Former Head, Division of Plant Breeding, Genetics and Biochemistry IRRI, Los Baños, the Philippines
Rice is the world's most important food crop and a primary source of food for more than half the world population. More than 90 percent of the world's rice is grown and consumed in Asia where 60 percent of the people on earth live. Rice accounts for 35 to 75 percent of the calories consumed by more than 3 billion Asians. It is planted to about 154 million ha annually or on about 11 percent of the world's cultivated land.
Rice is probably the most diverse crop. It is grown as far north as Manchuria in China and as far south as Uruguay and New South Wales in Australia. Rice grows at altitudes of over 300 m in Nepal and Bhutan and at 3 m below sea level in Kerala in India. Rice-growing environments are divided into four major categories: irrigated, rainfed, upland and flood-prone (Khush, 1984). The categorization is based on several criteria, including water regime, drainage, soils and topography.
Major advances have occurred in food production during the last four decades due to the adoption of green revolution technology. Between 1966 and 2000, the population of densely populated low-income countries grew by 90 percent, but rice production increased by 130 percent from 257 million tonnes in 1966 to 600 million tonnes in 2000. In 2000, the average per caput food availability was 18 percent higher than in 1966. The technological advance that led to dramatic achievements in world food production during the last 40 years was the development of high-yielding and disease- and insect-resistant varieties of rice. The adoption of green revolution technology was facilitated by:
The increase in per caput availability of rice and a decline in the cost of production per tonne of output contributed to a decline in real price of rice in international and domestic markets. The unit cost of production is about 20 to 30 percent lower from high-yielding varieties than from traditional varieties of rice (Yap, 1991). The cost of rice is 40 percent lower now than in the 1960s. The decline in food prices has benefited the urban poor and rural landless who spend more than half of their income on food grains.
THE RICE SCENARIO IN THE NEW MILLENNIUM
The world's capacity to sustain a favourable food production/population balance has again come under the spotlight in view of continued population increase and a drastic slowdown in the growth of cereal production (Brown, 1996). Rice production increased at a rate of 2.5 to 3.0 percent per year during the 1970s and 1980s. However, during the 1990s, the growth rate was only 1.5 percent. According to United Nations estimates, the world population will grow from 6.3 billion in 2003 to 8 billion in 2025. Most of this increase (93 percent) will occur in developing countries, whose share of population is projected to increase from 78 percent in the 1990s to 83 percent in 2020.
In spite of all the achievements of the green revolution, serious food problems still exist in the world. Every 3.6 seconds somebody dies of hunger. Chronic hunger takes the lives of 2 400 people every day. Currently there are more than 800 million undernourished people in the developing world, 300 million children under the age of five die of hunger and malnutrition and one out of five babies is born underweight.
FEEDING 5 BILLION RICE CONSUMERS IN 2025
According to various estimates, we must produce 40 percent more rice by 2025 to satisfy the growing demand without adversely affecting the resource base. This increased demand will have to be met from less land, using less water, less labour and fewer chemicals. If we are not able to produce more rice from the existing land resources, land-hungry farmers will destroy forests and move into more fragile lands, such as hillsides and wetlands, with disastrous consequences for biodiversity and watersheds. To meet the challenge of producing more rice from suitable lands, it is necessary to use rice varieties with higher yield potential and greater yield stability.
INCREASING THE YIELD POTENTIAL OF RICE
Strategies for increasing the yield potential of rice include:
Conventional hybridization and selection procedures
This is the time-tested strategy for selecting crop cultivars with higher yield potential. It has two phases. The first phase involves the creation of variability through hybridization between diverse parents. In the second phase, desirable individuals are selected on the basis of field observations and yield trials. It has been estimated that on average about a 1.0 percent increase has occurred per year in the yield potential of rice over a 3 5-year period since the development of the first improved variety of rice, IR 8 (Peng et al., 2000). The yields of crops where there is enough investment in research have become increasingly higher, and there is no reason why further increases cannot be attained.
