**3.1 Background and methods**

186 Agricultural Science

Fungus can be detected by microbiological methods involving visual, microscopy, and

Conventional methods of mold detection are based on direct observation by eye or by microscope of thalli, contaminated foodstuffs, or microbial cultures. These methods are time consuming and require viable samples and a good deal of expertise. Counting methods are difficult to apply to fungi because, during their reproduction, a spore generates a mycelium that can in turn divide itself into tens of individuals. Furthermore, a fungal contamination

Other methods are based on molecular biology or on the detection of antigens specific to given molds. Organisms, either dead or alive, can be detected by the polymerase chain reaction (PCR) by copying a large number of DNA sequences that are originally present in small quantities (with a multiplicative factor on the order of 109). By amplifying certain genes of toxigenic strains, PCR serves as a tool to determine the risk. Various researchers have tested PCR to detect *Fusarium* contamination in corn [Jurado et al., 2006; Jurado et al., 2005; Nicolaisen et al., 2009]. These methods are rapid, sensitive, and can be automated. They are good qualitative methods (e.g., good selectivity) but offer only average precision in quantitative terms (they are called "semiquantitative"). These techniques are thus very reliable, provided the fungal strain to be detected is known beforehand, and so are used as referential methods. With such methods, a grain is deemed of suitable microbiological

**2. Advantages of using infrared spectroscopy to manage fungal and** 

may not be visible at the surface of grains [Hirano, et al., 1998, Pearson, et al., 2001].

quality if less than 10 000 germs of the storage flora per gram of grain are detected.

types of compounds can be used as indicators of a fungal contamination.

Ergosterol is a component of fungal cell membranes.

molecule but also detects glucans by near-infrared spectroscopy.

New approaches are based on detecting constituents and fungal metabolites. Such approaches exploit the fact that molds have specific characteristics that distinguish them from other eukaryotes. These characteristics include the regulation of certain enzymes, the synthesis of lysine amino acid by a particular metabolic route, extremely structural characteristics (e.g., the Golgi apparatus), and genetic characteristics (e.g., haploid). From among these attributes, two

The secreted compounds are synthesized compounds such as soluble carbohydrates (e.g., disaccharide trehalose and polyhydric alcohols such as mannitol or arabitol) or products of the metabolization of complex carbons such as volatile aldehydes, alcohols, ketones, spores, primary metabolites, secondary metabolites (i.e., volatile compounds). The last item gives rise to the characteristic fungal odor and is often detected by an electronic nose. For nonvolatile compounds, other tools such as infrared spectroscopy seem better suited.

The structural compounds of mold can also be used for their detection. The main polysaccharides of the cell wall of mold are the α et β (1-3) glucans, as well as chitin.

Chitin may absorb infrared light, making it useful for infrared spectroscopy [Nilsson, et al., 1994; Roberts et al., 1991]. The main inconvenience in using this component as an indicator of fungal contamination is that chitin is not limited to fungi; it is found in insects, diatoms, arachnids, nematodes, crustaceans, and several other living organisms [Muzzarelli, 1977]. In addition, it may take different forms, each of which requires a specific detection technique. Roberts et al. [Roberts, et al., 1991] estimates the quantity of mold on barley by detecting this

**mycotoxic risk for wheat, barley, and corn** 

microbial-cultural methods.

The first application of infrared spectroscopy to detect microorganisms dates from the 1950s [Miguel Gomez et al., 2003]. In these applications, the spectrometers were calibrated depending on the method of dosing the fungi or mycotoxins. In the 1980s, Fraenkel *et al.* [Fraenkel et al., 1980] and Davies *et al.* [Davies et al., 1987] published their first works on the detection by infrared spectroscopy of fungal contamination (*Botrytis cinerea* and *Alternaria tenuissima*), but the application of this tool to detecting mold really grew in the 1990s. This growth was due to the fact the existing agronomic models required collecting a significant amount of data in the field, making this approach unsuitable for routine use. In addition, industry required nondestructive techniques to assess the health safety of crops. Therefore, several research teams used infrared spectroscopy to detect mold and mycotoxins on cereals, which could be done concomitantly with the quantification of other parameters such as protein content, humidity, etc.

One method proposed to determine the fungal or mycotoxin content is to quantify the total fungal biomass. Toward this end, ergosterol is used as a fungus marker [Castro et al., 2002; Saxena et al., 2001; Seitz et al., 1977; Seitz et al., 1979]. Very often, this type of study is coupled with a study of the mycotoxin content and fungal units (colony-forming units or CFU). Indeed, the quantity of fungi is not proportional to the quantity of mycotoxins; it is possible to have small quantities of fungi but large quantities of mycotoxins, and vice versa. Indeed, fungi may disappear after secreting its toxins, either because of the evolution of the mycoflora or because of the application of chemical treatments. In addition, certain strains are more toxic than others. Two conclusions exist from the work on this subject: some researchers find a correlation between the mycotoxin content, the ergosterol content, and/or the fungal units [Lamper et al., 2000; Le Bouquin et al., 2007; Miedaner et al., 2000; Seitz et al., 1977; Wanyoike et al., 2002; Zill et al., 1988] whereas the others find no correlation or cannot make categoric conclusions [Beyer et al., 2007; Diener et al., 1982; Gilbert et al., 2002; Nowicki 2007; Penteado Moretzsohn De Castro et al., 2002; Perkowski,et al., 1995].

