglobin (such as symbiotic leghemoglobin from soybean nodules) synthesis in grains of cereal crops in order to improve these crops as sources of Fe for people (personal communication, David Garvin, U.S. Plant, Soil and Nutrition Laboratory, Ithaca, NY). Expressing the gene for the synthesis of symbiotic leghemoglobin in staple seeds and grains is one approach that could be used, but it will be necessary to assure that this form of Fe once expressed in seeds and grains does not produce harmful effects on seed vigor and viability and that the leghemoglobin produced provides bioavailable Fe for humans in a safe and effective form.

Interestingly, within the last four years evidence has been reported showing that nonsymbiotic hemoglobin proteins are widely found in the plant kingdom (Arredondo-Peter et al., 1997; Hill, 1998). For an example, in barley, hemoglobin biosynthesis was induced under low oxygen tensions and was regulated by ATP or by actions of ATP on metabolism. Additionally, hemoglobin expression was demonstrated to be a normal consequence of barley seed germination (Duff et al., 1998). By using transformed maize cells, researchers were able to show that hemoglobin acted to improve the energy status of the cells when grown under low-oxygen stress conditions (Sowa et al., 1998). These authors suggested that nonsymbiotic hemoglobins act in plants to maintain energy status of plant cells growing in low-oxygen environments by promoting glycolytic flux via NADH oxidation and promoting substrate-level phosphorylation. Thus, hemoglobin proteins may play important functions in plant growth and in low-oxygen stress resistance.

Arredondo-Perter et al. (1997) reported the cloning and analysis of two nonsymbiotic hemoglobin genes (hb1 and hb2) from rice. They found at least three copies of the gene coding hemoglobin in rice DNA. These genes are expressed in the roots ( hb1) and the leaves (both hb1 and hb2) of rice. The cDNA for rice HB1 was expressed in Escherichia coli and the recombinant hemoglobin (rHB1) had a high affinity for O2. Further research identifying the genes responsible for hemoglobin protein synthesis in plants is needed in order to identify likely candidate genes to use in transforming staple plant foods.

Even though the exact metabolic function(s) of nonsymbiotic plant hemoglobin is (are) still to be elucidated, the natural occurrence of nonsymbiotic hemoglobin in plants is of interest because these genes that encode hemoglobin synthesis could be over expressed in edible portions of crop plants using modern molecular biology techniques and appropriate gene promoters for seed-endosperm expression. If successful, this could lead to dramatic improvements in the bioavailability of Fe in staple plant foods that would be less affected by the presents of antinutrients in these foods. Still, the bioavailability of Fe in nonsymbiotic hemoglobin proteins and the safety to humans of consuming this form of Fe will have to be determined in future studies to assure that such changes will have a significant and safe impact on human Fe status and health.

A cautionary note: Some groups of people are concerned about increasing the bioavailable levels of Fe in staple plant foods. Idiopathic hemochromatosis (a relatively rare disease affecting about 1 per 500 to 1 per 1000 people) is characterized by uncontrolled Fe absorption and Fe overload disease in individuals afflicted with this genetic disorder which can result in death from Fe toxicity in mid-life if not treated (Barisani et al., 1996). Additionally, homozygous thalassemia, a relatively rare hereditary hemolytic anemia, requires numerous blood transfusions for survival in individuals expressing this gene. Increased Fe intake in these individuals adds an undesirable additional Fe load to their body. Both idiopathic hemochromatitic individuals and individuals being treated for homozygous thalassemia want to avoid extra iron in their diets (Dallman, 1990). However, iron deficiency anemia afflicts over 2 billion people globally, having serious impacts on human health, cognitive function and productivity. It would seem prudent and necessary to improve the Fe content of staple foods to deal with this immense Fe-deficiency crisis in the general population, keeping in mind that relatively few people will be adversely affected by doing so. The profound benefits to massive numbers of people out weigh the risks to relatively few individuals who can easily avoid such foods.

