| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Centro Internacional de Agricultura Tropical A.A. 6713, Cali, Colombia
* Corresponding author (j.miles{at}cgiar.org).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSummaryREFERENCES |
|---|
Abbreviations: IAC, Agronomic Institute of Campinas • RRS, reciprocal recurrent selection • RS-SCA, recurrent selection on specific combining ability
Received for publication April 9, 2007.
Centro Internacional de Agricultura Tropical A.A. 6713, Cali, Colombia
* Corresponding author (j.miles{at}cgiar.org).
Apomixis—asexual reproduction through seed—provides a convenient means to faithfully propagate even heterozygous genotypes and hence exploit heterosis, in several naturally apomictic, warm-season forage grasses. Inheritance of apomixis has been shown to be monogenic dominant in at least four economically important panacoid grasses. Previously proposed breeding schemes for apomicts do not provide a means to accumulate genes contributing to nonadditive, heterotic effects over cycles of selection and recombination. Following the development of successful brachiariagrass [Brachiaria (Trin.) Griseb] cultivars by ecotype selection, artificial hybridization of brachiariagrasses began in the late 1980s with the development of a sexual tetraploidized biotype of the natural diploid, sexual ruzigrass (Brachiaria ruziziensis Germain and Evrard). A breeding scheme—recurrent selection for specific combining ability—designed to accumulate nonadditive effects, originally proposed for sexual maize (Zea mays L.), is suggested as an appropriate scheme for improvement of apomictic tropical grasses. Recurrent selection on specific combining ability or interpopulation selection schemes such as reciprocal recurrent selection should be appropriate for other asexually propagated crops.
Abbreviations: IAC, Agronomic Institute of Campinas • RRS, reciprocal recurrent selection • RS-SCA, recurrent selection on specific combining ability
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSummaryREFERENCES |
|---|
Apomixis is generally understood to mean asexual reproduction by seed, although its broader literal meaning encompasses all forms of asexual reproduction (apo = "away from" and mixis = "act of mixing" [Asker and Jerling, 1992]). As apomixis is considered a suppression of sexual reproduction, leading to an absence of genetic recombination, it was at one time thought to be an evolutionary "dead end" (Darlington, 1939). However, given that some of the world's most ecologically successful and widespread plants are apomicts (e.g., many species of dandelion [Taraxacum F.H. Wigg.] and hawkweed [Hieracium L.]), the evolutionary and plant breeding potential of apomictic reproduction has been reconsidered more recently. Although some authors still seem to view apomixis as an impediment to plant breeding (e.g., Jessup, 2005) numerous authors (e.g., Hanna and Bashaw, 1987; Hanna, 1995; Savidan, 2000) list a number of advantages of apomictic reproduction in crop improvement. These basically relate to facilitating the faithful propagation of unique heterozygous genotypes. Hence, apomixis, if transferred to crop plants, would allow exploitation of heterosis in crops where this is currently not economically feasible (e.g., wheat [Triticum aestivum L.]). In crops such as maize (Zea mays L.), where hybrid cultivars are currently used, apomixis would eliminate the complexity of producing and maintaining inbred parental lines and the cost of annual production of F1 hybrid seed.
So great is the perceived potential advantage of apomictic reproduction that significant efforts have been expended to transfer the trait to sexually reproducing crop species (e.g., maize, pearl millet [Pennisetum glaucum (L.) R. Br.], and wheat; reviewed by Savidan, 2000, 2001). However, the transfer of apomixis to naturally sexual crop species has proved to be a formidable task and none of the programs has so far been successful (Savidan, 2000, 2001).
Where apomixis is obligate at the species level, lack of genetic recombination may indeed be an evolutionary handicap since new genetic variation can arise only from mutations. However, where even a low level of sexuality allows the creation of new genetic combinations, the advantages of asexual seed propagation of one or a few successful hybrid genotypes clearly has immense advantages. Where natural apomictic species have been studied in detail, total absence of sexuality is found to be rare or nonexistent (Asker and Jerling, 1992).
Several types of apomixis are recognized (Asker and Jerling, 1992; Savidan, 2000), and many more are theoretically possible depending on the origin of the unreduced embryo sac (female gametophyte) and the details of the embryo and endosperm development in such unreduced sacs (Crane, 2001). However, in the tropical grasses of the subfamily Panicoideae, the source of several economically important warm-season forage grasses, only two types of apomixis have been reported. In both cases a female gametophyte is formed (gametophytic apomixis), originating either from the unreduced megaspore mother cell following a failure of meiosis (diplospory) or from a somatic, nucellar cell (apospory). In either case, the unreduced egg cell develops without fertilization (parthenogenesis) into an unreduced embryo, genetically identical to the mother plant. However, as pointed out emphatically by Nogler (2006), apomixis is not simply a loss of sexuality. While with diplospory, multiple embryo sacs in the same ovary are precluded, sexual and apomictic sacs can occur on the same plant. With apospory, a functional, meiotically derived female gametophyte may coexist with the aposporic sac in the same or in different ovaries on the same plant. Hence, different degrees of facultative apomixis can occur at the individual plant level.