Ideotype breeding aimed at modifying the plant architecture is a time-tested strategy for achieving increases in yield potential. Thus, selection for short-stature cereals, such as wheat, rice and sorghum, resulted in a doubling of yield potential. Yield potential is determined by the total dry matter or biomass and the harvest index (HI). Tall and traditional rices had an HI of around 0.3 and a total biomass of about 12 tonnes/ha, with a maximum yield of 4 tonnes/ha. Their biomass could not be increased by the application of nitrogenous fertilizers as the plants grew excessively tall, lodged badly and the yield decreased (rather than increasing). To increase the yield potential of tropical rice, it was necessary to improve the HI and nitrogen responsiveness by increasing lodging resistance. This was accomplished by reducing the plant height through incorporation of a recessive gene sd1 for short stature.
The first short-stature variety, IR 8, developed at the International Rice Research Institute (IRRI), also had a combination of other desirable traits, such as profuse tillering, dark green and erect leaves for good canopy architecture and sturdy stems. It responded to nitrogenous fertilizer much better, had a higher biomass (about 18 tonnes) and an HI of 0.45. Its yield potential was between 8 and 9 tonnes/ha (Chandler, 1969).
To increase the yield potential of rice further, a new plant type was conceptualized in 1988. Modern semi-dwarf rices produce a large number of unproductive tillers and excessive leaf area which cause mutual shading and reduce canopy photosynthesis and sink size, especially when they are grown under direct sowing conditions. To increase the yield potential of these semi-dwarf rices, IRRI scientists proposed further modifications of the plant architecture with the following characteristics:
This proposed ideotype became the "new plant type" (NPT) highlighted in IRRI's strategic plan (IRRI, 1989), and breeding efforts to develop NPT were initiated in 1990. The objective was to develop improved germplasm with 15 to 20 percent higher yield than the existing high-yielding varieties. Numerous breeding lines with the desired ideotype were developed (Khush, 1995) and shared with the national rice improvement programmes. Three NPT lines have been released in China and one in Indonesia. Other National Agricultural Research Systems (NARS) are evaluating and further improving NPT lines.
Yield improvement in maize has been associated with hybrid development. Yields of maize in the United States of America remained basically unchanged from the mid-nineteenth century until 1930 and then accelerated following the introduction of commercial double-cross hybrids. The subsequent replacement of double-cross hybrids by single-cross hybrids in 1960 is associated with a second acceleration in maize yields. The average yield advantage of hybrids over cultivars is approximately 15 percent (Tollenaar, 1994).
Rice hybrids with a yield advantage of about 10 to 15 percent over the best inbred varieties were introduced in China in the mid-1970s and are now planted to about 45 percent of the riceland in that country. Rice hybrids adapted to the tropics have now been bred at IRRI and by NARS and show similar yield advantages. The increased yield advantage of tropical rice hybrids is due to increased biomass, higher spikelet number and, to some extent, higher grain weight. Increased adoption of hybrids in the tropics should contribute to increased productivity.
Crop gene pools are widened through hybridization of crop cultivars with wild species and weedy races, as well as through intrasubspecific crosses. Such gene pools are exploited for improving many traits including yield. For example, Lawrence and Frey (1976) reported that a quarter of lines from BC3-BC4 segregants from the Avena sativa x Avena sterilis crosses were significantly higher in grain yield than the cultivated recurrent parent. Nine lines from this study (when tested over years and sites) had agronomic traits similar to the recurrent parent and 10 to 29 percent higher grain yield. The higher yield potential of these interspecific derivatives was attributed to higher vegetative growth rates or early seedling vigour.
Xiao et al. (1996) reported that some backcross derivates (from a cross between an Oryza rufipogon accession from Malaysia and cultivated rice) outyielded the recurrent parent by as much as 18 percent. They identified two quantitative trait loci (QTL) from wild species with a major contribution to yield increase. These QTL are now being transferred to several modern semi-dwarf varieties.
Since protocols for rice transformation are well established (Christou, Ford and Kofron, 1991), it is now possible to introduce single alien genes that can selectively modify yield-determining processes. In several crop species, incorporation of slower leaf senescence (the "stay green" trait) has been a major achievement of breeders in the past decade (Evans, 1993). In some genotypes with slower senescence, the rubico degradation is slower, resulting in longer duration of canopy photo-synthesis and higher yields. The onset of senescence is controlled by external and internal factors. Plant hormones (e.g. ethylene and abscisic acid) promote senescence. On the other hand, cytokinins are senescence antagonists; therefore, over-production of cytokinins can delay senescence. The ipt gene from agrobacterium tumefacions encoding an isopentenyl transferase (Akiyoshi et al., 1984) was fused with a senescence-specific promoter, SAG 12 (Gan and Amasino, 1995) and introduced into tobacco plants. Leaf and floral senescence in the transgenic plant were markedly delayed, biomass and seed yield were increased, but other aspects of plant growth and development were normal. This approach appears to have great potential for improving the crop yields through slowing senescence and rubico degradation and thus improving canopy photosynthesis.