b-carotene. Recently, several reports by Layrisse and colleagues (Garcia-Casal et al., 1998; Garcia-Casal and Layrisse, 1999; Layrisse et al., 1997; Layrisse et al., 1998) indicated that fortifying cereal-based diets with vitamin A or b-carotene and Fe(II)-fumarate enhanced the bioavailability of the Fe to humans dramatically (e.g., b-carotene increased Fe bioavailability more then three fold in rice-based meals and more then 1.8 fold in wheat and corn-based meals). The researchers also suggested that the effect of b-carotene and vitamin A on promoting the Fe bioavailability in a meal containing high levels of phytate and polyphenols was the result of complexation of Fe(III) with b-carotene or vitamin A in the gut during digestion. They suggested that this prevented the precipitation of Fe in the gut by phytate or polyphenolics from foods in the meals used in their study. If true, b-carotene and vitamin A may not promote Fe bioavailability in plant foods that are not rich in the antinutrients, phytic acid and polyphenolics. Preliminary results form our Laboratory indicate that this may the case because Fe bioavailability to rats fed intrinsically Fe⁵⁹- labeled cauliflower curds from high and low b-carotene cauliflower lines (containing negligible amounts of phytate and polyphenols) was not related to b-carotene levels in the cauliflower curds (unpublished results). Highest and lowest levels of bioavailable Fe were found for white-curded cauliflower lines neither of which had measurable b-carotene levels.

Interestingly, Potrykus and his colleagues (see discussion above) also reported producing yellow endosperm transgenic rice grain containing high levels of b-carotene along with the previously discussed elevated Fe levels, the heat-stable phytase and elevated metallothionein levels. They added three genes to the rice line to accomplish this, two from daffodil and one from the bacterium Erwina uredovora. The resulting transgenic rice line synthesized enough b-carotene in its grain-endosperm to meet the vitamin A requirements of people dependent on rice as a staple in South Asia.

Importantly, experiments should be conducted to determine if human nutriture with respect to vitamin A and iron status is improved significantly when fed the transformed rice grain produced by Potrykus and his associates. If this type of transgenic rice grain is found to improve the Fe and vitamin A status of targeted populations at risk of developing deficiencies of these micronutrients, attempts should be made to perform similar genetic modifications of other important plant foods including wheat, maize and beans.

It is virtually impossible to design human clinical studies under laboratory conditions to determine the micronutrient bioavailability in micronutrient-enriched lines of staple plant foods that give unequivocal conclusions applicable to free-living populations in various environments. There is no one technique that is universally applicable for all micronutrients and all crops, and that can encompass all the factors known to affect bioavailability. Clinical trials can be performed using enriched lines, but these types of controlled experiments performed under laboratory conditions cannot include all the variables that can affect micronutrient bioavailability to humans (see Figure 6). Just carrying out clinical studies might lead to conclusions that are misleading or even erroneous, either showing or not showing significant nutritional benefits to subjects eating the enriched lines under the conditions of the laboratory experiment. This could lead either to the elimination or the inclusion of an enriched line for future development, or even to the conclusion that breeding for micronutrient-enriched staples is without value. For this reason, the only ways to ascertain with certainty if micronutrient-enriched lines of staple food crops actually improve human nutriture is to carry out long-term feeding trials comparing the enriched lines with traditional lines of staple foods. Unfortunately, these types of studies are expensive and long-term in nature. They must be performed with groups of people that have been rigorously selected in ways that will assure reliable results. Some examples of factors to consider in developing Fe and Zn feeding trials using human subjects are discussed below for Fe and Zn.

Iron bioavailability feeding trials. Selecting human subjects for field trials to determine the efficacy of Fe-enriched staple foods for improving Fe status must meet certain criteria. They must assure that the subjects to be studied are dependent on the staple food as a major constituent of their diet, that the subjects are free of disease, inflammatory conditions and parasites, and that their body stores of iron are significantly depleted (e.g., serum-ferritin values are < 12 µg L-1, transferrin saturation < 16% for adults), but not so depleted that they are severely iron deficiency anemic (e.g., hemoglobin levels < 12 g L-1) (Dallman, 1990). Additionally, data concerning their typical diets, food preparation techniques and household food distribution patterns must be collected (i.e., a full knowledge of their food system must be acquired). The staple lines to be studied should be grown in the region where the study will take place by traditional means. The length of the study will depend on the amount of Fe enrichment obtained in the staple food produced, lasting as long as several months to a year or more. Traditional lines relatively low in Fe should be included as a control group. A dietary supplement should be provided that assures that all subjects are adequate in all micronutrients except Fe during the course of the study. Subjects should be monitored closely to ascertain compliance to the study protocols, to determine how much of the staple food they consume, how it is prepared and distributed among family members, and what foods are consumed with the test line. Iron status should be determined before the study ensues, during the study and when the study culminates. Anthropometric data (e.g., height, weight, arm circumference and triceps skinfold thickness) should also be collected during the course of the study.

