- Email: [email protected]
Optical detection of aflatoxins in maize using one- and two-photon induced fluorescence spectroscopy
Food Control 51 (2015) 408e416
Contents lists available at ScienceDirect
Food Control journal homepage: www.elsevier.com/locate/foodcont
Optical detection of aflatoxins in maize using one- and two-photon induced fluorescence spectroscopy L. Smeesters*, W. Meulebroeck, S. Raeymaekers, H. Thienpont Vrije Universiteit Brussel, Faculty of Engineering, Dept. of Applied Physics and Photonics (TONA), Brussels Photonics Team (B-PHOT), Pleinlaan 2, B-1050 Brussel, Belgium
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 August 2014 Received in revised form 24 November 2014 Accepted 3 December 2014 Available online 17 December 2014
The presence of aflatoxins in food and feed products is considered as one of the most important food safety problems in the world. Aflatoxins occur in a wide range of food products and can cause serious health risks. Moreover, they can nowadays only be detected by the use of destructive, time-consuming and expensive chemical analyses. We investigate the use of one- and two-photon induced fluorescence spectroscopy as nondestructive detection methods for the identification of aflatoxins. Particularly, as the samples under test, we consider the aflatoxin-contamination of different maize batches since maize is the staple food in many countries and cultivates in climates that show an extensive presence of the fungi. We first characterize the one- and two-photon induced fluorescence spectrum of pure aflatoxin B1, when excited with 365 nm and 730 nm laser light respectively. Subsequently, we experimentally investigate the fluorescence spectrum of various healthy and aflatoxin-contaminated maize samples, when excited with 365 nm, 405 nm, 730 nm, 750 nm, 780 nm and 810 nm laser light. For all excitation wavelengths, an intrinsic fluorescence signal of the maize grains is observed. However, for the contaminated maize grains, the present aflatoxin B1 significantly influences the intrinsic fluorescence. Depending on the excitation wavelength, we observe a different spectral contrast between the healthy and contaminated samples. The largest optical difference is observed for excitation with 365 nm and 780 nm, during the one- and two-photon induced fluorescence measurements respectively. The comparison of the measured fluorescence signals allows us to define a detection criterion for the optical identification of the contaminated maize samples. We can conclude that fluorescence spectroscopy can be a valuable tool for the measurement of aflatoxin-contents in maize, paving the way for real-time nondestructive industrial scanning-based detection. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Fluorescence spectroscopy Two-photon induced fluorescence Spectroscopy Optical sensing Maize Aflatoxin
1. Introduction The presence of several mycotoxins, secondary metabolites of toxic fungi, in food and feed commodities is a long standing food safety problem. These naturally toxic chemical compounds are observed on a wide range of agricultural commodities and under a diverse range of environments, resulting in a worldwide problem (Hruska et al., 2013). The Food and Agriculture Organization (FAO) estimates that 25% of the world's food crops are affected by mycotoxin producing fungi. Moreover, the accumulation of mycotoxins in foods and feeds represents a major threat to human and animal health, because they can induce cancer, liver diseases,
* Corresponding author. Tel.: þ32 2 629 10 19. E-mail address: [email protected] (L. Smeesters). http://dx.doi.org/10.1016/j.foodcont.2014.12.003 0956-7135/© 2014 Elsevier Ltd. All rights reserved.
immune-system suppression, mutagenicity and nervous disorders € hmannsro € ben, 2010b). (Rasch, Kumke, & Lo One of the most dominant mycotoxins in agriculture is aflatoxin. Aflatoxins are produced by the fungi Aspergillus flavus and Aspergillus parasiticus and occur in more than ten varieties, of which € ttcher, & aflatoxin B1 is the most dominant and toxic one (Rasch, Bo Kumke, 2010a). Its ingestion is associated with serious health risks and possibly death in many animal species and humans. Moreover, aflatoxins can contaminate the food both before and after harvest and cannot be destroyed by any form of food processing (Hruska et al., 2013). They may be present in a wide range of food products, like maize, cheese, pistachios, fruits, almonds and peanuts. However, the aflatoxin-contamination of maize is considered as a major threat, because maize is the staple food in many countries and cultivates in climates that show an extensive presence of the fungi, giving rise to permanent high aflatoxin-contamination levels.
L. Smeesters et al. / Food Control 51 (2015) 408e416
In this paper, we therefore focus on the presence of aflatoxin B1 contamination in maize grains. The presence of aflatoxins in food and feed products is regulated in over 100 nations (Unnevehr & Grace, 2013). The European Commission for example states the maximum allowed total aflatoxin concentration in maize to be 10 ppb, while the USA food safety regulations included a limit of 20 ppb of total aflatoxins in all food products (Romer Labs, 2014). To fulfill these limitations, the presence of aflatoxins is nowadays identified with the use of chemical analyses, like high-performance liquid chromatography (HPLC) and liquid chromatography-tandem mass spectrometry (LC-MS/MS). However, these analytical methods are generally expensive, time-consuming and destructive (Krska & Molinelli, 2007). Because of the uneven presence of the toxin in both the food products and the crops, the sample-based analysis often gives a limited view on the degree of contamination. The true aflatoxin concentration cannot be determined with 100% accuracy, causing difficulties in the elimination of aflatoxin-contaminated food and feed without destroying a large amount of products and incurring significant economic losses. The FAO estimates the global loss of foodstuff due to mycotoxins in the range of a billion tons per year (Rasch, Kumke, et al., 2010b). A real-time, non-destructive, cost-efficient and accurate method for the detection of aflatoxins is highly needed. We therefore propose the use of optical spectroscopic detection techniques to identify the aflatoxins. In the food industry, different spectroscopic techniques have already gained popularity for the real-time quality evaluation of food products (Meulebroeck & Thienpont, 2006, 2012; Saranwong, Sornsrivichai, & Kawano, 2014). With the use of absorption and fluorescence spectroscopy, it is for example possible to measure the water content of the products, to detect color deviations and to identify foreign objects in food streams. However, toxic contaminants in food or feed products can nowadays hardly be detected optically. The use of NIR spectroscopy for the determination of the mycotoxin deoxynivalenol is discussed by Petterson et al., but requires further improvements to lower the limit of detection (Petterson & Aberg, 2003). The current published fluorescence measurements only allow the identification of toxins €ttcher, et al., 2010a). In in liquids, like beer or wine (Rasch, Bo addition, these fluorescent toxins can only be identified if no or very low background fluorescent elements are present. Maize grains contain various fluorescent proteins which difficult the aflatoxin detection. Nevertheless, because aflatoxin is a fluorescent substance, fluorescence spectroscopy seems the most promising non-destructive optical detection technique. To detect the aflatoxins, we study both the one- and two-photon induced fluorescence spectra of healthy and contaminated maize grains. One-photon induced fluorescence is a linear process, giving rise to strong fluorescence signals, which can result in a strong aflatoxin signal of the solid food products and therefore the detection of low contamination levels. The two-photon induced fluorescence process generates weaker fluorescence signals, but features a more selective excitation of the aflatoxins, which can minimize the influence of the natural background fluorescent signals. Considering the practical implementation of one- and two-photon induced fluorescence spectroscopy, different excitation and detection criteria need to be taken into account. One-photon induced fluorescence (OPIF) can be obtained with smaller excitation power densities than two-photon induced fluorescence (TPIF). However, TPIF is obtained after excitation with near-infrared (NIR) laser excitation, which is more widely commercially available than the required ultraviolet (UV) laserline (of 365 nm) used during OPIF. In addition, compact NIR lasers generally feature larger output powers than the UV lasers, enhancing the fluorescence signal. To account for the natural variation within the maize samples, different sets of
409
maize grains are investigated. Moreover, because of the local presence of the aflatoxin inside the maize grains and the crops, a careful selection of the maize grains and the illumination position on the maize surface is indispensable to obtain accurate fluorescence data. For both the OPIF and TPIF measurements, we observe a significant difference between the fluorescence spectra of the healthy and contaminated samples. We investigate this optical difference in order to determine and optimize the optimum excitation and detection conditions for the optical identification of the aflatoxin-contaminated maize grains. The fluorescence excitation and detection conditions can afterwards be implemented into an industrial scanning setup, to obtain a real-time detection of the contaminated products in a product stream. Summarized, our main goal is the investigation of one- and twophoton induced fluorescence spectroscopy as a non-destructive optical detection technique for the identification of aflatoxins in solid maize grains. We pursue an optical detection implementable in industrial ultra-fast, laser-based optical scanning machines usable immediately after the harvest without preprocessing or grinding of the maize. Moreover, in contrast to the traditional sample-based chemical analyses, we want to evaluate the contamination of each maize grain individually to minimize food waste. This paper is structured as follows. In Section 2 we give an overview of the examined maize samples and discuss the fluorescence measurement setup. We validate our measurement setup by the characterization of the pure aflatoxin B1 powder. In Section 3 we investigate the fluorescence spectra of different healthy and aflatoxin-contaminated maize samples, when excited with 365 nm, 405 nm, 730 nm, 750 nm, 780 nm and 810 nm laser light. The intrinsic fluorescence of the maize grains is studied for each excitation wavelength, after which the difference between the fluorescence spectra of the healthy and aflatoxin-contaminated maize samples is quantified. Finally, a comparison of the optical performance of OPIF and TPIF is given. 2. Materials and methods To study the one- and two-photon induced fluorescence spectra of healthy and contaminated maize grains some challenges need to be tackled. First of all, the maize grains contain natural fluorescent proteins, like trypthophan and riboflavin, which can disturb the fluorescence measurement of the aflatoxin (Held, 2013). Secondly, the optical characterization of individual maize grains is much harder than optical measurements on homogenous solutions and powders. Natural maize grains feature a large internal variation in density, texture and substituents concentration. The investigation of different independent maize samples is required to minimize the influence of the maize type and the environment in which it was cultivated. In addition, since the aflatoxin contamination is mostly only locally present in the maize grains, we need to carefully select our samples as well as the areas on the maize grains that need to be illuminated. Finally, to be able to measure the weak aflatoxin fluorescence of the solid maize grains, an optimization of the excitation spot size, the excitation laser power, the excitation wavelength and the detector system of our fluorescence measurement setup is indispensable. In this section, we first give an overview of the investigated maize grains. Subsequently, we study the theoretical concepts and challenges of the OPIF and TPIF measurements. We discuss the fluorescence measurement setup and its optimization to obtain an optimal fluorescence spectrum. At the end, we validate the operation of our setup by the measurement and characterization of the one- and two photon induced fluorescence spectra of pure aflatoxin B1.
