Applications of Molecular Biology and Genomics to Genetic Enhancement of Crop Tolerance to Abiotic Stress - A Discussion Document

SDR/iSC:IAR/02/10 Rev.1

CONSULTATIVE GROUP ON INTERNATIONAL AGRICULTURAL RESEARCH
interim SCIENCE COUNCIL

by Mike Gale

interim Science Council SECRETARIAT
FOOD AND AGRICULTURE ORGANIZATION OF THE UNITED NATIONS
August 2003

Summary

Only some 10% of the world’s 13 billion ha is farmed, although one third of the total land area is considered as potentially suitable for arable agriculture to some degree. Even so, abiotic stress in one form or another, still limits production on most of the world’s 1.4 billion farmed hectares. This is a problem that is not going to go away. For example, yield reductions due to drought stress are already serious, and they will increase. Irrigation will cease to be a practical solution as water becomes scarcer, and the irrigation already in place will continue to lead to yet more soil salinisation. High and low temperatures, acid soils and soils with high levels of metal ions continue to reduce productivity over vast tracts of land and will remain an agricultural challenge for the foreseeable future.

Solutions to the problem will be as diverse as the lands affected. However new, locally adapted and improved varieties will always be a central component in any package of engineering, agricultural management, sociological and political solutions. Moreover in these times of surpluses in developed countries, solutions to the problem of abiotic stress are laid firmly at the door of developing country agriculture. It is here that the most severe stresses are found and here the need for increased food production to feed an increasing population is greatest.

The significance of abiotic stress has not been lost on CGIAR plant breeders. There is considerable work aimed at stress tolerant crop improvement already going ahead. Together with NARS and ARIs, the Centers are working towards an understanding of the genetic and physiological control of tolerance to the key stresses in their regions for their mandate crops, and are beginning to apply the results in breeding programmes. Progress, albeit incremental, is real and demonstrates that the problem is tractable to a genetic approach. In short, breeding is a viable option.

However developments elsewhere tell us that the future will not be the same as the past. Our science is becoming more generic on one hand, and more expensive on the other. The pressure is building for more centralization. The science of abiotic stress resistance in the CGIAR could be the test-bed for a new way of working. Genetics itself is one such area, where a new science is emerging from the masses of DNA sequence and the associated novel and high throughput technologies. This new ‘genomics’ promises more rapid and more spectacular returns, but with expensive equipment, much of which has a short ‘shelf life’. Some of these massively parallel genomics and gene manipulation technologies are already, and with some success, being turned on the abiotic stress tolerance problem in ‘model’ organisms by researchers at ARIs in developed countries. Some Centers are already tooling-up for plant genomics research. Another development is the discovery of ‘synteny’, where genome organization has been found to be much more conserved over species than was previously thought. Application of synteny will allow advances in our knowledge about stress tolerance and the underlying genetics to be transferred between crop species. Synteny will similarly allow CGIAR and NARS scientists to apply the array of genomics resources already available in the models arabidopsis and rice to their mandate crops.

However, in order to harness synteny and the new genomics in a cost-effective and efficient manner it will be necessary to develop new ways of doing science. These could involve: more rationalization and centralization and sharing of expensive and rapidly improving technologies; more outsourcing to providers of standard scientific services and the sharing of skills by assembling multidisciplinary teams and networks in virtual centres that will work on a range of crop species. DNA science and the expensive equipment it begs, is identical for all organisms. Suddenly there is obvious potential for economies of scale in major collaborations.

The time could be right for a full-blooded assault on abiotic stress. Ongoing work shows that the motivation is already there. The question is not whether the work is needed, rather only when, how and what firepower should be brought to bear. A really effective collaboration will involve: the NARS with their germplasm collections, their knowledge of and, access to, stressed agricultural environments, and their plant breeders; ARIs with their experience of technology and model systems; the CGIAR Centers with their comparative advantage with the mandate crops, their collections and their networks to the developing world; and possibly industry as well.

Many of these ideas have already been incorporated in a Global Challenge project, ‘Unlocking genetic diversity in crops for the resource poor’, and, apart from a recommendation to compile lists of potential alternative crops for use in sub-optimal soils and climates, are not dealt with further at length.

Optimal organization of genomics science within the System is relevant and not dealt with elsewhere. It has become clear that the efficient application of genomics and the provision of genomics services within the CGIAR and for NARS partners will require a co-ordinated approach that is not in place today. This paper looks towards a time when basic genomics resources are available for all the mandated crops, and when all CGIAR and NARS researchers have access to sustainably state-of-the-art genomics platform technologies.

The conclusions are that an increasing amount of work will be outsourced to specialist companies, leading ARIs or other Centers. There will likely be strong financial and infrastructural reasons for centralizing other technologies, possibly micro-arrays today and soon the next generation of high throughput genotyping for marker-aided selection and germplasm characterization. There are also scientific reasons for sharing intellectual resources that are in short supply or unevenly spread around the System, such as bioinformaticists and physiologists. There is an opportunity to cost-effectively appoint a central Genomics Facilitator who will carry out market testing, organize key facilities and link groups of researchers around the System so that we can best exploit the new generic aspects of our science. The existing and active CGIAR Task Force on Genomics, with iSC oversight, will provide an ideal forum to discuss these developments.

Introduction

The potential application of molecular biology to genetically enhance crop tolerance to abiotic stress - a discussion document.

Only some 10% of the world’s 13 billion ha is farmed. Apart from urban areas much of the remaining 11.5 billion ha are lands too hostile for any sort of agriculture¹. Moreover almost all the land that is farmable is under conditions sub-optimal, often to a considerable degree, for plant growth. Alongside losses due to pests and diseases, a further 70% of yield potential has been calculated to be lost to unfavourable physiochemical environments, even in developed agricultures².

It is acknowledged that, in order to feed the eight billion mouths we expect by 2030, we will need to double world food production yet again. And we will. One component of that achievement will be the breeding of new varieties of food crops that will both improve yields on land presently being farmed on sub-optimal soils and extend our productive agriculture into lands which are currently barren.

Abiotic stress - extent of the problem

Drought

Unpredictable drought is the single most important factor affecting world food security and the catalyst of the great famines of the past. Moreover, because the world’s water supply is fixed, increasing population pressures will ensure that the effects of successive droughts are more severe³ because competition from industry will increasingly limit the water available for agriculture. Crops are voracious consumers, for example, for paddy rice 5000 l of water is needed to produce 1 kg of grain. At present an unsustainable 70% of the world’s water is used for agriculture. By 2025 it is expected that most Asian countries will join those that already have water shortages. Uncertainties over global warming raise yet further concerns.

Drought stress is a concern for most crops at most Centres for most regions. These include, IITA Cowpea in the Sahel, soybean and tropical maize in the Dry Savanna, ICRISAT Sorghum, pearl millet, chickpea, groundnut and pigeon pea, CIAT Bean in Mexico, C America and NE Brazil, IRRI Rice in Bangladesh, E India, Thailand and Indonesia. CIP, Potatoes in China, India, Southern Africa, Kazakhstan and Afghanistan. CIMMYT, Wheat in C and W Asia and N Africa and maize in sub-Saharan Africa. ICARDA All crops (except faba bean which is only grown under irrigation) in N Africa and Asia. ICRISAT all crops in India and the Sahel.

Salt

Some 380 million ha, almost a third of the area farmed, is affected by salt, and the associated water logging and alkalinity⁴. Sixty million ha are a direct result of over-irrigation, where a raised water table brings underground salt, particularly NaCl, to the surface. It is probable that this agricultural salinisation now degrades as much land as is put under new irrigation each year. Pressures on water use will ensure that the net productive irrigated land will go negative very soon and that secondary salinisation will become critical in Asia, Africa and S America⁵.

Salt stress is of particular significance for rice. IRRI Coastal salinity in Bangladesh, Orissa, Vietnam, Philippines and inland salinity in the Indogangetic plain and Thailand. ICARDA Secondary salinisation is a problem for all crops in C Asia.