The structures of diplosporous and aposporous gametophytes differ. In diplospory, the unreduced megaspore mother cell undergoes three mitoses and produces an eight-nucleate embryo sac (with one ovule, two synergids, two polar nuclei, and three antipodals) that is indistinguishable cytologically from a normal sexual sac. In apospory of the Panicum type (first reported in guineagrass [Panicum maximum Jacq.]; Warmke 1954) the unreduced embryo sac develops from a somatic, nucellar cell following only two mitoses. Hence, the mature embryo sac has only four nuclei (one ovule, two synergids, and a single polar nucleus, with no antipodals), and it is readily distinguishable cytologically from the normal, eight-nucleate sexual sac. This peculiarity is the most important practical difference between the types of apomixis encountered in the tropical forage grasses as the four-nucleate aposporous embryo sac allows relatively efficient phenotyping of individuals in a segregating hybrid progeny as to whether or not they have a capacity to reproduce aposporically. Panicum type apospory occurs fairly commonly in the subfamily Panicoideae, and is the predominant type of apomixis among the economically important warm-season forage grasses.
Regardless of the embryological details of the different types of apomixis, the genetic consequence of all the types of apomixis is the same: faithful reproduction—cloning—of the maternal genotype in seedling progenies.
In spite of detailed cytological descriptions of apomictic processes (e.g., Asker and Jerling, 1992), Bashaw's (1980b:57) statement of 26 years ago ("There is yet no biochemical or physiological information on the factors responsible for apospory.") remains essentially true today ("...the genetic mechanisms underlying apomixis in nature have not been determined..." [Ozias-Akins, 2006:199]). This lack of mechanistic understanding does not, however, represent a serious obstacle to the utilization of naturally occurring apomixis in a plant breeding program. The critical needs are an understanding of the inheritance of reproductive mode and a means to distinguish, in hybrid-derived populations, between individual segregants that reproduce sexually or by apomixis.
While apomixis is fairly common and widely distributed among plant families (Asker and Jerling 1992), it is rare among the domesticated crops, where it has been found only in a few fruit species and in several forage grasses. The fruits where apomixis is found are generally propagated vegetatively (e.g., citrus (Citrus L.), mango (Mangifera indica L.), and several species of Malus Mill. [crabapples] and Rubus L. [blackberries]). Hence, the only apomictic crops where the advantages of seed propagation are combined with the fidelity of asexual reproduction are several economically important forage grasses, generally recent domesticates or semidomesticates, for example, guineagrass, buffelgrass [Pennisetum ciliare (L.) Link (syn. Cenchrus ciliaris L.)], bahiagrass (Paspalum notatum Flüggé), and several commercial species in the genus Brachiaria (Trin.) Griseb, collectively the brachiariagrasses.
Until recently, the commercial cultivars of these apomictic tropical forage grasses were apomictic "lines" (clones) selected directly from collections of natural germplasm. This approach to cultivar development has been highly successful. The natural apomictic cultivars are true breeding, facilitating commercial seed production and allowing direct exploitation of whatever heterotic effects they may contain, which are likely substantial given the natural selection that would have occurred before plant collection and to the fact that the cultivars were then selected for superior agronomic performance (largely forage yield) from large germplasm collections.
However, unmodified natural germplasm does not provide ideal cultivars even in forage plants. For example, it is unlikely that evolution under uncontrolled grazing in the African savannas would elicit plants with high ruminant nutritional quality. Hence, interest in breeding of natural apomicts to rectify perceived cultivar defects has been expressed persistently over the years: bahiagrass (Burton and Forbes, 1960), buffelgrass (Bashaw, 1962; Bashaw et al., 1970), guineagrass (Smith, 1975; Pernès et al., 1975), and the brachiariagrasses (Ferguson and Crowder, 1974; Gobbe et al., 1983). In the early years, breeding of apomictic species was perceived as a serious challenge (e.g., Bashaw, 1975), due to the perceived difficulty in achieving genetic recombination. It is only in the past two to three decades that the special features of apomictic reproduction began to be appreciated as a distinct advantage, rather than an obstacle, in a plant breeding program. The change in perception came with the application of new cytological tools that facilitated the detection of apomixis (Savidan, 1975; Young et al., 1979). These cytological tools allowed reliable phenotyping of reproductive mode in large populations, leading to advances in understanding the genetics of apomixis. This understanding, at least in principle, allowed increased control of the process in a breeding program.