410
L. Smeesters et al. / Food Control 51 (2015) 408e416
2.1. Sample preparation We consider two different independent maize batches, each with a healthy and contaminated subsample. One healthy and one contaminated maize batch were harvested in 2012 and provided by an Italian company. The second set of healthy and contaminated maize samples was collected from Croatian farmers, after the harvest in 2013. Of each maize sample, a subsample of 25 g was drawn for the analytical determination of the aflatoxin contamination level. Each sample was chemically analyzed using the ToxiQuant mycotoxin testing system of ToxiMet (Toximet, 2014). Considering the Italian maize grains, the contaminated sample shows 72.1 ppb aflatoxin and the healthy one 0.0 ppb. The maize samples from Croatia show approximately the same aflatoxin contamination level, namely 78.9 ppb for the contaminated sample and 0.8 ppb for the healthy one. After our measurements, the contamination of the samples was confirmed by the CODA-CERVA, the Belgian Reference Laboratory for Mycotoxins. Before the start of the fluorescence spectrum measurements, the samples are first investigated by illumination with a UV light source. When illuminated with a UV light source, the healthy maize samples show no significant fluorescence signals. In contrast, the contaminated samples show localized fluorescent regions. However, with the use of a UV light bulb, only the regions of strongest fluorescence can be visualized. To be able to measure and characterize weak fluorescence signals, a sensitive setup is indispensable. During our fluorescence measurements, we investigate the fluorescence spectra of 45 healthy and contaminated Croatian maize grains. To observe the variation between the sample types and the harvest environments through the fluorescence spectra, we also measure the fluorescence spectra of 15 healthy and contaminated Italian maize grains. The Italian company provided only a small maize sample, which limits the number of studied maize grains but is sufficient to monitor the environmental influences onto the data variation. Because aflatoxin is sensitive to light, the samples are permanently stored in a dark enclosure to minimize the environmental influences onto the measurements. 2.2. Fluorescence measurement setup One- and two-photon induced fluorescence are based on the absorption of laser light, which excites an electron from the ground state to a higher energetic state and results in the emission of a new fluorescence photon during the relaxation process of the excited electron (Lakowicz, 1999, chap. 1). To generate a OPIF photon, only one excitation photon needs to be absorbed, while during the generation of a TPIF photon two excitation photons need to be absorbed simultaneously to excite the electron and generate the
fluorescence signal (Fig. 1). In the case of OPIF, the energy of the incident photon equals the energy difference between the electronic states, such that the electron obtains sufficient energy to bridge the energy gap between the ground state and the excited state (Fig. 1a). In the case of TPIF, the excitation occurs in two steps with two simultaneously incoming photons (Fig. 1b). The first incident photon excites the molecule to a virtual state, while the second incident photon excites the molecule from the virtual state to its higher-energy excited state. Because for both OPIF and TPIF the electron is excited to the same energetic state, they will give rise to the same fluorescence wavelengths, but with another excitation wavelength. Typically, OPIF occurs after excitation with UV laser light, while TPIF occurs after excitation with NIR laser light. NIR laserlines are more widely commercial available than the UV laser lines. Moreover, when using UV laser light in an optical setup, the optical mirrors and lenses need to be coated or fabricated in fused silica, resulting in a more expensive configuration. Two-photon absorption is a non-linear process that is generally many orders of magnitude weaker than the linear one-photon absorption occurring during the OPIF process. The OPIF intensity increases linearly with the excitation power, while the TPIF intensity increases with the square of the excitation power. For the same excitation laser power, the TPIF intensity is thus proportional to the square root of the OPIF intensity. To obtain a significant TPIF signal during our measurements, we need a large excitation power density, which is defined as the ratio of the excitation laser power to the excitation spot area. As a result, large excitation powers and small excitation spot sizes are indispensable to obtain a strong TPIF signal. We pursue a fluorescence measurement setup that is suited for both OPIF and TPIF measurements. However, before we could design our measurement setup, we first had to determine the interesting excitation wavelengths. To detect the aflatoxincontamination inside the maize grains, we need an excitation wavelength which is strongly absorbed by the aflatoxin, since strong fluorescence intensities are theoretically obtained after excitation with a wavelength for which the sample shows a strong absorbance (Rasch, Kumke, et al., 2010b). The stronger the absorbance of a sample, the more electrons will be excited and the more fluorescent photons will be emitted. By exciting the maize grains with an excitation wavelength that is strongly absorbed by the aflatoxin, we want to maximize the influence of the aflatoxin on the natural fluorescence spectrum of the maize grains. The absorbance spectrum of aflatoxin B1 shows the strongest absorbance in the range between 200 nm and 250 nm and around 365 nm (Fig. 2). We designed our measurement setup such that we are able to emit wavelengths around 365 nm, to study the OPIF spectrum. Consequently, to study the TPIF spectrum excitation wavelengths around
ski diagram for (a) OPIF and (b) TPIF. Fig. 1. Jablon
L. Smeesters et al. / Food Control 51 (2015) 408e416
Fig. 2. Absorbance and fluorescence spectrum of aflatoxin B1.
