Anal. Chem. 2003, 75, 1211-1217
Development of a Method for the Determination of Fusarium Fungi on Corn Using Mid-Infrared Spectroscopy with Attenuated Total Reflection and Chemometrics Gregor Kos, Hans Lohninger,† and Rudolf Krska*
Center for Analytical Chemistry, IFA-Tulln, Konrad Lorenz Strasse 20, A-3430 Tulln, Austria
A novel method, which enables the determination of fungal infection with Fusarium graminearum on corn within minutes, is presented. The ground sample was sieved and the particle size fraction between >250 and 100 µm was used for mid-infrared/attenuated total reflection (ATR) measurements. The sample was pressed onto the ATR crystal, and reproducible pressure was applied. After the spectra were recorded, they were subjected to principle component analysis (PCA) and classified using cluster analysis. Observed changes in the spectra reflected changes in protein, carbohydrate, and lipid contents. Ergosterol (for the total fungal biomass) and the toxin deoxynivalenol (DON; a secondary metabolite) of Fusarium fungi served as reference parameters, because of their relevance for the examination of corn based food and feed. The repeatability was highly improved by sieving prior to recording the spectra, resulting in a better clustering in PCA score/score plots. The developed method enabled the separation of samples with a toxin content of as low as 310 µg/kg from noncontaminated (blank) samples. Investigated concentration ranges were 880-3600 µg/kg for ergosterol and 310-2596 µg/kg for DON. The percentage of correctly classified samples was up to 100% for individual samples compared with a number of blank samples. Fusarium fungi are among the most important agriculturally toxigenic fungi occurring in the moderate climatic zones of North America and Europe.1,2 Its main secondary metabolite is the toxin deoxynivalenol (DON), which has been reported to have several negative side effects on animal and human health. These include feed refusal (mainly by swine3,4), skin damage, vomiting, and * Corresponding author: (e-mail)
[email protected]; (tel) +43-2272-66280401; (fax) +43-2272-66280-403. † Present address: Institute for Chemical Technologies and Analytics, Analytical Chemistry Division, Getreidemarkt 9/151, A-1060 Vienna, Austria; (e-mail)
[email protected]; (fax) +43-1-58801-15199. (1) Miller, J. D. In Mycotoxins in Grain: Compounds Other than Aflatoxins, 1 ed.; Miller, J. D., Trenholm, H. L., Eds.; Eagan Press: St. Paul, MN, 1994; Chapter 2. (2) GIPSA, Grain Fungal Diseases & Mycotoxin Reference, 1 ed.; USDA, U.S. Department of Agriculture; USDA/GIPSA,Grain Inspection, Packers and Stockyards Administration, Technical Services Division: Kansas City, 2002; Chapter 2. (3) Pittet, A. Rev. Me´ d. Ve´ t. 1998, 149, 479-92. 10.1021/ac0260903 CCC: $25.00 Published on Web 01/31/2003
© 2003 American Chemical Society
hemorrhage of animals, which have consumed contaminated feed.5,6 Therefore, authorities Europe and North America have set limits for toxin concentrations in food and feed. The European Union has no standardized limits for DON in food and feed yet. However, the European Commission is about to harmonize limits for DON and has already proposed limits for cereals and corn at 0.50 (for cereal products as consumed) and 0.75 mg/kg (for flour as raw material in food processing).7 The Food and Drug Administration (FDA) has set Advisory Levels for DON in the United States at 1.00 mg/kg for products ready for human consumption and from 5.00 to 10.0 mg/kg depending on the type of commodity.8 New methods have to match those limits in order to be relevant for consideration. There are ongoing programs that deal with mycotoxin research on a national and international level.9 Several methods have been developed to assess the toxicity of moldy food and feed. The determination of the toxin itself10 and the determination of the ergosterol content,11-13 which serves as an indicator for the total fungal biomass,13 are among the most frequently employed methods of analysis. However, application of these methods is tedious and requires a lot of time. Several sample preparation steps such as extraction and cleanup are necessary before separation and detection can take place using gas chromatographic14 or liquid chromatographic techniques.15 Considering the time necessary to obtain a result (up to 90 min for a sample batch: 30-min extraction, 30-min cleanup, and 30 min for HPLC separation and detection), the need for a rapid (4) Marasas, W. F. O. In Mycotoxins and Animal Foods; Smith, J. E., Henderson, R. S., Eds.; CRC Press: Boca Raton, 1991. (5) D’Mello, J. P. F.; Placinta, C. M.; Macdonald, A. M. C. Anim. Feed Sci. Technol. 