Acid and degraded soils

Some 40% of the world’s arable land is associated with acid soils, with pH less than 5, where growth is hindered by high aluminium or manganese content. This is particularly important in S America where some 380 million ha are affected, including almost the whole of the Amazon basin⁶. Other excess metal ion contents reduce the agricultural potential of other soils. For example iron toxicity is a major problem affecting rice production in W Africa.

Acid soils are a widespread problem. IITA Cowpea and soybean in the humid rain forest. CIAT Bean in Africa and both bean and Brachairia in L America. IRRI Rice in Bangladesh, Indonesia and Philippines. CIMMYT maize in L America, SE Asia and Africa. Wheat in CWANA.

Other metal toxicities and deficiencies. IITA Low P for soybean. CIAT Low P for bean and Brachairia. IRRI Zn, P deficiency in Bangladesh, Indonesia and the Philippines and Fe deficiency in Sri Lanka and the Philippines. WARDA Fe deficiency is widespread in Africa. CIP Low P in China, Africa and in the Andes.

Low and high temperatures

Temperature also limits the range and production potential of many of our crops, even at tropical latitudes⁷. Occasional and unpredictable periods of low temperature can be devastating to yields. For example, in the Andes 70% of land devoted to potato production is prone to cold stress⁶.

Cold stress is a rice problem for IRRI in Korea and Nepal. CIP, potatoes in the Andes. ICARDA Low temperature tolerance has become a problem associated with the shift from spring to autumn sowing for barley, lentils and chickpeas.

Excessive heat is a problem for cowpea. IITA in the Sahel. CIP, for potatoes in S Asia.

In fact, abiotic stress tolerance, particularly drought, is the priority target trait for most of the CG Centers dealing with crop plants. In the present economic and agricultural climate, with food surpluses in developed countries, the focus of the private sector will continue to be protection from disease and improvements in aspects of quality. Even given the extent of the problem and although abiotic stress is a significant factor for production in developed countries, it is unlikely that genetic solutions will be actively sought by commercial breeding companies. If the problem is to be tackled at all, abiotic stress tolerance mechanisms and their genetic application in the crops of the developing world will have to be addressed by the public sector working in the developing world.

Gene mapping and marker development for genetic analysis and MAS in breeding

Genetic mapping as a prerequisite to genetic analysis is now part of standard plant breeding. Annex 1 shows clearly that base molecular maps are now available for most of the CGIAR’s crops. Those that are the focus of international effort, e.g. rice, wheat, potatoes, can use the well-developed public maps. Base maps for many of the ‘orphan crops’, in which there is little international trade, have been made at Centers or in Center-ARI collaborations. Only a few very minor mandated species remain unmapped.

The mapping of quantitative traits where there is often little knowledge of the genetic control in advance of the analysis, such as is usually the case with stress tolerance, is usually carried out by ‘QTL mapping’. This requires a scan of the genome, with markers every 10 cM or so to identify those regions where segregation of the trait is associated with segregation for the markers. The reason much denser base maps are needed is that only a subset of the available markers will segregate in any single population. These locations are the basis for establishing a marker aided selection (MAS) breeding programme for tolerance and for eventual map-based cloning of the genes underlying the QTLs. Annex 1 shows that key stresses in several crops are already being addressed in this way.

Breeders’ markers that are closely linked to the target gene may be derived straight from the base molecular map. Today the ideal marker system will be micro satellites, also known as simple sequence repeats (SSRs), although over the next few years single nucleotide polymorphisms (SNPs), which are more amenable to high throughput methods, will take over as the ideal marker.

For rice, which will soon have the benefit of a full genome sequence, markers will never be a problem again. The sequence has been found to contain some 40,000 SSRs⁸ and SNPs and, base pair deletions or insertions indels, are found in unique sequence at a rate of about 1%⁹, which works out at about 24 in every gene! However, apart from in the major cereals, an adequate supply of good quality markers for all applications is still a problem for most CGIAR mandate crops. The status of the genetic maps and markers available for the CGIAR crops is outlined in Annex 1.

Genetic and physiological mechanisms that control stress tolerance

The physiological mechanisms underlying crop responses to stress and potential biochemical, physiological and architectural modifications that will allow crops to escape, avoid or tolerate stress are the subject of a vast literature. Two general approaches are taken in relation to varietal improvement, and both have their place. The ‘empirical’ approach proceeds from genotypic differences associated the best sources of tolerance in the cultivated crop or its wild relatives. Typically sources of tolerance are identified and then the underlying genetic control is investigated by QTL analysis in lines segregating for high and low tolerance. Although the identification of a predominant causal physiological mechanism is helpful, transfer of the improved trait to an already otherwise adapted variety can proceed simply for selecting for and accumulating ‘beneficial’ alleles. The second approach is often described as ‘ideotype’ breeding, in which specific morphologies or physiologies that might be expected to contribute to improved performance under stress are identified in diverse cultivated or wild germplasm and transferred to otherwise adapted varieties. The crossbreeding and, these days, marker-aided pyramiding of the underlying alleles is progressed in the same way in both approaches.

There is already considerable work underway at all CGIAR Centers to improve their mandate crops for stress tolerance. Almost all these breeding projects are being carried out in collaboration with NARS to address the major problems affecting their own agricultures. However many of these projects are crop- or geographical area-specific, even though the target tolerance and the technologies used to address the problem beg a collaborative, pan-stress, pan-crop, pan-Center approach.

For example drought, salt stress and cold temperature stress are all physiologically linked because all three stress environments result in limiting the crops’ physiological access to water. Thus many of the strategies for improved tolerance are likely to be multiply applicable. These will include osmotic adjustment in roots and leaves to retain water, erecting hydrophobic barriers in roots and leaves to retain water, and improving aquaporin efficiency to speed water movement in the plant. Although tolerance mechanisms might be expected to overlap, escape or avoidance mechanisms are more likely to be stress specific. For example reducing time to flowering may escape late season drought but will not help in a chronic saline situation. Deeper roots may be able to reach the last of the water in a drought but would only aggravate salt stress where the salt is being brought to the surface by a rising water table.

With this background one would intuitively expect genetic control to be multigenic and complex, but to overlap somewhat in tolerance to the different stresses. This is exactly the situation found. Consider, for example, wheat and barley where the various reported genetic effects regulating responses to drought, salt and cold have been assembled on one comparative chromosome map, Fig 1¹⁰. While ten or more QTL s are found for each trait, many overlap so that a few chromosomal regions are home to controlling factors for all three traits.

	Fig.1 Abiotic tolerance QTLs and major genes mapped on a composite Triticeae chromosome map. Salt tolerance in orange, cold tolerance in blue, drought tolerance in red.

The empirical approach to aluminium tolerance, where the selection screen is usually for improved growth of roots and shoots in Al supplemented nutrient solution at low pH, also reveals complex control. However, the network of genes is often dominated by one locus which accounts for a major proportion of the genetic variation e.g.^11,12. These results identify MAS breeding priorities and also cry out to be followed by isolation of the key gene, either by cloning or the production of isogenic lines. Gene isolation and knowledge of the gene sequence can be critical steps in a project elucidate an understanding of the underlying mechanism.

A few stress tolerances do usually appear to be under the control of major genes, as revealed by genetic analysis. Submergence tolerance is a prime example, and it may be no coincidence that this trait lends itself to a straightforward and definitive selection screen. This is a key trait in SE Asia where some 25 million ha are prone to flash flooding which can completely submerge the rice crop for several days. Here a single locus, Sub1, has been shown to provide substantial tolerance^13,14.

Mention must be made of the potential of wild relatives as potential donors. Wild species, where the raison d’être is survival rather than yield, are likely to retain useful variation that may have been bred out of the cultivated crop. There are many examples where genes for tolerance have been identified in wild relatives and have been used to transfer useful variation to cultivated crops. In some cases the wild species themselves have been used directly to create new crop species, e.g. Tritipyrum incorporating salt tolerant Thinopyrum bessabaricum, Fig 2¹⁵. Annex 1 shows that CGIAR breeding programmes are accessing this source of variation. The germplasm collections, which are mostly still relatively uncharacterized, will be central to any future stress tolerance initiative.