Genetic Control of Apomixis
The genetic control of apomictic reproduction has been reviewed recently by Savidan (2000), Sherwood (2001), and Ozias-Akins (2006). In spite of many genetic studies, unequivocal results long remained elusive in several species, testifying to the difficulties involved in studying the inheritance of a trait whose phenotype can be difficult to assess and one that interferes with the hybridizations on which inheritance studies depend. Perhaps the most consistent pattern of inheritance of apomixis is among the panacoid grasses that exhibit monopolar (four-nucleate), Panicum type embryo sacs (guineagrass, buffelgrass, bahiagrass, and several brachiariagrasses). There is good reason for the relative success in this group: the typically four-nucleate Panicum type embryo sac is readily recognized microscopically, using efficient pistil clearing techniques developed in the 1970s (Young et al., 1979), allowing large segregating populations to be reliably phenotyped for reproductive mode.
The earliest attempts to understand the inheritance of apomixis in a tropical forage species were the pioneering studies reported by Burton and Forbes (1960) for bahiagrass. As is common in apomictic species, apomixis was found in polyploid (tetraploid) biotypes of bahiagrass, while the natural sexually reproducing plants were diploid. By doubling the chromosome number of several sexual diploid clones, sexual tetraploids were produced and these were found to be cross compatible (as females) with the tetraploid apomicts, which produce normal pollen with reduced (n = 2x) chromosome number. Reproductive mode of hybrid clones was assessed by progeny test and the somewhat ambiguous results suggested that at least two genes were involved in determining reproductive mode phenotype.
A more recent study (Martínez et al., 2001) showed sexuality to be recessive in tetraploid bahiagrass and apomixis to be conditioned by a single, dominant Mendelian factor. Significant divergence from the expected 1:1 segregation was observed in sexual x apomictic hybrid progenies. The distorted segregation ratios were attributed to either a partially lethal factor linked to the dominant apospory allele or to a pleiotropic lethal effect with incomplete penetrance of the apospory allele itself. Monogenic dominant inheritance of apospory was subsequently confirmed by the identification of molecular markers completely linked to the apospory phenotype (Martínez et al., 2003).
The discovery of a natural mutant, sexual buffelgrass plant, B-1s, provided the opportunity for genetic studies in this species. Segregation of reproductive mode was assessed in selfed progenies of B-1s and in hybrid progenies produced by crossing B-1s (as female) to two apomictic clones. Reproductive mode phenotype was determined in these early inheritance studies by progeny test (Taliaferro and Bashaw 1966). It was concluded that apomixis was recessive since fully apomictic segregants were found in selfed progenies of the sexual B-1s.
A subsequent study of inheritance of apomixis in buffelgrass (Sherwood et al., 1994), based on different, hybrid-derived sexual material and cytological assessment of reproductive mode in hybrid progenies concluded that apomixis is conditioned by a single, dominant Mendelian factor (either a single gene or several tightly linked genes inherited as a unit). These authors suggest that the B-1s sexual clone used in early genetic studies (Taliaferro and Bashaw 1966) was probably a highly sexual, facultative apomict.
Extensive genetic and breeding studies of apomictic guineagrass were conducted by French researchers in Ivory Coast, using the comprehensive germplasm collections assembled in East Africa during the 1960s (Combes and Pernès 1970). As in bahiagrass, diploid sexual biotypes were used to derive sexual tetraploids, which then were crossed with the natural tetraploid apomicts to generate populations segregating for reproductive mode. Reproductive mode phenotype was assessed by cytological observation of cleared embryo sacs. These were the first genetic studies of apomixis where modern, cytological techniques were used to assess reproductive mode phenotype (Savidan 1975), allowing more precise assessment of the capacity for apospory and the analysis of much larger populations. Clear evidence was found for a dominant Mendelian factor controlling apospory (Savidan 1983). These studies analyzed several types of progenies and propose a simplex genotype for tetraploid apomicts (Aaaa) and a homozygous recessive genotype for sexuals, whether diploid (aa) or tetraploid (aaaa).
The initial genetic study of apomixis in the brachiariagrasses was conducted in Belgium, where a tetraploid sexual biotype of B. ruziziensis was developed from this naturally diploid sexual species (Swenne et al., 1981). The first small hybrid populations were produced by crossing the tetraploidized sexual biotype, as female, to the closely related natural tetraploid apomictic species B. decumbens Stapf (signalgrass) and B. brizantha (A. Rich.) Stapf (palisadegrass) (Ndikumana, 1985). Cytological examination of embryo sacs of these hybrids was consistent with a single, dominant gene conferring apomixis, as for guineagrass. Much more extensive crossing studies using tetraploid sexual B. ruziziensis derived from the Belgian material were subsequently conducted by Valle and colleagues at Embrapa's Beef Cattle Center in Brazil (Valle et al., 1991, 1993a, 1994; Valle and Savidan, 1996). They fully confirmed Ndikumana's preliminary results.
The studies of the inheritance of reproductive mode in the natural aposporous tropical forage grass species (buffelgrass, guineagrass, bahiagrass, and the brachiariagrasses) indicate monogenetic control. The simple genetic model suggests the feasibility of managing apomixis in a breeding program.
