HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Google Scholar Articles by Pfeiffer, W. H. Articles by McClafferty, B. PubMed Articles by Pfeiffer, W. H. Articles by McClafferty, B. Agricola Articles by Pfeiffer, W. H. Articles by McClafferty, B. Related Collections Crop Genetics

HarvestPlus: Breeding Crops for Better Nutrition

^a HarvestPlus, c/o CIAT, A.A. 6713, Cali, Colombia
^b HarvestPlus, c/o IFPRI, 2033 K St., NW, Washington, DC., 20006

^* Corresponding author (w.pfeiffer{at}cgiar.org).

	ABSTRACT

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
Micronutrient malnutrition, the so-called hidden hunger, affects more than one-half of the world's population, especially women and preschool children in developing countries. Despite past progress in controlling micronutrient decencies through supplementation and food fortification, new approaches are needed to expand the reach of food-based interventions. Biofortification, a new approach that relies on conventional plant breeding and modern biotechnology to increase the micronutrient density of staple crops, holds great promise for improving the nutritional status and health of poor populations in both rural and urban areas of the developing world. HarvestPlus, a research program implemented with the international research institutes of the CGIAR, targets a multitude of crops that are a regular part of the staple-based diets of the poor and breeds them to be rich in iron, zinc, and provitamin A. This paper emphasizes the need for interdisciplinary research and addresses the key research issues and methodological considerations for success. The major activities to be undertaken are broadly grouped into research related to nutrition research and impact analysis, and research considerations for delivering biofortified crops to end-users effectively. The paper places particular emphasis on the activities of the plant breeding and genetics component of this multidisciplinary program. The authors argue that for biofortification to succeed, product profiles developed by plant breeders must be driven by nutrition research and impact objectives and that nutrition research must understand that the probability of success for biofortified crops increases substantially when product concepts consider farmer adoption and, hence, agronomic superiority.

Abbreviations: G x E, genotype by environment interaction • HPLC, high performance liquid chromatography • HTMs, high-throughput screening methods • ICP, inductively coupled plasma spectrometer • NIRS, near infrared reflectance spectrophotometry • TLC, thin-layer chromatography

	ACKNOWLEDGMENTS

Received for publication April 4, 2007.

HarvestPlus: Breeding Crops for Better Nutrition

^a HarvestPlus, c/o CIAT, A.A. 6713, Cali, Colombia
^b HarvestPlus, c/o IFPRI, 2033 K St., NW, Washington, DC., 20006

^* Corresponding author (w.pfeiffer{at}cgiar.org).

Micronutrient malnutrition, the so-called hidden hunger, affects more than one-half of the world's population, especially women and preschool children in developing countries. Despite past progress in controlling micronutrient decencies through supplementation and food fortification, new approaches are needed to expand the reach of food-based interventions. Biofortification, a new approach that relies on conventional plant breeding and modern biotechnology to increase the micronutrient density of staple crops, holds great promise for improving the nutritional status and health of poor populations in both rural and urban areas of the developing world. HarvestPlus, a research program implemented with the international research institutes of the CGIAR, targets a multitude of crops that are a regular part of the staple-based diets of the poor and breeds them to be rich in iron, zinc, and provitamin A. This paper emphasizes the need for interdisciplinary research and addresses the key research issues and methodological considerations for success. The major activities to be undertaken are broadly grouped into research related to nutrition research and impact analysis, and research considerations for delivering biofortified crops to end-users effectively. The paper places particular emphasis on the activities of the plant breeding and genetics component of this multidisciplinary program. The authors argue that for biofortification to succeed, product profiles developed by plant breeders must be driven by nutrition research and impact objectives and that nutrition research must understand that the probability of success for biofortified crops increases substantially when product concepts consider farmer adoption and, hence, agronomic superiority.

Abbreviations: G x E, genotype by environment interaction • HPLC, high performance liquid chromatography • HTMs, high-throughput screening methods • ICP, inductively coupled plasma spectrometer • NIRS, near infrared reflectance spectrophotometry • TLC, thin-layer chromatography

	INTRODUCTION

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
To be healthy, human beings require more than 20 mineral elements and more than 40 nutrients, particularly vitamins and essential amino acids, all of which can be supplied by an appropriate diet. However, human diets often lack one or more of these essential nutrients. Poor quality diets, characterized by high intakes of staple foods and low consumption of animal and fish products, fruits, legumes, and vegetables (all rich sources of bioavailable minerals and vitamins), cause micronutrient malnutrition.

Micronutrient malnutrition, the so-called hidden hunger, affects more than one-half of the world's population, especially women and preschool children in developing countries (UN SCN, 2004). Even mild levels of micronutrient malnutrition may damage cognitive development and lower disease resistance in children, and reduce the likelihood that mothers will survive childbirth. The costs of these deficiencies, in terms of diminished quality of life and lives lost, are staggering (see www.harvestplus.org). Despite past progress in controlling micronutrient deficiencies through diet supplementation and food fortification, new approaches are needed to expand the reach of food-based interventions to the rural poor and contribute to sustainable micronutrient deficiency alleviation in burgeoning urban populations.

Biofortification, a new approach that relies on conventional plant breeding and modern biotechnology to increase the micronutrient density of staple crops, holds great promise for improving the nutritional status and health of poor populations in both rural and urban areas of the developing world (Graham and Welch, 1996; Graham et al., 1999, 2001; Pinstrup-Andersen, 2000; Underwood, 2000; Bouis, 2003). Plant breeding to increase micronutrient density began to gain legitimacy when deficiencies in micronutrients such as Fe, I, Zn, and vitamins were recognized as an issue of overwhelming global public health significance and one of the major development challenges of the 21st century. To capitalize on agricultural research as a tool for public health, in July of 2003 the Consultative Group on International Agricultural Research (CGIAR) established HarvestPlus: the Biofortification Challenge Program, thus adding food quality to its agricultural production research paradigm.

