The Science of Managing Genetic Resources
During the past 25 years there has been a dramatic increase in the number of germplasm collections in the world as well as in the number of samples stored in these germplasm collections. This growth has occurred principally at the national level as more and more countries seek to conserve those genetic resources of historic or strategic importance to their agricultural systems. Much of this growth in number of collections has been by the partial or complete duplication of existing collections.
Although this growth reflects a growing commitment to genetic resources conservation, it has also raised concerns. In addition, base collections of many of the major crops have now grown so large and diffuse that they may inhibit, rather than promote, effective management of genetic resources and their use by plant breeders. Because of increased emphasis on the collection of wild and weedy relatives of crop plants and the continuing development of new elite gene combinations by plant breeders, collections that are already large may quickly grow progressively larger. The problems faced by curators charged with characterizing, monitoring, and periodically regenerating these collections are also increasing. The various operations described in this chapter must all function well if germplasm banks are to be able to encourage germplasm use rather than merely acquire and maintain specimens.
Germplasm banks may contain seeds or plant materials that are vegetatively propagated as clones. This chapter is concerned largely with the special problems of managing genetic resources in the form
of seeds. Chapter 7 discusses some new developments in propagating and maintaining clonal materials.
Regeneration practices must be related not only to the life cycles (including cultural needs) and the reproductive biology of individual crops and species but also to the cost of obtaining sufficient quantities of good-quality seed. The problems (and costs) vary greatly with the breeding system; they also depend on other features, such as the reproductive rates of individual genotypes. There are special problems for wild species, including seed dormancy, seed shattering, and high variability in flowering time and seed production.
Seed regeneration is thus costly in terms of resources and time, while the risks of genetic drift (change in allelic frequencies due to sampling accidents in small populations) and genetic shift (change in
allelic frequencies due to selection during regenerations) are compounded over each regeneration event. Consequently, the most cost-effective way of minimizing the loss of genetic integrity is to keep the frequency of regeneration to a minimum. Earlier, a distinction was made between regeneration needs for the rejuvenation of stored accessions and multiplication for distribution and use. If sufficient numbers of seeds are stored in long-term base collections, samples may be drawn off periodically as required for multiplication in active collections to satisfy user demands. Such seeds would not be used for conservation purposes thereafter, thus making it possible to carry out their regeneration under less rigid requirements.
By judicious planning, the number of generations required for both rejuvenation and multiplication could be limited to as few as two in the foreseeable future for many species with orthodox seed. Thus, for cost-effectiveness, as many seeds as possible should be collected during sampling or the first regeneration should be maximized to provide adequate seed for immediate use and long-term storage. A better integration of long-term (base) and medium-term (active) storage facilities may be an advantage, and several schemes have been proposed (for example, Linnington and Smith, 1987).
The nature of the breeding system is of paramount importance in establishing the tactics of regeneration. Seed bank managers must therefore be thoroughly familiar with the breeding systems of their material and must pay particular attention to the extent that the breeding system varies within the species (Widrlechner, 1987). This section describes the general characteristics of the most commonly encountered breeding systems.
Breeding systems determine to a considerable extent the patterns of ecotypic differentiation and, consequently, dictate subsequent regeneration procedures.
Complete self-pollination is the most extreme form of inbreeding. It results in the rapid fixation of allelic combinations into homozygous, true-breeding genotypes (those in which the offspring and parents are genetically alike). This promotes the buildup and maintenance of multilocus complexes (genotypes) that are preserved intact over several generations and which may be adapted to particular habitats.
Sexual reproduction in outbreeders (those that cross-pollinate) involves crossing, segregation, and recombination. Outbreeder populations thus retain high levels of potential heterozygous and homozygous
variability. Because of gene flow, neighboring populations are often less sharply differentiated ecogeographically. Outbreeding populations also show less distinct multilocus structural organization within their populations than inbreeding populations do (Allard, 1988). Important adaptive allelic combinations are, however, preserved over sexual generations through chromosomal linkage, whereas phenotypic correspondence is often further secured by dominance and epistasis (where one gene masks the effect of another) of favored alleles. In preserving the genetic integrity of populations during regeneration, it is important not to disrupt this multilocus genetic organization of the population. The break-up of genetic organization through segregation nearly always leads to loss in vigor and reduced reproductive fitness.
Apomixis (asexual reproduction) often allows highly heterozygous individuals to produce seeds or other propagules that breed true. However, apomixis is frequently facultative with the result that all progency do not necessarily have genotypes identical to that of the parental individual.
