HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Vaylay, R. Articles by van Santen, E. Search for Related Content PubMed Articles by Vaylay, R. Articles by van Santen, E. Agricola Articles by Vaylay, R. Articles by van Santen, E. Related Collections Turfgrass Plant Analysis

FORAGE & GRAZING LANDS

Application of Canonical Discriminant Analysis for the Assessment of Genetic Variation in Tall Fescue

Dep. of Agronomy and Soils, 202 Funchess Hall, Auburn University, Auburn, AL 36849-5412

* Corresponding author (evsanten{at}acesag.auburn.edu)

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Tall fescue pastures are perennial in habit and are continually exposed to intensive natural selective forces. Therefore, the genetic composition of tall fescue cultivars changes with time and the selection is cumulative. The objective of this study was to investigate the genetic diversity of tall fescue cultivars and age groups within cultivars in response to natural selective forces using a multivariate statistical method, canonical discriminant analysis. Ungrazed survivors from four cultivars (GA-5 EF, GA-5 EI, Johnstone, and KY-31) were collected randomly from four 1-yr-old paddocks. These paddocks were then grazed at 2.50, 3.75, 5.00, and 6.25 Angus yearling steers ha^-1 for 2 yr, and grazed survivors were collected in a similar fashion as that of ungrazed survivors. The ungrazed and grazed groups were compared in a 2-yr study along with plants grown directly from the original seed lots. Significant genetic diversity was noticed among these four tall fescue cultivars and clearly depicted the relationship of cultivars by ancestry and endophyte status. Significant genetic diversity between GA-5 EF and GA-5 EI indicated that the removal of endophyte resulted in a different cultivar with altered morphological and agronomic characteristics. Significant genetic diversity was observed among the age groups in GA-5 EF, GA-5 EI, and Johnstone. The age groups of KY-31 were stable in terms of genetic variation. These results indicated that changes in genetic variation occurred in a short period of time, that is, 3-yr old paddocks of GA-5 EF, GA-5 EI, and Johnstone.

Abbreviations: D², Mahalanobis distance • NDF, neutral detergent fiber

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
AN INSIGHT INTO THE MAGNITUDE of variability present in crop species is of utmost importance, as it provides the basis for effective selection. Phenotypic variation present in a population arises due to genotypic and environmental effects. Phenotypic variability is the observable variation present in a character in a population; it includes both genotypic and environmental components of variation and, as a result the magnitude of phenotypic variability, differs under different environmental conditions. Genotypic variation, on the other hand, is the component of variation that is due to the genotypic differences among individuals within a population or among populations within a species, and is the main concern of a plant breeder. The phenotype is based on quantitative characters and has a strong genotypic basis, though it often cannot be directly related to genotype (Loos, 1993). If phenotype observations are based on sufficiently large sample sizes and the traits measured show significant differences among populations, they can provide a reasonable representation of overall genetic performance (Humphreys, 1991).

Genetic variation may be measured in several ways. Considerable overlapping may occur in univariate analysis, since each variable is viewed separately. In canonical discriminant analysis, a multivariate statistical technique, all independent variables (traits) are considered simultaneously in the differentiation of cultivars. The resulting differentiation of populations is more distinct compared with univariate analysis. It extracts components so that the among population variability (genetic) is maximized compared with the within-population (environmental) variability. Therefore, canonical discriminant analysis can separate among population effects from within population effects (Riggs, 1973; Tai, 1989). Essentially, it maximizes the overall heritability of canonical variates and places very large weight on traits with low levels of environmental variability (Humphreys, 1991). After extraction of the among population variability (genetic), the genetic differentiation between populations can be measured by the Mahalanobis distance (D²) statistic (Humphreys, 1991; Loos, 1993).

Tall fescue (Festuca arundinacea Schreb.) is a perennial, self-incompatible, and thus highly cross-pollinated species. Even improved cultivars are highly heterogeneous and have abundant genetic variation. The within cultivar/population variation is not only due to environmental effects of cultivar, but also due to phenotypic (genotypic and environmental) effects of the genotypes of a cultivar or population. In canonical discriminant analysis the discrimination is obtained by the ratio of variance among cultivars to the variance within cultivars (Rencher, 1992); the differentiation of cultivars is due to genetic variation present among the cultivars. Moreover, canonical discriminant analysis summarizes among group variation, unlike principal component analysis, which summarizes total variation.