730 nm are required. In addition, we investigate the fluorescence spectrum of the maize grains when excited with 750 nm and 780 nm, to study the influence of the matrix constituents, the natural occurring fluorescent proteins, onto the fluorescence spectrum. The designed fluorescence measurement setup consists of two main parts, namely the excitation and detection side (Fig. 3a). For the excitation of the sample, a tunable titanium-sapphire laser (Spectra-Physics Tsunami laser), pumped by a frequency doubled Nd:YAG laser (continuous wave 532 nm laser), is used during the TPIF measurements (Fig. 3b). The wavelength of the tunable titanium-sapphire laser can be tuned between 710 nm and 835 nm. The maximum output power ranges from 1.20Watt to 1.50Watt, depending on the selected wavelength. During the OPIF measurements, the laser light of the titanium-sapphire laser is frequencydoubled with the use of a second harmonic generating crystal, to generate the necessary UV excitation wavelengths (Fig. 3b). The second harmonic generating crystal is able to generate wavelengths between 355 nm and 417 nm, with a maximal output power between 200 mW and 450 mW. Behind the output of the titanium-
411
sapphire laser and the frequency-doubling unit, the excitation light is directed towards the sample. To obtain a high excitation power density during the TPIF measurements, an additional focusing lens is positioned in front of the sample to minimize the excitation spot size (Fig. 3c). The sample holder is positioned on a translation stage, which allows the scanning of the sample. After the excitation of the sample, the fluorescent signals are captured by a collimating lens, coupled into a broadband optical fiber (UVIR600 fiber of Avantes) and guided towards the spectrum analyzer. In front of the detecting fiber, we position a long-wave pass filter during the OPIF measurements and a short-wave pass filter during the TPIF measurements to absorb the excitation light. Otherwise the excitation signals would saturate the measured fluorescence spectrum, leading to incorrect measurement data. The used spectrum analyzer (AvaSpec2084 spectrum analyzer of Avantes) is able to measure the spectrum between 300 nm and 1100 nm with a resolution of 8 nm. The measured spectra are corrected by a transfer function to account for the wavelength dependent transmittance of the optical fiber and the sensitivity of the detector inside the spectrum analyzer. 2.3. Validation of the measurement setup: characterization of the aflatoxin B1 fluorescence To validate the operation of our fluorescence measurement setup, we measure the OPIF and TPIF spectrum of pure aflatoxin B1 powder (Fig. 4). We purchased aflatoxin B1 powder, produced by the fungus A. flavus with 98% or better purity, from SigmaeAldrich. It is a white to yellow crystalline powder that we measure in its solid state. We present the OPIF spectrum that is obtained after excitation with 365 nm, with an excitation power density of 42 mW/mm2 (excitation power of 30 mW and an excitation spot diameter of
Fig. 3. Measurement setup, which allows the investigation of both OPIF and TPIF: (a) schematic representation of the setup; (b) tunable titanium-sapphire laser (710 nme835 nm) and harmonic generating unit that contains the frequency-doubling crystal (355 nme417 nm); (c) optical path for the excitation of the sample, after which the fluorescence spectrum is captured by the detecting fiber. The focusing lens minimizes the spot size during the TPIF measurements.
L. Smeesters et al. / Food Control 51 (2015) 408e416
0.2 1.5 0.15
OPIF
1
0.1
TPIF 0.5
0.05
0
400
428nm
450
500 Wavelength (nm)
550
Absolute TPIF intensity (uW/cm²/nm)
Abs olute OPIF intens ity (uW/c m ²/nm )
412
0 600
Fig. 4. OPIF and TPIF spectra of the pure aflatoxin B1 powder, after excitation with 365 nm and 730 nm respectively.
951 mm). The TPIF fluorescence spectrum is measured after excitation with 730 nm, with an excitation power density of 14317 mW/ mm2 (excitation power of 600 mW and an excitation spot diameter of 231 mm). Both fluorescence spectra show their maximal fluorescence intensity at 428 nm, which corresponds with the expected aflatoxin B1 fluorescence maximum indicated in Fig. 2. The maximal OPIF intensity, at 428 nm, is equal to 1.68 ± 0.93 mW/cm2, while we observe a maximal TPIF intensity of 0.20 ± 0.09 mW/cm2. The measured TPIF intensity is hence 10 times weaker than the OPIF intensity. Comparing the excitation power density and the resulting fluorescence intensity, the efficiency of the OPIF process is approximately 3000 times larger than the efficiency of the TPIF process. The shape of the TPIF spectrum shows a narrower peak than the OPIF spectrum, due to the more selective excitation during two-photon absorption than during one-photon absorption. To characterize the OPIF and TPIF spectra, we measure the fluorescence intensity as function of the excitation laser power, while maintaining a constant excitation spot size (Fig. 5). Subsequently, for each excitation power, we integrate the measured fluorescence spectrum to obtain the integrated fluorescence intensity. For the different excitation powers, we present the intensity of the mean integrated fluorescence spectrum. The variation of the measured data is less than 10% and 12%, during the OPIF and TPIF measurements respectively. Studying the integrated OPIF intensity as function of the excitation laser power, we observe a linear
relationship, which demonstrates the linear one-photon absorption. However, at higher excitation powers, starting from 35 mW onwards, the integrated OPIF intensity deviates from the linear relationship and starts saturating. In the saturated region, the maximum fluorescence intensity is reached since no more electrons can then be excited to the higher energy state. The TPIF intensity shows a quadratic dependence on the excitation laser power, confirming the occurrence of non-linear two-photon absorption. Both the linear and exponential fits show an adjusted rsquare value of 0.98 and 0.97 respectively, ensuring that the fitted function is a good representation of the measured data. Furthermore, considering the measured fluorescence intensities, we observe that TPIF requires higher excitation powers to obtain a measurable fluorescence spectrum. Specifically, taking the spot size into account, the TPIF process requires an excitation power density of 2000 mW/mm2, while the OPIF process can be generated with an excitation power density of 7 mW/mm2 (Smeesters, Meulebroeck, Raeymaekers, & Thienpont, 2014). The fluorescence measurements of the pure aflatoxin B1 powder correspond with the theoretical characteristics of OPIF and TPIF. Consequently, the above measurements validate the correct operation of our fluorescence measurement setup. In the next section, we investigate the one- and two-photon induced fluorescence spectra of the different sets of maize batches. We give an overview of measured fluorescence spectra, after which we quantitatively evaluate the optical difference between the healthy and contaminated samples. 3. Results and discussion To investigate the optical detection of aflatoxin B1 in maize, we study the OPIF and TPIF spectra of healthy and contaminated maize grains, of both the Croatian and Italian maize batches. We measured the OPIF spectrum after excitation with 365 nm, with an excitation laser power of 225 mW (Fig. 6a). The TPIF spectra are measured after excitation with 730 nm, 750 nm and 780 nm, with an excitation power of 1100 mW, 1220 mW and 1510 mW respectively (Fig. 6bed). During the TPIF measurements, we investigate the fluorescence spectrum after excitation with multiple excitation wavelengths to monitor the influence of the illumination wavelength onto the fluorescence of the maize grains. To maximize the signal to noise ratio, we illuminate the maize grains with the maximal output powers of the titanium-sapphire laser and the harmonic generating unit.
Fig. 5. Intensity of the integrated OPIF spectrum increases linearly with the excitation power, while the intensity of the integrated TPIF spectrum shows a quadratic dependence on the excitation power.
L. Smeesters et al. / Food Control 51 (2015) 408e416
413
Fig. 6. Fluorescence spectra of the healthy and contaminated maize grains of both the Croatian and Italian maize sample: (a) OPIF spectrum after excitation with 365 nm; (b) TPIF spectrum after excitation with 730 nm; (c) TPIF spectrum after excitation with 750 nm; (d) TPIF spectrum after excitation with 780 nm.
Both the healthy and the contaminated samples show a fluorescence signal, due to the intrinsic fluorescence of the maize grains (Fig. 6). For every maize batch, we display the mean fluorescence spectrum. Considering the different maize batches, the fluorescence spectra of the Croatian and Italian maize correspond well, within the variances of the measurements. For all excitation wavelengths, the fluorescence spectra of both maize types have the same shape and are present in the same wavelength regions. The displayed fluorescence intensity differences between the Croatian and Italian maize are within the variances of the measurement data (Table 1). These intensity variances are caused by the differences within the molecular structure of the maize grains. Because maize is a natural product, the maize grains show a large internal variation in the surface shape, density and natural composition. Comparing the fluorescence spectra of the healthy and contaminated maize grains, we observe significant differences in Table 1 Measured maximum fluorescence intensity, and its variation, for the different excitation wavelengths.