Biofortification research, a comprehensive program that spans from genetic crop improvement to research on the impact of biofortified crops on human health (Haas et al., 2005; Van Jaarsveld et al., 2005), is conducted mainly under the auspices of HarvestPlus. HarvestPlus targets a multitude of crops that are a regular part of the staple-based diets of the poor. However, discussing all progress achieved in all crops is beyond the scope of this paper. We will, therefore, give an overview of interdisciplinary research activities currently underway, introduce the underlying principles of breeding micronutrient dense crops, and address the key issues along the HarvestPlus impact pathway (Fig. 1 ). We will also place great emphasis on the current objectives, strategies, results, and activities of the plant breeding component of this multidisciplinary program.

View larger version (31K): [in this window] [in a new window]   Figure 1. HarvestPlus Impact Pathway.

HarvestPlus program objectives are established through a series of interdisciplinary activities. The major activities to be undertaken are allocated into research-related groups, including nutrition research, plant breeding and nutritional genomics research, research on delivering biofortified crops to end users effectively, and communication activities to support internal project constituents and external audiences.

Full-fledged plant breeding programs are underway for six staple foods (Phase I crops)—rice (Oryza sativa L.), wheat (Triticum aestivum L.), maize (Zea mays L.), cassava (Manihot esculenta Crantz), orange-fleshed sweetpotato [Ipomoea batatas (L.) Lam.], and common beans (Phaseolus vulgaris L.)—that are consumed by the majority of the world's poor in Africa, Asia, and Latin America. Feasibility studies on these crops have already been completed, and prebreeding feasibility studies are being undertaken on 10 additional staples (Phase II crops): bananas/plantains (Musa acuminata x M. balbisiana Colla), barley (Hordeum vulgare L.), cowpeas [Vigna unguiculata (L.) Walp.], groundnuts [Vigna subterranea (L.) Verdc.], lentils (Lens culinaris Medik.), millet (Panicum miliaceum L.), pigeon peas [Cajanus cajan (L.) Millsp. syn. Cajanus indicus Spreng.], potatoes (Solanum tuberosum L.), sorghum [Sorghum bicolor (L.) Moench], and yams (Dioscorea spp.).

There are marked differences between conventional plant breeding and the biofortification breeding process. Conventional breeding seeks mainly to improve traits of known economic value and develop product concepts for existing markets. Breeding for biofortification, on the other hand, focuses on making an impact on human micronutrient status and maintains close links with nutrition and socioeconomics research to be able to develop concepts for products with traits whose value is measured in terms of health outcomes. At the core of any biofortification breeding program is a product pathway driven by nutrition and impact research.

The Role of Nutrition
For biofortification to be successful, micronutrient levels targeted by breeding programs must be set by nutritionists who understand the complexities of making a measurable impact on human health. To set the target levels and determine the likely contribution to nutritional status, critical information is needed on the bioconversion and bioavailability of ingested nutrients; retention of the micronutrient after storage, processing, and cooking; human micronutrient requirements; and potential levels of consumption by the target population (Nestel et al., 2006). Hence, biofortification requires direct linkages between plant science research and the human health and nutrition sectors (Bouis, 2003), which, thus, become an integral part of crop improvement and product development.

Understanding the Needs of Farmers and End Users
The acceptance of biofortified crops by producers and consumers hinges on developing attractive trait packages without compromising agronomic and end-use characteristics. Product concepts for biofortified crops rely on feedback and a continuous flow of information from socioeconomics, impact assessment studies, and marketing and consumer behavior research. The linkages between plant breeding and socioeconomics allow the exchange of information to determine micronutrient target levels and the micronutrient burden to identify and define target areas, and to quantify micronutrient trait values in economic terms. The micronutrient burden is imposed by micronutrient deficiencies and related diseases on individuals and can be quantified using the disability-adjusted-life-years approach (Stein et al., 2005).

Issues related to trait visibility in micronutrient-dense biofortified crops are crucial for developing product concepts. Higher levels of provitamin A carotenoids will turn endosperm, seed, or tuber color from white or light yellow to dark yellow and orange. Color is important because consumers often prefer, for example, wheat flour, maize, and cassava from white-seeded or white-fleshed varieties. For consumers and producers to accept biofortified (yellow or orange) versions of these crops, they would have to be convinced of the health benefits of the new biofortified varieties. Therefore, diagnostic studies on the feasibility of achieving behavioral changes in a target population are crucial to developing product concepts for biofortified crops.

In contrast to carotenoid levels, high mineral concentrations are not visible and do not affect sensory traits. However, trait invisibility is linked to obstacles such as how to distinguish biofortified crops and issues relating to product identity, branding, and procurement. In addition, HarvestPlus, as an impact-focused delivery project, has to consider effective formal and informal seed systems, development of markets and food products, and demand creation.

	Crop Improvement

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
General Objectives and Strategies
In conventional breeding, traits are targeted for selection based on whether they can provide better crop and/or utilization options to farmers; however, nutritional value has been largely ignored. Existing genetic variation, trait heritability, gene action, associations among traits, and the availability of screening techniques and diagnostic tools are criteria commonly used to identify candidate traits and estimate potential genetic gains. A range of other criteria are applied to determine the probability of success and the feasibility of biofortified germplasm product concepts.

Breeding for enhanced micronutrient content requires improving a spectrum of traits essential for varieties to be adopted by farmers. Given that high levels of bioavailable Fe and Zn are not visible, farmer adoption of mineral-dense varieties will most likely depend on their concomitant value-added agronomic or sensory traits. Adoption of biofortified crops with visible traits will require that both producers and consumers accept the sensory changes, along with equivalent productivity and end-use features. Factors (e.g., those related to genetic advance and the probability of success) that culminate in product concepts and the methodologies used to achieve them are the subject of the ensuing sections.

Conceptual Framework
The key activities of biofortified germplasm development, according to HarvestPlus, are outlined in Fig. 2 . Different research categories reflect sequentially arranged stages and milestones, and are superimposed on a decision tree that allows monitoring progress and making strategic and go/no-go decisions when goals and targets cannot be achieved. Figure 2 illustrates how nutrition, food technology, and socioeconomics are integrated in product development. When dealing with a novel trait, the product concept needs to be modified and adjusted as part of an iterative process that builds on results generated during recurrent crop development cycles and associated nutrition research.

View larger version (37K): [in this window] [in a new window]   Figure 2. HarvestPlus Breeding Framework.