Perennial pasture species must adapt to ever-changing abiotic and biotic disturbances. Therefore, genetic variation is a prerequisite for adaptation. Rapid genetic changes in response to natural selection were observed in tall fescue (Vaylay and van Santen, 1997), perennial ryegrass (Lolium perenne L.) (Charles, 1970), and orchardgrass (Dactylis glomerata L.) (Tsurumi et al., 1985). In all of these studies, cultivar differentiation was studied with respect to individual characters. Our objectives were to use canonical discriminant analysis to study the differences among tall fescue cultivars subjected to natural selection.

	MATERIALS AND METHODS

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Plant Material
The experimental material consisted of a factorial combination of four cultivars and three age groups (Table 1). Age groups were plants established from original seed, ungrazed, and grazed survivors from a grazing experiment. These ungrazed and grazed survivors from four paddocks were collected at random from plots established in autumn 1988, and grazed from 1989 to 1991 (van Santen, 1992). Following collection of the ungrazed survivors in autumn 1989, paddocks were grazed by Angus yearling steers (Bos taurus spp.) for 2 yr at stocking rates of 2.50, 3.75, 5.00, and 6.25 animals ha^-1 (van Santen, 1992). Each paddock in the original grazing trial consisted of six complete blocks. Ten plants were collected for each cultivar in each block. Because recruitment of new individuals within plots was prevented by removal of inflorescences before anthesis, plants were of the same generation but differed in age. A detailed description of the collection procedures can be found in Vaylay and van Santen (1997).

Data Analysis
Canonical discriminant analysis was used for data analysis. This is a dimension-reduction technique related to principal component analysis and canonical correlation. Given a classification variable, such as population or age group, and several quantitative variables, the canonical discriminant analysis derives canonical discriminant functions (linear combinations of these quantitative variables) that have the highest possible multiple correlation with groups and summarizes among-class variation in much the same way that principal component analysis summarizes total variation. It facilitates differentiation of groups by taking into account the interrelationships of the independent variables (traits) and the dependent variables (cultivars and age groups). An important property of canonical variables is that they are uncorrelated even though the underlying quantitative variables may be highly correlated.

The groups to be differentiated in our research are known a priori. In such cases, canonical discriminant analysis is a very powerful tool in determining genetic distances among the groups. Three and two canonical variates were derived for differentiation of cultivars and age groups within cultivars, respectively. Unlike analysis of variance and multiple regression, the dependent variable in canonical discriminant analysis is categorical and the independent variable is metric. Cultivars and age groups within cultivars are dependent variables, whereas the traits that were measured are independent variables in canonical discriminant analysis.

The data were transformed on a single genotype basis into mean absolute deviations to achieve multivariate normal distribution. Mean absolute deviation was calculated as follows:

where M_ijkl is the mean absolute deviation of ith genotype from the mean of jth plot in the kth replicate in the lth year, µ_jkl = mean of jth plot of kth replicate of lth year, and X_ijkl = actual value of ith genotype of jth plot of kth replicate of lth year.

The mean value of the canonical discriminant function is referred to as group centroid. The difference between centroid values of two groups is the D² distance and is calculated as:

where ₁ and ₂ are the estimated mean vectors in the respective groups, and S^-1 is the inverse of the pooled sample variance-covariance matrix (Dillon and Goldstein, 1984).

	RESULTS AND DISCUSSION

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Univariate statistical techniques and analysis of variance do not show how cultivars or age groups within cultivars differ when all variables are considered together. Canonical discriminant analysis simultaneously examines differences in the morphological variables and indicates the relative contribution of each variable to cultivar discrimination. Multivariate procedures based on morphological and agronomic characters have been used in the assessment of genetic divergence in Glycine max (L.) Merr. (Bains and Sood, 1984), F. arundinacea, (Veronesi and Falcinelli, 1988), L. perenne (Humpreys, 1991), and Lolium spp. (Loos, 1993).

Differentiation of Cultivars
The first two canonical variates were significant (P < 0.01) and accounted for 89% of the among cultivar (genetic) variance (Table 2). Each canonical variate is the linear combination of the independent variables (traits) and is orthogonal to the other. Canonical correlation measures the strength of the overall relationships between the linear composites of predictor (canonical discriminant variate) and criterion (cultivars) sets of variables. The significant (P < 0.01) canonical correlation between the cultivars and the first canonical variate (r_c = 0.71) and cultivars and second canonical variates (r_c = 0.42) indicates that the canonical variates can explain the differentiation of the cultivars.