365 730 750 780
nm nm nm nm
excitation excitation excitation excitation
Maximum fluorescence intensity Croatian maize (mW/cm2/nm)
Maximum fluorescence intensity Italian maize (mW/cm2/nm)
Healthy
Healthy
15 0.09 0.07 0.06
± ± ± ±
8 0.05 0.03 0.04
Contaminated 17 0.02 0.02 0.02
± ± ± ±
10 0.01 0.01 0.01
25 0.08 0.05 0.03
± ± ± ±
17 0.05 0.02 0.01
Contaminated 13 0.02 0.02 0.02
± ± ± ±
9 0.01 0.01 0.01
intensity and emission wavelength. We do not directly observe the aflatoxin fluorescence, but we measure its influence onto the intrinsic fluorescence of the maize. For both the OPIF and TPIF measurements, the healthy and contaminated maize batches show different fluorescent intensities due to the different molecular structure of these maize grains (Table 1). In the contaminated samples, the aflatoxin is bonded to the different natural constituents of the healthy maize, changing the molecular structure of the constituents and therefore influencing the natural fluorescence of the maize. Moreover, this intensity contrast is the largest for the TPIF spectra, since the different bonds inside the maize are more selectively excited during the TPIF process than during the OPIF one. The largest intensity differences between the healthy and contaminated samples are observed after excitation with 730 nm, where the mean fluorescence intensity of the healthy maize is 4 times stronger than the mean fluorescence intensity of the contaminated maize. In contrast, the minimum intensity differences are observed after excitation with 365 nm. However, as we expect, the OPIF intensity is much stronger than the TPIF intensity. Specifically, the TPIF signal is approximately 500 times weaker than the OPIF signal. In addition to the fluorescence intensity differences, we also observe a wavelength shift between the fluorescence maxima of the healthy and contaminated maize samples (Table 2). Generally, the OPIF and TPIF spectra show a wavelength shift of approximately 50 nm. To compare the obtained wavelength shift for the different excitation wavelengths, we calculate the class difference for both the Croatian and Italian maize. The class difference (D) is a measure for the difference between the average values (m) of two product types, taking the standard deviation (s) and the
414
L. Smeesters et al. / Food Control 51 (2015) 408e416
amount of measured samples (N) into account (Downie & Heath, 1970, chap. 12):
mcontaminated mhealthy D ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi s2contaminated Ncontaminated
þ
(1)
s2healthy Nhealthy
The OPIF measurements show the largest class differences, indicating the largest optical contrast between the healthy and contaminated maize grains. The variation of the class differences between the OPIF and TPIF measurements are mainly caused by their different variances. The TPIF spectra show a larger variance than the OPIF spectra, which decreases their class difference. TPIF signals show a weaker intensity, resulting in a stronger relative noise signal and therefore a lower signal to noise ratio. The fluorescence spectrum of the contaminated maize shows weaker fluorescence intensities and a wavelength shift of the fluorescence maximum towards longer wavelengths. To validate the influence of the aflatoxin onto the maize intrinsic fluorescence spectrum, we sweep the excitation wavelength towards the wavelengths for which aflatoxin B1 shows a weaker absorbance. We measured the fluorescence spectrum of the healthy and contaminated maize samples after excitation with 405 nm and 810 nm, during the OPIF and TPIF measurements respectively (Fig. 7). Both the OPIF and TPIF measurements show a small difference between the fluorescence spectrum of the healthy and contaminated samples. We measure a wavelength shift of 8 ± 6 nm and 26 ± 12 nm, after excitation with 405 nm and 810 nm respectively. Moreover, we observe only small differences between the fluorescence intensities. Aflatoxin B1 shows a weak absorbance at 405 nm (Fig. 2), resulting in a minor influence onto the intrinsic fluorescence spectrum of the maize grains. So far, we compared the emission wavelength and intensity of the maize fluorescence spectra. We did not account for the shape of the fluorescence spectra. To obtain a complete quantitative
comparison between the measured fluorescence spectra, we study the integrated fluorescence intensity as function of the sample number. For both healthy and contaminated maize grains of the Croatian and Italian maize batches, the normalized fluorescence intensity from 475 nm until 550 nm is compared with the range from 400 nm until 475 nm, for excitation with 365 nm, 730 nm, 750 nm and 780 nm. In both wavelength intervals the fluorescence spectrum is integrated, after which we plotted the intensity ratio of these integrals (Fig. 8). For all OPIF and TPIF measurements the integral ratio shows a clear distinction between the healthy and contaminated samples. The contaminated samples show generally a larger variance because of the local presence of the aflatoxin and the variable aflatoxin-concentration in the maize grains. Moreover, the variance of the contaminated samples is larger for the TPIF measurements than for the OPIF ones. The largest variance is obtained after excitation with 780 nm (Fig. 8d). However, also the largest difference between the mean ratio of the healthy and contaminated samples is observed for excitation with 780 nm. Considering the class differences of the integral ratio between the healthy and contaminated samples, we obtain a class difference of 116.2, 111.1, 48.8 and 54.9, for excitation with 365 nm, 730 nm, 750 nm and 780 nm respectively. The largest contrast can thus be found after excitation with 365 nm. However, the contaminated samples can still be properly identified with TPIF, allowing using NIR excitation laserlines, instead of the UV laserlines for OPIF. Our measurement results clearly indicate the use of fluorescence spectroscopy as a promising detection technique for the indentification of aflatoxins in maize grains. Moreover, the commercial availability of high power NIR laserlines, optical bandpass filters and sensitive detectors allow its integration in practical systems. In comparison to the traditional chemical analyses, the optical detection techniques feature some important advantages enabling to improve our food safety. Traditional chemical analyses are destructive and measure the mean contamination level of a certain number of maize grains. In contrast, optical spectroscopy is
Table 2 Emission wavelength at maximum fluorescence intensity, and its variation, for the different excitation wavelengths. Dominant emission wavelength Croatian maize (nm) Healthy 365 730 750 780
nm nm nm nm
excitation excitation excitation excitation
443 441 444 450
± ± ± ±
5 3 7 10
Class difference Croatian maize
Contaminated 497 496 497 503
± ± ± ±
15 20 24 19
Dominant emission wavelength Italian maize (nm) Healthy
9.7 6.1 3.8 5.2
442 441 440 450
± ± ± ±
5 3 5 7
Class difference Italian maize
Contaminated 486 496 490 498
± ± ± ±
6 16 17 10
10.8 4.1 2.7 4.8
Fig. 7. Fluorescence spectra of the healthy and contaminated maize grains: (a) OPIF spectrum after excitation with 405 nm; (b) TPIF spectrum after excitation with 810 nm.
L. Smeesters et al. / Food Control 51 (2015) 408e416
415
Fig. 8. Contrast between the fluorescence spectra of the healthy and contaminated maize samples, expressed by the intensity ratio of the integrated fluorescence spectrum from 475 nm until 550 nm to the integrated fluorescence spectrum from 400 nm until 475 nm, after excitation with (a) 365 nm; (b) 730 nm; (c) 750 nm; (d) 810 nm.
an ultra-fast and non-destructive technique, since it does not use chemicals and does not requires grinding of the maize. Moreover, laser-based sorting systems are able to scan the sample, such that each maize grain can be individually evaluated and the possible local aflatoxin-contamination of the sample can be identified. 4. Conclusion We demonstrate the use of one- and two-photon induced fluorescence spectroscopy for the non-destructive detection of carcinogenic aflatoxins in maize. We measured the fluorescence spectrum of different healthy and contaminated maize batches, when excited with 365 nm, 405 nm, 730 nm, 750 nm, 780 nm and 810 nm. Both the healthy and contaminated maize grains show an intrinsic fluorescence signal. However in the contaminated samples (with 72.1 ppb and 78.9 ppb of aflatoxins) the intrinsic fluorescence of the healthy maize grains is influenced by the presence of the aflatoxin. In comparison to the healthy maize samples, the contaminated samples show lower fluorescence intensities and a wavelength shift of the fluorescence maximum. Both the measured one- and two-photon induced fluorescence spectra show a wavelength shift of approximately 50 nm. The largest intensity differences are observed after excitation with 730 nm, where the mean fluorescence intensity of the healthy maize is 4 times stronger than the fluorescence intensity of the contaminated maize. Moreover, to
consider both the shape and wavelength region of the fluorescence spectrum, we investigate the intensity ratio of the integrated fluorescence spectrum from 475 nm until 550 nm to the integrated fluorescence spectrum from 400 nm until 475 nm. The mean fluorescence ratio shows the largest contrast for excitation with 365 nm and 780 nm. Because the two-photon induced fluorescence spectra show a larger variance than the one-photon induced fluorescent ones, the largest class difference is obtained after excitation with 365 nm. In comparison to one-photon induced fluorescence, two-photon induced fluorescence requires higher excitation powers, smaller excitation spot sizes and a more sensitive detector. The measured one-photon induced fluorescence intensity is 500 times larger than the measured two-photon induced fluorescent one. However, the near-infrared excitation wavelengths for two-photon induced fluorescence are more widely commercially available than the required UV excitation wavelengths for one-photon induced fluorescence. Furthermore, when using UV laser light in an optical setup, the optical mirrors and lenses need to be coated or fabricated in fused silica, resulting in a more expensive configuration. The investigation of the optical differences between the fluorescence spectra of the healthy and contaminated products, together with the determination of the illumination and detector requirements, paves the way to real-time, industrial, non-destructive detection of the carcinogenic aflatoxins in food products.