The source of genetic variation is essential for the next breeding steps. If variation is present in the strategic gene-pool, prebreeding is required. If variation is present in the tactical gene pool, the materials may be used directly to develop competitive varieties. Most breeding programs simultaneously conduct prebreeding and final product enhancement to develop germplasm combining high levels of one or more micronutrients. If the available genetic variation suggests that target micronutrient increments are unlikely to be reached, it is still possible to find genetic variation through transgressive segregation or by exploiting heterosis. When variation is not available, a transgenic approach may be the only remaining option (Khush, 2002; Bouis et al., 2002; Al-Babili and Beyer, 2005). Further, options are currently being explored that could enhance micronutrient uptake, translocation, and sequestration in edible plant parts. Also, the search is on for mutants with favorably altered biochemical pathways (e.g., carotenoid synthesis; Matthews et al., 2003).

The next breeding steps involve developing and testing micronutrient-dense germplasm, conducting genetic studies, and developing molecular markers to facilitate breeding. Also, the influence of the growing environment on micronutrient expression, (i.e., genotype x environment [G x E] interaction) is studied in experiment stations and farmers' fields. Subsequent end-use activities include delivering seed to farmers and cooperators and, more specifically to HarvestPlus, building research capacity, and establishing productive research networks with national program partners and advanced research institutes. Although the same principles and methodologies are applied when breeding for enhanced levels of minerals, provitamins A, or micronutrient promoters and inhibitors as for conventional traits, the development of enabling technologies such as trait diagnostics for micronutrient screening pose a challenge to applied breeding and largely drive decisions on effective breeding methodologies.

Product Concepts
Methodologies and procedures for breeding micronutrient-dense crops follow the same principles applied to other traits with analogous characteristics, but that have been tailored according to breeding objectives and trait diagnostics. For example, product concepts for maize embrace a range of biofortified germplasm products for various target countries or maize-producing areas. These include germplasm products enriched with individual micronutrients or combinations of Fe, Zn, and provitamins A. Both conventional and quality protein maize adapted genetic backgrounds are used as platforms to add micronutrient density. Breeding efforts are focused on developing hybrids, but also consider synthetics and open-pollinated varieties.

Two factors are critical to farmer adoption: trait visibility and infrastructural development. Due to a strong cultural preference in Africa, Asia, and Latin America for white maize for human consumption, product concepts for provitamin A–dense maize have to consider acceptability to producers and consumers, and feasibility studies are required to guide breeding decisions. If crops are new or nontraditional, color preferences have not usually been established. Provitamin A–dense maize with orange-colored grain may be perceived as a new product and accepted. For visible traits, added nutritional value can constitute an incentive for adopting biofortified varieties, if crop and utilization options are comparable to those of current varieties.

For Fe and Zn, invisible traits that do not affect sensory characteristics, added economic value due to higher productivity and/or improved end-use quality may be essential for adoption. Technically, genotypes and germplasm products can be distinguished via morphological, biochemical, or molecular characteristics; thus, breeders could incorporate a marker trait for distinguishing micronutrient-dense genotypes. However, these markers are impractical for identifying, procuring, and labeling or branding a product, since they are not directly linked to a mineral and not apparent to growers and consumers. In addition, breeding for these types of markers is not viable, given the added costs and negative impact on genetic progress due to breeding for additional traits. Hence, breeding for micronutrient density must consider strategies to keep pace with rates of progress for value-added traits, particularly yield, in non-biofortified germplasm, while simultaneously incorporating additional traits for micronutrient density.

Product concepts must also address achieving agronomic superiority by exploiting the entire spectrum of genetic and nongenetic options. While the first challenge is of a more generic nature (and related to gains from selection G_s = i_ph²; G_s is a function of selection intensity i, genetic variation _p, and heritability h²), product concepts are specific to the infrastructure and environmental conditions in the target area, and to specific crops, with their own adaptive traits and end-use quality features.

Assessing Genetic Variation
Establishing germplasm screening and developing enabling technologies is a prerequisite for effectively assessing genetic variation. Given the large number of materials to be analyzed and the short turnaround time for doing sample analysis of crops with two or more cycles per year, breeding effectively for minerals and provitamin A carotenoids depends on the availability of low-cost, quick high-throughput screening methods (HTMs). Further, rapid postharvest deterioration, in particular of tuber crops (which are harvested with high moisture content and whose carotenoids experience varying degrees of degradation during storage, drying, milling, and processing), challenge sampling and trait diagnostics.

Rapid techniques for screening cereals, legumes, and tubers for minerals and provitamins A are currently being developed, validated, and implemented at various CGIAR centers and national research institutes. These research efforts include developing protocols for conventional analytical methods, given that sample preparation, as well as digestion, extraction, and milling procedures need to be standardized, and that the accuracy of participating laboratories has to be assessed by external quality assurance programs. In view of the lack of published information on micronutrient analysis and its critical role in breeding, we have included the following section, which describes in greater detail micronutrient analysis and research that has been conducted to date.

Micronutrient Analysis
Mineral micronutrients make up a minuscule fraction of the physical mass of a grain, tuber, or fruit, with concentrations in the ppm range. Typical Fe and Zn concentrations in major crops range from 5 µg g^–1 to 150 µg g^–1. In crops that show genetic variation for provitamins A or β-carotene, typical concentrations range from >1 µg g^–1 to >400 µg g^–1. Consequently, sensitive methods are required for conducting accurate micronutrient analyses.

Special attention is being given to developing experimental procedures that can eliminate sources of contamination. Iron contamination from soil, dust, metal parts, rubber products (particularly silicon and neoprene), paint, and gloves during harvest or grinding or milling processes can be a major problem. Thus the prevention of crop lodging to avoid Fe contamination from the soil, and suitable spatial experimental designs with replicated standards or checks to estimate error are warranted. In the literature, contamination is, in general, not addressed, and erroneously high Fe concentrations are reported that are likely due to contamination.