Early Breeding of Apomicts
Most of the published proposals for breeding apomicts seem to have a relatively short-term perspective, i.e., by appropriate crosses, given the available germplasm and knowledge in the particular species regarding the inheritance of apomixis, large hybrid populations segregating for reproductive mode are generated in which superior apomictic segregants are identified and directly multiplied for testing, selection, and commercial release. Most remark on the supposed simplicity and rapidity of breeding apomicts, given that inbred lines are not required to produce seed of a successful hybrid cultivar, but often overlooking, it seems, that the major investment of time and effort in any cultivar development program is in finding that best genotype in a population of candidate genotypes, rather than inbreeding to isolate and stabilize the new cultivar or parental lines. For example, even when the new cultivar genotype is pre-existing, as in a collection of germplasm accessions, finding the best genotype and developing it to a successful commercial forage grass cultivar can easily take 15 to 20 yr. Brachiaria brizantha cv. Toledo, the first and so far only cultivar from the germplasm collected in 1984–1985 in East Africa, was released 15 yr later, in 2000 (Argel et al., 2000). An additional candidate cultivar (Piatã) from the 1984–1985 collection is still not commercially available, more than 20 yr after the collection became available.
Almost without exception, the proposals for breeding apomictic tropical forage grasses highlight the advantages of asexual propagation in the utilization of heterosis or hybrid vigor. However, none of these proposals suggests a breeding scheme that will deliberately exploit and accumulate heterotic effects over selection cycles (e.g., Pernès et al., 1975; Hanna, 1995; Savidan, 2000).
Breeding of tetraploid bahiagrass was pursued at Tifton, GA (Burton, 1992), following the early genetic studies of apomictic reproduction (Burton and Forbes, 1960). Selection in a hybrid-derived sexual population led to improved performance, but owing to the very low frequency of apomictic genotypes in the population, the project was dropped (Burton, 1992). Genetic gain from selection in several additional, more or less sexual populations was recorded (Burton, 1992), but apomictic cultivars have apparently not resulted from any of these projects. The failure to produce commercially useful, improved apomicts in these selection programs may be due, in part, to the incomplete information existing at the time on the genetic control of apomixis in tetraploid bahiagrass. It is curious that these breeding activities, covering a period of more than three decades, are apparently undocumented, except for the brief review in the proceedings of the USDA workshop on apomixis (Burton, 1992).
Three cultivars resulting from the early buffelgrass genetic studies were named and released (Bashaw, 1968, 1980a). The first of these (cv. Higgins) was an apomictic selection that resulted from selfing the sexual clone B-1s. The other two were derived from hybrid populations produced by crossing B-1s as female with the existing commercial buffelgrass apomictic clones ‘Blue’ and ‘Common’. None of these bred cultivars achieved significant commercial success owing either to a lack of sufficient agronomic merit to warrant large-scale adoption or to poor seed production (M.A. Hussey, personal communication, July 2006). Simple breeding schemes, including selfing sexual plants or crossing with apomicts, were proposed for buffelgrass based on the early hybridization experiences (Taliaferro and Bashaw, 1966). More recent studies continue to develop tools for buffelgrass breeding (Jessup, 2005; Jessup et al., 2000). However, the only recent cultivar to be released, Frio, originated as a chance off-type plant in an open-pollinated progeny of a facultative apomictic buffelgrass accession (Hussey and Burson, 2005). No recent cultivar resulting from intentional hybridization has been released (Jessup, 2005; B.L. Burson, personal communication, July 2006).
Following the early genetic and breeding studies in Ivory Coast, the French guineagrass germplasm collection, including tetraploid sexual germplasm, was transferred to Embrapa's Beef Cattle Center in Brazil, where guineagrass is a commercially important forage plant. The intention was to initiate a hybridization program to add specific characters to elite apomictic clones selected directly from the germplasm collection (Savidan et al., 1989).
The main thrust of the initial cultivar development efforts in Brazil were logically focused on the selection of useful commercial genotypes directly from the collection of apomictic accessions. No bred cultivar has been released by the Embrapa program (Muir and Jank, 2004). Guineagrass breeding activities have also been underway in Brazil, at the Agronomic Institute of Campinas (IAC), in India, and in Japan (Muir and Jank, 2004). IAC named and released at least two bred cultivars (Centauro and Centenário) and at least one named cultivar (Natsukaze) was released by the Japanese program. It seems that the economic impact to date of guineagrass breeding has not been comparable to that achieved by the simpler strategy of direct selection and commercialization of superior apomictic accessions from comprehensive germplasm collections (Muir and Jank, 2004). The main Brazilian commercial cultivars in terms of seed sales (Tanzânia-I and Mombaça) are direct germplasm selections released by Embrapa. When the existing germplasm collections no longer yield useful cultivars directly and plant breeding is required for continued gains, breeding strategies for guineagrass have been proposed (Pernès et al., 1975; Savidan et al., 1989).