416
L. Smeesters et al. / Food Control 51 (2015) 408e416
Acknowledgments This work was supported in part by FWO (G008413N), IWT (IWT120528), the COST Action (MP1205), the Methusalem and Hercules Foundations and the OZR of the Vrije Universiteit Brussel (VUB). The authors would also like to thank dr. ir. Alfons Callebaut and dr. ir. Bart Huybrechts of the CODA-CERVA, the National Reference Laboratory for Mycotoxins, for their valuable feedback and know-how about mycotoxins. We are very grateful to them for the execution of the liquid chromatography-tandem mass spectroscopy analytical technique. References Downie, N. M., & Heath, R. W. (1970). Basic statistical methods. New York: Harper. Held, P. (April 2013). Quantitation of peptides and amino acids with synergy™ HT using UV fluorescence. http://www.biotek.com/resources/docs/Synergy_HT_ Quantitation_of_Peptides_and_Amino_Acids.pdf. Hruska, Z., Yao, H., Kincaid, R., Darlington, D., Brown, R. L., Bhatnagar, D., et al. (2013). Fluorescence imaging spectroscopy (FIS) for comparing spectra from corn ears naturally and artificially infected with aflatoxin producing fungus. Journal of Food Science, 7(8), 1313e1320. Krska, R., & Molinelli, A. (2007). Mycotoxin analysis: state-of-the-art and future trends. Analytical and Bioanalytical Chemistry, 387(1), 145e148.
Lakowicz, J. R. (1999). Principles of fluorescence spectroscopy. New York: Kluwer Academic/Plenum Publishers. Meulebroeck, W., & Thienpont, H. (2006). Optical detection techniques for laser sorting machines. Proceedings of SPIE, 6189, 337e348. Meulebroeck, W., & Thienpont, H. (2012). The study of vegetation indices for the monitoring of differences in chlorophyll and carotenoid composition in green vegetables. Proceedings of SPIE, 8439, 843920. Petterson, H., & Aberg, L. (2003). Near infrared spectroscopy for determination of mycotoxins in cereals. Food Control, 14(4), 229e232. €ttcher, M., & Kumke, M. (2010). Determination of aflatoxin B1 in Rasch, C., Bo alcoholic beverages: comparison of one- and two-photon induced fluorescence. Analytical and Bioanalytical Chemistry, 397(1), 87e92. €hmannsro € ben, H. G. (2010). Sensing of mycotoxin proRasch, C., Kumke, M., & Lo ducing fungi in the processing of grains. Food and Bioprocess Technology, 3(6), 908e916. Romer Labs. (June 2014). Mycotoxins affect everyone: Regulations for food. http:// www.mycotoxins.info/myco_info/consum_regu.html#usa. Saranwong, S., Sornsrivichai, J., & Kawano, S. (2014). Prediction of ripe stage eating quality of mango fruit from its harvest quality measured nondestructively by near infrared spectroscopy. Postharvest Biology and Technology, 31(2), 137e145. Smeesters, L., Meulebroeck, W., Raeymaekers, S., & Thienpont, H. (2014). The use of one- and two-photon induced fluorescence spectroscopy for the optical characterization of carcinogenic aflatoxins. Proceedings of SPIE, 9198, 919803. Toximet. (June 2014). Our award-winning mycotoxin testing system. http://www. toximet.com/. Unnevehr, L., & Grace, D. (2013). Aflatoxins finding solutions for improved food safety. International food policy research institute.
Contents lists available at ScienceDirect
Food Control journal homepage: www.elsevier.com/locate/foodcont
Optical detection of aflatoxins in maize using one- and two-photon induced fluorescence spectroscopy L. Smeesters*, W. Meulebroeck, S. Raeymaekers, H. Thienpont Vrije Universiteit Brussel, Faculty of Engineering, Dept. of Applied Physics and Photonics (TONA), Brussels Photonics Team (B-PHOT), Pleinlaan 2, B-1050 Brussel, Belgium
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 August 2014 Received in revised form 24 November 2014 Accepted 3 December 2014 Available online 17 December 2014
The presence of aflatoxins in food and feed products is considered as one of the most important food safety problems in the world. Aflatoxins occur in a wide range of food products and can cause serious health risks. Moreover, they can nowadays only be detected by the use of destructive, time-consuming and expensive chemical analyses. We investigate the use of one- and two-photon induced fluorescence spectroscopy as nondestructive detection methods for the identification of aflatoxins. Particularly, as the samples under test, we consider the aflatoxin-contamination of different maize batches since maize is the staple food in many countries and cultivates in climates that show an extensive presence of the fungi. We first characterize the one- and two-photon induced fluorescence spectrum of pure aflatoxin B1, when excited with 365 nm and 730 nm laser light respectively. Subsequently, we experimentally investigate the fluorescence spectrum of various healthy and aflatoxin-contaminated maize samples, when excited with 365 nm, 405 nm, 730 nm, 750 nm, 780 nm and 810 nm laser light. For all excitation wavelengths, an intrinsic fluorescence signal of the maize grains is observed. However, for the contaminated maize grains, the present aflatoxin B1 significantly influences the intrinsic fluorescence. Depending on the excitation wavelength, we observe a different spectral contrast between the healthy and contaminated samples. The largest optical difference is observed for excitation with 365 nm and 780 nm, during the one- and two-photon induced fluorescence measurements respectively. The comparison of the measured fluorescence signals allows us to define a detection criterion for the optical identification of the contaminated maize samples. We can conclude that fluorescence spectroscopy can be a valuable tool for the measurement of aflatoxin-contents in maize, paving the way for real-time nondestructive industrial scanning-based detection. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Fluorescence spectroscopy Two-photon induced fluorescence Spectroscopy Optical sensing Maize Aflatoxin
1. Introduction The presence of several mycotoxins, secondary metabolites of toxic fungi, in food and feed commodities is a long standing food safety problem. These naturally toxic chemical compounds are observed on a wide range of agricultural commodities and under a diverse range of environments, resulting in a worldwide problem (Hruska et al., 2013). The Food and Agriculture Organization (FAO) estimates that 25% of the world's food crops are affected by mycotoxin producing fungi. Moreover, the accumulation of mycotoxins in foods and feeds represents a major threat to human and animal health, because they can induce cancer, liver diseases,
* Corresponding author. Tel.: þ32 2 629 10 19. E-mail address: [email protected] (L. Smeesters). http://dx.doi.org/10.1016/j.foodcont.2014.12.003 0956-7135/© 2014 Elsevier Ltd. All rights reserved.
immune-system suppression, mutagenicity and nervous disorders € hmannsro € ben, 2010b). (Rasch, Kumke, & Lo One of the most dominant mycotoxins in agriculture is aflatoxin. Aflatoxins are produced by the fungi Aspergillus flavus and Aspergillus parasiticus and occur in more than ten varieties, of which € ttcher, & aflatoxin B1 is the most dominant and toxic one (Rasch, Bo Kumke, 2010a). Its ingestion is associated with serious health risks and possibly death in many animal species and humans. Moreover, aflatoxins can contaminate the food both before and after harvest and cannot be destroyed by any form of food processing (Hruska et al., 2013). They may be present in a wide range of food products, like maize, cheese, pistachios, fruits, almonds and peanuts. However, the aflatoxin-contamination of maize is considered as a major threat, because maize is the staple food in many countries and cultivates in climates that show an extensive presence of the fungi, giving rise to permanent high aflatoxin-contamination levels.
