Previous PageTable Of ContentsNext Page


Leonardo Velasco1, Christian Möllers, and Heiko C. Becker

Institut für Pflanzenbau und Pflanzenzüchtung, Georg-August-Universität, Von-Siebold-Str. 8, D-37075 Göttingen, Germany. 1Present address: Instituto de Agricultura Sostenible (CSIC) Apartado 4084, E-14080 Córdoba, Spain. E-mail:


The objective of the present research was to investigate the potential of near-infrared reflectance spectroscopy (NIRS) for estimating the total glucosinolate (GSL) content and the concentration of individual GSLs in a wide range of Brassica species. High-performance liquid chromatography was used as reference method. A calibration equation for total GSL content was developed from a calibration set of 486 intact-seed samples from 21 species, resulting in a coefficient of determination (r2) of 0.98 and a standard error of crossvalidation (SECV) of 6.64 µmol g-1 (mean ± SD of the calibration set = 78.1 ± 51.27 µmol g-1). Calibration equations for individual GSLs, expressed as percentage of the total GSL content, were developed from a calibration set containing 363 intact-seed samples from 21 species. Reliable calibration equations (SECV/SED<0.35) were developed for progoitrin, sinigrin, and gluconapin. Promising calibration equations were also developed for glucoiberin, epiprogoitrin, glucoraphanin, 4-hydroxyglucobrassicin, and glucoerucin, although they could not be adequately tested because of an insufficient number of entries having intermediate or high concentrations of these GSLs. The availability of further variability will probably allow more reliable calibration equations for a wider range of individual GSLs to be developed.

KEYWORDS: Brassica spp., germplasm, NIRS, screening, individual glucosinolates, total glucosinolate content.


Breeding for low glucosinolate (GSL) content in Brassica oilseeds has been the subject of intensive research during recent decades because of the toxic and antinutritive effects of these compounds (Downey and Röbbelen 1989). Success in this field was considerably facilitated by the development of rapid and accurate methods which permit large screenings to be made in a short time and at a low cost (e.g. Thies 1982). These methods were designed to estimate the total GSL content, and do not provide simultaneous information on the GSL profile (Sřrensen 1985). The development of NIRS calibration equations to estimate the total GSL content in intact seeds represented a further advance, since it can be estimated in parallel with the analysis of other traits such as oil and protein content (Renard et al. 1987).

Nowadays there is increasing interest on alternative uses of glucosinolates in pest and disease control (Mithen 1992) or as substances possessing anticancer activity (Giamoustaris and Mithen 1996). The biological activity of the GSLs is determined by both the GSL concentration and composition (Rosa et al., 1997). Nevertheless, the current use of NIRS for the analysis of GSLs in Brassica is exclusively quantitative, and no information is obtained from the GSL profile. NIRS is a multi-trait technique, i.e., a large number of traits can be measured simultaneously, and the development of reliable NIRS equations for individual GSLs would considerably increase the value of this technique in the analysis of Brassica samples. The objectives of this work were to develop and characterise a multi-species NIRS calibration equation to analyse total GSL content in intact-seed samples of 21 Brassica species, and to study the potential of this technique to estimate simultaneously the relative amount of individual GSLs.


A total of 486 entries were used in this study. They included breeding materials of B. napus, B. rapa, and B. carinata as well as germplasm accessions from 21 different Brassica species. Intact-seed samples from the 486 entries were used for calibration for total GSL content. Calibration for individual GSLs was performed using 363 samples from 21 species. The reference method was high-performance liquid chromatography of desulphoglucosinolates, as described by Velasco and Becker (1998).

About 300 mg intact seeds were scanned on a NIRS monochromator instrument (NIR Systems model 6500), and reflectance spectra (log 1/R) from 400 to 2500 nm were recorded at 2 nm intervals. Calibration equations were developed by using the spectral information from 1100 to 2500 nm and modified partial least squares (MPLS) regression, after applying second derivative transformation (2,5,5,1), SNV, and De-trend scatter correction. Calibration equations were developed for the total GSL content (µmoles g-1 dry seed weight) and for the concentration (% [mol/mol] of the total GSL content) of the following glucosinolates: glucoiberin, progoitrin, epiprogoitrin, sinigrin, glucoraphanin, gluconapin, 4-hydroxyglucobrassicin, and glucoerucin.


Calibration for total GSL content from a set containing 486 samples of 21 different species of Brassica resulted in a close relationship between NIRS estimated and actual HPLC values, with R2 of 0.98 and a ratio of the standard error of cross validation (SECV) to the standard deviation (SD) of the population of 0.13 (Fig. 1). These statistics demonstrate that the total GSL content of Brassica seeds can be accurately analysed by NIRS following an strategy based on the incorporation of samples from many different species within the calibration set. The performance of this multi-species equation is similar to that of single-species equations for total GSL content. For example, Biston et al. (1988) obtained a ratio of the standard error of calibration (SEC) to the SD of 0.13 in B. napus.