The breeding system, chromosomal organization, and reproductive mode (asexual versus sexual) all interact in the release of genetic variability. Together they constitute the genetic system that allows species and populations to adjust to the twin evolutionary demands of preserving and propagating the successful genotype while retaining the genetic flexibility to meet environmental changes or colonize new areas, that is, the opposing demands of short-term fitness and long-term adaptability. Consequently, they are themselves under genetic selection to meet the life-style and life-cycle needs of different organisms.
Principles Involved in Maintaining Genetic Integrity During Regeneration
The initial supply of seeds generally is inadequate for distribution and must itself be multiplied. The viabilities of seeds or other propagules held in storage eventually fall below acceptable levels. Distributions of seeds ultimately deplete stocks to levels at which regeneration becomes necessary. The genetic structures of accessions can be altered during the regeneration process by contamination because of accidental migration or hybridization between accessions, differential survival of alleles or genotypes within accessions, or to random drift within accessions. This section considers methods for protecting the genetic integrity of accessions and minimizing losses in allelic variability for these reasons.
Contamination from Outcrossing
One of the most serious threats to the genetic integrity of accessions arising during the regeneration process is contamination resulting from hybridization between different accessions. Regeneration of self-pollinated species is usually carried out in field nurseries in which different accessions are planted in closely adjacent rows. No species is completely self-pollinated, however. Under close planting conditions, outcrosses occasionally occur even between accessions of species that are considered to be very heavily self-pollinated. Visibly recognizable outcrosses can sometimes be discarded before maturity during the next regeneration cycle. However, the technical expertise to recognize outcrosses is often lacking, and variants may not be eliminated. This results in a genetic shift that has caused accessions in many collections of inbreeders to lose much of their original distinctiveness. The problems of genetic shifts for outcrossed species are even more severe. In many cases, crossed accessions are mixed to the point that the collections are complex hybrid swarms.
Isolation of accessions from one another during regeneration prevents genetic mixing (see Chapter 4). Fortunately, with most inbreeders, minimal isolation, such as that provided by planting accessions in rows sufficiently separated to prevent physical contact between different accessions, virtually eliminates interaccession outcrossing. In other cases, for example, when bees serve as pollen vectors, techniques such as caging or larger separation distances may be necessary.
Interaccession outcrossing is usually more difficult to eliminate with cross-pollinated species than with self-pollinated ones. The preferred method of regenerating outcrossers is in open-pollinated field plots sufficiently distant from one another to provide isolation adequate to protect against pollen migration. This method is relatively less expensive and more convenient than constructing pollen-proof isolation cages for each accession. For large collections, however, it may not be possible to provide a sufficient number of well-isolated plots or pollen-proof cages to prevent crossing of accessions. In such cases, controlled hand-pollination may be necessary. To compensate for increased costs and greater labor requirements, the number of individuals in each accession that is regenerated may be reduced to an extent that the potential for genetic drift may arise. With fewer plants there also may be insufficient seed for distribution.
The differential survival or selection of gametes or genotypes that occurs during regeneration is also a threat to the genetic integrity of
accessions. Selection that occurs during regeneration can substantially alter the genotypes and hence the phenotypes of accessions. Counter selection for desired individuals is at least partially successful in preserving the original phenotype. Monitoring of isozymes and restriction fragment loci, however, has shown that allele frequencies often shift dramatically or may be lost or fixed in a very few generations in standard, densely planted regeneration nurseries where competition among individuals is high. Changes in allelic or genotypic frequencies are, however, usually much smaller when regeneration is carried out under husbandry conditions that minimize the effects of selection. If individuals are widely spaced so that a high percentage of them germinate, survive to reproductive age, and contribute approximately equal numbers of gametes in the reproductive process and approximately equal numbers of seeds or other propagules to the next generation, only small changes are likely to occur in allelic frequencies.