Pernès et al. (1975) proposed a coherent breeding scheme with a view to realizing continuing genetic gains over the long term. The scheme was based on a simple (monogenic) model of inheritance of apomixis, and a thorough study of the patterns of distribution of the morphological variation in natural guineagrass germplasm. Following an initial phase of selection among available diploid sexual germplasm and chromosome doubling, the proposed breeding scheme involves infusion of germplasm and gradual upgrading of the available set of sexual tetraploid genotypes, not by selection and recombination in a sexual breeding population (which would be feasible given the homozygous recessive condition of sexuality found in guineagrass), but by adding selected sexual segregates from sexual x apomictic hybrid populations each cycle to the collection of sexual clones. Fixation of heterosis is mentioned as an advantage of asexual (apomictic) reproduction, and the proposed scheme should produce ever better hybrids. However, any portion of the superior performance of selected hybrid clones that is owing to heterosis will be lost on recombination of selected clones. Intrapopulation selection is unable to accumulate these heterotic effects, and the proposed breeding scheme of Pernès et al. (1975) includes no mechanism to accumulate nonadditive heterotic effects nor deliberately to exploit heterosis by breeding over the long term.
The breeding plans proposed by Savidan et al. (1989) are essentially backcrossing schemes. The authors apparently failed to consider that even when the recurrent parent is heterozygous, repeated backcrossing leads to high levels of inbreeding. Hence, the proposed breeding schemes are unlikely to produce superior heterotic commercial cultivars.
The Belgian team that developed the tetraploidized B. ruziziensis also proposed breeding schemes that utilized this synthetic tetraploid biotype and natural tetraploid apomicts (Gobbe et al., 1983). The two breeding schemes outlined, like those proposed by Savidan et al. (1989) are essentially backcrossing schemes, one to introduce apomixis into sexual B. ruziziensis, the other to introduce sexuality into apomictic B. decumbens. It seems that both schemes, as presented, would quickly lead to high levels of homozygosity.
History of the Brachiaria Improvement Programs
The Brachiaria spp. that are of greatest current commercial value were among the last of the important African forage grasses to be introduced to the American Tropics (Parsons 1972). Indeed, they barely rate a passing mention in Parsons's paper, as being "considered to have great promise" (Parsons, 1972:16). They probably now cover more hectares of improved, sown pasture than the combined hectarage of the six species discussed by Parsons (Miles et al., 2004).
As with other apomictic tropical forage grasses, the early brachiariagrass cultivars are direct selections of unmodified natural germplasm. There are seven of these, in four different Brachiaria spp. (Table 22–2 in Miles et al., 2004). Only one of them, the diploid B. ruziziensis cv. Kennedy, reproduces sexually. Given the apomictic reproduction of the others, cultivar development was essentially a process of reproducing superior genotypes identified in germplasm collections assembled by various institutions around the world (Keller-Grein et al., 1996). The first of these naturally apomictic cultivars to be commercialized in tropical America was ‘Basilisk’ signalgrass derived from a Ugandan accession that reached Australia in 1930 (Oram 1990). Early plantings in tropical America, mainly Brazil, were established with commercial seed from Australia beginning in the 1970s (Santos Filho, 1996). ‘Tully’ [B. humidicola (Rendle) Schweick. (koroniviagrass]) is another natural cultivar commercialized first in Australia and subsequently introduced to tropical America (Keller-Grein et al., 1996). More recent cultivars (Marandu, Llanero, and La Libertad) were introduced to one or another Latin American country initially as germplasm accessions and were selected locally during the 1980s, either by public or private Latin American institutions before their commercialization (Keller-Grein et al., 1996; Miles et al., 2004). In the mid-1980s the International Center for Tropical Agriculture (CIAT, its Spanish acronym) led a germplasm collecting mission focused specifically on the brachiariagrasses (Keller-Grein et al., 1996). To date, this collection has yielded a single commercial cultivar, which has been released in Costa Rica (Argel et al., 2000) and Colombia (Lascano et al., 2002) under the name ‘Toledo’. It is registered in Brazil under three different cultivar names (‘Vitoria MG-5’, ‘Toledo’, and ‘Xaraés’). Three additional accessions from this germplasm collection are in advanced stages of evaluation by Embrapa researchers and future commercial releases are expected.
Cultivar development by direct selection and commercialization of natural germplasm accessions has been very successful. The six natural apomictic cultivars of three Brachiaria species support a seed industry, mostly based in Brazil, whose annual sales a decade ago were estimated to be on the order of US$100 million (Santos Filho, 1996). However, specific deficiencies of these cultivars motivated interest in directed modification of Brachiaria germplasm by hybridization and selection as early as the 1970s (Ferguson and Crowder, 1974).
