Quantitative Trait Loci Associated with Agronomic and Fiber Traits of Upland Cotton

Identificationofquantitative traitloci (QTLs) for agronomic and fiber traits in upland cotton (Gossypium hirsutum L.) and their allelic association with molecular markers would be useful in cotton breeding. We used the mixed model approach of Zhu and Weir (1998) to analyze for QTLs associated with 19 agronomic and fiber traits across 96 F 2-derived families from the cross of two cotton lines, MARCABUCAG8US-1-88 x‘HS46’ (female parent). In the mixed model, molecular markers are random variables and QTLs are fixed variables. Thus, with the mixed model analysis, the QTLs are not dependent upon a particular fixed set of markers being inthe model. The model also provides estimates of additive and dominance genetic effects as well as the direction of the effects of alleles from both parents. The fiber and agronomic traits, except seed indexandbloomrate, were measuredinF 2-derivedF 5 families. We mapped 100 QTLs to 60 maximum likelihood positions in 24 linkage groups. Several QTLs influence more than one trait. The most frequentassociationofQTLs withmultiple traits was for fiber traits related to maturity and fineness. A positive correlation among traits would be beneficial formarker-assistedselectioninplantbreeding as well as forcloning genes fortransformation. Forexample, in linkage group 14 near markers C117C5RI and F26ERI, a QTL is located that affects micronaire, arealometer high pressure reading, weight fineness, and wall thickness. In linkage group 19, four closely linked QTLs, located in an 8 cMregion near marker C80F1RV, influence strength, fineness, and maturity of fiber. Maximumlikelihood locations such as those obtained in this study do not necessarily represent physical distances, thus, a physical map of linkage groups is also needed. Z.W. Shappley, Dep. of Plant and Soil Sciences, Mississippi State Univ., Mississippi State, MS 39762, Current address: Monsanto Company, 700 Chesterfield Village Parkway, St. Louis, MO 63198; J.N. Jenkins, USDA-ARS, Crop Science Research Lab., P.O. Box 5367, Mississippi State, MS 39762; J. Zhu, Zhejiang Agricultural Univ., Research Center of Biomathematics, Hangzhou 310029, Zhejiang, China; and J.C. McCarty, Jr., USDA-ARS, Crop Science Res. Lab., P.O. Box 5367, Mississippi State, MS 39762. Received 14 Apr. 1998. *Corresponding author ([email protected]). Abbreviations: LOD, log to the base 10 of the ratio of the odds of linkage to no linkage; QTL(s), quantitative trait locus (loci); RFLP, restriction fragment length polymorphism. 154 SHAPPLEY ET AL.: QTLs ASSOCIATED WITH UPLAND COTTON The identification and characterization of genes controlling traits of use in plant improvement has long been a focus of scientists in the agricultural community. Recent advances in molecular biological techniques have helped to hasten the realization of these goals. The association of molecular markers with desirable quantitative traits should contribute to the discovery of genetic variability and aid in the selection of desirable parents and progeny. The absence of environmental influence on molecular markers adds to their usefulness in marker-assisted selection for QTLs. The identification of multiple QTLs with varying genetic effects for an individual trait provides evidence of the quantitative nature of the genes influencing the trait. When the QTLs are also closely linked with molecular markers, the opportunity exists for marker-assisted selection for the trait. The MAPMAKER\ QTL method (Paterson et al., 1988) has been the standard for interval mapping for several years. This method uses a model that considers only two loci at a time for the calculations. The method of composite interval mapping that Zeng (1993, 1994) developed includes marker information for controlling background noise while searching for the QTL. The marker effects, as well as the QTL effects, in this model are treated as fixed effects. Therefore, the estimated QTL effects could be affected by the markers included in the model. Zhu and Weir (1998) proposed a new method that uses a mixed model approach for composite interval mapping of QTLs. In their mixed model, QTLs are fixed variables while molecular markers are random variables. Thus, the estimates of the QTLs will not depend upon a particular fixed set of markers being in the model. The model also provides estimates of additive and dominance effects of QTLs. Meredith (1992), in a study of heterosis and varietal origins, reported on the first RFLP evaluations in upland cotton, G. hirsutum L. Reinisch et al. (1994) developed a detailed RFLP map of cotton with 41 linkage groups by using an interspecific F2 population from the cross of G. hirsutum L. race “palmeri” x G. barbadense L. accession K101. Shappley