Published online 2 December 2005
Published in Crop Sci 46:1-5 (2006)
© 2005 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
CROP BREEDING, GENETICS & CYTOLOGY
Early Generation Testing in Upland Cotton
Darren G. Jonesa,* and
C. Wayne Smithb
a Monsanto, 4006 Old Leland Rd., Leland, MS 38756
b Dep. of Soil and Crop Science, Texas A&M University, 2474 TAMUS, College Station, TX 77843-2474
* Corresponding author: darren.g.jones{at}monsanto.com
 |
ABSTRACT
|
|---|
Early generation testing (EGT) is often used to identify segregating populations that are expected to contain the greatest frequency of favorable genotypes and to eliminate inferior populations with limited promise. The goal of this study was to determine if EGT methods utilized in cotton (Gossypium hirsutum L.) at College Station (CS) and Weslaco (WS), TX, since 1990 and 1992, respectively, and through 2002, could be used to predict advanced strain performance. Data were collected from 283 unique F2 populations at CS and 146 unique F2 and F3 populations at WS. Correlations of F2 population yield and individual plant selection fiber data suggests that the method used at CS limits the use of EGT for lint yield, yet enhances the use of EGT for the highly heritable fiber traits. Conversely, correlations of F2 population yield and population fiber data suggest that the method used at WS supports EGT for lint yield, but limits the use of EGT for fiber traits. However, by using F3 population fiber yield means and individual plant selection fiber data, EGT is validated at WS. Also, correlations of F2 yield and fiber data with combined yield and fiber data of advanced strains in standard performance trials across multiple locations in central and south Texas indicated that EGT reasonably predicted advanced generation performance.
Abbreviations: CC, Corpus Christi • CH, Chillicothe, CH • CS, College Station • DL, Dallas • DP50, Deltapine 50 • EGT, early generation testing • TH, Thrall • UV, Uvalde • WS, Weslaco
|
INTRODUCTION
|
|---|
THE PRACTICE of breeding cotton as a commercial enterprise was instituted in the early 1900s (Bowman, 2000) in the United States. This led to the development of cotton breeding programs within both the private and public sectors. These programs utilize different breeding schemes to develop germplasm, not only to improve yield but fiber quality characteristics also. During the past century of scientific plant improvement in upland cotton, breeders have sought to identify the most efficient generation advancement procedures.
One such method is EGT, first illustrated by Immer (1941), to select exceptional populations for generation advancement. Breeding efficiency within a program can be improved if the potential of future progeny can be ascertained during early generations (Barut, 1998). The method of EGT is used to identify segregating populations that are expected to contain the greatest frequency of high yielding genotypes and to eliminate inferior populations that show limited promise.
Success with EGT relies on the ability of the breeder to distinguish among genotypic families in early generations and the persistence of these contrasts in later generations (Hegstad et al., 1999). Ntare (1999) stated "if performance of progenies in early generations accurately reflects the genetic potential of the cross, then identification and selection of superior crosses in early generations will allow the breeder to increase the number of selections per cross retained, thereby increasing the probability of identifying superior lines," which is important when resources such as land, money, and time are limited and numerous lines must be evaluated in multiple generations and multiple locations.
Minimal information is available about EGT in cotton although it is practiced within the cotton breeding community (Bowman, 2000). Using 30 hybrid populations re-created for the purpose of comparing F2 bulk testing with pedigree selection, Percy (2003) reported that a poor association of F2 population traits was measured when correlated to the pedigree selection records. Barut (1998) found correlations for yield ranging from 0.49 to 0.69 between F2, F3, and F4 generations and their F4:5 lines. In examining F2 and F3 population means from a five-parent diallel, Meredith and Bridge (1973) reported a nonsignificant correlation of 0.48 between the generation means for lint yield. They did report, however, significant correlations between the F2 and F3 populations for lint percentage, seed index, and fiber length, strength, and elongation.