Isolation methods depend on the reproductive biology of the species and on the pollination mechanisms. Studies have shown that pollen dispersal by wind-and insect-pollinated species can result in cross-contamination of up to 5 percent over large distances (Breese, 1989). As a consequence, isolation distances in excess of 200 m are often recommended during the commercial multiplication of allogamous (cross-fertilizing) cultivars. Thus, isolation by distance alone is not normally a feasible method, particularly if large numbers of accessions are to be regenerated on limited amounts of land. However, effective pollen dispersal decreases rapidly over the first 10 or 20 m. Pollen dispersal can be further interrupted and reduced by intervening barriers. This has led to the practice of growing isolation plots interspersed among tall-growing crops, where relatively safe isolation distances of 30 m or less may be achieved for some crops. Other practices help to dilute the amount of unwanted pollen, particularly the planting of square plots rather than rows or using as large a plot as possible and discarding borders rows. Plot arrangement is also important in securing effective intra-accession random pollination. Alternative methods use insect-proof cages, pollen-proof glasshouse chambers for wind-pollinated species, or hand-pollination.
Population Size in Relation to Inbreeding Depression and Drift
Genetic integrity during seed regeneration is also threatened by
genetic drift. In infinitely large populations the frequencies of neutral alleles do not change from generation to generation. If only limited numbers of individuals participate in producing the next generation, however, random fluctuations in allelic frequencies occur because of sampling. The theoretical consequences of this phenomenon, known as random genetic drift, are well known (Breese, 1989).
Maintaining Effective Population Sizes
In regeneration, the total number of individuals, the nominal population size (N) , is less important than the average number of actively breeding individuals, the effective population size (Ne) . The practical aim in regeneration is to maximize Ne by ensuring, as far as practicable, complete random mating and equal contribution of as many individuals as possible to the next generation. Hand-pollination and open-pollination by natural vectors are examples of methods used to accomplish this goal.
For amenable species, such as maize and sunflower, pollination control by hand crossing and storage of equal amounts of maternal seed can maintain effective population size at, or above, the nominal population number and, thus, minimize the effects of drift and selective shift. For many crop species, hand-pollination is impractical, and an effective population size can only be maintained by ensuring that random mating is achieved as efficiently as possible by natural pollen vectors (wind or insect) and by equalizing maternal seed sources. The use of natural vectors is critical in these regeneration procedures. Nevertheless, their use will not overcome environmental and genetic variation in pollen production between the different genotypes and its profound effect on decreasing the effective population size.
The effect of numerical differences between the two sexes is very pronounced in dioecious species (those in which male and female flowers grow on separate plants) and the sex ratio of breeding plants. For an extreme case of 2 males and 20 females, the effective population size is 3.6. The calculations are slightly different for monoecious plants (those that have both sexes on the same plant), but the results are much the same.
The Effect of Genetic Drift and Natural Selection
Random loss of alleles due to genetic drift is directly related to the effective population size. Its effect, therefore, can be predicted within statistical limits. The effect of natural selection varies from environment to environment (including different years at a given location)
and hence its effects can be determined only through experimentation. In general, random drift causes rare alleles, whether adaptive or not, to drift out of small populations, whereas selection tends to eliminate the less adaptive alleles and to increase the frequency of alleles that are better adapted in the regeneration environment.
Evidence of marked phenotypic changes is now well documented in outbred populations, especially when multiplication is performed in population growing in regions that differ from their origins (Breese, 1989). Selection operates on individual phenotypes through differential survival or fecundities. Populations are maintained in their natural habitats through the pressures of natural (and farmer) selection. In ex situ situations, the selection pressures may change, and there is almost certainly a change in individual phenotypes as a result of genotype-environment interactions. Such interactions may lead to an increase in the phenotypic heterogeneity of the population. The aim is to minimize genotype-environment interactions and reduce selection pressures. Therefore, important operational considerations are the location of regeneration, growing conditions, effective cross-pollination, and harvesting of equal quantities of seed (Breese, 1989).
Regeneration of Predominantly Inbreeding Species
Completely self-pollinated and genetically homogeneous, accessions present no genetic drift or shift problems, and population numbers are determined entirely by the yield capacity of individual genotypes.
Maintenance of Landraces or Wild Species
Landraces and wild populations are usually genetically heterogeneous and have complex genetic structures, even when the degree of self-pollination is virtually complete. The most effective way of preserving the gene and genotype constitution of highly self-pollinated populations is to maintain the accessions as N subsets of homozygous inbred lines. When N > 100, chances are small that no allele or genotype present in the initial sample number will be lost during regeneration. The initial sample number needed to capture a copy of an allele for an inbred population with a similar frequency and probability is double that required for completely outbred populations (on the order of 100 in inbred populations compared with 50 in outbred populations) (Marshall and Brown, 1975). As Gale and Lawrence (1984) show, however, because allele loss (drift) continues in outbreeders, the same efficiency of conservation is achieved over four or five regeneration
events for similar constant population sizes of outbreeders and inbreeders when the inbreeders are effectively maintained as inbred lines.