A possible alternative to varietal improvement is crop replacement. This is probably a viable socio-economic strategy only in extreme stress environments. Nevertheless NARS and their extension services should have reliable information about those crops which generally perform best in stress environments. These differences are exemplified by investigations into the responses of tropical grasses to aluminium stress. Signalgrass (Brachairia decumbens) was found to be far more tolerant than both close relatives and Al-resistant varieties of wheat, triticale and maize¹⁶ Fig 3. It is probable that world-wide multi-centre, multi-crop studies, which are unlikely to be carried out by ‘one-crop specialists’, might reveal interesting alternatives in many stress situations, possibly even pushing back the borders of lands presently considered to marginal for agriculture at all. Overall the genetic dissection of stress tolerance for developing countries is receiving considerable attention, particularly in crops that already have advanced genetic maps. Initial understanding of the physiological and genetic controls most certainly informs breeding programmes. Marker-aided pyramiding of several genes is probably the only way forward for the transfer of improved phenotypes that have been shown clearly to be under control of multiple loci. However these methods have yet to impact stress tolerance breeding programmes. Although a few endogenous genes have been identified which are likely to have major beneficial effects when used as a transgene, these have not yet been applied to practical breeding.

Mapping by NARS and CGIAR Centres has identified a number of genes, usually as anonymous QTLs. Most of these genes have not yet been associated with physiological mechanisms. Linkages between geneticist-breeders and physiologists could now pay dividends. The potential is there for application of these QTLs through MAS but this has not yet been generally successful. There is considerable scope for more collected wild and cultivated germplasm characterisation for stress tolerance. There is also a need to identify any potential alternative crops for severe or chronic stress environments. Any Systemwide initiative to quantify crop yield potential under stress should use common genotypes across stresses and regions.

Progress in breeding - slow but real

Some progress has been made in breeding for drought, salt and aluminium tolerance or avoidance. The CGIAR Center breeding programmes have played a major role in these advances, a small selection of which are listed below. In general these have involved incremental, rather than quantum jump, improvements and have been achieved by empirical selection and not, as yet, by MAS.

• Many releases of drought and acid soils tolerant tropical maize varieties released worldwide, e.g. the acid soils tolerant CORPOICA H-108 and H-111 for Colombia, the Pool 25 and population 28 lines for acid soils used the Brazilian programme, ZM421, 521 and 621 recently released in Southern Africa which are both tolerant of low nitrogen and mid-season drought. (CIMMYT)

• Rice releases listed as salt tolerant for Bangladesh, e.g. PSBRc 84, 86 and 88. PSBRc 88 has good eating quality and is planted even in non-saline areas (1999, IRRI)

• Drought tolerant banana variety, FHIA01, bred and released in Honduras, now released in Tanzania and in trial in 50 other countries (INIBAP)

• Heat tolerant potato variety, Unica, released in Peru (1997, CIP)

• Series of durum wheat and barley varieties that have extended the range of these crops in Syria. Chickpea varieties which have facilitated the switch from spring to autumn sown crops (ICARDA)

• Release of drought tolerant Nerica lines, first in Cote d’Ivoire and now particularly in Guinea (1998, WARDA)

• Release of Mulato, a Brachairia Al tolerant variety for Mexico and C America (2001, CIAT)

Progress has however been hampered by the perception that, in some situations, stress tolerance and high yields are incompatible. The view that higher yields under stress conditions are incompatible with higher yields under good conditions^17,18,19 invokes the need for independent targeted breeding programmes of specialized varieties. In particular it has been argued that, as drought ‘stress intensifies, high yield potential and drought resistance become mutually exclusive’¹⁷. Counter to this is the conclusion that the improved yield of hybrid maize in the US, where there have been steady improvements since the 1930s, is mainly all the result of selection for response to tolerance to stress²⁰ and potential yields have not changed over this period, Fig 4.

Breeding for stress tolerance will proceed more efficiently once it is clear whether, for individual crops and specific stresses, yield potential under stress is controlled by the same genes as yield under optimal conditions. The conclusion will dictate breeding strategy.

Appropriate section screens

Selection screens appropriate to the field conditions that new varieties might experience are often problematic. Field trial climatic factors such as drought and temperature are often unpredictable, while uniform stress conditions are difficult to achieve in trials for edaphic stresses. Also different stresses are often found together, for example salinity problems are rarely all due only to common salt, NaCl. In fact stress the field is rarely due to a single factor. For example multiple metal toxicities and deficiencies are often found simultaneously, and these are farmers’ field conditions that are very difficult to match in the laboratory or glasshouse.

Of course, in the later stages of a breeding programme, empirical selection under field stress conditions is still probably the available approach for most breeders working in developing countries. Usually breeders will simply select those lines that remain the greenest after a period of stress, even though it is well known that plants retain chlorophyll even after all growth has ceased Fig 5²², i.e. that survival is not the same as productivity. The recent developments in thermal imaging and chlorophyll fluorescence imaging^23,24 may provide rapid, economic, non-invasive selection criteria applications over a range of crops and stress programmes.

Good uniform trial sites for stress tolerance selection are not common and where possible should be shared over breeding programmes. The involvement of physiologists with experience of appropriate imaging technologies could benefit a range of CGIAR stress tolerance programmes.

Genomics - the new genetics

Developments over the past decade, arising particularly from the human genome programme, have led to a new phase of plant genetics. ‘Plant genomics’ is the application of the newly available vast amounts of genomic DNA sequence, using a range of novel high-throughput, parallel and other technologies. In plants a ‘whole genome’ DNA sequence is available as yet only for arabidopsis, which was ‘finished’ in 2000. A ‘draft’ raw almost complete sequence of indica rice has been deposited in the public databases by the Beijing group⁹ and a similarly complete sequence of Japonica is available within a private company⁸. The fully annotated public DNA sequence of rice, 88% complete at the moment, will be finished later this year. Undoubtedly more species will follow. Possibly maize will be the next major crop plant to be sequenced, at least for ‘gene-rich’ regions of the genome. Technologies which are included under the umbrella of ‘genomics’ are: automatic DNA sequencing, where one machine can read two million base-pair a day; microarrays and DNA chips where tens of thousands of genes can be scanned for activity levels at the same time; automated genotyping machines that can assay tens of thousands of DNA diagnostic points a day. In fact it will soon be possible to monitor whole genomes for genetic markers or gene expression on single chips. Transformation technologies that allow the facile and efficient genetic modification of almost all crop plants can also be considered genomics technologies.

Genomics is still in its infancy. Genomics technologies, beyond the now conventional molecular biology technologies, are being taken up by CGIAR Centers and by NARS. High throughput capillary DNA sequencing machines and micro-arrayers are in place in the Centers, transformation as a research tool is available for most mandated species (see Annex 1).

The CGIAR Task Force for Genomics met in April 2002 to consider the Systemwide accountability and organisation needed for flexible, efficient, sustainable, cost effective genomics for the mandate crops. Some consensus was achieved and more can follow.

Synteny and comparative genomics

A second development, which has also emerged over the last decade, is the discovery that gene content and gene order is much more conserved over even quite distantly related species that was previously envisaged. This is known as ‘synteny’²⁵.

	Fig 6 Crop circles. The genomes of the three major cereals aligned syntenously so that homoeologous genes lie on radii. Three homoeoallelic series are shown, including the Rht (wheat), D8 & D9 (maize) and SLR (rice) dwarfing genes. M.D. Gale and K.M. Devos, unpublished

The key issue here for the application of genomics tools in the CGIAR crops is that all of the rice resources can be applied directly to the genetic analysis of wheat, maize, barley, pearl millet, finger millet and sorghum. First generation comparative maps have been published for rice and all these genomes.