C4 plants such as maize and sorghum are more productive than C3 rice and wheat, because C4 plants are 30 to 35 percent more efficient in photosynthesis. Ku et al. (1999) and Matsuoka et al. (2001) are trying to alter the photosynthesis of rice from the C3 to the C4 pathway by introducing cloned genes from maize to regulate the production of enzymes responsible for C4 synthesis. If successful, the yield potential of rice may increase by between 30 and 35 percent.
BREEDING FOR DURABLE RESISTANCE
The full yield potential of modern rice varieties is not realized because of the toll taken by disease and attacks by insect organisms. It is estimated that diseases and insects cause yield losses of up to 25 percent a year. Genetic improvement to incorporate durable resistance to pests is the preferred strategy for minimizing these losses. There is no cost to farmers and resistant cultivars are easily adopted and disseminated - unlike "knowledge-based" technologies. Also, concern for the environment has become an important public policy issue and pest management methods that minimize the use of crop protection chemicals are increasingly finding favour.
Diverse sources of resistance to major diseases and insects have been identified and rice varieties with multiple resistance to diseases and insects have been developed. However, no sources of resistance to sheath blight are available and there is a shortage of donors for resistance to virus diseases and stem borer. Recent break-throughs in cellular and molecular biology have provided tools to develop more durably resistant cultivars and to overcome the problem of lack of donors for resistance.
Wide hybridization for disease and insect resistance
Wild species of rice are a rich source of genes for resistance breeding. For example, no cultivated rice was found to be resistant to grassy stunt, while Oryza nivara, a wild species closely related to cultivated rice, was found to be resistant, and the dominant gene for resistance was transferred to improved germplasm through backcrossing. This gene for resistance has been incorporated into many widely grown varieties. When genes are to be transferred from more distantly related species, special techniques (e.g. embryo rescue) are employed to reproduce inter-specific hybrids. Jena and Khush (1990) transferred genes for resistance to three biotypes of brown planthopper from O. officinalis to an elite breeding line. Multani et al. (1994) transferred genes for resistance to brown plant-hopper from O. australiensis to cultivated rice. Similarly, genes for resistance to blast and bacterial blight have been transferred form O. minuta to improved rice germplasm (Brar and Khush, 1997).
Molecular marker-assisted breeding
Numerous genes for disease and insect resistance are repeatedly transferred from one varietal background to another. Most genes behave in a dominant or recessive manner and require time-consuming efforts to transfer. Sometimes the screening procedures are cumbersome and expensive and require large field space. If such genes can be tagged by tight linkage with molecular markers, time and money can be saved in transferring these genes from one varietal background to another. The presence or absence of the associated molecular marker indicates at an early stage the presence or absence of the desired target gene. A molecular marker very closely linked to the target gene can act as a "tag", which can be used for indirect selection of the target gene.
Two of the most serious and widespread diseases in rice production are rice blast (caused by the fungus Pyrcularia oryzae) and bacterial blight (caused by Xanthomonas oryzae pv. oryzae).
The development of durable resistance to these diseases is the focus of a coordinated effort at IRRI using molecular marker technology. Efforts to detect markers closely linked to bacterial blight resistance genes have taken advantage of the availability of near isogenic lines having single genes for resistance. Segregating populations were used to confirm co-segregation between RFLP (restriction fragment length polymorphism) markers and genes for resistance. Protocols for converting RFLP markers into PCR- (polymerase chain reaction) based markers and for the use of PCR markers in marker-aided selection have been established (Zheng et al., 1995). The PCR markers were also used for pyramiding genes for resistance to bacterial blight. Thus xa4, x5, xa13 and Xa21 were combined into the same breeding line (Huang et al., 1997). The pyramided lines showed a wider spectrum and higher level of resistance than lines with only a single gene for resistance. Marker-assisted selection has also been employed for moving genes from pyramided lines into new plant type (Sanchez et al., 2000) and into improved varieties grown in India (Singh l., 2001).