This method of maintaining separate subsets of inbred lines is, however, laborious and costly in administrative terms. Therefore, when large numbers of accessions are involved, it may be preferable to raise the population as a bulk. For highly self-pollinated populations no special isolation requirements are needed. At the same time, it should be remembered that few populations are completely self-pollinated in all environments, so that minimal isolation may require that accessions are sufficiently separated to avoid physical contact. When maintained as bulks, a doubling of the effective population size (Ne) is technically required for inbreeders compared with outbreeders for conserving the same allele frequencies with similar levels of probability (Bray, 1983). Because no crossing is involved, maintenance of effective population numbers (and avoidance of genetic drift) in a nominal population size (that is, the Ne/N ratio) is largely a question of equalizing maternal reproductive outputs. Problems with equalizing maternal reproductive outputs and minimizing genetic shift through natural selection are essentially the same as discussed above for outbreeders.
The mating systems of some populations feature intermediate levels of selfing versus outcrossing. Above particular threshold values of outcrossing (about 5 percent), problems of pollen contamination, crossing requirements, and potential genetic drift and shift require that the populations be handled essentially as described above for outbreeders. Normally, natural open-pollination methods would be used so that the population is maintained as much as possible at its natural level of outcrossing. Considerations for partial inbreeders thus include adequate isolation methods and the provision of the necessary effective pollen vectors.
Characterization is a systematic recording of selected morphologic and agronomic traits of an accession. It is typically restricted to those genetically controlled traits that are highly heritable and do not vary with environmental conditions (Frankel, 1989b; Williams, 1989b). Characterization data should be linked to passport and evaluation data. Passport data comprise the information about the site and environment from which an accession originated. Evaluation data for more variable traits, such as productivity, quality of product, and disease resistance are of greatest interest to users of germplasm. It
has been argued that (1) when traits such as these are included in data-base systems, such systems become a more useful tool for users who wish to identify accessions for specific purposes, and (2) it is consequently worthwhile for conservation units to assess the potential usefulness of accessions in their collections. For a given germplasm collection, the characterization data form a permanent set of biologic descriptors.
Characterization activities can be carried out concurrently with the regeneration process and therefore typically do not require an extra planting. When the process is divided between two alternating crop seasons, as it is implemented at the International Rice Research Institute, quantitative traits of agronomic importance are recorded in the main (wet) season, whereas qualitative traits can be scored in the off (dry) season.
Wide applicability of characterization data can be ensured by standardized descriptors (traits) and descriptor states (absolute or coded values). The challenges in setting up the descriptors and descriptor states are to choose between an exhaustively large set and a minimal set so as to render the task workable and data useful. Rice workers have arrived at a consensus, and the standardized descriptors (International Board for Plant Genetic Resources-International Rice Research Institute Rice Advisory Committee, 1980) are widely used. Characterization data are also useful in reidentifying an accession during seed preparation for planting or distribution and when accessions are regenerated. Such checks must be made routinely to ensure the maintenance of varietal identity.
With the advent of powerful electrophoretic and molecular diagnostic techniques, the scope of characterization can be vastly expanded to provide much greater resolution to the genetic differentiation among accessions (see Chapter 7).
Evaluation is a prerequisite for the use of conserved germplasm. Accessions that are not evaluated mostly remain curiosities. Evaluation has been described as including such fields as crop cytogenetics and evolution, physiology, and agronomy (Frankel, 1989b). It is essential to utilizing the genetic diversity of germplasm collections.
Large-scale systematic evaluation generally falls outside the domain of germplasm workers; rather, it is undertaken by a variety of biologists, including plant breeders, who specialize in crop improvement or production research. However, the initial phase of preliminary evaluation is usually performed by the germplasm workers during
the first cycle of multiplication-regeneration and observation. The observations are limited to a small number of morphological features, adaptability at the site of planting, and obvious economic worth. Nevertheless, such observations are crucial in enabling the germplasm worker to forward the materials to the appropriate researchers or breeders for further testing, leading to more systematic evaluation. Plant quarantine precautions should be observed at this stage.
The prerequisites for effective and efficient evaluation are described below.
