| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a College of Veterinary Medicine, Oklahoma State Univ., Stillwater, OK 74078
b Emergent Genetics, Inc., 6625 Lenox Park Dr., Suite 117, Memphis, TN 38115
c Texas Coop. Ext., Galveston Office, 5115 Highway 3, Dickinson, TX 77539
* Corresponding author (melaniebayles{at}starband.net)
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Abbreviations: ABS, absolute value • NRP, nonrecurrent parent • OV, observed value • RP, recurrent parent
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Backcross breeding is most frequently used to transfer monogenic traits into well adapted cultivars (e.g., see Harlan and Pope, 1922; Briggs and Allard, 1953; Allard, 1960; Verhalen et al., 2003). Though more difficult, it has also been used to transfer quantitative traits of moderate to high heritability (e.g., see Briggs and Allard, 1953; Allard, 1960; Knott and Talukdar, 1971; Duvick, 1974). Briggs and Allard (1953) point out that backcrossing to improve quantitative characters "is limited only by the ability of the plant breeder to select for a worthwhile intensity of the character." Reddy and Comstock (1976) used computer simulations (coupled with quantitative genetics theory) to estimate the effects of heritability and gene number on the backcross method. They defined effectiveness as "the probability of fixation ... of favorable alleles derived from the donor line" and found that it was higher when heritability was higher, but not as high as they had expected. The number of favorable alleles in the NRP had a greater impact on breeding success than did heritability. They derived substantial progress with heritability as low as 15% and with favorable alleles ranging from one to 16 in number.
Backcrossing in cotton was apparently first used in the development of ‘Griffin’ (released in 1867), over 50 yr before geneticists [i.e., Harlan and Pope (1922)] showed the method was scientifically sound for plant improvement (Ware, 1936). At the time, ‘Green Seed’, an older upland, was crossed with ‘Sea Island’ (Gossypium barbadense L.) and then backcrossed to Green Seed several times to produce an essentially green-seeded upland cotton with the long, fine fiber of Sea Island. Jenkins and Harrell (1950) described several case histories where backcrossing was used successfully in cotton to improve targeted characters. In addition, they described their own use of the method to transfer desirable fiber properties from Sea Island into upland cotton. Meredith (1977) used backcross breeding to derive improved combinations of lint yield and fiber strength in cotton. Backcross populations were not equal in strength to the NRP, but a satisfactory level (94% of the NRP) was maintained through three backcrosses. Lint yields of the Bc1F5, Bc2F4, and Bc3F3 were 78, 87, and 88% of the RP, respectively. Vroh Bi et al. (1999) developed several "low-gossypol seed and high-gossypol plant" cotton germplasms from trispecies hybrids using the backcross method. Employing amplified fragment length polymorphism techniques, they determined that an 80% genetic similarity existed between upland cotton and their Bc3 plants. Genetic similarity between the wild species used in the original hybrids and upland cotton ranged from 30 to 43%. The backcross method was used to develop all of the initial transgenic cotton cultivars with Bollgard (Monsanto Co., St. Louis, MO), Roundup Ready (Monsanto Co., St. Louis, MO), or both traits (released in 1996, 1997, and 1997, respectively) and almost all others since that time (Verhalen et al., 2003).
The backcross method has been used extensively in the past to transfer bacterial blight resistance in cotton. Knight (1945) used it in Africa to transfer four genes for blight resistance from three species into two G. barbadense cotton cultivars along with several gene transfers into G. hirsutum. He suggested that the technique could also be used to introduce genetic variability into existing cotton types with the eventual development of new cultivars. Bird (1960)(1962) developed upland cotton lines with immunity to bacterial blight by backcrossing G. barbadense with major resistance genes onto an upland line with an intermediate level of polygenic resistance. After each backcross, the segregating populations were screened with a mixture of races of the causal organism; and the resistant plants were then backcrossed to the RP. Immune plants were not detected until after several backcrosses. On the basis of their similar experience with bacterial blight resistance vs. immunity in cotton, Brinkerhoff et al. (1984) suggested that immunity in any disease–crop complex could be developed by a backcrossing program when combined with rigorous screening. They stated that the success of such an endeavor depended on the availability of pathogen genotypes which would permit consistent screening in segregating populations and on the proper choice of parental material. Ideally, the NRP should possess two or more qualitative genes for resistance while the RP should offer polygenic resistance.