The calibration equations for the individual GSLs progoitrin, sinigrin, and gluconapin were highly reliable, showing R2 above 0.90 and ratios SECV/SD below 0.34 (Fig. 2b,d,f). These results are very similar to those previously reported by Velasco and Becker (1998) using a calibration set with a lower number of samples (n=151). The calibration set used in the present study (n=363) included large variation for these three GSLs, with samples covering most of the potential range of values. However, the available variability for the other GSLs was much more scarce (Fig. 2). For example, the calibration equation for glucoiberin (Fig. 2a) showed similar statistics to those for the mentioned GSLs, i.e. R2 of 0.93 and ratio SECV/SD of 0.34. However, the number of samples having glucoiberin content between 40% and 80% was very low and the calibration equation will probably give a large error within this range. Similarly, the calibration set did not include enough variability for epiprogoitrin (Fig. 2c), glucoraphanin (Fig. 2e), 4-hydroxyglucobrassicin (Fig. 2g), and glucoerucin (Fig. 2h). Nevertheless, the good discrimination between low and high levels of these GSLs suggest the possibility of developing reliable calibration equations when further variability is available.

Fig. 1. Calibration plot (NIRS vs. HPLC) for total GSL content (µmoles g-1) from a calibration set containing 486 intact-seed samples from 21 different Brassica species. R2=coefficient of multiple determination in calibration, SECV=standard error of crossvalidation, SD=standard deviation of the population.


The integration of samples from several species of Brassica within a unique calibration set provided a range of values for total GSL content and concentration of individual GSLs that is not available in a single species. This strategy will permit the advantages of using NIRS to be extended to estimating the total GSL content in minor Brassica species with potential interest for plant breeding, for which the development of sufficiently accurate calibration equations would be impossible because of the lack of intraspecific variability.

Breeding Brassica oil crops for novel applications of GSLs, as for example in pest ad disease control (Mithen 1992, Giamoustaris and Mithen 1996), demands simultaneous analysis of the total GSL content and the GSL profile. Nowadays the analysis of the GSL profile requires chromatographic techniques, which are destructive, time consuming and expensive, and therefore unsuitable for the screening of large populations. The results of this work demonstrate that the relative amount of individual GSLs in Brassica intact-seed samples can be estimated by using NIRS. Such analyses involve a higher error than would be obtained with more accurate techniques, such as HPLC, but for most applications this error is acceptable and largely compensated for the obvious advantages of NIRS analysis. From the results of this work, it is concluded that the critical point for accurate discrimination among different levels of single GSLs is the availability of large variability within the calibration set. Such variability is not available within individual Brassica species, and hence a wide range of species, and also of intraspecific variability, has to be used to ensure the development of accurate calibration equations. One of the main advantages of NIRS technique is that the estimation of the GSL profile is simultaneous to the analysis of the total GSL content, and also of other traits, such as oil, protein, or the fatty acid composition of the oil.

Fig. 2: Calibration plots (NIRS vs. HPLC) for individual GSLs, expressed as % of total GSL content (mol/mol). For abbreviations see Fig. 1


The authors thank Christine Reuter and Uwe Ammermann for excellent technical assistance in HPLC analyses.


1. Biston, R., P. Dardenne, M. Cwikowski, M. Marlier, M. Severin, and J.-P. Wathelet, 1988. Fast analysis of rapeseed glucosinolates by near infrared reflectance spectroscopy. Journal of the American Oil Chemists’ Society 65:1599-1600.

2. Downey, R.K., and G. Röbbelen, 1989: Brassica Species. In: R.K. Downey, G. Röbbelen, and A. Ashri (eds), Oil Crops of the World, 339-362, McGraw-Hill, USA.

3. Giamoustaris, A., and R. Mithen, 1996: Genetics of aliphatic glucosinolates. IV. Side-chain modification in Brassica oleracea. Theoretical and Applied Genetics. 93:1006-1010.

4. Mithen, R., 1992: Leaf glucosinolate profiles and their relationship to pest and disease resistance in oilseed rape. Euphytica 63:71-83.

5. Renard, M., C. Bernard, M. Deschamps, V. Furtoss, M. Lila, A. Quinsac, J.-M. Regnier, and D. Ribaillier. 1987. Glucosinolate analysis in whole rapeseed by near infrared reflectance spectroscopy. In: J.-P. Wathelet (ed.), Glucosinolates in Rapeseed: Analytical Aspects, 173-176, Martinus Nijhoff, Dordrecht, The Netherlands.

6. Rosa, E.A.S., R.K. Heaney, G.R. Fenwick, and C.A.M Portas, 1997. Glucosinolates in crop plants. Horticultural Reviews 19:99-215.

7. Sřrensen, H., 1985: Limitations and possibilities of different methods suitable to quantitative analysis of glucosinolates occurring in double low rapeseed and products thereof. In: H. Sřrensen (ed.), Advances in the Production and Utilization of Cruciferous Crops, 73-84, Martinus Nijhoff/ Dr. W. Junk Publishers, Dordrecht, The Netherlands.

8. Thies, W., 1982: Complex-formation between glucosinolates and tetrachloropalladate (II) and its utilization in plant breeding. Fette Seifen Anstrichmittel 84:388-342.

9. Velasco, L., and H.C. Becker, 1998. Analysis of total glucosinolate content and individual glucosinolates in Brassica spp. by near-infrared reflectance spectroscopy. Plant Breeding 117:97-102.

Previous PageTop Of PageNext Page