Feeding trials for zinc bioavailability . Most of the information needed to conduct meaningful feeding trials with Fe also applies to Zn (Gibson and Ferguson, 1998). Normally, rich dietary sources of bioavailable Fe (i.e., animal meats) are also rich sources of bioavailable Zn, and people at risk of developing Fe deficiency are usually also at risk of developing Zn deficiency. Unfortunately, determining the Zn status of an individual cannot be performed using an available clinical procedure unless the person is severely Zn deficient because no reliable clinical test exists to determine marginal Zn status in humans. Plasma-Zn is the most frequently used method to determine Zn status, but a variety of factors affect plasma-Zn leading to misinterpretation of Zn status. A combination of Zn deficiency characteristics (e.g., growth retardation, delayed sexual and skeletal maturity, skin lesions such as orificial and acral dermatitis, diarrhoea, alopecia, and behavioral changes) and severe hypozincemia makes the detection of severe human Zn deficiency relatively easy (Hambidge et al., 1986; World Health Organization, 1996). In the severe Zn-deficient state, concentrations of Zn in the plasma are usually < 0.4 µg mL-1 and many times < 0.2 µg mL-1 and under moderate deficiency they fall between 0.4 to 0.6 µg mL-1. In Zn-adequate individuals plasma Zn values lie within the range of 0.65 µg mL-1 to 1.10 µg mL-1. Regrettably, in moderate Zn-deficient individuals, many features of Zn deficiency are nonspecific; factors other than Zn deficiency can cause these low Zn-plasma levels. Identification of mild Zn deficiency in humans is very difficult because plasma Zn levels can be within the normal range and other characteristics of Zn deficiency are not specific to this disease. Currently, the only sure way to determine mild Zn deficiency in humans is through determining the response of individuals to Zn supplementation of their diet (Hambidge et al., 1986). Therefore, any human feeding trial conducted to determine the effects of Zn-enriched staple foods on human Zn status will have to rely primarily on anthropometric data in children and associated Zn-deficiency-related diseases and characteristics. Selecting subjects with low Fe stores will aid in finding individuals with low Zn status, but cannot guarantee that Zn-deficient subjects will be included in the study. Zinc status is important in the selection criteria because individuals with adequate Zn status down regulate their absorption of Zn from food and may not respond to Zn-enriched staples to the extent that Zn-deficient individuals would (Cousins and Hempe, 1990).

Interestingly, Paik et al. (1999) recently published a paper indicating that serum extracellular superoxide dismutase can be used as a functional indicator of marginal Zn deficiency in humans. If this method proves to be reliable after further testing, a clinical method to determine mild Zn deficiency in humans may become available. Other recent papers have also reported that Zn clearance tests may be diagnostic of marginal Zn deficiency in children of small stature (Kaji et al., 1998). Further, mean plasma zinc concentration may be a useful indicator of population Zn status for children in low-income nations despite the high prevalence of common childhood infections in these children (Brown, 1998).

An underlying cause of and fundamental constraint to solving the micronutrient malnutrition problem is that non-staple foods, particularly animal products, tend to be the foods richest in bioavailable micronutrients, which the poor in developing countries cannot afford. Their diets consist mostly of staple foods, primarily cereals. In fact, per capita direct consumption of staple foods in the aggregate varies little by income level. For the poor, these staple foods already are primary sources of what micronutrients, particularly minerals, they are able to consume.

This is demonstrated by food intake data shown in Tables IX and X for survey populations in Bangladesh and the Philippines, respectively. Average incomes in these households range from U.S.$45 per capita per year in the poorest 20% of households to $250 in the richest 20% of households. Thus, they are typical of the middle to lower end of the income distribution in rural areas of these countries.

The first priority for these poor households in terms of food purchases is to obtain calories to satiate hunger. The most inexpensive sources of calories are food staples, rice and wheat in the case of Bangladesh and maize and rice in the case of the Philippines. Once a critical intake of calories is acquired from inexpensive food staples, as incomes increase consumers purchase non-staple foods at the margin, particularly animal products and fruits, and to some extent substitute more expensive but more preferred food staples for inexpensive staples.