While working in the cotton breeding program at Oklahoma State University, Samayoa-Armienta (1974) characterized 53 traits, including bacterial blight resistance, among 31 foreign and eight U.S. upland cultivars. Six cultivars from Africa in that study showed an appreciable level of resistance to bacterial blight, and each was subsequently used as a NRP in a backcrossing program to transfer its resistance into a blight-susceptible cultivar previously developed in Oklahoma. Remnant seed from those six family groups were used herein to study the backcross method itself through four backcrosses. The objectives of this study were to measure the degree and rate of recovery of RP traits through four backcross generations in upland cotton as well as to determine reactions to three diseases (including blight) in the Bc4F4 generation.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
|
|
Lint yield was determined from weights of hand pulled cotton (equivalent to machine stripped cotton) harvested from each plot, converted into kilograms of lint per hectare. Before harvest, 15 mature bolls were randomly sampled from each plot in each test primarily to measure fiber properties. The samples were ginned on an eight-saw laboratory-type gin, and the lint was processed in the Cotton Quality Research Laboratory then at Oklahoma State University. In the laboratory, 2.5 and 50% span lengths were determined on the digital fibrograph and converted into millimeters. Uniformity index was computed as the ratio of 50 to 2.5% span length and expressed as a percentage. Fiber fineness was estimated on the micronaire in standard curvilinear micronaire units. Fiber strength was measured on the stelometer using 0.0-mm (0-inch) gauge (T0) and 3.175-mm (1/8-inch) gauge (T1) measurements, converted into kilonewton meters per kilogram.
From various weights, measures, and counts derived while ginning the boll samples, seven agronomic characters were estimated. Picked lint percentage was calculated as lint weight converted into a percentage of seedcotton weight; pulled lint percentage as lint weight converted into a percentage of the combined weights of seedcotton and bur; boll size as the weight of seedcotton in grams per boll; bur size as weight of the empty bur in grams per boll; lint weight per boll in grams; lint index as weight of lint in grams from 100 seed; and seed index as weight of 100 fuzzy seed in grams.
Measurement of Disease Reactions in the Bc4F4
The six Bc4F4 populations and three check cultivars [Westburn 70—the RP, ‘Westburn M’ (Oklahoma Agric. Exp. Stn., 1976), and ‘Deltapine 55’] were grown in randomized complete-block experimental designs on the Plant Pathology Research Farm, Oklahoma State University, Stillwater, OK, in 1981 and 1982 and on the Agronomy Research Station near Perkins, OK, in 1983 to measure disease reactions. The soil types at those locations were a Norge loam (fine-silty, mixed, active, thermic Udic Paleustoll) and a Teller loam (fine-loamy, mixed, active, thermic Udic Argiustoll), respectively. Three replications were used in 1981 (Stillwater), four in 1982 (Stillwater), and three in 1983 (Perkins). Plots were single rows, 6.7 (Stillwater) or 11.0 (Perkins) m long, and 1.0 m apart. After emergence, seedlings within rows were thinned to approximately 15-cm intervals. The Stillwater plots received frequent overhead sprinkler irrigations to encourage disease development while the Perkins plots only received one irrigation because the material was planted relatively late.
Responses to a mixture of Races 1, 2, 7, and 18 of the bacterial blight causal organism (which collectively overcome all known individual genes for blight resistance) were determined in 1981 and 1983 after artificially inoculating plants at the six to eight true-leaf stage. The inoculum was an aqueous suspension containing approximately 5.0 x 105 viable bacterial cells per milliliter. It was applied to the abaxial side of leaves with a single-nozzle gun attached to a power sprayer at a pressure of 1.4 to 2.1 x 106 Pa. Fourteen days after inoculation, individual plants were scored for their disease reactions with the 0.0 (immune) to 4.0 (fully susceptible) grading system described by Brinkerhoff (1963). Those grades were converted into a whole-number scale of 0 (for his 0.0 grade), 1 (for 0.1), 2 (for 0.2), 3 (for 1.0), 4 (for 1.2), 5 (for 2.3), and 6 (for 4.0) before analyses using plot means. All plants within a replication were graded for blight resistance.