It turns out that the 240 million years that separate the grasses from the broad leafed plants, the eudicots, has degraded precise map correspondence to the point where the retained synteny does not have predictive utility^8,28,29. However the arabidopsis sequence and the arabidopsis genomic resources are available and applicable to broad-leafed crops, e.g. tomato³⁰. The genome relationships are close within the Crucifereae, which includes arabidopsis and the brassica crops. The Solanaceae and the Crucifereae are more distant from one another at an estimated 150 million years. Nevertheless the arabidopsis gene organization can still be used to aid genetic analysis in tomato³¹. The syntenic relationships between arabidopsis and the majority of the broad-leafed CGIAR crops, particularly the legumes, are not well established (Annex 1).

However other models are emerging which will aid genetic analyses in these crops. The DNA sequence of a legume model, probably Medicago (alfalfa), will be available in the foreseeable future. Also work is progressing rapidly with the tomato genome to bring this species up to model status for all Solanaceous crops.

Breeders and geneticists of cereal crops should become familiar with the relationship of the rice genome organisation and that of their own crop. Breeders of broad-leafed crops where the genome relationships with arabidopsis are not known should have access to the arabidopsis sequence, and should be beginning to establish the syntenic relationships. Novel bioinformatics applications will be are necessary for full application of synteny to crop improvement.

Genomics applications in CGIAR mandate crops in relation to abiotic stress tolerance

In order to employ genomics to address the problems of abiotic stress in mandate crops and to be able to exploit synteny, rather than simply rely on solutions formulated in the models, a basic genomics infrastructure in the crop itself is required. The bare minimum is probably a molecular framework map of the chromosomes, a large DNA insert library, and a facile (and reasonably efficient) transformation system capable of delivering relatively large numbers of engineered plants. The map will have markers every 2 or 3 cM and will have a number of anchor loci, RFLPs or ESTs marked with SSRs or SNPs, that will allow comparisons with the appropriate models. The library will probably be a better than 2-times coverage BAC library with insert sizes in excess of 100 kb.

Additional resources that will be of value include a collection of ESTs (transcribed gene sequences), a comparative map and some ‘knockout populations. The ESTs will often be comparative collections from stressed and non-stressed plant tissues. The comparative maps will align syntenous chromosome regions of the crop with the model. The knockouts will probably be mutation or deletion libraries in which genes have been disabled at random, although T-DNA tagged or transposon-tagged population, as are available or under development in arabidopsis and rice, are also possible when a good transformation system is available.

Extensive public sector genomics resources are available for maize, wheat and rice. Genomics resources for other CGIAR crops, such as pearl millet, sorghum and Musa, are being created in collaborations with ARIs. Other crops such as bean, cassava and Brachairia are being worked up within the CGIAR System and yet others, such as forage legumes, beans and chickpeas, have yet to be started (see Annex 1).

At the very least base genetic maps should be available for all CGIAR crops (a microsatellite-based map would today be the work of a 3-year post-graduate student). This will open up all species to MAS breeding. Further resources may easily be added at a later date. Links with ARIs or direct outsourcing should be explored for EST and BAC library production.

Gene cloning

There are many ways to clone genes, however if all that is available is a map location of a stress tolerance gene and there is no indication of its precise function, then the process will probably be either to identify possible candidate genes by their map position or to find genes whose expression is associated with the trait. Certainly in order to find out unequivocally what any gene actually does it will generally be necessary to clone it first.

Map-based cloning

A novel method of map-based cloning, known as ’deletion tiling’, involving the generation of a number of deletions that include the target gene in the crop, and using the minimum overlap to identify candidates in the model, is being pioneered in wheat, Fig 8³². This method has the advantage that variation is not required for either the target gene or the flanking DNA regions, and it could find many applications in CGIAR crops.

Microarrays

Candidate genes will also emerge from microarray analyses. Genes that are induced by stress are ideal for comparative microarray analysis. A typical experiment will be to challenge an array of ESTs with RNA extracted from stressed and unstressed tissues. A comparison of the two will identify genes that are up-regulated in the stressed tissues and those which are down-regulated or switched off, Figs 9 and 10. Whether these represent genetic cause or effect is another challenge. Ideally one would like these experiments to scan entire genomes because, until all genes are available, any microarray experiment will always be incomplete. This will be possible in the near future for the 25,000 arabidopsis genes and the 50,000 rice genes, but not for other crops for some time. Mini-array’s can however be built from collections of ESTs assembled from random cDNA libraries, or from more targeted collections made from cDNAs collected from stressed tissues. Even more targeted will be the special ‘stress arrays’ made up of all the expressed genes for which there is any evidence of implication. Stress arrays are being contemplated in several CGIAR Centers. Other technologies such as cDNA-AFLP and differential display can also identify critical gene sequences.

A major NSF grant³³ has recently been completed which has investigated the use of arrays to investigate salt tolerance. This project has made a commendable start to cataloguing stress inducible genes in halotolerant and salt-sensitive plants. The results are generally relatively complex For example, in rice 10% of the genes were significantly up- or down-regulated after 1 h of salt stress³⁴. An added complication is that almost half of the genes available as ESTs or as hypothetical genes in genomic DNA sequence have, as yet, completely unknown functions.

Fig 9 Micro-arrays can carry 20,000 genes on a 2 cm2 plate. When arrays are probed with RNAs from, say, stressed and unstressed plants computer enhanced imaging identifies genes that are under-expressed (in red) and over-expressed (in green).

Comparisons of the DNA sequence or the hypothetical protein product sequence with other isolated genes of known function can provide a lot of information. However while some half of the genes revealed by whole plant genome sequencing remain of unknown function, it will often be necessary to attempt to elucidate function by various reverse and forward genetics methods including transformation and knockout analysis, in both the crop and the model. The definitive experiment will usually involve the production of transgenics with an expected phenotype.

	Fig 10. Microarray data. Rice response 3 h drought using 6,400 ESTs. Research by Shinji Kawasaki. Figure kindly provided by Hans Bohnert, University of Illinois.

Knockout populations

Various T-DNA or transposon tagged populations are available in rice and arabidopsis. These reverse genetics ‘gene machines’ allow the identification of lines in which any gene of interest is disabled. These lines can then be investigated to identify a phenotype that may give clues as to the genes function. A recent development, TILLING (from ‘targeted induced lesions in genomes’)^35,36 allows production of targeted knockouts and also the creation of allelic series in any gene. TILLING populations are available for arabidopsis and are under investigation for rice.

Chemical or irradiation mutant populations in the crop itself will allow a forward genetics approach. Stress tolerant lines can be identified by the simple expedient of subjecting populations to drought, high salt, cold temperatures, submergence etc. and selecting vigorous survivors. The challenge then is to link the phenotype with a deleted gene for which the sequences of candidates will be a good starting point.

With the notable exception of rice, CGIAR germplasm curators have not entered the field of knock-out populations in mandate crops. CGIAR Centers and NARS have significant comparative advantage in having the facilities to grow and maintain large populations, and they have the expertise in the crop to recognize and screen for key knockout phenotypes. These populations will be central to functional genomics efforts and, especially for the minor CGIAR crops, their availability should encourage collaborations with ARI researchers working in model species. The opportunity costs of not taking an international lead of this sort should at least be evaluated for all of the mandate crops.

Transformation in the crop and the model

Genetic transformation is now possible for most crop species. However ‘possible’ is not the same as ‘efficient’. CGIAR Centers need to be able to produce at least tens, and ideally hundreds, of low copy insert transgenics for any construct (the need for many lines is demonstrated by the range of phenotypes produced in any single transformation experiment, see Fig 14). Good systems exist for many broad-leafed crops and for most cereals. Legumes are probably the most recalcitrant group of crop plants. The status of CGIAR Center transformation capability is shown in Annex 1. Transformation is often seen only as means of making transgenic crops. Indeed, the ideological debate surrounding transgenics notwithstanding, it is inconceivable that we will penetrate far into the 21^st Century and its looming food shortages without needing to use all the technology that we have available. However for the time being, transformation has another use as the ultimate test of function of any candidate gene, either in the model or the crop.