The centroid values for the first two canonical discriminant functions for the four cultivars were plotted (Fig. 1) . The extent of separation of the cultivars is measured by D². All pairwise distances between cultivars were significant (P < 0.05). The distance between GA-5 EF and GA-5 EI was only 1.29 but, was nevertheless significant (P 0.04). GA-5 EF and GA-5 EI are genetically similar except for the removal of the endophyte [Neotyphodium coenophialum (Morgan-Jones & W. Gams) Glenn, C.W. Bacon & Hanlin comb.nov.] from GA-5 EI. This relationship has been clearly revealed by the distance between GA-5 EF and GA-5 EI. Removal of the endophyte altered the genetic expression of morphological and agronomic characteristics of the EF cultivar, thereby making it significantly (P 0.04) different from the EI cultivar. This clearly demonstrates the influence of the removal of endophytes on the host's performance, thus making it a new cultivar with different morphological and agronomic traits (Vaylay and van Santen, 1997). Differences in the genetic variances of EI and EF cultivars have been reported by Rice et al. (1994). Our results also support the hypothesis of Pedersen and Sleper (1993), that the removal of endophyte magnifies the genetic differences among host plants. Had there been no confounding effect of endophyte, the genetic distance between EF and EI cultivars would not have been significant.

View larger version (26K): [in this window] [in a new window]   Fig. 1. Scatterplot of centroid values of the four cultivars on the two canonical discriminant functions. Mahalanobis distances and their probability values in parenthesis measure the extent of genetic diversity between the cultivars. The study was conducted at the Sand Mountain Substation in Crossville, AL, from 1993 to 1995.

Johnstone has Lolium in its genetic background (Table 1) and is maximally separated from the others (Fig. 1). Two of the parent strains of Johnstone are derived from Kenhy, which in itself derived from a KY-31 x L. multiflorum cross. This may explain why Johnstone more closely resembles KY-31 (D² = 4.46) than the other cultivars. Though GA-5 EI is derived from KY-31, Johnstone is quite different from GA-5 EI. This may be due to the selection process involved in the development of GA-5 EI. Johnstone and KY-31 were selected in the state of Kentucky, whereas GA-5 EI was selected in Georgia. These results indicate the influence of the selection environment on the genetic diversity of cultivars. The greater distance between Johnstone and GA-5 EI (D² = 6.98) compared with Johnstone and GA-5 EF (D² = 5.55) may be a further illustration of the role of endophyte removal on the genetic diversity among cultivars; both Johnstone and GA-EF are endophyte-free.

Differentiation of Age Groups within Cultivars
In this research, we used the concept of distance to study the changes in genetic variation of morphological and agronomic traits of age groups in response to natural selective forces. As the number of canonical variates is equal to the number of groups minus one, only two canonical discriminant functions were extracted from three age groups. The age groups of each cultivar are shown in Fig. 2 on the scatter plot of centroid values on the two canonical discriminant functions.

View larger version (33K): [in this window] [in a new window]   Fig. 2. Scatterplot of centroid values of the three age groups of the four cultivars on the two canonical discriminant functions. Mahalanobis distances and their probability values in parenthesis measure the extent of genetic diversity between age groups of cultivars. The study was conducted at the Sand Mountain Substation in Crossville, AL, from 1993 to 1995.

GA-5 EI
Both canonical variates were significant, and their correlation with the age groups were 0.77 and 0.62, respectively (Table 3). The highest loading on the first canonical function was by maturity (0.49) followed by plant diameter, tiller length, dry matter yield, and flag leaf area. Unlike GA-5 EF, the probability associated with the second discriminant function was much lower (P = 0.18). The second function was highly influenced by tiller number followed by node number, maturity, and dry matter yield. The variance explained by first and second canonical discriminant functions were 69 and 30%, respectively.

Mahalanobis distance (Fig. 2) was significant between original and ungrazed (P = 0.08), but unlike GA-5 EF, the ungrazed and grazed survivors were significantly different (P = 0.02). Animal preference (van Santen, 1992) of EI age groups at low stocking rate may be a possible explanation for the differences of genetic variation between ungrazed and grazed survivors. In contrast to GA-5 EF, the genetic distance between original and grazed populations was not significant, indicating similar genetic composition of original and grazed survivors. The difference in the calculated significance of similar D² values (D² = 6.59, P = 0.28 for original vs. ungrazed; D² = 5.59, P = 0.02 for ungrazed vs. original) is the result of hidden replication during the evaluation. The frequency for the original population is only four (a single plot, but two replicates, and two evaluation years), whereas those of ungrazed and grazed are 16 each (four plots due to the four stocking rates in the original trial, two replicates, and two evaluation years). The above contrasting results of GA-5 EF and GA-5 EI demonstrate the influence of the removal of endophyte on the host's genetic variability in response to natural selective forces.