Protocols for rice transformation have been developed to allow the transfer of foreign genes from diverse biological systems into rice. Direct DNA (desoxyribonucleic acid) transfer methods - e.g. protoplast-based (Datta et al., 1990) and biolistic (Christou, Ford and Kofron, 1991), as well as Agrobacterium-mediated (Hei et al., 1994) - are being used for rice transformation. Major targets for rice improvement through transformation are disease and insect resistance.
As early as 1987, genes encoding for toxins from Bacillus thuringiensis (Bt) were transferred to tomato, tobacco and potato, where they provided protection against Lepidoptern insects. A major target for Bt deployment in transgenic rice is the yellow stem borer. This pest is widespread in Asia and causes substantial crop losses. Improved rice cultivars are either susceptible to the insect or have only partial resistance. Thus Bt transgenic rice has much appeal for controlling the stem borer. Codon-optimized Bt genes have been introduced into rice and show excellent levels of resistance in the laboratory and greenhouse (Datta et al., 1997). Bt rices have also been tested under field conditions in China (Tu et al., 2000) and have excellent resistance to diverse populations of yellow stem borer. Besides Bt genes, other genes for insect resistance - e.g. those for proteinase inhibitors, á-amylase inhibitors and lectins - are also beginning to receive attention. Insects use diverse proteolytic or hydrolytic enzymes in their digestive gut for the digestion of food proteins and other food components. Plant-derived proteinase inhibitors or á-amylase inhibitors are of particular interest because these inhibitors are a part of the natural plant defence system against insect predation. Xu et al. (1996) reported transgenic rice carrying cowpea trypsin inhibitor (Cpti) gene with enhanced resistance against striped stem borer and pink stem borer.
Several viral diseases cause serious yield losses in rice. A highly successful strategy known as CP- (coat protein) mediated protection has been employed against certain viral diseases (e.g. tobacco mosaic virus) in tobacco and tomato. A CP gene from rice strip virus was introduced into two japonica varieties by electroporation of proto-plasts (Hayakawa et al., 1992). The resultant transgenic plants expressed CP at a high level and exhibited a significant level of resistance to virus infection and the resistance was inherited to the progenies.
BREEDING FOR ABIOTIC STRESS TOLERANCE
A series of stresses, such as drought, excess water, mineral deficiencies and toxicities in soil and unfavourable temperatures, affect rice productivity. The progress in developing crop cultivars for tolerance to abiotic stresses has been slow because of lack of knowledge of mechanisms of tolerance, poor understanding of inheritance of resistance or tolerance, low heritability and lack of efficient techniques for screening the germplasm and breeding materials. Nevertheless, rice cultivars with varying degrees of tolerance to abiotic stresses have been developed.
Rainfed rice is planted to about 40 million ha world-wide. Vast areas suffer from drought at some stage of the growth cycle. QTL for various component traits of drought tolerance have been mapped (Champoux et al., 1995) and the information is being utilized to develop improved cultivars with drought tolerance.
Genetic engineering techniques hold great promise for developing rice with drought tolerance. Garg et al. (2002) introduced ots A and ots B genes for trehalose biosynthesis from Escherichia coli into rice, and transgenic rices accumulated trehalose at three to ten times the rate of non-transgenic controls. Trehalose is a nonreducing disaccharide of glucose that functions as compatible solute in the stabilization of biological structures under abiotic stress. The transgenic rice lines had increased tolerance for abiotic stresses such as drought and salinity.
Accumulation of sugar alcohols is a widespread response that may protect the plants against environmental stress through osmoregulation. Mannitol is one of the sugar alcohols commonly found in plants. Tobacco plants lacking mannitol were transformed with a bacterial gene mtlD encoding mannitol (Tarczynski, Jensen and Bohnert, 1992). Mannitol concentrations exceeded 6 % mol/g (fresh weight) in the leaves and in the roots of some transformants, whereas this sugar alcohol was not detected in these organs of control tobacco plants. The growth of plants from control and mannitol-containing lines in the absence and presence of sodium chloride (NaCl) in culture solution was analysed. Plants containing mannitol had an increased ability to tolerate salinity (Tarczyznski, Jensen and Bohnert, 1993). After 30 days of exposure under concentrations of 250 mM NaCl in culture solution, transformed plants increased in height by a mean of 80 percent, whereas control plants increased by a mean of only 22 percent in the same interval. This approach is worth trying in rice.