Reports of EGT for both yield and quality exist in many other crops, but results vary. In testing F1–, F2–, and F3–derived soybean [Glycine max (L.) Merr.] families, St. Martin and Geraldi (2002) recommended testing F2–derived families to maximize genetic gain for yield unless the use of off-season nurseries permits the advancement to the generation without disruption. Boerma and Cooper (1975) compared effectiveness of EGT with the pedigree method and the single-seed descent method in soybean. They found that these methods did not give reliable differences in effectiveness. Brennan and O'Brien (1991) demonstrated that EGT led to a faster rate of progress for quality in wheat (Triticum aestivum L.), but a lesser rate of progress for yield as compared with a program without EGT. Halward et al. (1990) and Wynne (1976) concluded that EGT in peanut (Arachis hypogaea L.) for yield has limited value, but EGT for quality traits appeared to be useful. However, after consecutively testing populations in the F2, F3, F4, F5, and F6 generations, Coffelt and Hammons (1974) suggested that EGT might be an acceptable breeding procedure for evaluating and selecting for yield in peanut.
Early generation testing methods have been employed in College Station and Weslaco, TX, since 1990 and 1992, respectively, by the Cotton Improvement Laboratory at Texas A&M University. The goal of this study was to determine if lint yield and fiber quality of early generations are associated with advanced strain performance.
|
MATERIALS AND METHODS
|
|---|
Data from segregating populations developed from 1990 through 2002 at CS and from 1992 through 2002 populations developed in WS were used to evaluate EGT procedures used by the Cotton Improvement Laboratory at Texas A&M University. The EGT at CS was a standard EGT procedure where lint yield and quality of the F2 generation was used to select populations to advance for pedigree selection in later generations. The F2 populations were planted in two-row plots, 1 by 12 m, with four replications, in a randomized complete block design. One row from each plot was harvested with a spindle picker, modified for plot harvest, to obtain yield estimates and fiber samples. Grab samples of the machine-harvested seed cotton from the first and third replications were ginned on a laboratory saw gin and High Volume Instrument (HVI) fiber properties determined by the Texas Tech International Textile Research Center at Lubbock, TX. Initial selections for advancement were based on average yield, followed by additional selection for HVI fiber characteristics. Seed of each F2 population for the following F3 nursery were harvested from a separate planting each year. The following growing season, the F3 seed was planted to a spaced nursery, with plants thinned to approximately one per 45 cm. Individual F3 plants were selected visually for yield potential and subsequently reselected based on HVI fiber characteristics for advancement to the F3:4 generation. The F3:4 selections were planted the next growing season as progeny rows, and individual row selections were based on visual yield potential assessments and reselected on HVI fiber characteristics. The F3:5 strains were planted the following season in progeny rows with two replications at one or more locations. Again, progeny rows were selected on their average visual performance at all locations and reselected on fiber characteristics to advance to the F3:6. Once strains reached the F3:6 generation, they were evaluated in standard performance trials at multiple locations that resembled the aforementioned F2 trials.
The modified EGT used at WS began with a four-row, four-replication F2 performance test, planted in a randomized complete block design. One internal row of each plot was not thinned to individual plant configuration and was harvested with a modified spindle picker to obtain population yield and fiber quality estimates. The harvesting of one of the interior rows eliminated or reduced potential border effects. Fiber properties were determined from seed cotton grab samples as described previously. Individual spaced F2 plants were selected based on earliness of maturity and visual yield potential from the remaining three rows per plot and bulked to obtain seed for the F3 generation. Individual F2 plant selections were also based on morphological traits, for example, leaf shape, when appropriate for that population. The following season, the F3 populations, after reselection on F2 population fiber HVI characteristics, were planted in a randomized complete block design with four-row plots and four replications. Again, one unthinned interior row from each plot was harvested with a spindle picker to obtain yield estimates and fiber properties as described above. Individual spaced F3 plants were selected from the remaining three rows of each plot to produce F3:4 progeny. These F3:4 individual plant selections, based on apparent yield potential, earliness of maturity, and morphological traits when appropriate, were reselected for HVI fiber characteristics. The remainder of the pedigree selection and strain evaluation process for the WS EGT procedure was the same as described above for the CS EGT.
Fiber parameters were determined by HVI. These parameters included micronaire reading, upper half mean fiber length, fiber strength, fiber elongation, and length uniformity index. Micronaire reading is an estimate of fiber fineness and fiber maturity, measured by passing pressurized air through a standard volume of a standard weight of fiber (Anthony, 1999). Upper half mean length is the average length of the longest one-half of the fibers by weight in the sample. Fiber strength measures the amount of force (kN m kg–1) required to break a bundle of fibers of a specified weight. Fiber elongation is a measurement determined during the assessment of fiber strength by measuring the extension of the fibers before break time, and its value in predicting textile performance has yet to be fully understood. Uniformity index is the ratio of mean fiber length to upper half mean length (Anthony, 1999).