Johnstone
The canonical correlations between the age groups and first and second canonical variates were 0.70 and 0.67, respectively. Both canonical discriminant functions were significant at P 0.07 (Table 3). The variance explained by the first and second canonical discriminant functions were 53 and 46%, respectively. Dry matter yield had highest loading on first canonical variate followed by maturity, total cell wall content, and node number. On second discriminant function, plant diameter had highest influence followed by tiller number, tiller length, and node number.

All three age groups differed from one another at P < 0.12 (Fig. 2). Grazed survivors were more similar to the original population than the ungrazed survivors, as indicated by the somewhat smaller distance values. The same trend occurred in the first two populations discussed (GA-5 EF, GA-5 EI). The fact that these additional changes occurred in different dimension (first or second canonical variate) seems to indicate that different organizational levels were involved in the response.

KY-31
The principle of canonical discrimination is maximization of the among-group (genetic) variability and minimization of the within-group (environmental) variability. The canonical correlations between the age groups and canonical variates for KY-31 were low compared with the other populations, and not significant (Table 3). If the groups being analyzed are not very different for the variables being analyzed, then all of the correlations will be low, because we cannot create discrimination when none already existed (Klecka, 1980).

Not much genetic variability existed among the age groups of KY-31, and, therefore, the correlations were not significant. Also, the genetic distances between the age groups were not significant, indicating genetic similarity of age groups (Fig. 2). Unlike the previous three cultivars discussed, ungrazed and grazed survivors were virtually identical, as evidenced by the low distance estimate (D² = 0.62, P = 0.97) and the similar distance estimates to the original population (Fig. 2).

The lack of differences among age groups within KY-31 compared with GA-5 EF, GA-5 EI, and Johnstone reinforces experiences many fescue breeders have had in the past; given similar relative maturity, it is very difficult to improve performance relative to KY-31. This cultivar seems to be much more stable than other cultivars. On the other hand, GA-5 EF and GA-5 EI are experimental cultivars and still have to go through cultivar improvement. Johnstone is also an established cultivar, but not quite as stable as KY-31. However, the grazed populations of this research were only 2 yr old, and further research with older cultivars may offer better insight into the genetic variation among KY-31 age groups.

	CONCLUSIONS

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Significant genetic diversity was observed among the cultivars. The results of this study show the extent of the genetic relationship among the cultivars by ancestry and endophyte status. The concept of measuring genetic diversity has been extended in this research to study the changes in genetic variation of age groups in response to natural selective forces. Significant genetic distance among the age groups represented the change in genetic diversity of the age groups. The changes in the genetic composition of age groups indicated the stability of the cultivars in response to natural selective forces. The change in the genetic variability is permanent, indicating that selection has taken place in three out of four cultivars with increased pasture age. If the changes had been plastic, then the differences in genetic variability should have disappeared once all the age groups were brought to a common environment. The cultivar KY-31 was quite stable in terms of genetic variability. In other cultivars, considerable changes in genetic composition of cultivars had taken place. This indicated that selection is still taking place due to natural selective forces. These changes in genetic variation took place in a 3-yr period of time. From a plant breeding point of view, canonical discriminant analysis is useful in identifying the genetic variation and the most influential traits affecting genetic variation of plant populations. Canonical loadings of morphological and agronomic traits of an individual cultivar indicate the magnitude of genetic variation. The influential traits are the ones that change in response to natural selective forces. Knowledge of genetic variation of traits among age groups in response to natural selective forces will be useful for a plant breeder by focusing attention on those particular traits that have adaptive significance.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Contribution is research conducted in partial fulfillment of Ph.D. requirements for the senior author for the Dep. of Agronomy and Soils at Auburn Univ. Salary support provided in part by state and federal funds appropriated to the Alabama Agric. Exp. Stn. Project No. ALA 003-038.