By the late 1980s, it appeared that the brachiariagrasses had all the attributes necessary for a successful plant breeding program aimed at producing improved apomictic cultivars (Savidan, 2000): Panicum-type apospory (Warmke, 1954) characterized by readily distinguishable monopolar, four-nucleate embryo sacs (Pritchard, 1967); fully sexual tetraploid germplasm, developed by chromosome doubling of a natural diploid sexual (Swenne et al., 1981); simple (monogenic, dominant) inheritance of apomixis (Ndikumana, 1985; Valle and Savidan, 1996), and the availability of a large collection of natural tetraploid apomictic germplasm in species cross-compatible with the tetraploid sexual (Keller-Grein et al., 1996). This germplasm was under advanced agronomic evaluation (Valle et al., 1993b). In addition, the brachiariagrasses had (and have) significant economic importance, particularly in tropical America, and a reasonably straightforward breeding objective could be formulated based on recognized deficiencies in the existing commercial cultivars and results of evaluation of natural apomictic germplasm accessions. The major original motive for initiating brachiariagrass breeding projects in tropical America was to combine resistance to spittlebugs (Hemiptera: Cercopidae), found in apomictic tetraploid B. brizantha, with the superior edaphic adaptation characteristic of the apomictic tetraploid B. decumbens (Miles and Valle 1996).
Several breeding schemes have been proposed for species with simple genetic control of reproductive mode (Pernès et al., 1975; Hanna, 1995). The first breeding proposal to be based on concrete inheritance studies and a well-substantiated genetic model was that of Pernès et al. (1975). However, unlike the brachiariagrasses, guineagrass breeding began with a diversity of sexual germplasm so that the first step was the selection among diploid sexual genotypes before tetraploidization, and the very first cycle of crosses between sexuals and apomicts might be expected to produce useful hybrids.
In contrast, the brachiariagrass breeding programs began, in the late 1980s, with only a single source of sexuality at the tetraploid level—the Belgian tetraploidized B. ruziziensis. This germplasm has a very narrow genetic base (Swenne et al., 1981). Furthermore, it is extraordinarily susceptible to spittlebugs, like B. ruziziensis accessions in general (Cardona et al., 1999) as well as poorly adapted to Al toxicity and soil acidity (Rao et al., 1998), two fatal defects in the Latin American context. In general, the proposals for breeding apomicts seem to assume that merely by crossing a heterozygous apomict with a sexual plant "the genetic diversity in the apomictic (sic) [will be] released and expressed in the F1 progeny" (Vogel and Burson, 2004:80) and commercially useful apomictic hybrids would follow directly. Bashaw et al. (1970) and Savidan (2000) make similar statements. In contrast, we had little expectation of isolating commercially useful hybrids in the early cycles of hybridization, since both commercial apomicts would have to be crossed individually with the agronomically inferior, tetraploid sexual B. ruziziensis. The combination of genes from the two apomicts would require a longer-term breeding effort based on initial recombination of selections from early cycles of hybridization. It seemed that the most efficient means to combine high levels of expression of the desired characteristics, originally dispersed among different apomictic accessions, was to select recurrently in a segregating population containing genes from different parental apomictic genotypes from both commercial species.
We initially contemplated selection in populations segregating for reproductive mode (Miles and Valle, 1996). However, the expense of determining reproductive mode phenotype on large numbers of plants each cycle appeared formidable. Hence, the strategy adopted was to synthesize a fully sexual breeding population from sexual, first cycle hybrids formed by crossing the Belgian tetraploid sexual biotype with a diversity of apomicts (Miles and Valle, 1996; Miles and Escandón, 1997; Miles et al., 2004). Since sexuality was known to be conditioned by the homozygous recessive genotype at a single locus (aaaa), a synthetic tetraploid sexual population would remain fully sexual over generations and could be handled with simple mass or family selection methods. Given that the objective from the beginning was to release only apomictic cultivars, the original idea was to broaden the genetic base of, and subsequently improve the sexual tetraploid gene pool, later to be crossed to one or more elite apomicts to generate hybrid populations from which to extract superior apomictic segregants to be developed to cultivar status (Fig. 1 ; Miles and Escandón, 1997; Valle and Miles, 2001; Miles et al., 2004).
|
However, given that the intended cultivar type is a true breeding, heterozygous clone, the importance of implementing a breeding scheme deliberately designed to exploit heterosis by accumulating heterotic effects was increasingly evident (Miles, 1997; Miles and Escandón, 1997; Valle and Miles, 2001). Two breeding schemes proposed to improve F1 hybrids of sexually reproducing maize were considered: reciprocal recurrent selection (RRS) (Comstock et al., 1949) and recurrent selection on specific combining ability (RS-SCA) (Hull 1945). Both schemes involve selection on the performance of hybrid genotypes and both accumulate over cycles of selection nonadditive genetic effects contributing to heterosis. In RRS, two reproductively isolated, sexual breeding populations are formed. Selection in each population is based on the performance of hybrid families, generally full- or half-sib, formed by crossing individuals in one population with the opposing population. With RS-SCA, selection in the breeding population is based on the performance of testcross families formed by crossing individuals from a single breeding population with a tester (with maize, normally a homozygous line). Both breeding schemes are designed to enhance heterotic effects over cycles of selection.