Minerals
Table 1 summarizes precision analysis and HTMs that are applied and/or being tested for use in breeding different crops. The inductively coupled plasma spectrometer (ICP), atomic absorption spectrometer, and X-ray fluorescence spectrometer allow identifying a wide range of micronutrients including elements such as P, which is indicative of the antinutrient phytate. The ICP is the current method of choice to detect elements such as aluminum, which has been proposed as an indicator of contaminant Fe. The various HTMs will be applied in prescreening to reduce the large number of samples in population development to a more practical number that can be used in high precision analyses. Depending on the method used, a 66% proportion for more qualitative colorimetric methods to a 75 to 85% proportion for semiquantitative methods of "lows" can be discarded. Correlations between Fe determined by ICP and by the 2,2 dipyridal colorimetric method in rice, wheat, maize, sweetpotato, and cassava are between 0.88 and 0.98 (J. Stangoulis, personal communication, 2006). Elements indicative of contamination are not determined with color staining techniques, but need to be quantified as part of subsequent precision mineral analysis.

View this table: [in this window] [in a new window]   Table 1. Analytical methods for micronutrients.

Milling and Polishing
Minerals in rice, wheat, maize, and other cereals are concentrated in the aleurone; mineral concentration decreases sharply toward the center of the kernel and is much lower in the endosperm. The mineral-containing aleurone layer is removed by polishing or milling; small differences in polishing can thus have a dramatic effect on micronutrient concentration. Extensive studies at the International Rice Research Institute (IRRI) have revealed a poor association between mineral concentrations in brown rice and polished rice (G. Barry, personal communication, 2005), while recent studies by Sison et al. (2006) resulted in a close correlation. Research on how the degree of polishing or milling affects micronutrient concentration is complicated, as researchers at the same time are trying to identify suitable noncontaminating equipment for milling small samples (such as seed from individual plants). Standardized screening protocols have been developed and are now being validated and implemented by breeding projects to achieve a level of standardization that would permit data comparison.

Micronutrient Concentration versus Content
Special care in milling is warranted when assessing seed of, for example, wild relatives of crop species, genetic stocks, inbred lines, and unadapted germplasm that may have small, shriveled seed and/or incomplete seed set. The same effect on seed and seed set occurs in adapted genotypes as a result of biotic and abiotic stress or other production constraints. If the remaining portions of a seed fraction have a high micronutrient concentration, this can inflate concentration levels. Further, shriveled seed may require a disproportionate degree of milling or polishing to make processed products that satisfy commercial or laboratory standards.

Seed shriveling, wrinkling, and weathering can have dramatic effects on micronutrient density, given that micronutrient concentration in the embryo and seed coat is much higher than micronutrient levels in the endosperm. The seed coat to endosperm ratio is high, which can result in elevated micronutrient concentrations, the "concentration" effect. Likewise, a concentration effect can result when few grains have been produced due to sterility or poor seed set, and fewer grains act as the micronutrient sink. In plump seed, the seed coat to endosperm ratio is much lower, causing a "dilution" effect. Because micronutrient concentration is generally determined on whole grain, concentration levels in shriveled seed can be overestimated (Cakmak et al., 2000; Imtiaz et al., 2003). Figure 3 displays the Fe concentration and content of 58 Triticum monococcum L. accessions and suggests that, in such cases, it is crucial to determine micronutrient content (µg seed^–1 or, in certain cases, µg plant^–1), rather than micronutrient concentration (µg g^–1 = mg kg^–1 = ppm). There are few reports in the literature regarding content within the context of mineral accumulation in conventional germplasm and transgenic materials.

View larger version (26K): [in this window] [in a new window]   Figure 3. Iron concentration and content measured on 58 Triticum monococcum accessions grown in field experiments at Cd. Obregon, Sonora, Mexico. Data source: I. Ortiz-Monasterio, unpublished data, 2006.

Contamination in Mineral Analyses
In assaying micronutrient concentration, the detection of contamination (for example, by Fe from soil, dust, threshing equipment, sample preparation, seed handling, or digestion procedures) is complicated, and references that could guide researchers are not yet available. Zinc is less subject to contamination than Fe. Research to establish protocols and guidelines for determining approximate thresholds and, in particular, for developing corrective measures is currently underway. However, diagnostics for contamination cannot substitute for validating micronutrient concentrations of selected genotypes through additional screening. Past research has rarely considered mineral contamination.

One approach for detecting contamination entails using indicator elements that are (i) abundantly found in soil, dust, or equipment; (ii) uniform in concentration of contaminating fractions (e.g., of soil); (iii) reproducibly released and easily measured. However, they must be absent in plants or seed, or present in trace concentrations. Earlier research considered Al, Ti, and Cr (or a combination of these) as indicator elements and attempted to identify the most suitable element and establish ideal threshold levels or bands. Other elements may also be suitable, but good indicator elements for soil are, by definition, very hard to determine accurately in plant tissues, if any soil at all is present. To date, research in this area focuses on Al as the indicator element of choice. Potential indicator elements have frequently not been measured or they are not reported in the literature.

Statistical approaches use population parameters to identify values that fall outside the range expected for an assumed normal distribution and to detect values that can be erroneously high. Their accuracy increases when data from replicated trials or replicated check varieties are available. The two approaches are complementary and should be used in combination.

Data on mineral analysis provided by HarvestPlus crop leaders from micronutrient screening of cereal, legume, and tuber crops have been analyzed to identify tentative Al thresholds for Fe contamination. The criteria used to examine contamination consisted of a combination of statistical outlier tests, comparison of data subsets with different Al ranges, comparison of replicated check data, and correlations among elements. Results suggest that Al concentrations of more than 5 to 10 µg g^–1 are frequently associated with contaminant Fe. Further, analyses revealed that significant correlations between Fe and Al generally indicate Fe contamination. Eliminating data with Al values >5 µg g^–1 from the analyses reduced the average correlation between Fe and Al across datasets or crops from r = 0.35 to r = 0.11; these findings coincided with results from statistical analyses.

Provitamin A Carotenoids
Spectrometric Measurements
The quantification of major carotenoids is a challenging task. One of the difficulties is the variation in the carotenoid composition of different crops (Rodriguez-Amaya and Kimura, 2004; Kimura et al., 2007). A crop's carotenoid composition is essential for determining the quantification method to use. Currently high performance liquid chromatography (HPLC) is the method of choice due to its sensitivity and selectivity. Although HPLC allows individual quantification of all provitamin A carotenoids, it is elaborate, time-consuming, and costly, making it unsuitable for rapid screening purposes. The cost per sample varies between US$50 and US$70, and the sample throughput is 15 to 45 samples a day.