Increase of Seed or Plant Parts
Sufficient materials in the form of either seed or plant parts are imperative to evaluation experiments. For materials of a heterogeneous and heterozygous nature, a sample size larger than that for the pure-line cultivars will be needed. Wild species generally produce little seed and are difficult to handle. For users of wild species, guidelines should be provided to break strong dormancy, to raise slow-growing seedlings, to provide appropriate solar radiation levels and photo period requirements, to follow a taxonomic key for reidentification, and to collect the precious seeds (Chang, 1976).
Germplasm workers are likely to be asked for a second supply of seed or materials by the same or other researchers. Therefore, maintenance of the genetic integrity and the correct nomenclature and accession numbers are essential. It is preferable to maintain a large seed stock rather than making repeated rejuvenations, because the latter leads to genetic drift, outcrossing, and errors in labeling and handling.
Multidisciplinary Approach to Evaluation
Most evaluation operations now require expertise beyond that of one or two related disciplines. A multidisciplinary approach is essential in planning, execution, and assessment of results. The Genetic Evaluation and Utilization Program of the International Rice Research Institute is an example of such multidisciplinary collaboration and interdisciplinary interaction. Effective evaluation experiments often lead to more refined and rewarding research.
Planning and Implementing Evaluation Tests
Every evaluation test should be treated as a scientific investigation. It should embody well-defined objectives, the experimental design
should be efficient, and there should be judicious choice treatments and control varieties. The environmental factors should be controlled, if necessary, and complete statistical analysis is essential. Data should be entered in computerized data files for easy access and monitoring. Results from repeated tests should be compared with previous data before entering the information into the files for updating purposes. The usual progression in the tests is from simple, large-scale screening to verification tests and then to more critical studies.
Choice of Representative Environments or Sites
It is imperative that evaluation experiments be carried out in environments that represent the major target production area(s). For many adverse environments, off-station test sites in stressed environments are necessary. To have a heavy epidemic, disease and insect nurseries need to be planted at "hot spots" where the pest is endemic; artificial infestations should be arranged when possible. The control of temperature, relative humidity, and light intensity in greenhouses or growth chambers may be necessary. Efficient experimental designs are needed to make maximum use of available resources.
Preventing Exotic Germplasm from Becoming a Serious Pest
The extreme seed shattering and strong rhizomatous characteristics of some wild or primitive germplasm pose a threat if such vigorously growing plants or seeds were allowed to escape into water canals or farmers' fields. Measures should be taken to prevent such accidents. At the International Rice Research Institute, wild rices are grown in large pots to control their spread, to ensure their identity, and to obtain harvests from ratooned plants.
Verification of Evaluation Data
Disease- and insect-susceptible accessions may be recorded as such without retesting, whereas populations identified as resistant or tolerant need to be retested under more precisely controlled conditions to establish the validity of the preliminary evaluation. Appropriate control varieties or treatments should be included in each experiment. This step should also apply to the evaluation of other traits, especially those being subject to the effect of variable environments or planting densities in field trials. Sampling of environmental variance
and the genotype-environment interaction component adds scope to the evaluation experiments.
Communication Among Plant Germplasm Users
Open and continuous communications among conservationists, scientists in related research areas, such as entomologists and physiologists, and plant breeders is essential to sustain evaluation efforts and subsequent use. Crop advisory committees that bring together a variety of interests from public and private sectors facilitate cooperation and communication. They are important sources of crop-specific expertise for the U.S. National Plant Germplasm System, for example (National Research Council, 1991a).
When evaluation efforts are coupled with in-depth research by members of allied disciplines, further refinements in testing techniques are accelerated. Scientific advances in the discipline(s) concerned are also accelerated as an outcome of the multidisciplinary interaction. Breeding programs benefit from both areas of endeavor. There are many examples of rice breeding programs that illustrate the rewarding results from such collaborations and interactions (Chang, 1985a,c).
Despite the accelerated efforts on different fronts, inadequate documentation still constrains the use of exotic or unimproved germplasm. A lack of general adaptation to commercial production areas and poor agronomic and quality characteristics are common complaints aired by breeders as the main drawbacks of exotic germplasm. Low crossability and sterility barriers are additional drawbacks voiced by users. Lack of adequate classification systems, poor passport data sets and sometimes uncertain fertility relationships add to the reluctance in using exotics.