The soil in the area where these experiments were grown at Stillwater was highly infested with the causal organism for Verticillium wilt. The first 20 plants in each plot were scored for their Verticillium wilt responses in 1981 and 1982. Plants were evaluated in mid-October on the basis of their gross external symptoms and on vascular discoloration in cut stems of those plants without external symptoms. Grades were assigned by the 1 (no visible leaf symptoms; no vascular discoloration in stems) to 10 (defoliated; stems dead down to ground level) scale utilized by Verhalen et al. (1971a). Analyses of this trait were also based on plot means.
The six Bc4F4 populations, Westburn 70, and Westburn M were evaluated for resistance to the Fusarium wilt–root-knot nematode complex as part of the 1981 and 1985 then Regional (now National) Cotton Fusarium Wilt Testing Program at Tallassee, AL (Kappelman, 1982; Johnson and Williams, 1985). The soil type at that location is a Cahaba loamy fine sand (fine-loamy, siliceous, semiactive, thermic Typic Hapludult). The experimental design for each contributor in these tests is a randomized complete-block with four replications and with the center two rows in each replication occupied by a susceptible check (‘Rowden’) and a resistant check (‘McNair 235’). Wilting percentage was determined in each plot.
Statistical Analyses
Because the crucial comparisons for lint yield, the fiber properties, and the agronomic characters were made within family groups and not between them, each family group was treated as a separate experiment arranged in a randomized complete-block design. Our interest was in making an overall assessment of the backcross method, rather than in detailed analyses of individual traits. Therefore, analyses of variance for each trait within each family group were conducted over the seven replicated experiments. F tests for each source of variation were performed with the appropriate error term, assuming a random model. For a trait within a family group to be included in subsequent assessments, two criteria were established. One, the generations mean square in its analysis had to be significant at the 0.05 probability level to demonstrate that at least some of the differences observed in that data set were probably real. Two, a protected LSD test (also at the 0.05 level) had to indicate a significant difference between the NRP and the RP to permit the study of mean shifts from generation to generation with repeated backcrossing. When both criteria were fulfilled, all generation means for that trait in that family group were compared with the RP by the protected LSD. To this point, all traits were expressed in their original units of measurement.
For some characters, the NRP exhibits the more desirable performance; for others, the RP is superior. Shifts over generations of backcrossing are toward the RP, regardless of its superiority or lack thereof. Some characters are measured in the same units as others (e.g., percentages); others are in different units. When comparing trends in different parental combinations and/or traits, such complications are avoided by setting the RP for each trait in a family group equal to 1.00 and expressing all other generation means as decimal fractions relative to it. The significant differences from the RP (derived from the original units) were retained.
A linear regression was used to compare the observed vs. the theoretical rate of recovery of RP traits over experiments, family groups, and traits. The theoretical model predicts 0.50, 0.75, 0.875, 0.9375, and 0.96875 for the average recovery of RP genes in the F1, Bc1F1, Bc2F1, Bc3F1, and Bc4F1 generations, respectively (Fehr, 1987). In these studies, the F4 (not the F1), the Bc1F4 (not the Bc1F1), etc., were used. Also, during the development of these materials, intentional selection was practiced only for bacterial blight resistance (Table 1), not for any of the characters used in this regression. Admittedly, that selection could have indirectly influenced these traits by linkage and/or pleiotropy. To convert the observed data (which conformed to the two criteria previously described) to a common scale with the above theoretical expectations, the means over experiments for each trait, generation, and family group were adjusted by the following formula:
 |
Analyses of variance were combined over years for each disease reaction. In each case, if the entry x year interaction was not significant, means over years were utilized; if significant, analyses were conducted separately for each year. F tests for each source of variation were performed at the 0.05 probability level with the appropriate error term, assuming a mixed model with entries considered as fixed and years as random.
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Lint Yield
Significant differences in lint yield were detected among generations and between the parents in each of the family groups studied (Table 3). In Oklahoma over the seven environments, the NRPs were very low in lint yield relative to the RP, ranging from 29 to 51% for ‘AH(67)M’ and ‘HL-1’, respectively. All F4 populations were significantly lower yielding than the RP, as were three Bc1F4 and two Bc2F4 populations. Only one Bc3F4 mean (in the ‘HG9’ family group) differed in lint yield from the RP, and it was 20% higher! No statistical differences in yield were detected between the Bc4F4 and the RP in any group. Thus, in these experiments, the fourth backcross provided no significant "added value" in lint yield. In retrospect, that should not have been a surprise because on the average the Bc3 and Bc4 theoretically resemble the RP by 93.75 and 96.875%, respectively (Fehr, 1987), a difference of only 3.125%. In our experience, field experiments in cotton are incapable of discriminating between differences in lint yield that small (e.g., Bayles et al., 2003, 2004). Precision can be increased, but certain costs would be entailed to do so. For example, increasing the number of replications per experiment from the four utilized herein to 16 would double precision (Little and Hills, 1978) but would quadruple the work involved.