Carotenoid extraction, followed by spectrometric measurement, is a simpler method. However, a thin-layer chromatography (TLC) method only separates the three different carotenoid groups (β-carotene and -carotene; β-cryptoxanthin; and lutein and zeaxanthin), but does not give a quantitative estimate of a particular carotenoid. If one provitamin A carotene predominates—for example, β-carotene in cassava and orange-fleshed sweetpotato—the TLC method can be used for quantification (Kimura et al., 2007).

Zeaxanthin and lutein are the major carotenoids in maize, which has much smaller amounts of β-carotene and β-cryptoxanthin (Kimura et al., 2007). Both lutein and zeaxanthin are vitamin A inactive, but have important roles in human health in terms of preventing macular degeneration and cataracts. When TLC is used, the absorption of lutein and zeaxanthin will completely mask, for example, β-carotene absorption; thus HPLC should be used both to separate provitamin A carotenoids from lutein and zeaxanthin, and to quantify them. Since β-cryptoxanthin and -carotene have about half the provitamin A activity of β-carotene, actual crop-specific levels are relevant in deciding whether their concentrations should be quantified and addressed through breeding.

Near Infrared Spectroscopy
Carotenoids show absorption in the visible and infrared regions of the electromagnetic spectrum. Hence, NIRS offers great potential for screening for total carotenoids and provitamins A. Brenna and Bernardo (2004) applied NIRS to determine carotenoids and the relevant vitamin A precursors β-carotene and β-cryptoxanthin in maize, and cross-validation procedures indicated close associations between HPLC values and NIRS estimates. In sweetpotato and cassava, β-carotene predominates among total carotenoids, and NIRS screening of the two crops is currently being validated at CIP and CIAT. Correlations between total carotenoids and β-carotene determined by HPLC and NIRS in tubers range from 0.80 to >0.90 (W. Grüneberg, CIP, personal communication, 2006). In addition to the high potential of NIRS for screening for total carotenoids and β-carotene, research revealed its effectiveness for prescreening for Fe and Zn, as well as precision screening for promoters, antinutrients, and value-added traits such as protein. Hence, NIRS could catalyze a paradigm shift in breeding for micronutrients by allowing inexpensive screening of large numbers of genotypes in early generations and the use of selection indices to simultaneously select for micronutrients, bioavailability, and relevant value-added traits.

Visual Screening of Provitamin A Carotenoids
The crop-specific variation in carotenoids translates into differences in the association between provitamin A concentration and visual color intensity. It also determines the suitability of using color intensity for visual grading or selection with color charts when prescreening for provitamins A. Color charts can be used for cassava and orange-fleshed sweetpotato, two crops in which β-carotene constitutes the major portion of provitamins A. For maize, visual color is dominated by the non-provitamin A precursors zeaxanthin and lutein, and inexpensive high-throughput visual selection can only be applied to separate white or light yellow maize grains from dark yellow and orange color types.

	Genetic Variation as a Prerequisite For Genetic Advance

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
The availability of genetic variation for micronutrient density is essential for determining the feasibility of achieving meaningful increments through conventional breeding. When there is sufficient genetic variation, breeders can exploit additive gene effects, transgressive segregation, and heterosis to improve micronutrient density. When the required genetic variation is not available, transgenic approaches can provide additional sources of variation from which to introgress provitamins A or Fe via ferritin in the endosperm and an alternative for achieving the target micronutrient density (Bouis et al., 2002; Taylor et al., 2004; Al-Babili and Beyer, 2005, Paine et al., 2005; Shewry and Jones, 2005; Sautter et al., 2006; Khalekuzzaman et al., 2006). In the future, breeding will likely combine both conventional and transgenic approaches.

Screening objectives entail assaying representative samples of the genetic diversity for micronutrient density that is available in breeding programs and gene banks. Agronomic and end-use quality features of trait-source genotypes are also evaluated. To date, only a relatively small portion of the existing genetic diversity for micronutrients, antinutrients, and promoters has been assayed. Evaluating all accessions of individual species conserved in gene banks would be beyond the scope of HarvestPlus due to the huge number of samples involved. Gene banks at CGIAR centers alone preserve more than 530,000 accessions in-trust (see http://www.cgiar.org/impact/accessions.html). For this reason, screening under HarvestPlus focuses on core collections of the different crops, which are representative of existing gene pool variation for target micronutrients.

Micronutrient concentrations are affected by numerous factors such as microenvironmental variation, G x E interaction, and germplasm type. Consequently, micronutrient data reported in the literature reveal significant variation in average values and genotypic variation per se among crops and within crop species. Thus, when interpreting these data, one must consider differences in sampling, milling, analytical protocols, and experimental screening design used, and, hence, error.

Ranges in micronutrient concentrations reported in the literature reveal significant genetic variation for numerous crops, including barley (Ma et al., 2004), beans (Beebe et al., 2000; Nunez-Gonzalez et al., 2002; Wissuwa, 2005), cassava (Maziya-Dixon et al., 2000; Chavez et al., 2000, 2005), maize (Bänziger and Long, 2000; Mi et al., 2004), rice (Gregorio et al., 2000), sorghum (Reddy et al., 2005; Kayodé et al., 2006), and wheat (Cakmak et al., 2000; Monasterio and Graham, 2000). White and Broadley (2005) provided a recent review of genetic variation for minerals. Maximum micronutrient levels are frequently present in genetically distant sources such as wild relative species or landraces. Accessing the genetic variation present in these sources and transferring it to adapted genetic backgrounds usually require prebreeding and parent building, depending on the extent of "linkage drag," which adds to product development time.

Figure 4 displays average (baseline) and maximum values for Zn in adapted germplasm resulting from HarvestPlus screening and allows predicting progress in the shorter term. The short-term exploitable variation in adapted germplasm is of similar magnitude for cereals and legumes for both Zn and Fe (not shown). Maximum values reported in the literature for Fe in certain crops can be up to 10 times higher than those encountered in later studies and coincide, in some cases, with high values for Al.

View larger version (25K): [in this window] [in a new window]   Figure 4. Typical average and maximum zinc concentrations for adapted genotypes evaluated in field experiments for major cereals, legumes, and tubers. Data source: HarvestPlus database.