Commercial cultivars were included within each generation for comparative purposes within both EGT methods. In general, the commercial cultivars were planted every 20th row for ease of visual comparison when the populations were planted to space-planted or progeny-row nurseries. Checks were included in the randomized complete block design used in the F2 and/or F3 tests and later generation standard performance trials. Individual plant fiber data from the commercial cultivar Deltapine 50 (DP50) were used to check the variability of fiber across years at CS and WS. Yield data were not collected from DP50 individual plants.
To measure the efficiency of EGT, F2 or F3 lint yields and fiber characteristics at CS and WS were correlated with their respective lint yields and fiber characteristics in later generation standard performance trials at CS, WS, Corpus Christi (CC), Thrall (TH), Uvalde (UV), Dallas (DL), and Chillicothe (CH), TX.
|
RESULTS AND DISCUSSION
|
|---|
Mean squares for individual plant fiber characteristics from the commercial cultivar DP50 indicated differences among years for micronaire reading, fiber length, fiber strength, fiber elongation, and uniformity index at CS from 1991 to 1999 (Table 1) and differences among years for lint percentage, micronaire reading, fiber length, fiber strength, fiber elongation, and uniformity index at WS from 1994 to 1999 (Table 2). Such data point out a major obstacle of EGT, namely that populations are selected on the basis of performance from a single year while subsequent selections must be evaluated across multiple years to validate their merits. However, breeders must devise methodology such as EGT to help reduce the number of populations or inbred lines carried forwarded in pedigree systems. The issue with EGT is whether or not data from a single location and year are associated with the performance of inbred lines in later generations. The study reported herein attempts to determine if such correlations exist in two sets of multiple year data, one from WS where maximum selection pressure was applied for yield potential with secondary selection on acceptable fiber quality and the other from CS where the primary selection pressure was for fiber quality rather than yield.
View this table:
[in this window]
[in a new window]
|
Table 1. Mean squares for individual plant lint percentage (LP), micronaire reading (MIC), length (LG), strength (STR), elongation (EL), and length uniformity index (UI) from the commercial cultivar DeltaPine 50 (DP50) rom 1991 to 1999 at College Station, TX.
|
|
View this table:
[in this window]
[in a new window]
|
Table 2. Mean squares for individual plant lint percentage (LP), micronaire reading (MIC), length (LG), strength (STR), elongation (EL), and uniformity index (UI) from the commercial cultivar DeltaPine 50 (DP50) from 1994 to 1999 at Weslaco, TX.
|
|
Correlations of F2 population lint yields with F6, F7, and F8 inbred line lint yields were determined at CS between 1990 and 2002 (Table 3). F2 lint yields were negatively correlated with F6 lint yields (r = –0.36; P < 0.01) but positively correlated with F8 lint yields (r = 0.51; P < 0.01) at CS. The change from a negative correlation for the F2:6 populations to a positive correlation in the F2:8 populations, suggests that EGT was not effective in identifying yield potential when the primary selection pressure was for fiber quality and secondary selection pressure in both the F2 and inbred lines was for lint yield at CS. However, lint percentages of F2 populations were positively correlated with F6 and F7 lint percentages, 0.36 and 0.39, respectively, but not F8 line lint percentages, although a positive correlation value was obtained. Lint percentage tended to be less affected by environment than yield or fiber quality traits (Table 1). Improvements in fiber bundle strength was a major goal of the Texas Agricultural Experiment Station program at CS during the 1990s and is reflected in the correlations for strength between the F2:6, F2:7, and F2:8 generations. The F2 populations were correlated with F6 (r = 0.24, P < 0.05) and F7 (r = 0.57, P < 0.01) line performance, while the F2:8 generations were weakly correlated (r = 0.28, P < 0.11). Although considerable pressure was applied to the selection of F2 populations and to derived F3 plant selections and subsequent progeny rows, EGT appears to be of little value in identifying populations that will enhance the probability of improving fiber length. Our findings for micronaire reading should be regarded with caution because selection pressure is only applied to keep it within the acceptable limits of 3.5 to 4.9. Other fiber traits (i.e., elongation and uniformity index) were not selection criteria although the authors concede that selection for upper half mean length can impact both of these measurements.