Not only are food staples poor (low-density) sources of trace minerals, but anti-nutrient (e.g. phytate) levels are high which may reduce the bioavailability of the trace minerals consumed. Nevertheless, for poor populations such as those represented in Tables IX and X, food staple consumption so dominates diets (their low incomes preclude the consumption of desired levels of non-staple foods) that primary food staples provide in the range of 40-55% of total iron intakes for lower income households, as shown in Tables IX and X.

If a single food staple provides 50% of total iron intakes for a poor population (e.g. for rice in Bangladesh), then doubling the iron density in that food staple will result in a 50% increase in total iron intakes, and tripling the iron density will mean a doubling of total iron intakes.

A strength of a plant breeding approach which focuses on food staples, then, is that it is based on existing consumer behavior. The poor consume large amounts of food staples on a daily basis. If a high proportion of the domestic production of food staples can be provided by nutritionally improved varieties, nutritional status can be improved without resort to programs that depend on behavioral change. Trace minerals constitute such a small physical part of the grain (at most a few dozen parts per million) that increasing density is not expected noticeably to change consumer characteristics (appearance, taste, odor, texture, cooking qualities).

For the lower income households in Tables IX and X, iron intakes for women range between 50-75% of recommended daily allowances. Despite well-known difficulties with determining useful benchmarks for recommended daily allowances of iron, it would seem evident that a 50% increase in intakes of bioavailable iron would be of considerable benefit to anemic women with such low iron intakes. Nevertheless, human studies still need to be undertaken to measure effects of increased iron (or zinc) density in food staples on iron (or zinc) status and consequent improvements in health and productivity.

Similar arguments apply to those staples in which provitamin A content may be enriched by plant breeding (wheat, maize, and cassava, for example). Some differences apply, however, as compared with trace minerals. First, no agronomic advantages accrue to higher provitamin A content, so that high density will need to be bred into varieties that are otherwise high yielding. Second, the color of the final food product may change so that consumers may need to be educated as to the improved nutritional content, although if education programs are successful, the color change becomes an advantage in that it identifies those particular varieties of superior nutritional quality.

If plant breeding costs substantially more per benefit received than existing interventions to improve human nutrition (e.g. supplementation and fortification programs), then it makes no sense to pursue such strategy. Analysis summarized below in fact shows, however, that plant breeding is highly cost-effective. The underlying factors resulting in high benefit-cost ratios for a plant breeding strategy are easy to understand. First, breeding for trace mineral dense seeds has a double benefit - it can improve human nutrition and at the same time it can improve plant nutrition as discussed in section VI. Existing interventions address only human nutrition - they have no direct effect on agricultural productivity as well. Second, once the initial investment is made in developing a more nutrient-dense genotype, the plants in some sense ìfortify themselves.î There is no need to undertake and enforce legislation to fortify specific foods. Costs of adding the fortificant to the food vehicle during processing are unnecessary. The only drawback to pursuing a plant breeding strategy is the lag time between initiation of breeding research and the point at which improved varieties become available after being adopted by farmers. No benefits accrue during this period which may range anywhere between five and ten years on average.

1. Costs of Plant Breeding

To obtain a rough estimate of plant breeding costs, the example of the CGIAR Micronutrients Project may be used. The general objective over five years has been to assemble the package of tools that plant breeders will need to produce mineral and vitamin-dense cultivars. The target crops are wheat, rice, maize, phaseolus beans, and cassava. The target micronutrients being studied are iron, zinc, and vitamin A. For these crops and nutrients, this project is conceived as a pre-breeding study to determine:

(1) the range of genetic variability available for exploitation by future breeding programs;

(2) The stability of expression of the high-density trait across environments (G*E)

(3) the bioavailability of the micronutrients contained in the grain (or seeds or other storage tissue) of the best selections;

(4) the genetics and physiology/biochemistry of the selected traits;

(5) screening protocols for use in subsequent breeding programs.

The plant breeding effort can be seen as a two-stage process. The first five-year phase primarily involves research at central agriculture research stations, at an estimated $2 million per year for research on all five crops. During this initial phase, promising germplasm is identified and the general breeding techniques are developed for later adaptive breeding.

During the second phase, the research needs to shift to national agricultural research centers. Total costs and duration of this second phase are difficult to estimate, but will depend on the number of countries involved and the number of crops worked on in each country. Certainly, the annual cost for each country should not be more than the $2 million per year per environment, estimated for the first phase.