L. Smeesters et al. / Food Control 51 (2015) 408e416
In this paper, we therefore focus on the presence of aflatoxin B1 contamination in maize grains. The presence of aflatoxins in food and feed products is regulated in over 100 nations (Unnevehr & Grace, 2013). The European Commission for example states the maximum allowed total aflatoxin concentration in maize to be 10 ppb, while the USA food safety regulations included a limit of 20 ppb of total aflatoxins in all food products (Romer Labs, 2014). To fulfill these limitations, the presence of aflatoxins is nowadays identified with the use of chemical analyses, like high-performance liquid chromatography (HPLC) and liquid chromatography-tandem mass spectrometry (LC-MS/MS). However, these analytical methods are generally expensive, time-consuming and destructive (Krska & Molinelli, 2007). Because of the uneven presence of the toxin in both the food products and the crops, the sample-based analysis often gives a limited view on the degree of contamination. The true aflatoxin concentration cannot be determined with 100% accuracy, causing difficulties in the elimination of aflatoxin-contaminated food and feed without destroying a large amount of products and incurring significant economic losses. The FAO estimates the global loss of foodstuff due to mycotoxins in the range of a billion tons per year (Rasch, Kumke, et al., 2010b). A real-time, non-destructive, cost-efficient and accurate method for the detection of aflatoxins is highly needed. We therefore propose the use of optical spectroscopic detection techniques to identify the aflatoxins. In the food industry, different spectroscopic techniques have already gained popularity for the real-time quality evaluation of food products (Meulebroeck & Thienpont, 2006, 2012; Saranwong, Sornsrivichai, & Kawano, 2014). With the use of absorption and fluorescence spectroscopy, it is for example possible to measure the water content of the products, to detect color deviations and to identify foreign objects in food streams. However, toxic contaminants in food or feed products can nowadays hardly be detected optically. The use of NIR spectroscopy for the determination of the mycotoxin deoxynivalenol is discussed by Petterson et al., but requires further improvements to lower the limit of detection (Petterson & Aberg, 2003). The current published fluorescence measurements only allow the identification of toxins €ttcher, et al., 2010a). In in liquids, like beer or wine (Rasch, Bo addition, these fluorescent toxins can only be identified if no or very low background fluorescent elements are present. Maize grains contain various fluorescent proteins which difficult the aflatoxin detection. Nevertheless, because aflatoxin is a fluorescent substance, fluorescence spectroscopy seems the most promising non-destructive optical detection technique. To detect the aflatoxins, we study both the one- and two-photon induced fluorescence spectra of healthy and contaminated maize grains. One-photon induced fluorescence is a linear process, giving rise to strong fluorescence signals, which can result in a strong aflatoxin signal of the solid food products and therefore the detection of low contamination levels. The two-photon induced fluorescence process generates weaker fluorescence signals, but features a more selective excitation of the aflatoxins, which can minimize the influence of the natural background fluorescent signals. Considering the practical implementation of one- and two-photon induced fluorescence spectroscopy, different excitation and detection criteria need to be taken into account. One-photon induced fluorescence (OPIF) can be obtained with smaller excitation power densities than two-photon induced fluorescence (TPIF). However, TPIF is obtained after excitation with near-infrared (NIR) laser excitation, which is more widely commercially available than the required ultraviolet (UV) laserline (of 365 nm) used during OPIF. In addition, compact NIR lasers generally feature larger output powers than the UV lasers, enhancing the fluorescence signal. To account for the natural variation within the maize samples, different sets of
409
maize grains are investigated. Moreover, because of the local presence of the aflatoxin inside the maize grains and the crops, a careful selection of the maize grains and the illumination position on the maize surface is indispensable to obtain accurate fluorescence data. For both the OPIF and TPIF measurements, we observe a significant difference between the fluorescence spectra of the healthy and contaminated samples. We investigate this optical difference in order to determine and optimize the optimum excitation and detection conditions for the optical identification of the aflatoxin-contaminated maize grains. The fluorescence excitation and detection conditions can afterwards be implemented into an industrial scanning setup, to obtain a real-time detection of the contaminated products in a product stream. Summarized, our main goal is the investigation of one- and twophoton induced fluorescence spectroscopy as a non-destructive optical detection technique for the identification of aflatoxins in solid maize grains. We pursue an optical detection implementable in industrial ultra-fast, laser-based optical scanning machines usable immediately after the harvest without preprocessing or grinding of the maize. Moreover, in contrast to the traditional sample-based chemical analyses, we want to evaluate the contamination of each maize grain individually to minimize food waste. This paper is structured as follows. In Section 2 we give an overview of the examined maize samples and discuss the fluorescence measurement setup. We validate our measurement setup by the characterization of the pure aflatoxin B1 powder. In Section 3 we investigate the fluorescence spectra of different healthy and aflatoxin-contaminated maize samples, when excited with 365 nm, 405 nm, 730 nm, 750 nm, 780 nm and 810 nm laser light. The intrinsic fluorescence of the maize grains is studied for each excitation wavelength, after which the difference between the fluorescence spectra of the healthy and aflatoxin-contaminated maize samples is quantified. Finally, a comparison of the optical performance of OPIF and TPIF is given. 2. Materials and methods To study the one- and two-photon induced fluorescence spectra of healthy and contaminated maize grains some challenges need to be tackled. First of all, the maize grains contain natural fluorescent proteins, like trypthophan and riboflavin, which can disturb the fluorescence measurement of the aflatoxin (Held, 2013). Secondly, the optical characterization of individual maize grains is much harder than optical measurements on homogenous solutions and powders. Natural maize grains feature a large internal variation in density, texture and substituents concentration. The investigation of different independent maize samples is required to minimize the influence of the maize type and the environment in which it was cultivated. In addition, since the aflatoxin contamination is mostly only locally present in the maize grains, we need to carefully select our samples as well as the areas on the maize grains that need to be illuminated. Finally, to be able to measure the weak aflatoxin fluorescence of the solid maize grains, an optimization of the excitation spot size, the excitation laser power, the excitation wavelength and the detector system of our fluorescence measurement setup is indispensable. In this section, we first give an overview of the investigated maize grains. Subsequently, we study the theoretical concepts and challenges of the OPIF and TPIF measurements. We discuss the fluorescence measurement setup and its optimization to obtain an optimal fluorescence spectrum. At the end, we validate the operation of our setup by the measurement and characterization of the one- and two photon induced fluorescence spectra of pure aflatoxin B1.
410
L. Smeesters et al. / Food Control 51 (2015) 408e416
2.1. Sample preparation We consider two different independent maize batches, each with a healthy and contaminated subsample. One healthy and one contaminated maize batch were harvested in 2012 and provided by an Italian company. The second set of healthy and contaminated maize samples was collected from Croatian farmers, after the harvest in 2013. Of each maize sample, a subsample of 25 g was drawn for the analytical determination of the aflatoxin contamination level. Each sample was chemically analyzed using the ToxiQuant mycotoxin testing system of ToxiMet (Toximet, 2014). Considering the Italian maize grains, the contaminated sample shows 72.1 ppb aflatoxin and the healthy one 0.0 ppb. The maize samples from Croatia show approximately the same aflatoxin contamination level, namely 78.9 ppb for the contaminated sample and 0.8 ppb for the healthy one. After our measurements, the contamination of the samples was confirmed by the CODA-CERVA, the Belgian Reference Laboratory for Mycotoxins. Before the start of the fluorescence spectrum measurements, the samples are first investigated by illumination with a UV light source. When illuminated with a UV light source, the healthy maize samples show no significant fluorescence signals. In contrast, the contaminated samples show localized fluorescent regions. However, with the use of a UV light bulb, only the regions of strongest fluorescence can be visualized. To be able to measure and characterize weak fluorescence signals, a sensitive setup is indispensable. During our fluorescence measurements, we investigate the fluorescence spectra of 45 healthy and contaminated Croatian maize grains. To observe the variation between the sample types and the harvest environments through the fluorescence spectra, we also measure the fluorescence spectra of 15 healthy and contaminated Italian maize grains. The Italian company provided only a small maize sample, which limits the number of studied maize grains but is sufficient to monitor the environmental influences onto the data variation. Because aflatoxin is sensitive to light, the samples are permanently stored in a dark enclosure to minimize the environmental influences onto the measurements. 2.2. Fluorescence measurement setup One- and two-photon induced fluorescence are based on the absorption of laser light, which excites an electron from the ground state to a higher energetic state and results in the emission of a new fluorescence photon during the relaxation process of the excited electron (Lakowicz, 1999, chap. 1). To generate a OPIF photon, only one excitation photon needs to be absorbed, while during the generation of a TPIF photon two excitation photons need to be absorbed simultaneously to excite the electron and generate the
fluorescence signal (Fig. 1). In the case of OPIF, the energy of the incident photon equals the energy difference between the electronic states, such that the electron obtains sufficient energy to bridge the energy gap between the ground state and the excited state (Fig. 1a). In the case of TPIF, the excitation occurs in two steps with two simultaneously incoming photons (Fig. 1b). The first incident photon excites the molecule to a virtual state, while the second incident photon excites the molecule from the virtual state to its higher-energy excited state. Because for both OPIF and TPIF the electron is excited to the same energetic state, they will give rise to the same fluorescence wavelengths, but with another excitation wavelength. Typically, OPIF occurs after excitation with UV laser light, while TPIF occurs after excitation with NIR laser light. NIR laserlines are more widely commercial available than the UV laser lines. Moreover, when using UV laser light in an optical setup, the optical mirrors and lenses need to be coated or fabricated in fused silica, resulting in a more expensive configuration. Two-photon absorption is a non-linear process that is generally many orders of magnitude weaker than the linear one-photon absorption occurring during the OPIF process. The OPIF intensity increases linearly with the excitation power, while the TPIF intensity increases with the square of the excitation power. For the same excitation laser power, the TPIF intensity is thus proportional to the square root of the OPIF intensity. To obtain a significant TPIF signal during our measurements, we need a large excitation power density, which is defined as the ratio of the excitation laser power to the excitation spot area. As a result, large excitation powers and small excitation spot sizes are indispensable to obtain a strong TPIF signal. We pursue a fluorescence measurement setup that is suited for both OPIF and TPIF measurements. However, before we could design our measurement setup, we first had to determine the interesting excitation wavelengths. To detect the aflatoxincontamination inside the maize grains, we need an excitation wavelength which is strongly absorbed by the aflatoxin, since strong fluorescence intensities are theoretically obtained after excitation with a wavelength for which the sample shows a strong absorbance (Rasch, Kumke, et al., 2010b). The stronger the absorbance of a sample, the more electrons will be excited and the more fluorescent photons will be emitted. By exciting the maize grains with an excitation wavelength that is strongly absorbed by the aflatoxin, we want to maximize the influence of the aflatoxin on the natural fluorescence spectrum of the maize grains. The absorbance spectrum of aflatoxin B1 shows the strongest absorbance in the range between 200 nm and 250 nm and around 365 nm (Fig. 2). We designed our measurement setup such that we are able to emit wavelengths around 365 nm, to study the OPIF spectrum. Consequently, to study the TPIF spectrum excitation wavelengths around
ski diagram for (a) OPIF and (b) TPIF. Fig. 1. Jablon
L. Smeesters et al. / Food Control 51 (2015) 408e416
Fig. 2. Absorbance and fluorescence spectrum of aflatoxin B1.