While Comstock et al. (1949) showed RRS to be more efficient than RS-SCA for loci with partial to complete dominance (rather than overdominance), certain logistical considerations militated against its use, and in favor of RS-SCA, for apomictic Brachiaria. To implement RRS, two sexually recombining populations are required. Given that the objective in breeding an apomictic species is to produce a superior apomictic hybrid, at least one of the breeding populations would have to be segregating for reproductive mode, with the gene(s) for apomixis at a sufficiently high frequency that apomictic hybrids would be found in the hybrid progenies tested each selection cycle, but not so high as to unduly restrict sexual recombination. With a simple and reliable screening tool to distinguish apomictic from sexual genotypes in a segregating population, it might be feasible to manage a population segregating for apomixis. However, such is not yet the case for Brachiaria, where even cytological techniques are time consuming and require trained laboratory personnel, and existing molecular markers are not fully reliable for distinguishing between plants that effectively reproduce sexually or by apomixis (as determined by progeny test) (unpublished data, 2002, 2006), possibly owing, at least in part, to variability in the expression of the apomixis "allele."
RS-SCA was perceived to be much simpler logistically. A single sexually reproducing breeding population would be selected on performance of hybrid progenies formed by crossing each of a series of sexual clones with a single apomictic tester genotype (Miles, 1997). The proposed selection scheme involves a 3-yr cycle (Fig. 2 ). While this is an additional year (3 vs. 2) as compared with the previously used intrapopulation selection scheme (Fig. 1), RS-SCA has the advantage of imposing selection in the sexual population based directly on the performance of hybrid genotypes. An additional advantage is that a cohort of sexual x apomictic hybrids is generated each selection cycle as an integral component of the scheme, that is, population improvement and cultivar development are directly linked (e.g., Baudouin et al., 1997). When the best testcross families are identified, the respective sexual mother plants, maintained as living, perennial plants, are propagated to an isolated crossing block and recombined to regenerate an improved version of the sexual breeding population. Within the best testcross families superior individual plants, each the equivalent of a unique single-cross hybrid, are candidates for development to cultivar status, following determination of reproductive mode (by progeny test or other means) and detailed assessment for all attributes required in a successful commercial cultivar.
|
The choice of a tester genotype was our first important decision in implementing RS-SCA. We considered several aspects: first, an obvious requirement was that the tester be an apomict as this will ensure that testcross progenies contain approximately 50% apomictic hybrids.
While variance among testcross progenies will be greater for a homozygous than for a heterozygous tester (Bernardo 2002), it seems that it would be impossible to achieve complete homozygosis in an apomict, except possibly by anther culture, for which no protocol currently exists for Brachiaria. Successive backcrossing to an apomict would lead to the generation of apomictic clones with high levels of homozygosis (assuming that loss of viability from increasing inbreeding, by itself, did not frustrate the backcrossing program). This may be a productive approach in the longer term to produce a more useful tester genotype.
In our breeding program, an apomictic tester—the highly successful, but spittlebug-susceptible, commercial cultivar Basilisk—was chosen based on observations of hybrid progenies produced by crossing the sexual breeding population with a number of candidate apomictic tester genotypes. Although use of a tester with a high frequency of dominant favorable alleles may not, in theory, maximize genetic gain achieved in the selected population (Bernardo 2002), it seemed a logical choice in the short term in that, in addition to improving the sexual breeding population, we were also interested in direct commercial exploitation of individual hybrids isolated from the testcross progenies. Further, the choice made sense given the perceived complementarity in characteristics between the tetraploid sexual breeding population (high levels of spittlebug resistance, but relatively poor edaphic adaptation) and Basilisk (spittlebug susceptible but excellent edaphic adaptation and strongly stoloniferous growth habit).
The first testcross progenies were formed during 2005, by transplanting a single propagule of each of approximately 500 sexual clones from the sexual breeding population on 5-m centers in a square grid within a uniform field of the tester genotype. Open-pollinated seed harvested from the sexual plants is expected to be mostly hybrids with the tester owing to self-incompatibility of sexual brachiariagrasses (Ngendahayo et al., 1988) that will minimize selfing and to the distance between sexual plants and the intervening tester plants that should minimize crossing between different sexual clones.