View this table:
[in this window]
[in a new window]
|
Table 3. Correlation coefficients for F2 lint yield and fiber properties and F6, F7, and F8 mean lint yield and mean fiber properties at College Station, TX, from 1990 to 2002.
|
|
Correlations of F2 and F3 population yields with F6, F7, and F8 yields across all years at WS were significant (P < 0.01) for each generation (Table 4). The correlations of 0.47, 0.59, 0.41, and 0.62 for F2:6, F2:7, F3:6, and F3:7, respectively, show that F2 and F3 yields under the WS selection method could reasonably predict yield in later generations. Although the correlations of F2:8 and F3:8 of –0.62 and –0.54, respectively, are troubling, overall these data were encouraging in that performance trials in four of six comparisons revealed positive and significant correlations of F2 and F3 lint yields with the lint yield of subsequent strains developed from those populations at WS. The negative correlations at the F8 generation are not easily explained but could reflect a limited range in yields among a smaller set of F8 lines compared with a wider range in yields among early generation populations.
View this table:
[in this window]
[in a new window]
|
Table 4. Correlation coefficients for F2 lint yield and fiber properties and F6, F7, and F8 mean lint yield and mean fiber properties, and F3 lint yield and fiber properties and F6, F7, and F8 mean lint yield and mean fiber properties at Weslaco, TX, from 1992 to 2002.
|
|
Selections at WS were advanced from the F2 primarily based on their yield and secondarily on their fiber quality. The small correlations between the F2 to F6, F7, and F8 generations relative to the F3 population for fiber properties confirmed that yield was the major factor in determining advancement of F2 generation populations at WS (Table 4). The F3 to subsequent generation correlations at WS were generally larger and a greater number of the correlations were significant for the corresponding F2 correlations, since the F2 used the population mean, while the F3 was the individual plant that gave rise to subsequent generations. These correlations were similar to those obtained at CS (Table 3) and thus support that EGT identified populations from which advanced lines were extracted with improved fiber strength, but with less consistent results relative to the use of EGT for other traits.
Performance of F6, F7, and F8 strains were averaged by generation across multiple locations throughout central and south Texas and correlated with EGT data from CS or WS (Table 5). By averaging lint yield performance from CS, WS, CC, and TH, significant correlations were detected between the F2 populations and average performance of their F6 progeny. Similar results were found when utilizing F2 populations and their respective F8 progeny (r = 0.60, P 
0.01) across seven locations. However, a nonsignificant correlation (P = 0.36) was detected between the F2 populations and the average performance of their F7 progeny evaluated at CS, WS, CC, UV, TH, DL, and CH. The lack of correlation between the F2 and the F7 generation is difficult to explain since there were more F7 strains being evaluated than F8 strains and genotypes should segregate for very few loci in these advanced generations. A logical assumption is that such change in the power of the correlation analysis to detect associations must result from a large environmental component of variation. This assumption is verified by the results in Tables 1 and 2 and thus significant correlations of advanced inbred lines with their respective F2 populations in two of three years across multiple locations is strong evidence that EGT will help plant breeders identify desirable populations from which to make selections for yield potential.
View this table:
[in this window]
[in a new window]
|
Table 5. Correlation coefficients for F2 lint yield and fiber properties with respective averages of F6, F7, and F8 performance in standard yield trials from College Station (CS), Weslaco (WS), Corpus Christi (CC), Uvalde (UV), Thrall (TH), Dallas (DL), and Chillicothe (CH), TX, from 1990–2002.
|
|
Lint percentages of the F2 populations at WS or CS were correlated (P = 0.04) with F6, F7, and F8 lint percentages when averaged across multiple locations (Table 5), verifying that lint percentage is a stable trait that becomes fixed very early in the selection process and is influenced little by environment. Data presented in Table 5 suggest that fiber bundle strength and fiber elongation before break, a product of the measurement of fiber bundle strength in HVI testing, also became fixed with single plant selections in the F3 generation following identification of F2 populations with acceptable to enhanced fiber strength. Average fiber strengths of the F6 and F7 inbred lines were highly correlated with F2 population performance when the inbred lines were evaluated in multiple environments, while the F2 and F8 generations were weakly correlated (P 0.09). The same was observed for elongation, which is defined as the amount of stretch of the fibers before breakage occurs in the measurement of fiber bundle strength.