2. Benefits to Improved Human Nutrition

The World Bank (1994) estimated that at the levels of micronutrient malnutrition existing in South Asia, 5% of gross national product is lost each year due to deficiencies in the intakes of just three micronutrients: iron, vitamin A, and iodine. For a hypothetical country of 50 million persons burdened with this rate of malnutrition, deficiencies in these three nutrients could be eliminated through fortification programs costing a total of $25 million annually, or 50 cents per person per year. The monetary benefit to this $25 million investment is quite high in terms of increased productivity -- estimated to be $20 per person per year, or a forty-fold return on an investment of 50 cents. These benchmark numbers will be used below as a basis of comparison with the benefits of a plant breeding strategy.

3. Calculation of Benefit-Cost Ratios

Calculation of benefit-cost ratios inherently requires making a number of assumptions. In order to tie the assumptions made here to a concrete example, it will be helpful to use the country of Turkey as a case in point. The analysis will focus specifically on the problem of wheat production on zinc-deficient soils in Turkey and its relationship to human nutrition.

Using numbers rounded off for illustrative purposes, about 10 million hectares of wheat are harvested each year in Turkey. Average yields are about two tons per hectare, giving a production of 20,000 million kilograms for a population of 60 million people. International trade in wheat is negligible, giving a per capita annual availability of 330 kilograms, of which about two-thirds is directly consumed as wheat and bread products (FAO, 1995). This establishes that wheat is the primary food staple consumed.

In deriving benefit-cost estimates, it will be useful to make assumptions which will likely understate benefits and overstate costs, as discussed below.

About one-half of the ten million hectares harvested to wheat are planted on zinc-deficient soils (Cakmak, 1999). Braun (personal communication) has estimated that development of zinc-efficient varieties (the seeds of which are more viable and vigorous as discussed in section VI) will allow a lowering of seeding rates from 250 to 150 kilograms on these five million hectares. A savings of 80 kilograms per hectare in seed is assumed below.

Zinc-efficient varieties with zinc-dense seeds are higher yielding in zinc-deficient soils for reasons discussed in section VI. Data from Cakmak (1999) show that wheat yields for present (non-efficient) genotypes be increased by between 1-2t/ha on zinc-deficient soils through applications of zinc fertilizers. In the benefit-cost analysis, an increase in yield of 300 kilograms (about 20% of the possible yield increase of 1.5 t/ha suggested by the data for zinc fertilizer applications) is assumed for improved zinc-efficient genotypes grown on zinc-deficient soils.

It is assumed that zinc-efficient cultivars eventually are adopted on a maximum of only four million out of five million hectares. This is forty percent of total area harvested for all wheat production in Turkey. It is assumed that these improved cultivars are consumed by the equivalent of 20 million people, or one-third of the population.

Between 1993 and 1996, international prices of wheat varied between $120 and $263/t (FAO, 1997). A value of $125/t is used in the benefit-cost analysis here.

Although there is increasing evidence that zinc deficiencies are a major public health problem in developing countries (Gibson, 1994), no estimates of the benefits of zinc intervention programs exist simply because zinc fortification and supplementation programs are not yet in place. The World Bank analysis of a set of comprehensive fortification programs for three nutrients provides an estimate of benefits of $20 per person per year, or about $7 benefit per person per nutrient deficiency eliminated.

For lack of a basis for a more precise estimate, the benefit-cost analysis here simply assumes a benefit of only $1 per person equivalent consuming the zinc-dense improved wheat cultivars. As discussed in section VIII.A, if zinc densities in food staples and other foods are similar to those of iron, and if zinc densities in wheat seeds were doubled and percent bioavailable zinc were to remain constant, this could represent the benefit derived from a daily increase of 50 percent in the intake of bioavailable zinc.

Turning to costs, it is initially assumed that all first-phase central research costs (for all five crops for all three nutrients - $10 million in total over five years) accrue to only one country, Turkey, and that this research is very narrowly successful -- only for zinc for wheat. Second phase costs (over five years) for Turkey are also assumed to be $10 million, again successful only for zinc for wheat. Maintenance breeding costs are assumed to be $200,000 per year after adaptive breeding (phase 2) is completed.