Efficient documentation should extend across all phases of conservation. The availability of evaluation data is crucial for arousing the interest of potential users (for example, breeders, entomologists, plant pathologists) of the germplasm. However, passport data and characterization information are the most important to have. Free and widespread exchange of evaluation results also leads to enhanced exchange of germplasm. The use of standardized descriptors and descriptor states (International Rice Research Institute, 1975) has proved its usefulness in international evaluation programs. A huge backlog of data accumulated by germplasm workers needs to be located, interpreted,
transcribed, and entered into the central files. With the constraint of time and space and the death or retirement of earlier workers, a substantial portion of the invaluable data may never be made available to other interested workers.
MONITORING OF SEED VIABILITY AND GENETIC INTEGRITY DURING STORAGE
An important aspect of the viability of stored seeds is the vigor of the germinated seeds. Seed banks should also monitor the appearance of abnormal seedlings, some of which could be mutations that have arisen during storage. Seed production is costly and the viability test destroys stored seeds, however. Other measures to prolong the longevity of stored seeds and reduce the quantity of seeds used in germination tests should be fully exploited to reduce the amount of seeds lost in the process and to decrease workloads.
Nondestructive tests to monitor seed viability should therefore be explored and developed. Processing operations and storage conditions that would prolong viability of stored seeds should also be developed.
The International Board for Plant Genetic Resources has recommended that 85 percent be the threshold viability level for rejuvenation of stored seeds (National Research Council, 1991a). Germplasm banks of the People's Republic of China and the International Rice Research Institute use a 50 percent threshold level, whereas the U.S. National Seed Storage Laboratory has, until recently, used a 60 percent threshold (National Research Council, 1991a). In practice, it may not be possible to obtain high germination rates for some species or accessions. Thus, the viability standards for collections must be a percentage of the optimum attainable germination with carefully produced fresh seeds.
REDUNDANCY AMONG COLLECTIONS
Many national collections (and the collections maintained by international centers) have become very large, and on a worldwide basis, collections of most major crops and some minor crops have become enormous (Holden, 1984). For wheat, for example, there are a total of more than 400,000 accessions in 37 collections worldwide (Plucknet et al., 1987). How many unique accessions exist is not known for any major crop. Large size in itself is not necessarily a problem if the germplasm bank has sufficient storage facilities and management expertise as well as resources—factors that lead to low
operation costs (see, for example, Chang, 1989, 1992). Size can, however, limit important activities such as regeneration and evaluation. Management strategies, such as the identification of core subsets and increased efforts to capture passport and characterization data, have been proposed to address the problem of managing large collections (International Rice Research Institute, 1991).
If all duplicate accessions in individual collections were eliminated, the gains in efficiency from a reduction in entry numbers would likely be modest for most collections. The costs of eliminating duplicates may exceed the cost of keeping them, particularly if passport data are not adequate to help in their identification. Isozyme or restriction fragment length polymorphism analyses are far too costly for routine germplasm bank use. Greater gains could perhaps be made by reducing redundancy among collections to the minimum necessary for insurance against catastrophic loss. This is far more difficult to achieve, however, because it requires a high level of cooperation and coordination among national as well as international germplasm banks. The size of collections is seen to be of importance in terms of prestige and funding. Furthermore, many countries have strict quarantine laws, and accessions are retained to avoid the inconvenience and cost of repeated quarantines. When countries heavily depend on particular crops, germplasm collections of those species acquire strategic importance, and larger national collections are seen as insurance against an uncertain future, especially in the light of the growing controversy surrounding the ownership and control of crop genetic resources.
The major problem in redundancy is not within collections, but between them. This redundancy has arisen from the repeated distribution of the same sets accessions. However, for breeders and curators in countries isolated by distance or quarantine barriers, some duplication between collections offers the advantage of ready access and the opportunity of becoming familiar with a wide range of crop species. Unfortunately, recipient collections do not always retain the original accession identifiers and, thus, it may be difficult to subsequently identify duplicates between collections.
Core Subsets of Collections
The limited scope for eliminating redundancy within collections led to a proposal for developing core subsets in germplasm banks (Frankel and Brown, 1984). Under this concept, the accessions in large germplasm banks would be sampled to form a core subset that contains the genetic diversity in the crop species and its relatives in a
base collection, with minimum redundancy. The base collection would include those materials in the core subset plus all other materials and would continue to be maintained (Brown, 1988; Frankel and Brown, 1984; National Research Council, 1991a).