|
Within the six family groups, only one backcross population displayed significant transgressive segregation for lint yield, i.e., performance significantly outside the parental range. The Bc3F4 population in the HG9 group mentioned earlier yielded 20% more than the RP over all seven experiments. In six of those seven experiments, it outyielded the RP by 10, 14, 15, 27, 56, and 60%; but it was 20% lower yielding in the seventh. This degree and consistency of yield increase leads us to conclude that the difference over experiments was real and not a Type I error. Clearly, that population should be investigated further with eventual release as a possibility. The backcross of that Bc3F4 to the RP resulted in a Bc4F4 closer in yield to the RP. An additional backcross is normally considered a desirable action; but in this particular family group, it had a detrimental effect on yield.
Fiber Properties
Six fiber properties were studied in each of the six backcross family groups. In 30 of those 36 combinations, significant differences were detected between the NRP and the RP (Table 3). ‘CA(68)36’ and the RP did not differ for micronaire. ‘BJA 592’ and the RP were similar for 2.5 and 50% span lengths and for T0 and T1 as were HL-1 and the RP for uniformity index. Those six family group–trait combinations were eliminated because if the endpoints of a progression from the NRP toward the RP cannot be demonstrated to differ statistically, an examination of the intermediate steps between them would lack conviction. The remaining 30 combinations were examined in detail.
Some NRPs were superior in certain fiber properties to the RP; e.g., 2.5% span length in AH(67)M as indicated by its significant decimal fraction >1.00 (Table 3). Others were inferior in certain traits to the RP such as the same trait in HG9 as denoted by its significant fraction <1.00. Regardless of the starting point in the NRP for any character, the progression with successive generations of backcrossing is expected to be a closer and closer approximation to the 1.00 value assigned to the RP. The number (and percentage) of significant differences from the RP in the 30 combinations involving fiber properties in the F4, Bc1F4, Bc2F4, Bc3F4, and Bc4F4 were 23 (77%), 19 (63%), 13 (43%), 11 (37%), and 5 (17%), respectively. Neither fiber length measurement nor uniformity index was significantly different from the RP in the corresponding Bc4F4. The same was true for 50% span length in all Bc3F4 populations. One of five Bc4F4 vs. RP comparisons was significant for micronaire as were four of 10 for fiber strength.
Five cases of transgressive segregation were noted in the backcross generations for fiber properties. Three were losses of 2 to 3% in 2.5% span length relative to the RP. One was an increase of 2% in uniformity index relative to the NRP; the last was an increase of 5% in micronaire compared with the NRP.
Agronomic Characters
Seven agronomic characters were examined in each backcross family group. In 33 of those 42 combinations, the NRP and the RP exhibited significant differences (Table 4). Neither CA(68)36 nor HG9 differed from the RP in bur size. BJA 592 and the RP were similar for picked and pulled lint percentages, for bur size, and for lint and seed indexes. HL-1 and the RP were like one another for pulled lint percentage and seed index. Those nine family group—trait combinations were discarded on the basis of the same rationale as in the previous subsection. The 33 remaining combinations were scrutinized further.
|
Numerous instances of transgressive segregation were detected in the backcross generations among the agronomic characters. All were positive relative to the RP. Six values were noted for picked lint percentage (ranging from 2–5%), five for pulled lint percentage (from 2–5%), five for boll size (from 6–13%), one for bur size (18%), four for lint weight per boll (from 12–18%), five for lint index (from 5–7%), and one for seed index (12%).