The other fiber traits appear to be much more influenced by the environment as indicated by smaller correlations. The F2 EGT micronaire readings were correlated with the average of their F6 strains across these seven locations, but not with the F7 nor F8 strain values (Table 5). A small but significant correlation was observed between the F2 and the F7 generations for fiber length, but not between the F2 and F6 or F8. Advanced progeny performance trials at three of these locations, CC, TH, and DL, were grown without supplemental irrigation and several years during the 1990s were characterized by severe droughts, with all other years characterized by moderate drought periods during boll fill. Apparently, fiber length is influenced more by moisture stress than fiber bundle strength and, thus, when environments were combined, the variation resulting from varying degrees of drought stress in these dryland environments would have impacted length to a greater degree than fiber bundle strength. However, it should be pointed out that this study was not designed to identify environmental factors influencing fiber properties, but only to ascertain if EGT could identify superior populations in the cotton breeding program of the Texas Agricultural Experiment Station for central and south Texas.
|
CONCLUSIONS
|
|---|
Correlations between EGT yields and later generation performance trial yields did not support EGT at CS probably due to the emphasis on fiber quality development by the breeder at this location. However, the correlation data reported herein, especially at CS where the major emphasis was on fiber quality, did support the value of EGT for fiber quality traits, especially fiber bundle strength. At WS, EGT lint yield data did correlate with later generation performance trial lint yield data, thereby supporting the use of EGT where lint yield was the primary selection criteria.
Using data from the EGTs at both CS and WS, correlations obtained for F2 yield and fiber properties, and yield and fiber properties of advanced strains in standard performance trials were interesting and informative. Early generation test data did provide the breeder with information relative to the lint yield and fiber length, strength, and elongation of advanced strains when the strains were evaluated at multiple locations. Therefore, the use of EGT, as a population selection tool, should be considered and evaluated by individual cotton breeders for use across specific locations and selection goals.
Received for publication September 2, 2004.
|
REFERENCES
|
|---|
- Anthony, W.S. 1999. Postharvest management of fiber quality. p. 293–333. In Amarjit S. Basra (ed.) Cotton fibers: Developmental biology, quality improvement, and textile processing. Food Products Press, Binghamton, NY.
- Barut, A. 1998. Early generation bulk testing for predicting F4:5 line performance in cotton. M.S. thesis. Mississippi State Univ., Starkville, MS.
- Boerma, H.R., and R.L. Cooper. 1975. Comparison of three selection procedures for yield in soybeans. Crop Sci. 15:225–229.[Abstract/Free Full Text]
- Bowman, D.T. 2000. Attributes of public and private cotton breeding programs. J. Cotton Sci. 4:130–136.
- Brennan, J.P., and L. O'Brien. 1991. An economic investigation of early-generation quality testing in a wheat breeding program. Plant Breed. 106:132–140.
- Coffelt, T.A., and R.O. Hammons. 1974. Early generation yield trials of peanuts. Peanut Sci. 1:3–6.
- Halward, T.M., J.C. Wynne, and E.J. Monteverde-Penso. 1990. The effectiveness of early generation testing as applied to a recurrent selection program in peanut. Peanut Sci. 17:44–47.
- Hegstad, J.M., G. Bollero, and C.D. Nickell. 1999. Potential of using plant row yield trials to predict soybean yield. Crop Sci. 39:1671–1675.[Abstract/Free Full Text]
- Immer, F.R. 1941. Relation between yielding ability and homozygosis in barley crosses. J. Am. Soc. Agron. 33:200–206.
- Meredith, W.R., Jr., and R.R. Bridge. 1973. The relationship between F2 and selected F3 progenies in cotton (Gossypium hirsutum L.). Crop Sci. 13:354–356.[Abstract/Free Full Text]
- Ntare, B.R. 1999. Early generation testing for yield and physiological components in groundnut (Arachis hypogaea L.). Euphytica 107:141–147.[CrossRef]
- Percy, R.G. 2003. Comparison of bulk F2 performance testing and pedigree selection in thirty pima cotton populations. J. Cotton Sci. 7:170–178.
- St. Martin, S.K., and I.O. Geraldi. 2002. Comparison of three procedures for early generation testing of soybeans. Crop Sci. 42:705–709.[Abstract/Free Full Text]
- Wynne, J.C. 1976. Evaluation of early generation testing in peanuts. Peanut Sci. 3:62–66.
TOP
ABSTRACT
INTRODUCTION