et al. (1996) established five linkage groups in a cross of two upland G. hirsutum L. cottons. Shappley et al. (1998) also developed a genetic linkage map with 31 linkage groups in upland cotton from a cross of two G. hirsutum L. lines. This map was based on segregation in 96 F2:F3 families scored for 129 probe-enzyme combinations that resulted in 138 RFLP loci (120 in linkage groups and 18 nonlinked, Shappley et al., 1998). These were established with an LOD (log to the base 10 of the ratio of the odds of linkage to no linkage) score of greater than 3.0. There were 84 codominant loci of which 76 segregated normally (1:2:1 ratio) for codominant alleles and 54 dominant loci at which only one allele was identified, of which 50 segregated normally (3:1 ratio). These 31 linkage groups covered 865 cM or an estimated 18.6% of the genome (Shappley et al., 1998). Shappley (1996) provided the first linkage map of QTLs in a cross of upland cotton. However, while carefully examining these data in preparation for writing this manuscript, we discovered a computer coding error in the QTL data of Shappley (1996). Thus, no correct linkage map with QTLs and associated molecular markers has been reported in crosses of two G. hirsutum lines. Such maps may be especially valuable for analysis and detection of variability in G. hirsutum including elite germplasm. A map showing a QTL for several fiber traits from a cross of G. hirsutum x G. barbadense was published recently (Jiang et al., 1998). Interspecific incompatibility usually complicates segregation in interspecific hybrids. Upland cultivars (G. hirsutum) comprise more than 90% of cotton acreage in the world. Identification of QTLs and their association with molecular markers in segregatinggenerations following crosses of upland cotton is of great interest to cotton breeders. The identification of QTLs controlling traits of interest to breeders of upland cotton and their association with RFLP molecular markers was the focus of this research. MATERIALS AND METHODS Material and Traits Analyzed QTLs affecting 19 agronomic and fiber traits were searched for among the 31 linkage groups established by Shappley (1996) and Shappley et al. (1998) in upland cotton. Molecular methods and mapping methods establishing the 31 linkage groups are given in Shappley et al. (1998). We used 155 JOURNAL OF COTTON SCIENCE, Volume 2, Issue 4, 1998 the same cross for the QTL analysis as Shappley (1994, 1996), Shappley et al. (1996), and Shappley et al. (1998) used to establish the RFLP linkage map in upland cotton. All measurements were made on 96 F2-derived families from the cross of two G. hirsutum L. lines, MARCABUCAG8US-1-88 as male parent x ‘HS 46’ female parent. These parents are very diverse in agronomic and fiber traits as well as diverse for RFLP markers. A cross was made in 1991, and in 1992 nine F1 plants were grown and analyzed to determine if restriction fragment length variability was observed among the plants. Some variability was observed, thus one plant was chosen to selfpollinate to produce the F2 population. One hundred F2 seed were planted in the greenhouse in the winter of 1992 and 96 plants grew and were allowed to self-pollinate. This planting was the beginning of successive generations of F2-derived families. Bulk samples of leaves were collected from F2:F3 families and analysis with RFLP probes was procured from Biogenetics Services Incorporated, Brookings, SD. Biogenetic Services Inc. developed the probes using cDNA cotton leaf and fiber libraries. Individual families were self-pollinated and seed bulked by families in the F3 and F4. In the spring of 1995 tworow plots of F5 seed were planted and agronomic and fiber data were collected for the QTLs study. Conventional and arealometer fiber measurements, as well as selected agronomic measurements, were made in the F5 generation. Blooming rates and seed indexes were measured in the F3 and F4 generations, respectively. Agronomic and fiber traits are listed in Table 1. Samples for lint percentage measurements, and all measurements of fiber traits were made from handpicked boll samples, ginned on a 10 saw gin, at Mississippi State, MS. Conventional and arealometer fiber measurements were conducted by Starlab Inc., Knoxville, TN, on samples from 25 individual F5 plants per family. Cottonseed for seed index measurements were collected from handpicked boll samples from each family in the F4 generation. One hundred fuzzy seed were counted and weighed to determine an average seed weight for each family. Seed index is the weight of 100 ginned, but not delinted seed and is an indicator of seed size or density. Lint percent, or lint fraction, is the ratio of lint to the total weight of unginned seed cotton expressed as a percentage. Micronaire is a