1999, 80, 183-205. (6) Hochsteiner, W.; Kahlbacher, H.; Schuh, M. Tiera ¨ rztl. Monatsschr. 2001, 342-6. (7) European Commission, DG-SANCO SANCO/1925/00-rev1, 2000; pp 1-6. (8) Trucksess, M. W. J. AOAC Int. 1995, 78, 135-41. (9) Pohland, A. E. In Mycotoxins and Phycotoxins: Developments in Chemistry, Toxicology and Food Safety, 1 ed.; Miraglia, M., Van Egmond, H. P., Brera, C., Gilbert, J., Eds.; Alaken Inc.: Fort Collins, 1998; Chapter 2. (10) Krska, R.; Baumgartner, S.; Josephs, R. D. Fresenius J. Anal. Chem. 2001, 371, 285-99. (11) Norme Fr. 1991, NF V18-112, 1-10. (12) Schwadorf, K.; Mu ¨ ller, H.-M. J. AOAC Int. 1989, 72, 457-62. (13) Seitz, L. M.; Mohr, H. E.; Burroughs, R.; Sauer, D. B. Cereal Chem. 1977, 54, 1201-17. (14) Weingaertner, J.; Krska, R.; Praznik, W.; Grasserbauer, M.; Lew, H. Fresenius J. Anal. Chem. 1997, 357, 1206-10. (15) ISO New Work Item Proposal 1994, 1-5.
Analytical Chemistry, Vol. 75, No. 5, March 1, 2003 1211
method is obvious, when thinking about the amounts of agricultural cereals and corn processed in the food and feed industry. Frequent checks are required by the authorities in Europe and the United States for imported commodities and processed foods. A small number of rapid methods has been proposed in order to assess the amount of fungal contamination in a given sample. Visual inspection is widely used for the assessment of corn cobs, which estimates the amount of damaged or invaded kernels on the cob.16 The method is labor-intensive (each single cob has to be inspected and evaluated) and requires a lot of experience. Furthermore, the amount of fungus present on the cob is not always visible as it often resides inside the kernel17 and the visual inspection and evaluation of ground samples is not possible. Some infrared spectroscopic techniques have been developed as a means to identify samples infected by fungi.18,19 Photoacoustic infrared spectroscopy (IR-PAS) has been used for single-kernel analysis in order to detect Aspergillus flavus infection on corn.18,20 However, the investigation of single kernels makes it difficult to draw conclusions for the sample as a whole. To overcome the inhomogeneity problem, samples are usually ground prior to analysis in order to obtain a representative sample.21 Additionally, no data for the fungal biomass (or an equivalent parameter quantifying the fungal content) or a toxic secondary metabolite were given in the study conducted by Gordon et al.18 They relied on a reference test that takes the fluorescence of A. flavus-infected kernels into account insteadsa unique property, which does not occur with Fusarium infected commodities. Furthermore, IR-PAS requires sensitive instrumentation, which is not suitable for usage in the field. Other approaches employing infrared spectroscopy included the use of diffuse reflectance spectroscopy (DRS) accessories.22 In principle suitable for the measurement of inhomogeneous samples, DRS usually requires dilution of the sample with KBr or NaCl in order to achieve better absorbance characteristics, especially for highly absorbing matrixes such as carbohydrates.23 Usually dilutions of a 1% sample in KBr/NaCl are used, making it difficult to obtain a homogeneously mixed sample. The water content of the sample might also result in further difficulties during measurements and requires constant purging of the sample chamber. It is necessary with DRS measurements to minimize interferences from scattering, which can be related to particle size effects, making DRS a method that requires a lot of experience, when it comes to interpreting the obtained spectra and the identification of possible artifacts.24 Greene et al. compared results (16) Bechtel, D. B.; Kaleikau, L. A.; Gaines, R. L.; Seitz, L. M. Cereal Chem. 1985, 62, 191-7. (17) Greene, R. V.; Gordon, S. H.; Jackson, M. A.; Bennett, G. A.; McClelland, J. F.; Jones, R. W. J. Agric. Food Chem. 1992, 40, 1144-9. (18) Gordon, S. H.; Schudy, R. B.; Wheeler, B. C.; Wicklow, D. T.; Greene, R. V. Int. J. Food Microbiol. 