	Setting Nutritional Target Levels

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
Available genetic variation influences the level of micronutrient increments that can be achieved through breeding, but contribution to nutritional status largely depends on factors related to bioavailability that have to be considered when setting nutritional target levels for breeding. Critical information is needed on how much nutrient is retained after storage, processing, and cooking; on micronutrient bioconversion and bioavailability once the nutrient is ingested; and on micronutrient requirements of the target population (Institute of Medicine, 2001; Nestel et al., 2006; White and Broadley, 2005). In addition, the daily micronutrient intake supplied by a crop to a given target population and target country must be considered when setting target levels. Many of these parameters are interrelated in a highly complex manner, since human micronutrient status, dietary composition, and health status affect bioavailability (for example, for β-carotene: β-carotene absorbed/β-carotene in food) and its components bioaccessibility (β-carotene freed/micronutrient in food), bioconversion (retinol formed/β-carotene absorbed), and bioefficacy (retinol formed/β-carotene in food). A more detailed discussion of these factors is beyond the scope of this paper.

As a starting point, when setting tentative target levels without detailed information at hand for accurate assessment, bioavailability of Zn can be assumed to be 25%, and bioavailability of Fe to be 5% for legumes (e.g., beans, lentil, cowpea) and cereals with significant concentrations of phytate {e.g., wheat, maize, sorghum, pearl millet [Pennisetum glaucum (L.) Lam.], barley}. For tubers (e.g., cassava, potato, sweetpotato, yams) and low-phytate rice, 10% bioavailability can be assumed.

Within this context, Fig. 5 illustrates the relationship between micronutrient intake and the micronutrient increment from the baseline concentration (for Fe [Fig. 5a], Zn [Fig. 5b], and β-carotene/provitamins A [Fig. 5c]) needed to make a measurable biological impact on women from a public health perspective. For Fig. 5, 100% retention has been assumed to allow generalizations across crops. Figures 5b and 5c imply, in combination with published information on micronutrient variation and HarvestPlus data, the feasibility of reaching nutritional target increments of Zn and provitamins A through conventional breeding for most crops in the shorter term. The genetic variation for Zn concentration in varieties, germplasm lines, and parental stocks, especially of cereal and legume crops, is high (Fig. 4; White and Broadley, 2005). However, due to the substantially lower bioavailability of Fe when compared with Zn, significantly higher micronutrient increments have to be added to obtain a measurable impact on human health and achieve nutritional target levels. The bioavailability of Fe and Zn is associated with the presence of antinutrients and/or the lack of promoter substances for micronutrients (see White and Broadley, 2005). Since an increase in bioavailability translates into a proportional decrease in the nutritional target increment (increasing Fe bioavailability from 5 to 10% reduces the target increment by 50%), breeding strategies for micronutrient density should consider indirect breeding for increased bioavailability, increased retention, or reduced postharvest micronutrient deterioration. Although not well understood, breeding for increased bioavailability offers tremendous potential (Welch et al., 2000; Lucca et al., 2001; Oikeh et al., 2003; Hambidge et al., 2004; Drakakaki et al., 2005).

View larger version (73K): [in this window] [in a new window]   Figure 5. Micronutrient increments from baseline concentrations needed for making a measurable biological impact on women from a public health perspective for various intake levels. For calculating micronutrient increments, we assumed 100% retention and (a) for iron, 5% and 10% bioavailability, and an 8 mg d–1 requirement; (b) for zinc, 25% bioavailability and a 3 mg d–1 requirement; (c) for β-carotene/provitamins A, assumed β-carotene/provitamins A to retinol bioconversion rates of 3:1, 6:1, and 12:1. We further assumed the crop provides 50% of the estimated average requirement (EAR) and a requirement of 500 µg EAR d–1.

	Breeding for Increased Bioavailability

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

Significant genotypic differences in retention that could be exploited in breeding have been found, for example, in cassava and yams for provitamins A and, to a lesser extent, for minerals (B. Maziya-Dixon, IITA, personal communication, 2006), and evaluation of the genetic variation for micronutrient retention in other crops is warranted. Also, micronutrient retention has been related to factors associated with flour extraction rate in wheat (e.g., grain hardness, texture, grain shape) and degree of polishing in rice (J. Peña, CIMMYT; G. Barry, IRRI, personal communications, 2006).

	Genetics

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
Knowledge of heritability as it relates to genetic progress (G_s = i_ph²) and associated genetics is crucial for decisions concerning screening and breeding methodologies, the scale of breeding operations, and G x E testing strategies. Growing evidence from HarvestPlus research supports findings that Fe and Zn concentration is controlled by several (2–5) relevant genes, and that mineral heritability is of intermediate magnitude (Philip and Maloo, 1996; Maloo et al., 1998; Long et al., 2004; Cichy et al., 2005). Provitamins A appear to be controlled by a few (2) major genes, and trait heritability is high (Egesel et al., 2003b; Menkir and Maziya-Dixon, 2004; Grüneberg et al., 2005). However, heritability can be overestimated if studies contrast non-provitamin A and high provitamin A genotypes. For both minerals and provitamins A, additive gene action and general combining ability predominate. Further, data on temperate maize have revealed significant reciprocal effects for provitamins A (Egesel et al., 2003a). Transgressive segregation for provitamins A has been encountered, for example, in cassava crossed with wild relatives, for Fe in beans, and for Zn in wheat.

Correlations among Minerals and Value-Added Traits
Screening of HarvestPlus crops has revealed a generic positive association between Fe and Zn concentrations that facilitates raising levels of both micronutrients simultaneously by selecting for both or via capitalizing on indirect selection response. Further, results exhibited positive correlations of a general nature for Fe and Zn with a range of nutritionally important minerals and trace elements (Ca, Cu, K, Mg, Mn, P, and S).

Significant associations between mineral concentration (in particular Zn) and grain protein concentration have been reported for wheat (Peterson et al., 1986; Feil and Fossati, 1995; A. Morgounov, CIMMYT, personal communication, 2006) and for maize, but of lower magnitude (Bänziger and Long, 2000). Due to a general negative association between grain protein and grain yield, particularly in wheat, these correlations could result in a significant negative association between mineral concentration and grain yield, as well as traits positively correlated with grain yield. However, as stated in the section on micronutrient concentration versus micronutrient content, correlations can be overestimated. Data on HarvestPlus crops (except wheat) have not revealed relevant negative associations between micronutrients and productivity traits (Menkir and Maziya-Dixon, 2004; Mi et al., 2004), but knowledge is still incomplete. Associations with sensory characteristics can be of relevance. In sweetpotato, dry matter content and β-carotene concentration are negatively associated (Zhang and Xie, 1988; Grüneberg et al., 2005), a fact that complicates breeding, since adult African consumers prefer sweetpotato with high dry matter content (Tomlins et al., 2004).