Initial attempts will usually employ constructs of the beneficial allele of the gene linked to a constitutive promoter, such as CMV35S, in over-expression experiments. Early transgenic trials will also usually involve antisense constructs that will provide information by negating the effects of the gene. Later trials in the crop itself will probably employ specific alleles of the gene in constructs with promoters that target the effects to specific tissues, such as roots or developing seeds, at particular developmental stages. Transgenics in the crop itself will likely be in an already otherwise adapted genetic background, and these may serve as breeders’ lines for eventual introduction into the main stream breeding programmes.

Leads from model species

Research into the molecular basis of abiotic stress tolerance is being carried out mainly in model species, particularly arabidopsis. Although this area of our science is still in its infancy there are some 200 references and claims in the reputable scientific press. Genetic transformation experiments to improve stress tolerance are beginning to yield some promising results.

A particularly encouraging approach is the use of transcription factors, regulatory elements that control batches of genes, including those which are induced by stress. One such is CBF1 in arabidopsis, which is the likely regulator of the cold acclimation response. Over-expression of CBF1 enhances the levels of a swathe of cold-regulated genes to mimic the effect of cold acclimation that provides subsequent resistance to freezing, and provides protection against cold temperature damage, Fig 11³⁷. Transcription factors act as ‘master switches’ and provide one means of rationalizing and exploiting the information obtained from gene expression microarray analyses.

DREB1A is another transcription factor that regulates expression of a further range of stress tolerance genes. Over-expression of DREB1A, again in arabidopsis, activates expression of a range of genes and results in improved drought, salt and freezing tolerance³⁸.

Fig 11 Effect of CBF1 over-expression in arabidopsis, Left: Non- acclimated controls after freezing for 5 days; middle: Non-acclimated transgenics after freezing, right: Acclimated controls after freezing. Reprinted with permission from Science 280, p 105, fig 3 “Freezing survival of RLD and A6 Arabidopsis plants”, Jaglo-Ottosen et al. Copyright 1998 American Association for the Advancement of Science

Interestingly it was noted that, when DREB1A was driven by CaMV35S, a strong constitutive promoter, normal growth of the plants in an unstressed environment was severely retarded. However the simple expedient of driving DREB1A with a stress inducible promoter reduced adverse side effects and further improved tolerance. Negative pleiotropic effects on fruit yield were also seen in tomato with CaMV35S driven yeast HAL1 gene, which enhances K⁺/Na⁺ selectivity and maintenance of water status³⁹.

Specific ‘ideotype’ approaches have also been tried. For example, it has been argued that plants should be able to exploit ions to achieve osmotic adjustment and internally distribute these ions to keep sodium away from the sites of metabolism. To achieve just this a vacuolar Na⁺/H⁺ antiport, AtNHX1, was over-expressed to provide protection up to about half seawater salt levels, Fig 12⁴⁰. Similar effects have been demonstrated with AtNHX1 over-expression in oil seed rape, Brassica napus⁴¹.

There are many opportunities using the transgenic approach, including, eventually, to produce lines that can be entered into mainstream breeding programmes. Novel genes can be expressed with increasing precision, as more tissue and developmental time specific promoters become available. Endogenous genes can be over-expressed, or negated by the use of antisense constructs.

Interestingly, synteny can also be exploited to good effect. Once the molecular basis of beneficial alleles in any species, has been discovered, including in models like arabidopsis, it is possible to engineer the equivalent homoeologous genes in the target crop with the same alterations in DNA sequence. An excellent example of this approach is the recent targeted engineering of rice to produce a GA-insensitive dwarf phenotype. In 1997 the GAI gene, which produced a gibberellin insensitive dwarf phenotype was isolated from arabidopsis⁴². Soon it was it was possible to demonstrate that rice homoeologues (found in rice ESTs) of the arabidopsis gene mapped to locations in cereal genomes that coincided with the location of the ‘Green Revolution’ wheat semi-dwarfing genes. Moreover the allelic difference between tall and dwarf phenotypes of both rice and wheat were based on the same 51 base-pair (17 amino acid) deletion in homoeologous genes⁴³. Just last year, a Japanese group⁴⁴ added the last step when they were able to demonstrate that they could engineer the equivalent rice gene (SLR), for which no dwarf mutant had ever been found, with the same 51 bp deletion and produce GA-insensitive dwarf transgenic plants (Fig 13), and thereby avail rice of a completely new and potentially very valuable form of short straw. The same paradigm can, and undoubtedly will, be used to transfer alleles between quite distantly related plants and will no doubt be of value for transferring stress tolerance genes between crops and between wild species and crops.

	Fig 14. The same pSLRtr construct produces a range of transgenic ‘alleles’. From Ikeda et al (2001) Plant Cell 13:999

There appears to a tremendous potential in these results on stress genetics, either in the direct application of model plant gene constructs in crops or in the modification of endogenous crop genes to emulate the effective model plant alleles. The arabidopsis DREB genes have already been transferred to wheat with most encouraging effects on salt tolerance⁴⁵. Most CGIAR Centers have effective transformation systems for their own mandate crops and could enter into collaborations to obtain the necessary genes, at least for research purposes.

Nevertheless we should not forget Abraham Blum’s warning that ‘...any claim for a genetic modification of stress resistance that is presumed to impact crop performance in agriculture will remain on paper unless proven .... under field conditions’⁴⁵

Germplasm collections

The collections held in trust by the CGIAR Centers and by many NARS have major roles to play in both candidate gene discovery and, once genes have been identified as being important in the control of a trait such as stress resistance, identifying the range of alleles available at that locus. It is very clear that these collections will increase in importance. Everything that can be done should be done to provide added value over the next few years.

Allele mining

‘Allele mining’ will involve PCR-extraction and sequencing of the different versions of genes found in varieties, land races and wild relatives. Variation in gene sequence may then be correlated with the stress tolerance performance of the accession, and may well identify the best alleles for future transgenic experiments.

Association genetics

‘Association genetics’ is a collection-related development that the CGIAR Centers are particularly well placed to exploit. This new area of science derives from human genetics where analysis of large segregating populations is not possible. Genes associated with any trait are identified by correlation of phenotype with specific alleles at linked markers. In plants this involves scanning collection accessions for variation at marker loci dispersed over the genome (genotyping) and then correlating, for example, performance under stress of the genotypes with allele dis-equilibrium around the genome. This a young science in plant biology, however the potential is great for discovering novel genes of adaptive significance and for providing added value to the collections. It will not go unnoticed that, once the collection has been ‘genotyped’, the same data is applicable to any trait of interest.

CGIAR Centers are advantageously positioned to develop plant association genetics. The biodiversity discussion has increased public awareness of the value of ex situ collections and any genotyping and further characterization will increase their value still further. Again the technology and the analyses will be generic so there is much to be gained by close Systemwide collaboration

Conclusions

Abiotic stress is a major constraint to food production and one that will grow in significance as we approach the increasing world food shortages in the developing world that will characterize the first half of the 21^st Century. Aid and technology may be available from the North but the problem is one for the developing world alone. New crop varieties that will produce more in increasingly marginal agricultural environments will be desperately needed. These varieties will be bred only by the CGIAR System and their national agricultural programme partners.

Considerable collaborative work, usually with NARS and often with ARIs, is already underway for most mandate crops from most Centers. Levels of expertise and motivation are high. However the various projects are crop and region specific and are being carried out in relative isolation.

The new science of plant molecular biology is beginning to impact all of our work, including stress tolerance research and breeding. Comparative genomics in particular promises new opportunities and is developing fast. Researchers in ARIs are already making discoveries that will impact stress tolerance breeding, and they are highly motivated to work with developing world problems. CGIAR scientists have been quick to appreciate this and are collaborating with key academics and are beginning to import and install the technology. However the notion that any single Center can keep at the cutting edge of this fast moving and expensive field must be unsustainable. The science is generic and the need for centralization, rationalization and outsourcing will become obvious to everybody soon. The centralization, at least to virtual centres, could extend to personnel with key skills as well as technologies.