730 nm are required. In addition, we investigate the fluorescence spectrum of the maize grains when excited with 750 nm and 780 nm, to study the influence of the matrix constituents, the natural occurring fluorescent proteins, onto the fluorescence spectrum. The designed fluorescence measurement setup consists of two main parts, namely the excitation and detection side (Fig. 3a). For the excitation of the sample, a tunable titanium-sapphire laser (Spectra-Physics Tsunami laser), pumped by a frequency doubled Nd:YAG laser (continuous wave 532 nm laser), is used during the TPIF measurements (Fig. 3b). The wavelength of the tunable titanium-sapphire laser can be tuned between 710 nm and 835 nm. The maximum output power ranges from 1.20Watt to 1.50Watt, depending on the selected wavelength. During the OPIF measurements, the laser light of the titanium-sapphire laser is frequencydoubled with the use of a second harmonic generating crystal, to generate the necessary UV excitation wavelengths (Fig. 3b). The second harmonic generating crystal is able to generate wavelengths between 355 nm and 417 nm, with a maximal output power between 200 mW and 450 mW. Behind the output of the titanium-
411
sapphire laser and the frequency-doubling unit, the excitation light is directed towards the sample. To obtain a high excitation power density during the TPIF measurements, an additional focusing lens is positioned in front of the sample to minimize the excitation spot size (Fig. 3c). The sample holder is positioned on a translation stage, which allows the scanning of the sample. After the excitation of the sample, the fluorescent signals are captured by a collimating lens, coupled into a broadband optical fiber (UVIR600 fiber of Avantes) and guided towards the spectrum analyzer. In front of the detecting fiber, we position a long-wave pass filter during the OPIF measurements and a short-wave pass filter during the TPIF measurements to absorb the excitation light. Otherwise the excitation signals would saturate the measured fluorescence spectrum, leading to incorrect measurement data. The used spectrum analyzer (AvaSpec2084 spectrum analyzer of Avantes) is able to measure the spectrum between 300 nm and 1100 nm with a resolution of 8 nm. The measured spectra are corrected by a transfer function to account for the wavelength dependent transmittance of the optical fiber and the sensitivity of the detector inside the spectrum analyzer. 2.3. Validation of the measurement setup: characterization of the aflatoxin B1 fluorescence To validate the operation of our fluorescence measurement setup, we measure the OPIF and TPIF spectrum of pure aflatoxin B1 powder (Fig. 4). We purchased aflatoxin B1 powder, produced by the fungus A. flavus with 98% or better purity, from SigmaeAldrich. It is a white to yellow crystalline powder that we measure in its solid state. We present the OPIF spectrum that is obtained after excitation with 365 nm, with an excitation power density of 42 mW/mm2 (excitation power of 30 mW and an excitation spot diameter of
Fig. 3. Measurement setup, which allows the investigation of both OPIF and TPIF: (a) schematic representation of the setup; (b) tunable titanium-sapphire laser (710 nme835 nm) and harmonic generating unit that contains the frequency-doubling crystal (355 nme417 nm); (c) optical path for the excitation of the sample, after which the fluorescence spectrum is captured by the detecting fiber. The focusing lens minimizes the spot size during the TPIF measurements.
L. Smeesters et al. / Food Control 51 (2015) 408e416
0.2 1.5 0.15
OPIF
1
0.1
TPIF 0.5
0.05
0
400
428nm
450
500 Wavelength (nm)
550
Absolute TPIF intensity (uW/cm²/nm)
Abs olute OPIF intens ity (uW/c m ²/nm )
412
0 600
Fig. 4. OPIF and TPIF spectra of the pure aflatoxin B1 powder, after excitation with 365 nm and 730 nm respectively.
951 mm). The TPIF fluorescence spectrum is measured after excitation with 730 nm, with an excitation power density of 14317 mW/ mm2 (excitation power of 600 mW and an excitation spot diameter of 231 mm). Both fluorescence spectra show their maximal fluorescence intensity at 428 nm, which corresponds with the expected aflatoxin B1 fluorescence maximum indicated in Fig. 2. The maximal OPIF intensity, at 428 nm, is equal to 1.68 ± 0.93 mW/cm2, while we observe a maximal TPIF intensity of 0.20 ± 0.09 mW/cm2. The measured TPIF intensity is hence 10 times weaker than the OPIF intensity. Comparing the excitation power density and the resulting fluorescence intensity, the efficiency of the OPIF process is approximately 3000 times larger than the efficiency of the TPIF process. The shape of the TPIF spectrum shows a narrower peak than the OPIF spectrum, due to the more selective excitation during two-photon absorption than during one-photon absorption. To characterize the OPIF and TPIF spectra, we measure the fluorescence intensity as function of the excitation laser power, while maintaining a constant excitation spot size (Fig. 5). Subsequently, for each excitation power, we integrate the measured fluorescence spectrum to obtain the integrated fluorescence intensity. For the different excitation powers, we present the intensity of the mean integrated fluorescence spectrum. The variation of the measured data is less than 10% and 12%, during the OPIF and TPIF measurements respectively. Studying the integrated OPIF intensity as function of the excitation laser power, we observe a linear
relationship, which demonstrates the linear one-photon absorption. However, at higher excitation powers, starting from 35 mW onwards, the integrated OPIF intensity deviates from the linear relationship and starts saturating. In the saturated region, the maximum fluorescence intensity is reached since no more electrons can then be excited to the higher energy state. The TPIF intensity shows a quadratic dependence on the excitation laser power, confirming the occurrence of non-linear two-photon absorption. Both the linear and exponential fits show an adjusted rsquare value of 0.98 and 0.97 respectively, ensuring that the fitted function is a good representation of the measured data. Furthermore, considering the measured fluorescence intensities, we observe that TPIF requires higher excitation powers to obtain a measurable fluorescence spectrum. Specifically, taking the spot size into account, the TPIF process requires an excitation power density of 2000 mW/mm2, while the OPIF process can be generated with an excitation power density of 7 mW/mm2 (Smeesters, Meulebroeck, Raeymaekers, & Thienpont, 2014). The fluorescence measurements of the pure aflatoxin B1 powder correspond with the theoretical characteristics of OPIF and TPIF. Consequently, the above measurements validate the correct operation of our fluorescence measurement setup. In the next section, we investigate the one- and two-photon induced fluorescence spectra of the different sets of maize batches. We give an overview of measured fluorescence spectra, after which we quantitatively evaluate the optical difference between the healthy and contaminated samples. 3. Results and discussion To investigate the optical detection of aflatoxin B1 in maize, we study the OPIF and TPIF spectra of healthy and contaminated maize grains, of both the Croatian and Italian maize batches. We measured the OPIF spectrum after excitation with 365 nm, with an excitation laser power of 225 mW (Fig. 6a). The TPIF spectra are measured after excitation with 730 nm, 750 nm and 780 nm, with an excitation power of 1100 mW, 1220 mW and 1510 mW respectively (Fig. 6bed). During the TPIF measurements, we investigate the fluorescence spectrum after excitation with multiple excitation wavelengths to monitor the influence of the illumination wavelength onto the fluorescence of the maize grains. To maximize the signal to noise ratio, we illuminate the maize grains with the maximal output powers of the titanium-sapphire laser and the harmonic generating unit.