The first set of testcross progenies is currently in field trials. We transplanted testcross seedlings to five-plant plots with between one and four completely randomized replicates (5–20 sibs) per testcross family, depending on the number of seedlings obtained. A total of 2250 unique testcross genotypes in 233 families are being evaluated. Following culling on performance at two field sites, selected clones will be assessed for spittlebug resistance, nutritive quality (digestibility and crude protein content), and tolerance to Al (Wenzl et al., 2006). We expect that final family selections can be made by mid-2007, and sexual parental clones recombined between July and December 2007.
Reflections on Breeding Schemes for Asexually Propagated Crops
Aside from the few commercially important seed-propagated apomictic forage grasses, a number of crop species are propagated asexually by vegetative means. These include the potato (Solanum tuberosum L.), the tropical root crop cassava (Manihot esculenta Crantz), sweet potato [Ipomoea batatas (L.) Lam. var. batatas], and sugarcane (Saccharum officinarum L.) and a number of forestry and horticultural tree species. Given that asexual propagation faithfully perpetuates even a highly heterozygous genotype it would seem that asexually propagated crops where sexual recombination can be achieved, would be ideal candidates for reciprocal recurrent selection or some other recurrent selection scheme designed to accumulate nonadditive genetic effects contributing to heterosis and hence, superior performance. In fact, Hull, in his 1945 paper proposing recurrent selection for specific combining ability, explicitly points out that it could appropriately and profitably be applied to an asexually propagated crop using the example of sugarcane.
The nearly complete absence of any consideration of such recurrent selection schemes for asexually propagated, annual crops is very curious. It seems that the only vegetatively propagated species for which reciprocal recurrent selection is being investigated are tree species, either forest or horticultural (Baudouin et al., 1997; Kopp et al., 2001; Pâques, 2004; Lindegaard, 2001).
Possibly the most critical step in initiating a RRS program is identifying or synthesizing the two mutually heterotic populations. In some cases these pools will exist and be recognized or they might be inferred from diversity of geographic origin or genetic markers.
Even in the absence of recognized heterotic pools, it should be fairly simple to set them up. A review of past breeding records would reveal the cross-combination that produced the highest yielding hybrid family. The two parental clones of this cross would be taken as the "founder" genotypes of the two pools. A decision on assignment of additional candidate clones to the two populations would be made as follows: each candidate would be crossed to each of the two founder clones to produce two hybrid (full-sib) families. Yield (or any other criterion of merit) of the two hybrid families would be evaluated. A candidate clone would then be assigned to the pool belonging to the founder clone with which it produced the lower yielding of the two hybrid families. Hence, only 2n crosses would be required to assign n clones rationally to one or the other of the two populations. For any number of candidate clones above n = 5, the scheme proposed would be more economical than evaluating a half-diallel series of crosses. Where 20 candidate clones were available, only 40 crosses would have to be made in the proposed scheme, vs. 190 in a half-diallel.
Once the two base populations are formed, the logistics of conducting RRS would be straightforward by forming and evaluating half-sib or full-sib families. Half-sib families could be formed (e.g., in potato) using bulk pollen. Parents of the best families (maintained vegetatively) would be recombined in their respective population, likely by controlled crosses.
The best individuals in the best families would enter directly into advanced evaluations as candidates for eventual commercial release. Each selection cycle would produce a new cohort of hybrids. With each cycle of selection, the hybrid clones would improve.
|
Summary |
|---|
|
TOPABSTRACTINTRODUCTIONSummaryREFERENCES |
|---|
Apomictic reproduction is the natural mode of reproduction in several economically important, seed-propagated, warm-season forage grasses, including several species of Brachiaria. Inheritance of reproductive mode in the most important of these grasses is simple and readily manageable in a plant breeding program. We have shown that recurrent, intrapopulation selection in a synthetic, fully sexual brachiariagrass population is effective. Selection on spittlebug resistance has produced very high levels of resistance to multiple species. Recent, unpublished data suggest that resistance can be recovered in hybrid populations produced by crossing resistant sexual clones with agronomically superior apomicts, even when the apomictic male parent is susceptible to spittlebugs.
Current breeding programs and nearly all previous proposals for breeding schemes for apomicts, although highlighting the fact that apomictic reproduction will perpetuate heterozygous genotypes, fail to include a means to exploit heterosis.
RS-SCA would appear to be appropriate to breeding apomicts where several conditions occur. While in theory RS-SCA is never superior and may be inferior to RRS, RRS will be logistically impractical until a very simple, high throughput screening methodology exists (e.g., a reliable, linked molecular marker) to discriminate between apomictic and sexual segregants in hybrid populations, or until a means is found to induce (or suppress) apomixis exogenously.
Given the ease of propagating heterozygous genotypes, and hence exploiting heterosis in vegetatively propagated crops, it is curious that (aside from tree breeding) essentially no attention seems to have been given to reciprocal recurrent selection in the common vegetatively propagated field crops such as potato and cassava. A simple scheme to set up the initial base populations for RRS is proposed.
Received for publication April 9, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSummaryREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