	G x E Interaction

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
Micronutrient trait expression and the extent of G x E interactions across different environments largely determine screening, breeding, and testing methodologies used, and reflect trait heritability, genetic variation, and, hence, potential genetic gains. Early biofortification efforts were hindered by knowledge gaps regarding site suitability for trait assessment and the effect of permanent and variable environmental factors, production constraints, and crop management practices on micronutrient concentration. Similarly, microenvironment variation had often not been investigated. Mineral traits were perceived as qualitative rather than quantitative until results from multi-environment experiments revealed significant G x E interactions and substantial differences in the suitability of test sites for micronutrient selection in terms of expressing variation and discriminating among genotypes (Reynolds et al., 2005).

An increasing body of evidence suggests that the expression of provitamins A across crops is relatively stable under different growing conditions, with crop-specific differences (Egesel et al., 2003b; Menkir and Maziya-Dixon, 2004). Cassava, maize, and sweetpotato genotypes with high and stable expression across environments were identified, with G x E interactions predominantly of the non-crossover type (H. Ceballos, CIAT; A. Dixon, IITA; W. Grüneberg, CIP; personal communications, 2006). These results agree with findings that provitamins A are controlled by relatively few genes and more simply inherited. The expression of Zn and, to a lesser extent, of Fe is related to and affected by permanent and variable environmental factors, and the higher variation due to G x E interactions when compared with provitamins A reflects the more complex inheritance of Fe and Zn, particularly in cereals and legumes. However, results from multi-environment trials identified genotypes of cereals, legumes, and tubers with high, stable trait expression in the presence of high G x E interaction.

Differences in genotypic variation for minerals, even at proximate test sites can be large. For cassava in Colombia, site means for Zn varied two- to threefold, while Zn standard deviations at sites varied two- to fourfold in multilocation experiments (H. Ceballos, CIAT, personal communication, 2005). For Fe, site mean values were comparable to the respective standard deviations. Similar results for Zn have been obtained for wheat in multilocation trials in Kazakhstan (A. Morgounov, CIMMYT, personal communication, 2005). Particularly for Zn, these results are not unexpected, since soil Zn deficiency is a common problem in major agricultural areas. Given the complexity of soil mineral dynamics and the interaction with environmental factors, soil mineral status often explains only part of poor Fe or Zn expression in the plant. Research aimed at understanding the underlying factors of G x E interactions and the magnitude of micronutrient trait expression by analyzing soil and plant samples from various target environments is being conducted. This will also help to understand the association between soil micronutrient status and crop mineral concentration. Further, research on the effect of Zn fertilizer on crop Zn levels is underway in target areas to identify optimal environments for screening micronutrients. Synergistic fertilizer effects can be exploited via crop management recommendations for farmers to increase mineral density and reduce spatial and temporal fluctuations due to G x E interaction.

Microenvironment variation for minerals, particularly Zn, can be highly significant and cause false high positives in mineral screening, especially when concentration is measured on seed from space-planted individual plants. Critical plot sizes to sample the variation are necessary for evaluating microenvironment variation. The use in breeding of standards, repeated checks, replications, and spatial experimental designs that take on-site spatial variation into account is essential for comparing results from different environments and for different types of germplasm. Larger plot sizes to better sample the variation are necessary for evaluating microenvironment variation and comparing results from different environments and for different types of germplasm. Further, field trials revealed highly significant differences in average mineral concentration and _p between planting seasons for rice and pearl millet (G. Barry, P. Virk, IRRI, 2005; K. Rai, ICRISAT, personal communications, 2006), and for different planting dates for wheat. Hence, next to spatial and temporal variation, systems variation caused by differential crop management practices can have significant effects.

	Strategies and Approaches for Breeding Competitive Biofortified Crops

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 
The major target areas for biofortified crops are located in developing countries, where malnutrition and micronutrient deficiency frequently coincide. Recent studies predict that per capita land and water resources will diminish in the coming decades, and that production will have to increase to meet rising food demands at the global level. Increasing and stabilizing production under these circumstances poses one of the greatest challenges for agricultural research of the 21st century, given the fragile and highly variable nature of target areas and the continued deterioration of natural resources. Hence, at the same time as micronutrient concentration is being improved, production efficiency in the different agroecological cropping systems must be maximized and the natural resource base protected. Within this context, environmental, cultural, and political sustainability is what defines the focus of the research agenda.

Breeding for High Yield and Micronutrient Density
Breeding for additional traits not associated with crop productivity or economic yield and, in particular, for novel traits causes lower rates of progress for productivity traits. When resources are not limited, increasing the operational scale and scope of breeding activities can substantially increase both _p and selection intensity i (G_s = i_ph²) and avoids compromising yield by breeding for micronutrient density.

Under HarvestPlus, micronutrient density is initially being addressed as a specific trait to accelerate product development for fast impact. However, a sustainable biofortification concept requires considering micronutrient density as a generic trait present in all germplasm products with trait incorporation in the tactical gene-pool. Micronutrient traits are presumably not subject to genetic erosion (such as that caused by the evolution of pathogenic races) and require little maintenance breeding once genes have been incorporated. Hence, the cost of breeding for micronutrients decreases over time, and micronutrient density built into the gene pool will not affect future breeding for productivity traits.

Other routine factors thought to enhance genetic progress are breeding and testing in target environments (or in controlled environments that reliably simulate target environments) to (i) increase heritability or the genetic correlation between selection and target environments; (ii) intensify testing of experimental germplasm in target environments; and (iii) facilitate the use of molecular markers in selection. Recent progress in the development of quick, inexpensive methods for screening micronutrients and using molecular marker–assisted background selection (Guzmán-Maldonado et al., 2003; Wong et al., 2004), particularly to accelerate the introgression of increased micronutrient concentration from exotic sources into locally adapted elite germplasm, suggests substantial increases in micronutrient breeding efficiency in the future.