The sooner we start the better prepared we will be for the future. A pan-crop, pan-Center global collaboration involving NARS, key ARIs and even industry, to approach the problem of abiotic stress tolerance, building on what has been achieved already, could be the vehicle around which to begin to build this new way of doing science. The notes below will provide a basis upon which to open the discussion.

Options for the way forward

In the CGIAR there is already a wide understanding of the importance of stress tolerance. There is tremendous motivation, and considerable ongoing work, to breed new varieties of CGIAR crops with improved tolerance to stresses. Most Centers have incorporated molecular biology into their science and their breeding. Genomics technologies are beginning to be absorbed by the leading Centers. Links are already being made with the ARI scientists working in the area of abiotic stress tolerance. Technology transfer is continuing through the established CGIAR Center-NARS networks.

Even since the first draft of this discussion document in April 2002, progress has been made. The ‘fast-tracked’ Global Challenge project ‘Unlocking genetic diversity in crops for the resource poor’, already submitted by CIMMYT, IRRI and IPGRI, has incorporated the potential of molecular genetics and genomics to mine the CGIAR germplasm resources for novel genes and alleles that can be employed to improve the characteristics of varieties in developing countries, particularly for abiotic stress resistance. In fact an approach to drought tolerance is included as an example for ‘proof of concept’. So with regard to abiotic stress comments will be restricted to a list of goals for a successful co-ordinated approach and a few suggestions for further work that have not been incorporated into the GCP. More attention will be devoted to ways forward to build a pan-System corporate knowledge of genomics technologies and to provide a genomics infrastructure that will make the CGIAR System the preferred partner for national programmes and provide the System itself with state-of-the - art technology for the foreseeable future.

Abiotic stress research in the CGIAR Centers

The arguments developed above indicate that there is a need for a highly co-ordinated pan-crop approach that seeks out and exploits the comparative advantages of all the partners. The approach will also exploit the generic aspects the new genetics and comparative genomics.

The overriding goal must be to produce varieties that will extend the range of arable agriculture and yield more under stressed conditions to provide improved food security for the poor of the world.

Sub-goals will include:

• Implementation of an initiative which will underpin stress breeding programmes by providing good science and the best tools to address different problems, without duplication, including:

  - novel genes, and improved versions of genes, to address abiotic stress tolerance

  - improved breeding tools for more efficient incorporation of these genes in new varieties, e.g. better selection screens, better molecular markers, more efficient transformation protocols

  - information on the most appropriate breeding strategies, e.g. whether, for specific crops and specific stresses, breeding can be integral to the core programme, should comprise a separate specialized programme or would be better outsourced to participatory programmes

  - improved knowledge of the physiology and biochemistry underlying stress tolerance.

• Provision of a framework to ensure a continual flow of information between keys ARIs, Centers and NARS. Also the provisions of a forum where ARI scientists can be exposed to the problems encountered in developing countries, and at the same time allow Centers and NARS access to ARI stress science and generic technology.

• Exploitation of the generic experience in molecular genetics and genomics technology, and exploitation of the new opportunities arising from the discovery of synteny between crop genomes, and between crops and models.

• Bringing key skills together to link with breeders to address the problems, e.g. molecular physiology, bioinformatics, genomics technology, IP management etc., particularly where all these skills are not present in a single Center

A successful initiative will exploit the comparative advantages of all potential partners:

• NARS - national breeding programmes will have access to the most relevant stress trial sites, increasingly NARS will have molecular biology and genomics expertise, together with skills which are becoming uncommon in the CGIAR, such as physiology and biochemistry. NARS will know local market drivers and will be able to interact will local relevant industry, such as plant breeding and seed companies.

• ARIs - Advanced Research organizations will provide access to state-of-the - art genomics, and, notably, a few genomics laboratories specializing in abiotic stress. ARI’s will provide access to arabidopsis genomics, including the first whole genome arrays, bioinformatics, and biochemical pathway research.

• CGIAR Centers have a specialized knowledge of the mandate crops, including genetic transformation, availability of field and glasshouse space for trials, and a growing bioinformatics, molecular genetics and genomics capacity. CGIAR Centers hold the key germplasm collections. The Centers have special relationships with NARS and networks in place to transfer technologies

• Industry - The multi-national agbiotech industry does undertake some fundamental and strategic and controls some results relevant to Centers’ needs. The first whole genome rice arrays are likely to be available from industry. Smaller specialized, often local, companies can provide market-tested genomics service providers, e.g. BAC libraries, DNA sequencing, which will provide benchmarks for Centers planning their in-house/outsourcing research strategy.

The goals and objectives above are embodied in the ‘Unlocking genetic diversity in crops for the resource poor’ Global Challenge programme. One outstanding recommendation concerning crop replacement should be considered. Although crop improvement to tolerate local will probably be preferred it will be very valuable for agronomists to have available lists of alternative crops that are inherently more tolerant of particular sub-optimal soil types or particular adverse climatic conditions. The compilation of such information for global application would be very valuable indeed.

Genomics and genomics resources in the CGIAR

It is very clear that genomics will play an increasingly important role in CGIAR science and crop improvement. It is equally clear that genomics platform technologies are the most expensive, and probably the most rapidly advancing, that the System has ever had to accommodate. Therefore an early single pan-System policy for the acquisition and deployment of these technologies is imperative.

The goal should be to provide Centers and NARS partner's access to the appropriate genomics resources for all the mandated crops and sustainable access to state-of-the-art platform technologies and plant genomics capacity.

An initiative to achieve this would have a number of sub-goals:

• Should be cost-effective and achieve economies of scale while still being flexible and responsive to new developments.

• Providing the CGIAR Centers and NARS access to the rapidly changing state-of-the-art genomics technology.

• Allowing smaller Centers working with wider portfolios of marginal crops to learn from the experiences of the Centers concentrating on the major staples.

• Providing a framework to ensure a continual flow of information between keys ARIs, Centers, NARS and commercial technology suppliers.

• Providing Centers and NARS access to model plant, e.g. rice, Arabidopsis, Medicago, tomato etc., genomics resources.

• Developing ways of outsourcing between Centers

• Providing access to common negotiated sources for standard within-crop genomics services.

• Achieve a minimum basic ‘in house’ genomics infrastructure at the Centers for each crop, e.g. genetic maps, comparative maps with models, BAC libraries, EST collections, efficient transformation methods, genotyped QTL mapping populations (possibly also ‘knock-out’ libraries), to be shared across the System and with NARS and as vehicles for collaborations with ARIs.

The policy should enable the build-up of pan-System corporate knowledge and capacity of technologies, including:

• ESTs, cDNAs and BAC library production

• marker development - SNPs and, possibly still, SSRs

• high throughput (HTP) genotyping - both for SSRs and SNPs

• association genetics as applied to in-house germplasm collections

• map and comparative map construction and application in mandated crops

• fully genotyped segregating populations for QTL applications

• comparative genomics and bioinformatics, particularly between mandated crops and models

• insertion/mutation populations, TILLING

• handling and storing genomics resources, using laboratory information management systems

• microarrays - both built in-house and ‘bought in’ Affymetrix-type arrays

• proteomics

• high throughput DNA sequencing

• transformation protocols

All of the above technologies are, or will soon, be required by all Centers. The issue is whether any particular technology is best centralized within Centers, centralized somewhere within the System or outsourced altogether. Centralization and outsourcing between Centers will certainly become necessary as technology, and associated costs, evolve to soon become beyond the scope of any one Center. Costs are, of course, not the only determining factor. Convenience, service level, relationships between customers (particularly between Centers and NARS) and training considerations will all play their part.