Advantages of Core Subsets
The aim of identifying a core subset is to facilitate use and, in particular, to provide efficient access to the whole collection (Brown, 1989a; National Research Council, 1991a). Consider a breeder faced with a new virulent race of a pest of pathogen. The breeder initially searches for resistance in readily available material, usually a limited working collection. If no resistance is found in the working collection, the next step by the breeder is to screen material from an appropriate germplasm bank collection. If that germplasm bank has a well-defined core, this would be screened first. If no effective resistance is located in the core, then the breeder knows that resistance is relatively rare; the breeder is then faced with a substantial problem and the prospect of screening the other 90 percent of the collection. If resistance is found in the core collection, then the breeder can use that resistance immediately in the breeding program as well as screen additional accessions from the now identified geographic areas held in the base collection.
The core subset concept also has advantages for curators. It is envisaged that more seed of the core subset would be kept on hand, packaged, and ready for distribution to meet general seed requests. It is also envisaged that accessions in the core subset would receive priority in evaluation and characterization, so that in time many more characteristics would be evaluated for all of the core samples than would be evaluated for the remainder reserve samples. In this way, curators could better use a limited budget to promote the distribution of information and material and, hence, facilitate use of the entire collection. With good communication, core subsets may also effectively limit redundancy between collections. The typical accessions of the Latin American races of maize serve as a useful example of a core subset.
Disadvantages of the Core Concept
One disadvantage of the core concept is the possibility, or indeed, probability, that the remainder of the base collection, held in reserve, would erode away and disappear from neglect or, alternatively, would be seen by some administrators as being of less value and therefore
dispensable in the interests of economy (National Research Council, 1991a). To minimize this disadvantage, Frankel and Brown (1984) and Brown (1989a) have stressed that this reserve must remain an important and integral part of the base collection. Under the core concept, the reserve serves at least two important functions: (a) as an alternative to be screened if needed variation is not found in the core subset, and (b) as a source of additional diversity when many different genes or alternative alleles are required for the same trait. Nevertheless, the reserve may be more vulnerable to neglect and dismemberment, regardless of how strong a case is presented for its maintenance.
Practical Problems with Core Subsets
There are several practical problems that must be overcome before the core concept can be entirely implemented. One is establishing the appropriate size of a core subset and defining its composition so that it contains a representative sample within the defined size. These problems have recently been examined by Brown (1989a,b). Using the sampling theory of selectively neutral alleles, Brown (1989a) argued that a core should consist of about 10 percent of the collection, up to a maximum of about 3,000 accessions for each species. At this level of sampling, the core subset will generally include over 70 percent of the alleles in the whole collection. Brown (1989a) also argued that, as a general rule, a fixed proportion of 10 percent is more useful than a fixed-number upper limit.
Brown (1989b) examined alternative procedures for choosing core entries. He showed that stratified sampling, in which the collection is first divided into nonoverlapping groups and a sample is taken from each group, is more efficient in establishing a core than is random sampling, in which accessions are chosen from the whole collection at random. Brown also examined options for deciding the number of samples from each nonoverlapping group to include in the core. These included a constant strategy (an equal number of accessions from each defined group) and two variable strategies (with sample sizes correlated with sizes of the groups). Of these, the variable strategies gave broadly similar results and were better than the constant strategy. This leaves the problem of how to stratify the collection into nonoverlapping groups (Brown, 1989a). This can be done rationally only when reasonable knowledge about the materials in question is available (International Rice Research Institute, 1991). Even a knowledgeable curator would need access to the following:
The origin of the accession (collection site of area adaptation);
Characterization data, including taxonomic and agronomic data and information from the analysis of genetic marker loci; and
Evaluation data for economically important quantitative or qualitative traits (for example, yield, disease resistance, and cold tolerance).
It would be inappropriate for an inexperienced curator to attempt to define a set of core accessions or for anyone to define a core collection for a crop with inadequate passport or characterization data. For large collections, computer-based multivariate methods such as numerical cluster analysis, principal component analysis, or network analysis might be helpful to order the collection into /related groups.
Neither the reduction of collection size by elimination of accessions nor the use of core subsets is universally accepted (see, for example, Brown, 1989a,b; Chang, 1989; Marshall, 1989; National Research Council, 1991a). The best strategy for addressing increases in collection size depends on many factors, including the crop species being maintained, the accession data, management practices in use, the needs of users, and the resources available to conserve and manage the collection.
The capacity to perform regeneration of seed accessions when needed should be a major determinant for limiting the size of germplasm collections.