Assessment of Degree of Recovery over All Recurrent Parent Traits
In this study of an adapted RP and six unadapted NRP cultivars, no "magic" number of backcrosses resulted in the recovery of all RP traits. Lint yield and 50% span length were recovered in all family groups with three backcrosses. No statistically detectable improvement in either trait was obtained with a fourth backcross. With the fourth backcross, 2.5% span length, uniformity index, and pulled lint percentage were recovered in all family groups. One-third of the Bc4F4 micronaire and fiber strength readings were not yet equal to the RP over all family groups as were 31% of the other agronomic characters. If such differences are unacceptable to the breeder, additional backcrosses would still be required. When RP vs. NRP differences are greater than were used herein, the number of backcrosses required would likely be greater still (Isleib, 1999). Backcrossing is the plant breeding method used almost exclusively to develop transgenic cotton cultivars (Verhalen et al., 2003). The number of backcrosses used in those efforts is not always shared with others; but when given, the number is commonly three (Sheetz and Speed, 1997) or four (Sheetz, 1998).
Thirty-three instances of transgressive segregation were detected in the backcross generations in this material. Only three were in the Bc4F4 populations. Similar results may have been overlooked in the past because breeders, including these authors, usually conduct a backcross program and then test the end results of that program, ignoring populations developed during earlier backcrosses. Only if transgressive segregants are detected, can they be selected and utilized.
Rate of Recovery of Recurrent Parent Traits
The second objective of this study was to measure the observed rate of recovery of RP traits relative to the theoretical rate through four backcross generations. Their respective regression lines, formulas, and coefficients of determination are shown in Fig. 1. The theoretical rate was based on the percentages, converted to decimal fractions, provided by Fehr (1987). Observed values for the F4 and each backcross generation were derived for the 69 family group—trait combinations reported in Tables 3 and 4. The difference between the NRP and the RP served as the reference point for each combination while the difference between the NRP and the generation in question measured the degree of change in that combination to that point. Absolute numerical values were used because the relative "economic worth" of the NRP and RP for a combination was deemed irrelevant for this purpose. Means in all cases tended to shift away from the NRP toward the RP. The observed means over the 69 estimates for the NRP, F4, Bc1F4, Bc2F4, Bc3F4, Bc4F4, and the RP were 0.00 (by definition), 0.69, 0.83, 0.90, 0.96, 0.96, and 1.00 (by definition), respectively. The observed rate of recovery was only 4% less efficient than the theoretical rate. It accounted for 96% of the variation observed. An alternative regression that forced the intercept to pass through the origin resulted in a regression coefficient of 1.04 and an R2 of 0.95.
|
All six Bc4F4 populations were statistically more resistant to bacterial blight than was Westburn 70, the RP (Table 5). BJA 592 and HG9 had the highest overall levels of resistance among the cultivars used as NRPs in this study, followed by HL-1 (Table 2). The Bc4F4s in which BJA 592 and HL-1 were the NRPs were the most resistant to blight in both years they were tested (Table 5). CA(68)36 and ‘SATU 65’ had the lowest levels of blight resistance among the NRPs (Table 2); and likewise, the backcross populations derived from them tended to have the lowest levels of resistance (Table 5).
|
Family groups consisted of a different NRP (a cotton cultivar from Africa resistant to bacterial blight), the same RP (Westburn 70, susceptible to that disease), and the F4, Bc1F4, Bc2F4, Bc3F4, and Bc4F4 populations between them. Six such family groups were characterized for lint yield, six fiber properties, and seven agronomic characters, a total of 84 sets of comparisons. The NRPs differed significantly from the RP in 69 of the 84. The degree of recovery of RP traits was examined by statistically comparing the generations studied with the RP. Among those 69 combinations, the number (and percentage) of significant differences from the RP in the F4, Bc1F4, Bc2F4, Bc3F4, and Bc4F4 were 47 (68%), 35 (51%), 25 (36%), 25 (36%), and 14 (20%), respectively. No significant differences were detected between any Bc4F4 and the RP for lint yield, 2.5 and 50% span lengths, uniformity index, and pulled lint percentage. One or more such differences were noted for all other traits. The fourth backcross did not improve lint yield or 50% span length. The observed rate of recovery of RP traits was 4% less overall than the theoretical rate. The only intentional selection applied during the development of these materials was for bacterial blight resistance. The level of blight resistance in the Bc4F4 tended to reflect that of its NRP. The level of tolerance to Verticillium wilt and of resistance to the Fusarium wilt–nematode complex in the Bc4F4 tended to reflect that of the RP with three possible exceptions.
|
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Received for publication October 27, 2004.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Related articles in Crop Science:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