Strategies for Achieving Agronomic Superiority
Factors associated with probability of success that should be considered in product concepts mainly relate to adoption and, hence, agronomic superiority.

Production increases can originate from various sources: (i) genetic gains in yield potential; (ii) genetic gains in tolerance or resistance to abiotic and biotic stresses; (iii) productivity gains through improved, sustainable crop management techniques; and (iv) the synergistic effects of all these factors within the context of production economics. In practice, indirect selection for tolerance or resistance to key constraints is frequently more efficient for raising genotypic production potential (and, eventually, triggering adoption) than selecting for yield or a specific abiotic stress per se. Breeders can raise productivity by concentrating on improving resistance or tolerance to biotic and abiotic factors, particularly diseases for which they have known and repeatable variation.

In most developing countries where micronutrient deficiency is prevalent, agronomic superiority can be achieved more easily by replacing open-pollinated varieties (for example, of maize, sorghum, and pearl millet) with hybrids or synthetics. However, if a product concept entails deploying hybrid technologies, it must consider the feasibility of having sustainable seed systems in place.

In addition, higher economic returns (e.g., via reducing production costs or agronomic inputs rather than increasing yields, and price premiums for end-use quality), higher yield or production stability, end-use quality and sensory traits, and crop characteristics related to production (such as weed suppression or threshability) can provide incentives for variety adoption.

Adoption often entails implementing a technological package. Improved agronomic practices, such as direct seeding under zero tillage, along with germplasm adapted to these practices, allow capitalizing on synergies between genetic and agronomic solutions to achieve production and end-use quality objectives (Trethowan et al., 2005). Further, since the agronomic component can be a prerequisite for trait expression and/or stability, developing crop management recommendations is warranted.

Pleiotropic effects associated with high micronutrient content can affect agronomic performance and, hence, agronomic options. For example, seed Zn concentration and micronutrient-dense seeds in wheat are closely associated with greater seedling vigor, increased stand establishment, and higher grain yield, in particular in Zn-deficient soils (Cakmak et al., 1990). Product concepts can capitalize on these options.

CGIAR Challenge programs are time-bound, independently governed programs of high-impact research that target CGIAR goals in relation to complex issues of overwhelming global and/or regional significance, and rely on partnerships among a wide range of institutions to deliver their products. In the case of HarvestPlus, biofortification research is conducted by a global alliance of research institutions and implementing agencies in developed and developing countries, and co-convened by two CGIAR centers, the International Center for Tropical Agriculture (CIAT) and the International Food Policy Research Institute (IFPRI). This paper builds on the research accomplishments of HarvestPlus to date. The authors would like to recognize the intellectual contributions made by research partners within the HarvestPlus Alliance. They would also like to acknowledge support received from key members of the donor community, including the Asian Development Bank, the Danish Agency for International Development Assistance, the U.K. Department of International Development, the Bill and Melinda Gates Foundation, the United State Agency for International Development, and the World Bank. Our special thanks to Alma McNab for her editorial expertise.

Received for publication April 4, 2007.

	REFERENCES

TOP ABSTRACT INTRODUCTION Crop Improvement Genetic Variation as a... Setting Nutritional Target... Breeding for Increased... Genetics G x E Interaction Strategies and Approaches for... REFERENCES

 

• Al-Babili, S., and P. Beyer. 2005. Golden Rice—Five years on the road—Five years to go? Trends Plant Sci. 10:565–573.[CrossRef][ISI][Medline]
• Bänziger, M., and J. Long. 2000. The potential of increasing the iron and Zn density of maize through plant breeding. Food Nutr. Bull. 21:397–400.
• Beebe, S., A.V. Gomez, and J. Renfigo. 2000. Research on trace minerals in the common bean. Food Nutr. Bull. 21:387–391.
• Bouis, H.E. 2003. Micronutrient fortification of plants through plant breeding: Can it improve nutrition in man at low cost? Proc. Nutr. Soc. 62:403–411.[CrossRef][ISI][Medline]
• Bouis, H.E., S. Harris, and D. Lineback. 2002. Biotechnology-derived nutritious foods for developing countries: Needs, opportunities, and barriers. Food Nutr. Bull. 23:342–383.
• Bregitzer, P., and V. Raboy. 2006. Effect of four independent low phytate mutations on barley agronomic performance. Crop Sci. 46:1318–1322.[Abstract/Free Full Text]
• Brenna, O.V., and N. Bernardo. 2004. Application of near-infrared spectroscopy (NIRS) to the evaluation of carotenoids in maize. J. Agric. Food Chem. 52:5577–5582.[CrossRef][ISI][Medline]
• Cakmak, I., M. Kalayc, H. Ekiz, H.-J. Braun, Y. Klnc, and Y. Ylmaz. 1990. Zinc deficiency as a practical problem in plant and human nutrition in Turkey: A NATO science-for-stability project. Field Crops Res. 60:175–188.[CrossRef]
• Cakmak, I., H. Ozkan, H.-J. Braun, R.M. Welch, and V. Römheld. 2000. Zinc and iron concentrations in seeds of wild, primitive, and modern wheats. Food Nutr. Bull. 21:401–403.
• Chavez, A.L., J.M. Bedoya, T. Sanchez, C. Iglesias, H. Ceballos, and W. Roca. 2000. Iron, carotene, and ascorbic acid in cassava roots. Food Nutr. Bull. 21:410–413.
• Chavez, A.L., T. Sanchez, G. Jaramillo, J.M. Bedoya, J. Echeverry, E.A. Bolanos, H. Ceballos, and C.A. Iglesias. 2005. Variation of quality traits in cassava roots evaluated in landraces and improved clones. Euphytica 143:125–133.[CrossRef][ISI]
• Cichy, K.A., S. Forster, K.F. Grafton, and G.L. Hosfield. 2005. Inheritance of seed zinc accumulation in navy bean. Crop Sci. 45:864–870.[Abstract/Free Full Text]
• Dorsch, J.A., A. Cook, K.A. Young, J.M. Anderson, A.T. Bauman, C.J. Volkmann, P.P.N. Murthy, and V. Raboy. 2003. V. Seed