A benefit-cost analysis, shown in Table XI, is undertaken applying these assumptions. Note that no benefits are derived from plant breeding research until year 11 of the analysis, after all adaptive breeding is completed. Improved varieties are assumed to be adopted on 0.8 million hectares, 1.6 million hectares, 2.4 million hectares, 3.2 million hectares, in years 11 through 14 respectively, and on 4.0 million hectares thereafter.

All benefits and costs are expressed in terms of present value. A discount rate of 12% is used (see Gittinger, 1982, p. 314). Because of the long lag times between investment and benefits for agricultural research, benefits are discounted by 80-90% in years 15-20 of the analysis.

Expressed in present values, costs are about $13 million and benefits $274 million, giving a benefit-cost ratio of over 20, which is quite favorable despite the very conservative assumptions made and despite the long time lag between investments and benefits.

In Table XI, the present value of central research costs ($8.1 million) constitute 62% of total costs. Although the results of this central research may be applied in countries (developing and developed) all over the world, all costs are applied to Turkey. Also in Table XI, agronomic benefits of the improved seeds account for 90% of the benefits, partially due to the conservative assumptions made as to nutritional benefits.

Table XII undertakes an analysis of the sensitivity of benefit-cost ratios to some of these assumptions. Specifically, benefit-cost ratios are calculated assuming instead that: (i) 10% of central costs are apportioned to Turkey, (ii) agronomic benefits are reduced by half over those assumed in Table XI, (iii) only nutrition benefits are realized (no agronomic benefits), and (iv) an additional five years and $10 million in costs are required to realize agronomic and human nutrition benefits.

Under the more realistic assumption that only ten percent of central costs are applied to Turkey and maintaining the benefits structure applied in Table XI intact, the benefit-cost ratio approaches the extremely high value of 50, which compares well with the benefit claimed by the World Bank for supplementation and fortification programs directed at micronutrient malnutrition.

Delaying benefits by five years (and increasing costs) reduces benefit-cost ratios by a factor of about three. Still benefit-cost ratios remain at high levels, except in the extreme case where there are only human nutrition (and no agronomic) benefits and the realization of these benefits is pushed back to years 16-20.

Does this mean that a plant breeding strategy directed at improving human nutrition depends critically on the coexistence of agronomic benefits? Certainly not. Benefit-cost ratios for scenarios involving only benefits to human nutrition in Table XII reasonably could be multiplied by several multiples for the following reasons:

1. Size of population affected. Zinc-dense varieties are assumed to reach 1 out of 3 persons in Turkey (ignoring any population growth). Some agricultural research could and should be directed at incorporating zinc-density into high-yielding wheat varieties grown in parts of the country where zinc-deficient soils are not a problem. This would expand the population benefitting from improved zinc intakes.

Morever, Turkey is not a large country in terms of population. Turkey was selected for these calculations because of the availability of reasonably hard numbers for undertaking estimates of agronomic benefits (see Cakmak, 1999). For more populous countries (e.g. India), the underlying research costs are not necessarily substantially higher, but the benefits may accrue to a much larger number of people.

2. Size of benefit per person. According to World Bank estimates cited earlier, elimination of iron, iodine, and vitamin A deficiency is worth an annual benefit of about $7 per nutrient per capita. A figure of $1 per capita has been used here for zinc. There is no specific reason for lowering the benefit by a factor of seven other than a methodological strategy of undertaking conservative estimates of benefits. While the daily ìdosageî of zinc supplied through nutritionally-improved staple grains will be lower certainly than could be supplied through supplementation or perhaps commercial fortification (although no zinc supplementation or fortification programs are as yet in operation in developing countries), once these varieties enter the food production and marketing system they will provide this dosage day in and day out for an entire lifetime eventually.

This last point highlights an essential difference between investments in standard fortification programs and fortification through plant breeding strategies. Standard fortification programs must be sustained at the same level of funding year after year. If investments are not sustained, benefits disappear. Such investments apply to a single geographical area such as a nation-state. By contrast, investments in research in plant breeding have multiplicative effects -- benefits may accrue to a number of countries. Moreover, benefits are sustainable -- benefits from breeding advances typically do not disappear after initial investments and research are successful, as long as an effective domestic agricultural research infrastructure is maintained. Finally, these benefits are economic only and do not take into account improved health per se and quality of life, nor the potential of the latter to lower population growth rate.