measure of the fineness of the sample of fibers and is reported in standard micronaire units. Elongation is a measure of the elasticity of the fiber sample. The value is determined at the break point in the strength determination and is defined as a percent stretch of the fiber sample at the breaking point. Strength is the fiber strength of a bundle of fibers measured with two stelometer jaws holding the fiber bundle separated by 0.3175 cm (one-eighth inch). The digital fibrograph is an instrument for measuring fiber length. Span length is the distance spanned by a specific percentage of the fibers in the test specimen when the initial starting point of the scanning in the test is considered 100%. The 50% span length is the length on the test specimen spanned by 50% of the fibers scanned at the initial starting point. The 2.5% span length is the length on the test specimen spanned by the longest 2.5% of the cotton fibers scanned at the initial starting point. The 2.5% span length approximates the classer’s staple. The arealometer instrument measures the resistance a given mass of fibers offers to the flow of air at two pressures. From these data, other fiber properties such as fineness and shape can be determined and used to calculate immaturity ratio, percentage maturity, perimeter, weight fineness, and wall thickness. The measurement, A, describes the external surface of the fibers of a given volume of fibrous material under standard pressure, expressed in terms of square millimeters per cubic millimeter of fibrous material. The measurement, Ah, measures the same fibers as the A measurement, but under high pressure. The difference between A and Ah is an estimate of the flatness of the fiber ribbon. The greater the difference, the more ribbon-like are the fibers. The immaturity ratio is a dimensionless number that describes a physical characteristic of the fiber crosssection. It is defined as the ratio of the area that the fiber cross-section would have if its perimeter enclosed a circle compared to the area that the perimeter actually encloses. Measurement of fiber maturity is based on the simple linear regression prediction of the caustic soda percent maturity method (Hertel and Craven, 156 SHAPPLEY ET AL.: QTLs ASSOCIATED WITH UPLAND COTTON 1951). The prediction equation is M = 150.5 – 38.1I, where I = the calculated immaturity value. The perimeter is defined as the distance around the outside wall of the fiber section in micrometers. The weight fineness, or linear density, is defined as the mass per unit length of fiber expressed in micrograms per inch. The fiber wall thickness is the measurement in micrometers of the width of the wall of the cotton fiber. Equations for calculation of each of these traits and their relationships are given in the National Cotton Variety Test Report by Rayburn et al. (1996). The total number of nodes is a total of all nodes with the cotyledon node counted as one. Node of first fruiting branch is a physiological trait that gives an indication of earliness and is the node at which the plant develops its first nonvegetative branch. Plant height was measured from ground level to the top of the plant at harvest time. The height/node ratio is obtained by dividing the plant height by the total number of nodes on the plant. Lint percent measurements were calculated from cotton harvested from individual plants in the F5 generation. A mean was calculated from individual measurements of 50 plants in each tworow family plot. Fiber samples from 25 plants per F2-derived F5 family were measured twice for each of the fiber traits: micronaire, elongation, strength, 50% span length, 2.5% span length, A, Ah, immaturity, maturity, perimeter, weight-fineness, and wall thickness. A mean was then calculated for each of the traits in each family. For number of nodes, node of the first fruiting branch, and plant height in the F5 generation, all plants in the two-row plot were measured individually and a mean was taken from these measurements. White bloom counts were taken in the F3, once a week, over a 4 week period. A percentage of the plants flowering at a given date for each family was calculated. Statistical Analysis Methods To determine if trait data were normally distributed, the skewness and kurtosis values were calculated for each trait. When seeking to detect a QTL between two markers, the other markers linked with some other QTLs are likely to have marker effects, which should be considered in controlling background genetic variation. By employing a mixed model approach where effects of the QTLs are considered fixed and molecular markers are random (Zhu and Weir, 1998) the phenotypic value of a quantitative trait measured on the jth individual can be expressed as a mixed linear equation y ax dx e z j A D M M j