Covering the last 20 years, we count over 20 articles dealing with the use of infrared spectroscopy (primarily near-infrared) to detect molds and mycotoxins in wheat, barley, and corn. Because some of the work in infrared spectroscopy deals both with the detection and the identification of mycotoxins, we separate the articles into three groups. Table 1 is for molds, Table 2 compiles the trials dealing with deoxynivalenol (DON), fumonisins (FUMs), and B1 aflatoxins (AF1). Finally, Table 3 contains articles in which the authors worked


Table 1. Infrared spectroscopy applied to identification of mold species and gena (FTIR spectroscopy =Fourier transform infrared spectroscopy). 

Infrared Spectroscopy Applied to Identification and Detection

of Microorganisms and Their Metabolites on Cereals (Corn, Wheat, and Barley) 189

Table 2. Infrared spectroscopy applied to quantification of levels of deoxynivalenol (DON), fumonisins (FUM), and aflatoxins in

/ FTIR=Fourier-transform infrared spectroscopy)

cereals (NIR=Near Infrared Spectroscopy


Table 2. Infrared spectroscopy applied to quantification of levels of deoxynivalenol (DON), fumonisins (FUM), and aflatoxins in cereals (NIR=Near Infrared S pectroscopy / FTIR=Fourier-transform infrared s pectroscopy)

188 Agricultural Science

Table 1. Infrared spectroscopy applied to identification of mold species and gena (FTIR spectroscopy =Fourier transform infrared

spectroscopy).


Table 3. Infrared spectroscopy applied to fungal and mycotoxic quantification in cereals (NIR=Near Infrared Spectroscopy / FTIR=Fourier-transform infrared spectroscopy / ATR=Attenuated Total Reflection)

Infrared Spectroscopy Applied to Identification and Detection

principle conclusions, and the characteristic wavelengths.

**3.2.1 Identification and quantification of fungi** 

carbohydrate level (starch, cellulose, etc.).

Increasing the number of samples;

ranges are often very narrow);

multicontaminations);

mycotoxins;

(RPD).

for improvement of these studies may be the following:

are thus better than they would be for an external test);

**3.2 Principal conclusions** 

fungi.

results.

of Microorganisms and Their Metabolites on Cereals (Corn, Wheat, and Barley) 191

simultaneously on the fungal and mycotoxic aspects. Each table lists the matrix studied (wheat, barley, corn), the apparatus, the content ranges, the performance of the models, the

For identification of fungi, the performances, given in terms of percentage of correct classification, are very satisfactory because they exceed 77%. Each study identifies the peaks or spectral zones related to the growth of fungi or to the damage inflicted on the grains by

Moreover, the performance of the quantification of ergosterol always gives enticing

Mycotoxins are present in quantities too small (in the order of parts per million) for direct detection. Their detection is thus associated with a complex ensemble of information related to the growth of the fungus on the cereal; notably with modifications of the protein or

Regarding the capabilities of infrared spectroscopy to quantify mycotoxins, the conclusions differ from one author to another. In general, when dealing with deoxynivalenol, the performance is higher than when dealing with fumonisins. Yet despite this, even if the quantification of mycotoxins appears possible, it is not sufficiently precise to be used in the field. Indeed, the standard error of prediction (SEP) is too large with respect to the regulatory limits—notably European limits. This could be explained by a magnification of the non-negligible standard errors of the chemical benchmarks from which they are developed. Moreover, to work under conditions of realistic of toxin levels, the main avenues

**3.2.2 Quantification of mycotoxins deoxynivalenol, fumonisins, and aflatoxins** 

Increasing the annual variability (samples often come from a single harvest);

 Allowing a natural contamination of the grains (artificial contamination does not account for all the natural parameters of contamination, notably so for

Accepting a range of mycotoxin levels more adapted to the reality in the field (the

 Acquiring spectra not grain-by-grain, but from entire lots of grain. Indeed, it is more difficult to assign a global mycotoxin level to a lot, because the distribution of mycotoxins is very heterogeneous, as is the case for the molds that synthesize the

Using a set of test samples (the proposed performances are often cross validated and

Displaying the ratio of standard error of prediction to sample standard deviation

simultaneously on the fungal and mycotoxic aspects. Each table lists the matrix studied (wheat, barley, corn), the apparatus, the content ranges, the performance of the models, the principle conclusions, and the characteristic wavelengths.