IX. Conclusions

A. Biological

 

The breeding of nutrient-dense staple foods that give resource-poor populations a better balanced nutritional base to their diet is feasible. Results so far indicate that the breeding parameters are not difficult and are highly likely to be cost-effective. The following points are seminal:

• Adequate genetic variation in concentrations of b-carotene, other functional carotenoids, iron, zinc, and other minerals exists in the major germplasm banks to justify selection.
• The micronutrient-density traits are stable across environments.
• In all crops studied, it is possible to combine the high micronutrient-density trait with high yield, unlike protein content and yield that are negatively correlated.
• The genetic control is simple enough to make breeding economic
• It will be possible to improve the content of several limiting micronutrients together, thus pushing populations towards nutritional balance.
• High nutrient density not only can benefit the consumer but also produce more vigorous seedlings in the next generation.
•  
• Bioavailability of the extra nutrient in elite breeding lines is high for rats and where the density is high enough for the test, also to human colon cell lines. Tests on human populations is now a high priority.

The combining of benefits for human nutrition and agricultural productivity, resulting from breeding staple food crops which are more efficient in the uptake of trace minerals from the soil and which load more trace minerals into their seeds, results in extremely high ex ante estimates of benefit-costs ratios for investments in agricultural research in this area. This finding derives from the confluence of several complementary factors:

• Rates of micronutrient malnutrition are high in developing countries, as are the consequent costs to human welfare and economic productivity.
• Because staple foods are eaten in large quantities every day by the malnourished poor, delivery of enriched staple foods (fortified by the plants themselves during growth) can rely on existing consumer behavior.
• A significant percentage of the soils in which these staple foods are grown is deficient in one or more of these nutrients, which has kept crop yields low. In general, these soils in fact contain high amounts of trace minerals (sufficient for hundreds of crops). However, because of chemical binding to soil particles, these trace minerals are poorly available to the staple crop varieties under present use by farmers.
• Adoption and spread of nutritionally-improved cultivars by farmers can rely on profit incentives, either because of agronomic advantages on micronutrient-deficient soils or incorporation of nutritional improvements in the most productive cultivars being released by agricultural research stations.
• Benefits to relatively small investments in agricultural research may be disseminated widely, potentially accruing to hundreds of millions of people and hectares of crop lands.
• Benefits from breeding advances, derived from initial, fixed costs, typically do not involve high recurring costs, and thus tend to be highly sustainable as long as an effective domestic agricultural research infrastructure is maintained.

In treating iron deficiency in developing countries, Yip (1994) argues that if prevalence rates are above 25%, the best approach is to develop programs to improve the iron status of the entire population. In such situations, which for preschoolers and women in developing countries are the rule rather than the exception, this is cheaper than screening for iron-deficient individuals. By increasing the iron content of food staples through plant breeding, the entire iron status distribution curve can be shifted to the right, so that targeting a subsequently smaller group of iron-deficient persons could become feasible. On the other hand, since the cultivars under development contain phytate-bound iron of plant origins, their dissemination can be expected not to increase the risk of iron toxicity that may exist in a very low percentage of wealthy individuals in the population.

Unfortunately, much less is known about the prevalence of zinc deficiency in developing countries, or about the distribution curve for biochemical indicators of zinc status. Even less is known about the cost of interventions for the prevention and control of zinc deficiency simply because wide-scale testing of possible interventions awaits conclusion of the debate over whether zinc deficiency should be regarded as a major public health problem. Certainly, plant breeding is an option which should be exploited as quickly as possible to reduce zinc deficiency.

A plant breeding strategy, if successful, will not eliminate the need for supplementation, fortification, dietary diversification, and disease reduction programs in the future to combat micronutrient malnutrition. Nevertheless, this strategy does hold great promise for significantly reducing recurrent expenditures required for these higher cost, short-run programs by significantly reducing the numbers of people requiring treatment. Cost is not a key issue in the decision to pursue a plant breeding strategy to improve human nutrition. A relatively modest level of resources is required and the potential payoff is quite high.

Rather the two key issues are: (i) will staple varieties with mineral-dense seeds be widely adopted by farmers in developing countries, either because of agronomic advantages on trace mineral-deficient soils or through this characteristic being bred into the highest-yielding varieties?; and (ii) will the additional nutrients contained in the seeds be of a sufficient magnitude and sufficiently bioavailable so as to have an appreciable impact on micronutrient status?