In some areas, rice crop suffers from floods when it is submerged under water for up to 10 days. Rice cultivars cannot survive such prolonged submergence. A small number of rice cultivars have been identified which survive submergence for 8 to 10 days. Using FR13A, one of the submergence-tolerant donors, improved rice cultivars with submergence tolerance have been developed (MacKill et al., 1993).
Akiyoshi, D.E., Klee, H., Amasino, R., Nestor, E.W. & Gordon, M. 1984. TDNA Agrobacterum tumefaciens encodes an enzyme for cytokinin biosynthesis. Proc. Natl. Acad. Sci. USA, 81: 5994-5998.
Brar, D.S. & Khush, G.S. 1997. Wide hybridization for rice improvements: Alien gene transfer and molecular characterization of introgression. In M.P. Jones, M. Dingkhun, D.E. Johnson & S.O. Fagade, eds. Interspecific hybridization: Progress and prospect, p. 21-29. Bouake, Côte d'Ivoire, WARDA.
Brown, L.R. 1996. Tough choices: facing the challenge of food security. London, UK, Earthscan Publications.
Champoux, M.C., Wang, G., Sarkarung, S., MacKill, D.J., O'Toole, J.C., Huang, N. & McCouch, S.R. 1995. Locating genes associated with root morphology and drought avoidance in rice via linkage to RFLP markers. Theor. Appl. Genet., 90: 969-981.
Chandler, R.F. Jr. 1969. Plant morphology and stand geometry in relation to nitrogen. In J.D. Eastin, F.A. Haskin, C.Y. Sullivan & C.H.M. Van Baul, eds. Physiological aspects of crop yield, p. 265-285. Madison, WI, USA, ASA Publication.
Christou, P., Ford, T.L. & Kofron, M. 1991. Production of transgenic rice (Oryza sativa L) plants from agronomically important indica and japonica varieties via electric discharge particle acceleration of exogenous DNA into immature zygotic embryos. Bio/Tech., 9: 957-962.
Datta, S.K., Peterhaus, A., Datta, K. & Potrykus, I. 1990. Genetically engineered fertile indica rice recovered from protoplasts. Bio/Tech., 8: 736-740.
Datta, S.K., Torrizo, L., Tu, J., Oliva, N. & Datta, K. 1997. Production and molecular evaluation of transgenic rice plants. IRRI Discussion Paper Series No. 21. Manila, Philippines, IRRI.
Evans, L.T. 1993. Raising the ceiling to yield: key role of synergism between agronomy and plant breeding. In K. Muralidharan & E.A. Siddiq, eds. New frontiers in rice research, p. 103-107. Hyderabad, India, DRR.
Gan, S. & Amasino, R.A. 1995. Inhibition of leaf senescence by autoregulated production of cytokinin. Science, 270: 1986-1988.
Garg, A.K., JuKon Kim, Owens, T.G., Ranwala, A.P., Yang Do Choi, Kochian, L.V. & Wu, R.J. 2002. Trehalose accumulation in rice plants confers high tolerance levels to different abiotic stresses. Proc. Natl. Acad. Sci. USA, 10: 1073.
Hayakawa, T., Zhu, Y., Ito, K. & Kimura, Y. 1992. Genetically engineered rice resistant to rice stripe virus, an insect transmitted virus. Proc. Natl. Acad. Sci. USA, 89: 9865-9869.
Hei, Y., Ohta, S., Komari, T. & Kumashiro, T. 1994. Efficient transformation of rice (Oryza Sativa L.) mediated by Agrobacterium and sequence analysis of the boundaries of TDNA. Plant J., 6: 271-282.
Huang, N., Angeles, E.R., Domingo, J., Magpantay, G., Singh, G., Zhang, G., Kumaravadivel, N., Bennett, J. & Khush, G.S. 1997. Pyramiding of bacterial blight resistance genes in rice: marker assisted selection using RFLP and PCR. Theor. Appl. Genet., 95: 313-320.
IRRI. 1989. IRRI towards 2000 and beyond. Manila, Philippines, IRRI.
Jena, K.K. & Khush, G.S. 1990. Introgression of genes from Oryza Officinalis Well ex Watt to cultivated rice, O. Sativa L. Theor. Appl. Genet., 80: 737-745.