Received for publication January 12, 2001.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 

• Bains, K.S., and K.C. Sood. 1984. Resolution of genetic divergence for choice of parents in soybean breeding. Crop Improv. 11:20–24.
• Bouton, J.H., R.N. Gates, G.M. Hill, M. Owsley, and D.T. Wood. 1993. Registration of ‘Georgia 5’ tall fescue. Crop Sci. 33:1405.[Free Full Text]
• Buckner, R.C., J.A. Boling, P.B. Burrus II, L.P. Bush, and R.A. Hemken. 1983. Registration of ‘Johnstone’ tall fescue. Crop Sci. 23: 399–400.[Free Full Text]
• Charles, A.H. 1970. Ryegrass cultivars from intensively managed leys. I. Seedling and spaced plant character. J. Agric. Sci. (Cambridge) 75:103–107.
• Dillon, W.R., and M. Goldstein. 1984. Multivariate analysis-methods and applications. John Wiley and Sons, New York.
• Fergus, E.N, and R.C. Buckner. 1972. Registration of Kentucky 31 tall fescue. Reg. No. 7. Crop Sci. 12:714.[Free Full Text]
• Hair, J.F., Jr., R.E. Anderson, and R.L. Tatham. 1987. Multivariate data analysis. Macmillan Publishing Company, New York.
• Humpreys, M.O. 1991. A genetic approach to the multivariate differentiation of perennial ryegrass (Lolium perenne L.) cultivars. Heredity 66:437–443.
• Klecka, W.R. 1980. Discriminant analysis. Sage University paper series on quantitative applications in the social sciences. Sage Publications, Beverly Hills, CA.
• Loos, B.P. 1993. Morphological variation in Lolium (Poaceae) as a measure of species relationships. Plant Syst. Evol. 188:87–99.
• Pedersen, J.F., and D.A. Sleper. 1993. Genetic manipulation of tall fescue. Agric. Ecosyst. Environ. 44:187–194.
• Rencher, A.C. 1992. Interpretation of canonical discriminant functions, canonical variates, and principal components. Am. Stat. 46: 217–225.
• Rice, J.S., N.H. Ferguson, B.W. Pinkerton, and W.C. Stringer. 1994. Alteration of phenotypic variances by endophyte of tall fescue. Crop Sci. 34:1487–1489.[Abstract/Free Full Text]
• Riggs, T.J. 1973. The use of canonical analysis for selection within a cultivar of spring barley. Ann. Appl. Biol. 74:249–258.
• Tai, G.C.C. 1989. A proposal to improve the efficiency of index selection by ‘rounding.’ Theor. Appl. Genet. 78:798–800.
• Tsurumi, Y., T. Yoshiyama, K. Nakashima, and H. Yamaguchi. 1985. Characteristic differences of survival plants under grazing from the original varieties in orchardgrass (Dactylis glomerata L.). p. 286– 287. In T. Okubo and M. Shiyomi (ed.) Proc. International Grassland Congress, 15th, Nishi-nasumo, Japan. 24–31 August 1985. Science Council of Japan and the Japanese Society of Grassland Science, Nishi-nasuno and Tochigi-ken, Japan.
• van Santen, E. 1992. Animal preference of tall fescue during reproductive growth in the spring. Agron. J. 84:979–982.[Abstract/Free Full Text]
• Vaylay, R., and E. van Santen. 1997. Grazing pressure induces a patterned response in tall fescue. Crop Sci. 39:44–51.[Abstract/Free Full Text]
• Veronesi, F., and M. Falcinelli. 1988. Evaluation of an Italian germplasm collection of Festuca arundinacea Schreb. through a multivariate analysis. Euphytica 38:211–220.

This article has been cited by other articles:

				L. J. A. Capo-chichi, J. M. Goatley Jr., W. Philley, J. Krans, D. Davis, A. Kato, and E. van Santen Dinitroaniline-Induced Genetic Changes in Bermudagrass Crop Sci., June 24, 2005; 45(4): 1504 - 1510. [Abstract] [Full Text] [PDF]

				K. M. Yeater, G. A. Bollero, D. G. Bullock, A. L. Rayburn, and S. Rodriguez-Zas Assessment of Genetic Variation in Hairy Vetch Using Canonical Discriminant Analysis Crop Sci., January 1, 2004; 44(1): 185 - 189. [Abstract] [Full Text] [PDF]

				J. S. McElroy, R. H. Walker, and E. van Santen Patterns of Variation in Poa annua Populations as Revealed by Canonical Discriminant Analysis of Life History Traits Crop Sci., March 1, 2002; 42(2): 513 - 517. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Vaylay, R. Articles by van Santen, E. Search for Related Content PubMed Articles by Vaylay, R. Articles by van Santen, E. Agricola Articles by Vaylay, R. Articles by van Santen, E. Related Collections Turfgrass Plant Analysis