Regeneration of seeds in a germplasm bank collection becomes necessary for two purposes: for rejuvenation, when the viability of seeds in stored accessions falls below acceptable levels, or for multiplication, when high levels of distribution to other breeders or institutions lead to the depletion of seed inventories. Performing regeneration exposes accessions to a variety of risks that could lead to genetic depletion, genetic shift, or mixing with other accessions. With large collections, there may be insufficient resources to grow adequately large populations to preserve genetic integrity. Alternately, insufficient resources may necessitate deciding which of many accessions should be regenerated. In either case, the genetic integrity of a collection can be placed at risk because of the inadequacy to perform this fundamental management practice.
Collections should employ storage and management practices that minimize the need for regeneration.
The risk of allele loss is repeated during each cycle of regeneration. Consequently, the interval between regeneration events should
be made as long as possible. Improvements in long-term storage have aided this by reducing the frequency of regeneration needed for orthodox seed. Three other considerations emphasize the need to minimize regeneration: (1) The regeneration process is costly in terms of resources; (2) each regeneration event places accessions at risk from clerical and other human errors, and in danger of genetic shift due to selection or genetic drift; and (3) effective population size is bound to vary with each regeneration event, and over several generations will be much closer to the smallest single value. Even a single regeneration event in which the effective population size is small (less than 50) can result in serious genetic loss within the accession.
Core subsets should be identified for large collections to aid in managing, evaluating, and using accessions.
Core subsets should not be used as a justification for neglecting the maintenance of the other accessions in a collection. Further, identification of core subsets is not be viewed as a prelude to elimination of other accessions. A core subset should "include, within an acceptable level of probability and with minimum redundancy, most or much of the range of genetic diversity in the crop species and its relatives (typically no more than 10 percent of the whole collection)" (National Research Council, 1991a:125).
Although they should be selected to represent the diversity of the main collection, core subsets are unlikely to capture all of the genetic diversity within a collection. It is probable that if 10 percent of the accessions in a collection are assigned to a subset, no more than 70 percent of all the available alleles will be included (National Research Council, 1991a). Thus, the rest of the collection will continue to serve as a source of diversity. This larger portion may also serve as a resource for genetic traits not captured in the core subset.
Core subsets are management tools that can be used to develop priorities for evaluation and collection. They do not require the construction or establishment of separate collection facilities. Identifying cores does, however, require a basic amount of information about accessions. Minimally this should be the passport data regarding the origins and environment of the accessions. Such information is essential to linking core subsets to the potentially wider genetic diversity of the entire collection. An initial step in identifying cores, therefore, is the recording of basic passport information about each accession in a collection.
Redundancies within global collections should be minimized
The problem of redundancy has arisen because it is easier to add an accession than to discard one. Elimination of redundancy in existing
collections is not cost-effective. The costs for biochemical or molecular methods needed to identify duplicates are not justified by the relatively small number of accessions that would be eliminated. A more rational approach is to prevent exacerbation of the problem by a more cautions approach to the addition of new accessions. To accomplish this, basic information (passport data) for existing accessions and new ones is needed. Much of this may exist, but is not often found in the databases of germplasm collections. This information, coupled with the more detailed information that might come from study of selected core subsets would prevent or reduce the addition of duplicative accessions to germplasm collections.
Research is needed to develop long-term storage methods for short-lived, desiccation-sensitive, or cold-intolerant seeds.
At present, many important collections must be maintained as field-grown plants. Many crops are maintained as clones, including important ones such as yams, cassava, sweet potatoes, potatoes, bananas, sugarcane, rubber, cacao, coconuts, and oil palms. Understanding the mechanisms of dormancy in seeds and the physiology of these difficult-to-store seeds may lead to methods of improved storage. It also may be possible to develop methods of storing excised plant parts or embryos as in vitro tissue cultures. Storage of excised embryos at cryogenic temperatures (-150° to -195°C) may also prove feasible for some species (see Chapter 7). However, long-lived field collections are necessary for preserving clones and other plants that have short-lived seeds. For example, the seeds of rubber plants survive a few weeks, and attempts at in vitro storage and propagation have been unsuccessful.
Increased efforts should be instituted to train germplasm workers.
Germplasm activities require long-term commitment and continuity in personnel. Being a new and composite technology, the subject is not taught in most universities. The number of trained workers is small and inadequate for the heavy workload. The requisite expertise derives in part from direct experience beyond a training course or university degree. The development of trained professionals and germplasm workers is an imperative for sustaining the future of genetic resources and making advances in agriculture.