1997, 35, 179-86. (19) Wheeler, B. C.; Gordon, S. H.; DeWispelare, D. R.; Jackson, M. A.; Greene, R. V. Proceedings of the 15th International Conference of the Institute of Electrical and Electronic Engineers: Engineering, Medical and Biological Society; 1993; pp 7037-8. (20) Gordon, S. H.; Wheeler, B. C.; Schudy, R. B.; Wicklow, D. T.; Greene, R. V. J. Food Prot. 1998, 61, 221-30. (21) Park, D. L.; Pohland, A. E. J. AOAC Int. 1989, 72, 399-404. (22) Winson, M. K.; Goodacre, R.; Timmins, E. M.; Jones, A.; Alsberg, B. K.; Woodward, A. M.; Rowland, J. J.; Kell, D. B. Anal. Chim. Acta 1997, 348, 273-82. (23) Olinger, J. M.; Griffiths, P. R. Appl. Spectrosc. 1993, 47, 695-701. (24) Olinger, J. M.; Griffiths, P. R. Appl. Spectrosc. 1993, 47, 687-94.
1212
Analytical Chemistry, Vol. 75, No. 5, March 1, 2003
from PAS and DRS measurements, but again did not give any reference concentrations for ergosterol or toxic metabolites.17 By taking the above considerations as well as requirements from the food and feed industry into account, several criteria can be formulated that have to be matched by a rapid method: (a) minimal sample preparation and measurement time; (b) high sample throughput; (c) suitable for application in the field (robust instrumentation); (d) ease of use and automation possible; (e) quick overview and identification of samples, which require further investigation This paper describes the development of a rapid method using mid-infrared spectroscopy with attenuated total reflection (ATR). This setup was chosen for the study, because it matches the above-mentioned criteria well: Apart from grinding and sieving, which can be performed in a single step, no further sample preparation is required. The sample is placed on a horizontally mounted diamond crystal, and a pressure applicator ensures full contact between sample and crystal with reproducible pressure. With a standard DTGS detector, measurements are easy and no servicing is necessary. Together with a low maintenance and rugged spectrometer, in-the-field measurements are possible. A high sample throughput can be achieved with more than 200 measurements a day, because a sample scan takes less than 1 min (using 32 coadded scans for one spectrum). An earlier publication of the authors described first steps of the approach discussed here, but some of the investigated samples contained ergosterol at levels that were far too high in order to be of practical relevance.25 Furthermore, parts of the results were obtained using a model system of artificially inoculated and mixed corn, whereas samples in this study were solely obtained from naturally infected corn with toxin and ergosterol concentrations being a lot lower. All naturally contaminated samples chosen for modeling here were already used in the previous study in order to allow for a direct comparison of results and method performance. Corn was chosen as a commodity, because it is almost exclusively infected by Fusarium fungi during the growth period in moderate climates, thus minimizing contamination from other fungal species and making it a good model system to study.1 Although the near-infrared region usually is the method of choice for spectroscopic analyses in the food and feed industry,26 the mid-infrared range was investigated, because it contains a wealth of information from functional groups to vibrations of the whole molecule.23 Changes in protein, lipid, and carbohydrate contents can easily be observed by looking at the respective regions in the spectrum (see Figure 3 and Table 1).27 EXPERIMENTAL SECTION Sample Preparation. Corn samples of the same genotype (RWA2) that were predominantly naturally infected with Fusarium graminearum in the field during the growth period were used in this study. After harvesting, the samples were dried as whole cobs for 3 days using air at 40 °C. Thus, the moisture content was reduced to 710,