Precisely these issues were considered two years ago and reported in the TAC ‘Systemwide review of plant breeding methodologies in the CGIAR’⁴⁶. The review emphasized that outsourcing and, especially, outsourcing between Centers should become common for some technologies. The report also noted that centralization should be considered for technologies that had ‘broad utility for all Centers’, were ‘so expensive that individual Centers cannot afford it’, and where’ information transfer was synergistic’. In the intervening two years at least two of the CGIAR’s genomics technologies have moved into this group.

Below most of the technologies are briefly considered from this point of view.

ESTs, cDNAs and BAC library production should probably all be contracted out, and there are excellent suppliers out there. These are resources that will be revisited over and over, for which quality is paramount. Quality large insert BAC libraries in particular are very reliant on experience and the availability of the appropriate colony picking robots, which will not be present in any Center.

Molecular marker development. This is still a critical activity for many crops (see Annex 1). SSRs are generally identified in various enriched libraries or directly from EST sequences. Libraries will probably be made in collaboration with expert ARI labs and the sequencing contracted out directly. SNP detection in the non-staples (where e-detection is not possible) will again probably be by sequencing PCR copies of cDNA sequences from different varieties. A job for outsourcing.

Of course the marker only acquire real value once they have been located in a map framework. This mapping will probably be carried out within Centers (see genotyping below)

Genotyping - High throughput genotyping should by now be part of most CGIAR breeding programmes. Using SSRs, although there is trend among industrial groups for global centralization, I would still recommend development of relatively HTP systems within Centers. These should extend to liquid handling robots for mass PCR and automatic reading of fluorescent labeled products. Proximity to the users - breeders, germplasm curators and geneticists - and the availability of a local system for training purposes argue for in-Center systems. These facilities could also be a focus for specific crop NARS breeders’ use and training.

All indications are that MAS and germplasm collection characterization will soon move to HTP SNP genotyping. At present the favoured SNP format has not yet emerged, however when it does it will probably be beyond the financial reach of individual Centers. There are other factors mitigating for centralization of such a facility. These include large economies of scale and value in having collections of primers at a single site. Also Centers should be incentivised to use a centralized site which represents even a small part of their budget.

A watching brief should be kept on developments, particularly at international breeding companies with interests in multiple crops.

Map construction and development of genotyped QTL mapping populations will be carried out within Centers, although the development of crop-model comparative maps will likely be carried out in collaboration with ARIs with expertise in the models.

Mutation and deletion populations, other forms of ‘knock-out’ lines, possibly ‘targeted induced lesions in genomes’ (TILLING) populations have not, other than in rice at IRRI, been considered at the Centers. Such genetic stocks play a key role in functional genetics, i.e. assigning function to anonymous gene sequences. The Centers, with their specialized knowledge of the crops and, generally, the space and the manpower to grow large populations under good agronomic conditions, have a comparative advantage in the production of such stocks for the mandated crops. The leveraging power of such resources in the promotion of interactions with ARIs and even companies is obvious. The Centers should carefully consider the opportunity costs of not producing such resources.

Microarrays are of two main types. The first are the high quality, high-density GeneChips produced commercially by a commercial company, Affymetrix, using short 16-32-mer gene-specific oligonucleotides built up on the chip. These include the recently produced 24,000 gene (400,000 spot) whole arabidopsis genome arrays (and, in the near future, barley and rice whole genome arrays, and, in the foreseeable future, wheat, maize etc). The only option at the moment is to buy such chips in (about $700 each or $4,800 including sample processing for a minimal experiment) and process them on an Affymetrix Genechip system costing around $150,000. The second type usually use ‘spotted’ cDNAs or larger 50-70-mer gene specific oligos and can be made within academic labs. Modern arrayers can work up to 20,000 spots, and cost around $50-100,000. This sort of facility will usually be associated with significant liquid handling capacity to enable the large numbers of PCR reactions necessary. Also, significant -80°C freezer space will be needed to accommodate the growing amplified cDNA resource. Finally, in a perfect world, one will validate all amplification products by resequencing, so a HTP sequencer may also be required.

It is generally acknowledged that the value of special arrays, such as the rice stress arrays being developed at IRRI, is very dependent on their quality, and this means dedicated expert technical staff associated with a facility. Chip production could be outsourced or centralized within the System, with obvious advantages and disadvantages, but clearly the development of multiple facilities around the Centers is not the best option. Among the advantages the development of quality controlled libraries of cDNAs all in the public sector or for which IP issues are known to have been centrally negotiated. Since it will, for quality control purposes, probably also be necessary to run the hybridizations at the same site, gene expression databases will be developed which are available to all users. A Systemwide facility will be useful for training of other Center and NARS staff. Return on capital outlay is also a factor with an in-house facility. The thinking is that a top-flight arrayer purchased today would remain ‘current’ for three to four years. Bioinformatics support, both for the direct analysis of results and comparative analyses of the rapidly growing array result databases, is vital and must be factored in to an in-System facility.

Proteomics. Protein analysis, such as performed ‘time-of-flight’ mass spectrometers is currently outsourced. The cost of the equipment and the rapidly changing state of the art will probably ensure that outsourcing is favoured for the time being.

DNA sequencing. For relatively large-scale sequencing, e.g. several BACs, ESTs in the 1,000s, outsourcing will be the preferred route. This is a very competitive commercial market. Local sequencing capacity within Centers may still be justified for small jobs. Nevertheless, other issues, like the present Indian policy of not allowing DNA to leave the country, may also convince Centers to retain some sequencing capacity.

Laboratory management systems. It will soon become clear that laboratory information management systems (LIMS) are vital for tracking samples and maintain quality control in the laboratory. At present there are none in use around the System (although CIAT are exploring options). The eventual benefits will be large if all Centers were to use the same, or compatible, systems.

Genetic transformation is most definitely required in-house and for all crops, although development of the technology may be best carried out in collaboration with ARIs.

The way forward. For cost effective pan-System provision of state-of-the-art genomics infrastructure and technologies iSC might consider the appointment of an independent ‘CGIAR Genomics Facilitator’. The initial JD might include:

• a constantly updated trans-national review of outsourced providers and costs for DNA sequencing, DNA library production, proteomics analyses, micro-array facilities and high-throughput genotyping, i.e. constant market testing

• a constant review of outsourcing possibilities between Centers

• act as a clearing house for CGIAR and NARS genomics related queries

• a review of LIMS available, with a view towards harmonization across the System.

• undertake a special study of the advantages and costs associated with CGIAR centralized micro-array and HTP SNP genotyping services. and, if such facilities move ahead

• the collection of international genomics resources under appropriate MTAs for use with all CGIAR Centers and their stakeholders

• interaction with local managers to establish service level agreements and appropriate financial structures

• commission the development of a web-based tracking system whereby CGIAR and NARS customers can follow the progress of their samples in real time and automatically receive results. This is a key component of any effective and competitive service.

I believe that any Systemwide genomics service should also be overseen by an International Stakeholder Steering Group. This group will include technology experts (probably managers of service laboratories in developed countries), CGIAR representatives (probably at the DG or DDG level) and NARS representatives. The role of the group will be ensure that the service(s) are state-of-the-art, competitive and appropriate for the major CGIAR and NARS customers.

Line management and financial structures will of course need considerable discussion. Operational models like that of Central Advisory Service (CAS) for intellectual property matters at ISNAR should be explored. Also it is possible that the independent facilitator could be closely aligned with, or even be part of, the ISNAR Biotechnology Service (IBS). Plainly a Genomics Facilitator will regularly report and interact on activities through the existing CGIAR Genomics Task Force and the post could report to the System through the chair of that group.

Further activities. Yet another role of a Genomics Facilitator might also assemble and work with multidisciplinary teams, again in actual or virtual centres, across and over crops to address specific issues. The CGIAR Genomics Task-force is an excellent vehicle through which such groupings could meet:

• Crop type groups. Cereals, legumes and roots and tuber groups have already been initiated. Certainly the development of stress arrays and anchor markers over related genomes will be crop group activities. These groups will have also specific comparative bioinformatics needs

• Bioinformatics. CGIAR over crops bioinformaticists are already linked through Systemwide projects to ARIs in the US and the EU and have skills appropriate to all crops.