Khush, G.S. 1995. Breaking the yield barrier of rice. Geo J., 35: 329-332.
Khush, G.S. 1984. Terminology for rice growing environments. In Terminology of rice growing environments, p. 5-10. Los Baños, Philippines, IRRI.
Ku, M.B.S., Agarie, S., Nomura, M., Fukayama, H., Tsuchida, H., Ono, K., Horose, S., Toki, S., Miyao, M. & Matsuoka, M. 1999. High-level expression of maize phosphoenolpyruvate carboxylase in transgenic rice plants. Nature Biotech., 17: 76-80.
Lawerence, P.L. & Frey, K.J. 1976. Backcross variability for grain yield in species crosses (Avena Sativa L x A. sterilis L.). Euphytica, 24: 77-85.
MacKill, D.J., Amante, M.M., Vergara, B.S. & Sarkarung, S. 1993. Improved semi-dwarf rice lines with tolerance to submergence. Crop Sci., 33: 749-753.
Matsuoka, M., Fukayama, H., Ku, M.B.S. & Miyao, M. 2001. High level expression of C4 photosynthetic genes in transgenic rice. In G.S. Khush, D.S. Brar & B. Hardy, eds. Rice Genetics IV, p. 439-447. Los Baños, Philippines, IRRI.
Multani, D.S., Jena, K.K., Brar, D.S., Delos Reyes, B.G., Angeles, E.R. & Khush, G.S. 1994. Development of monosomic alien addition lines and introgressions of genes from Oryza australiensisDomin to cultivated rice, O. Sativa L. Theor. Appl. Genet., 88: 102-109.
Peng, S., Laza, R.C., Visperas, R.M., Sanico, A.L., Cassman, K.G. & Khush, G.S. 2000. Grain yield of rice cultivars and lines developed in the Philippines since 1966. Crop Sci., 40: 307-314.
Sanchez, A.C., Brar, D.S., Huang, N., Li, Z. & Khush, G.S. 2000. STS marker-assisted selection for three bacterial blight resistance genes in rice. Crop Sci., 40: 792-797.
Singh, S., Sidhu, J.S., Huang, N., Vikal, Y., Li, Z., Brar, D.S., Dhaliwal, H.S. & Khush, G.S. 2001. Pyramiding three bacterial blight resistance genes (xa5, xa13 and Xa21) using marker-assisted selection into indica rice cultivar PR 106. Theor. Appl. Genet., 102: 1011-1015.
Tarczynski, M.C., Jensen, R.G. & Bohnert, H.J. 1992. Expression of a bacterial mtl D gene in transgenic tobacco leads to production and accumulation of Mannitol. Proc. Natl. Acad. Sci. USA, 89: 2600-2604.
Tarczyznski, M.C., Jensen, R.G. & Bohnert, H.J. 1993. Stress protection of transgenic tobacco by production of osmolyte mannitol. Science, 259: 508-510.
Tollenaar, M. 1994. Yield potential of maize: impact of stress tolerance. In K.G. Cassman, ed. Breaking the yield barrier. Proc. of a workshop on rice yield potential in favorable environments, p. 103-109. Manila, Philippines, IRRI.
Tu, J., Zhang, G., Datta, K., Xu, C., He, Y., Zhang, Q., Khush, G.S. & Datta, S.K. 2000. Field performance of transgenic elite commercial hybrid rice expressing Bacillus thuringensis endoprotein. Nature Biotech., 18: 1101-1104.
Xiao, J., Grandillo, S., Ahn, S.N., McCouch, S.R. & Tanksley, S.D. 1996. Genes from wild rice improve yield. Nature, 384: 1223-224.
Xu, D., Xue, Q., McElroy, D., Mawal, Y., Hilder, V.A. & Wu, R. 1996. Constitutive expression of a cowpea trypsin inhibitor gene cpti in transgenic rice plants confers resistance to two major rice insect pests. Mol. Breeding, 2: 167-173.
Yap, C.L. 1991. A comparison of cost of producing rice in selected countries. FAO Economic and Social Development Paper 101. Rome, Italy, FAO.
Zheng, K., Huang, N., Bennett, J. & Khush, G.S. 1995. PCR-based marker-assisted selection in rice breeding. IRRI Discussion Paper Series No. 12. Manila, Philippines, IRRI.