International meetings, such as Plant and Animal Genome that is held at San Diego every January, provide excellent opportunities for CGIAR scientists to interact with international academics in:

• Crop groups. These already exist and many individual crops are already the subject of international meetings and annual meetings at PAG. A global Musa genomics consortium has also recently been formed⁴⁷

Annual meetings organized by CAS at ISNAR or elsewhere for:

• Intellectual property managers. Already in place with CAS at the hub, common systems and corporate knowledge, particularly in dealings with industry, are vital. Similarly common IP arrangements should be anticipated across the System for collaborative grants with ARIs, particularly IP for humanitarian use. The Systemwide IP group should also be the preferred partner for NARS.

So finally, rapidly moving research cusps, increasingly expensive technologies, more obvious links between traits and over crops, and increasing technological capacity in NARS are all indicative of more rationalization, more centralization, more outsourcing, and more virtual groupings over institutions. The time when we adopt new ways of working cannot be put off much longer.

An appropriate first step forward would be to convene a meeting of the key CGIAR stakeholders to formulate ways in which the new science can be brought to bear in the most efficient manner to deliver the new crops that developing country agricultures need. Such a workshop would be organized by the CGIAR Task Force on Genomics.

Reference List

1. FAO statistics, http://apps.fao.org/(2002)

2. Boyer, J.S. Science 218, 443-448 (1982).

3. McWilliam, J.R. Aus.J.Plant Phys. 13, 1-13 (1986).

4. Flowers, T.J., Hagibagheri, M.A. & Clipson, N.J.W. Q. Rev. Biol. 61, 313-337 (1986).

5. Flowers, T.J. & Yeo, A.R. Aus.J.Plant Phys. 22, 875-884 (1995).

6. Jaffé, W. & Rojas, M. Biotechnology and Development Monitor 18, 6-7 (1994).

7. Thomashow, M.F. Adv. Genet. 28, 99-131 (1990).

8. Goff, S.A. & and 30 other authors. Science 296, 92-100 (2002)

9. Yu, J. & and 96 other authors. Science 296, 79-92 (2002)

10. Cattivelli, L., Baldi, P., Crosatti, C., et al. Plant Mol.Biol. 48, 649-665 (2002)

11. Nguyen, V.T., Burow, M.D., Nguyen, H.T., Le, B.T. & Paterson, A.H. Theor.Appl.Genet. 102, 1002-1010 (2001).

12. Luo, M.C. & Dvorak, J. Euphytica 91, 31-35 (1996).

13. Nandi, S., Subudhi, P.K., Senadhira, D., Manigbas, S.-M. & Huang, N. Mol.& Gen.Genet. 255, 1-8 (1997).

14. Xu, K., Xu, X., Ronald, P.C. & Mackill, D.J. Mol.& Gen.Genet. 263, 681-689 (2000).

15. Forster, B.P., Gorham, J. & Miller, T.E. Plant Breed. 98, 1-8 (1987).

16. Wenzl, P., Patino, G.M., Chaves, A.L., Mayer, J., Rao, I. & Madhusudana. Plant Physiol. 125, 1473-1484 (2001).

17. Blum, A. in Increasing yield potential in wheat: Breaking the barriers (eds Reynolds, M.P., Rajaram, S. & McNab, A.) 90-100, CIMMYT, Mexico, D.F. (1996).

18. Ceccarelli, S. & Grando, S. Euphytica 57, 157-167 (1991).

19. Cox, T.S., Shroyer, J.P., Liu, B.-H., Sears, R.G. & Martin, T.J. Crop Sci. 28, 756-760 (1988).

20. Duvick, D. in Developing drought and low N-tolerant maize (eds Edmeades, G.O., Banziger, B., Mickelson, H.R. & Pena-Valdiva, C.B.) 332-335, CIMMYT, El Batan, (1997).

21. Ortiz, R., Ekanayake, I.J., Mahalakshmi, V., et al. 8th JIRCAS Sympoisium, 2Water for sustainable agriculture in developing regions”, Tsukuba, Japan, 27-28 November 2001 (2002).(in press)

22. Boyer, J.S. Plant Physiol. 46, 233-235 (1970).

23. Lawson, T., Oxborough, K., Morison, J.I.L. & Baker, N.R. Plant Physiol. 128, 52-62 (2002).

24. Reddy, A.R. & Strasser, R.J. in Molecular approaches for the genetic improvement of cereals for stable production in water-limited environment. A strategy planning workshop held at CIMMYT, El Batan, Mexico 21-25 June 1999 (eds Ribaut, J.-M. & Poland, D.) 90-91 International Maize and Wheat Improvement Centre, El Batan, (2000).

25. Kellogg, E.A. Proc.Nat.Acad.Sci. 95, 2005-2010 (1998).

26. Gale, M.D. & Devos, K.M. Science 4817, 1-7 (1998).

27. Dubcovsky, J., Ramakrrishna, W., SanMiguel, P.J., et al. Plant Physiol. 125, 1342-1353 (2001).

28. Devos, K.M., Beales, J.E., Nagamura, Y. & Sasaki, T. Genome Res. 9, 825-829 (1999).

29. Mayer, K. & 17 other authors. Genome Res. 11, 1167-1174 (2001).

30. Ku, H.-M., Vision, T., Liu, J. & Tanksley, S.D. Proc.Nat.Acad.Sci. 97, 9121-9126 (2000).

31. Ku, H.-M., Liu, J., Doganlar, S. & Tanksley, S.D. Genome 44, 470-475 (2000).

32. Roberts, M.A., Reader, S.M., Dalgliesh, C., et al. Genetics 153, 1909-1918 (1999).

33. Bohnert, H.J. & et.al. Plant Physiol. 39, 295-311 (2001).

34. Kawasaki, S., Borchert, C., Deyholos, M., et al. Plant Cell 13, 889-905 (2001).

35. McCallum, C.M., Comai, L., Greene, E.A. & Henikoff, S. Nat. Biotechnol. 18, 455-457 (2000).

36. Colbert, T., Till, B.J., Tompa, R., et al. Plant Physiol. 126, 480-484 (2001).

37. Jaglo-Ottosen, K.R., Gilmpour, S.J., Zarka, D.G., Schabenberger, O. & Thomashow, M.F. Science 280, 104-106 (1998).

38. Kasuga, M., Liu, Q., Miura, S., Yamaguchi-Shinozaki, K. & Shinozaki, K. Nat. Biotechnol. 17, 287-291 (1999).

39. Rus, A.M., Estan, M.T., Gisbert, C., et al. Plant Cell Environ. 24, 875-880 (2001).

40. Apse, M.P., Aharon, G.S., Sneddon, W.A. & Blumwald, E. Science 285, 1256-1258 (1999).

41. Zhang, H.-X., Hodson, J.A., Williams, J.P. & Blumwald, E. Proceedings National Academy of Science 98, 12832-12836 (2001).

42. Peng, J., Caro, P., Richards, D.E., et al. Genes & Dev. 11, 3194-3205 (1997).

43. Peng, J.R., Richards, D.E., Hartley, N.M., et al. Nature 400, 256-261 (2000).

44. Ikeda, A., Ueguchi-Tanaka, M., Sonoda, H., et al. Plant Cell 13, 999-1010 (2001).

45. Blum, A. in Molecular approaches for the genetic improvement of cereals for stable production in water-limited environment. A strategy planning workshop held at CIMMYT, El Batan, Mexico 21-25 June 1999 (eds Ribaut, J.-M. & Poland, D.) 29-35, International Maize and Wheat Improvement Centre, El Batan, (2000).

46. Report on the Systemwide review of plant breeding methodologies in the CGIAR, TAC Secretariat, FAO of the United Nations (2001)

47. The global Musa genomics consortium (2002). A strategy for the global Musa genomics consortium. Report of a meeting held in Arlington, USA, 17-20 July 2001. IPGRI, Rome. Italy.

Annex 1 Genomics resources for CGIAR mandate crops