Fig. 5. Intensity of the integrated OPIF spectrum increases linearly with the excitation power, while the intensity of the integrated TPIF spectrum shows a quadratic dependence on the excitation power.
L. Smeesters et al. / Food Control 51 (2015) 408e416
413
Fig. 6. Fluorescence spectra of the healthy and contaminated maize grains of both the Croatian and Italian maize sample: (a) OPIF spectrum after excitation with 365 nm; (b) TPIF spectrum after excitation with 730 nm; (c) TPIF spectrum after excitation with 750 nm; (d) TPIF spectrum after excitation with 780 nm.
Both the healthy and the contaminated samples show a fluorescence signal, due to the intrinsic fluorescence of the maize grains (Fig. 6). For every maize batch, we display the mean fluorescence spectrum. Considering the different maize batches, the fluorescence spectra of the Croatian and Italian maize correspond well, within the variances of the measurements. For all excitation wavelengths, the fluorescence spectra of both maize types have the same shape and are present in the same wavelength regions. The displayed fluorescence intensity differences between the Croatian and Italian maize are within the variances of the measurement data (Table 1). These intensity variances are caused by the differences within the molecular structure of the maize grains. Because maize is a natural product, the maize grains show a large internal variation in the surface shape, density and natural composition. Comparing the fluorescence spectra of the healthy and contaminated maize grains, we observe significant differences in Table 1 Measured maximum fluorescence intensity, and its variation, for the different excitation wavelengths.
365 730 750 780
nm nm nm nm
excitation excitation excitation excitation
Maximum fluorescence intensity Croatian maize (mW/cm2/nm)
Maximum fluorescence intensity Italian maize (mW/cm2/nm)
Healthy
Healthy
15 0.09 0.07 0.06
± ± ± ±
8 0.05 0.03 0.04
Contaminated 17 0.02 0.02 0.02
± ± ± ±
10 0.01 0.01 0.01
25 0.08 0.05 0.03
± ± ± ±
17 0.05 0.02 0.01
Contaminated 13 0.02 0.02 0.02
± ± ± ±
9 0.01 0.01 0.01
intensity and emission wavelength. We do not directly observe the aflatoxin fluorescence, but we measure its influence onto the intrinsic fluorescence of the maize. For both the OPIF and TPIF measurements, the healthy and contaminated maize batches show different fluorescent intensities due to the different molecular structure of these maize grains (Table 1). In the contaminated samples, the aflatoxin is bonded to the different natural constituents of the healthy maize, changing the molecular structure of the constituents and therefore influencing the natural fluorescence of the maize. Moreover, this intensity contrast is the largest for the TPIF spectra, since the different bonds inside the maize are more selectively excited during the TPIF process than during the OPIF one. The largest intensity differences between the healthy and contaminated samples are observed after excitation with 730 nm, where the mean fluorescence intensity of the healthy maize is 4 times stronger than the mean fluorescence intensity of the contaminated maize. In contrast, the minimum intensity differences are observed after excitation with 365 nm. However, as we expect, the OPIF intensity is much stronger than the TPIF intensity. Specifically, the TPIF signal is approximately 500 times weaker than the OPIF signal. In addition to the fluorescence intensity differences, we also observe a wavelength shift between the fluorescence maxima of the healthy and contaminated maize samples (Table 2). Generally, the OPIF and TPIF spectra show a wavelength shift of approximately 50 nm. To compare the obtained wavelength shift for the different excitation wavelengths, we calculate the class difference for both the Croatian and Italian maize. The class difference (D) is a measure for the difference between the average values (m) of two product types, taking the standard deviation (s) and the
414
L. Smeesters et al. / Food Control 51 (2015) 408e416
amount of measured samples (N) into account (Downie & Heath, 1970, chap. 12):
mcontaminated mhealthy D ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi s2contaminated Ncontaminated
þ
(1)
s2healthy Nhealthy
The OPIF measurements show the largest class differences, indicating the largest optical contrast between the healthy and contaminated maize grains. The variation of the class differences between the OPIF and TPIF measurements are mainly caused by their different variances. The TPIF spectra show a larger variance than the OPIF spectra, which decreases their class difference. TPIF signals show a weaker intensity, resulting in a stronger relative noise signal and therefore a lower signal to noise ratio. The fluorescence spectrum of the contaminated maize shows weaker fluorescence intensities and a wavelength shift of the fluorescence maximum towards longer wavelengths. To validate the influence of the aflatoxin onto the maize intrinsic fluorescence spectrum, we sweep the excitation wavelength towards the wavelengths for which aflatoxin B1 shows a weaker absorbance. We measured the fluorescence spectrum of the healthy and contaminated maize samples after excitation with 405 nm and 810 nm, during the OPIF and TPIF measurements respectively (Fig. 7). Both the OPIF and TPIF measurements show a small difference between the fluorescence spectrum of the healthy and contaminated samples. We measure a wavelength shift of 8 ± 6 nm and 26 ± 12 nm, after excitation with 405 nm and 810 nm respectively. Moreover, we observe only small differences between the fluorescence intensities. Aflatoxin B1 shows a weak absorbance at 405 nm (Fig. 2), resulting in a minor influence onto the intrinsic fluorescence spectrum of the maize grains. So far, we compared the emission wavelength and intensity of the maize fluorescence spectra. We did not account for the shape of the fluorescence spectra. To obtain a complete quantitative
comparison between the measured fluorescence spectra, we study the integrated fluorescence intensity as function of the sample number. For both healthy and contaminated maize grains of the Croatian and Italian maize batches, the normalized fluorescence intensity from 475 nm until 550 nm is compared with the range from 400 nm until 475 nm, for excitation with 365 nm, 730 nm, 750 nm and 780 nm. In both wavelength intervals the fluorescence spectrum is integrated, after which we plotted the intensity ratio of these integrals (Fig. 8). For all OPIF and TPIF measurements the integral ratio shows a clear distinction between the healthy and contaminated samples. The contaminated samples show generally a larger variance because of the local presence of the aflatoxin and the variable aflatoxin-concentration in the maize grains. Moreover, the variance of the contaminated samples is larger for the TPIF measurements than for the OPIF ones. The largest variance is obtained after excitation with 780 nm (Fig. 8d). However, also the largest difference between the mean ratio of the healthy and contaminated samples is observed for excitation with 780 nm. Considering the class differences of the integral ratio between the healthy and contaminated samples, we obtain a class difference of 116.2, 111.1, 48.8 and 54.9, for excitation with 365 nm, 730 nm, 750 nm and 780 nm respectively. The largest contrast can thus be found after excitation with 365 nm. However, the contaminated samples can still be properly identified with TPIF, allowing using NIR excitation laserlines, instead of the UV laserlines for OPIF. Our measurement results clearly indicate the use of fluorescence spectroscopy as a promising detection technique for the indentification of aflatoxins in maize grains. Moreover, the commercial availability of high power NIR laserlines, optical bandpass filters and sensitive detectors allow its integration in practical systems. In comparison to the traditional chemical analyses, the optical detection techniques feature some important advantages enabling to improve our food safety. Traditional chemical analyses are destructive and measure the mean contamination level of a certain number of maize grains. In contrast, optical spectroscopy is
Table 2 Emission wavelength at maximum fluorescence intensity, and its variation, for the different excitation wavelengths. Dominant emission wavelength Croatian maize (nm) Healthy 365 730 750 780
nm nm nm nm
excitation excitation excitation excitation
443 441 444 450
± ± ± ±
5 3 7 10
Class difference Croatian maize
Contaminated 497 496 497 503
± ± ± ±
15 20 24 19
Dominant emission wavelength Italian maize (nm) Healthy
9.7 6.1 3.8 5.2
442 441 440 450
± ± ± ±
5 3 5 7
Class difference Italian maize
Contaminated 486 496 490 498
± ± ± ±
6 16 17 10
10.8 4.1 2.7 4.8
Fig. 7. Fluorescence spectra of the healthy and contaminated maize grains: (a) OPIF spectrum after excitation with 405 nm; (b) TPIF spectrum after excitation with 810 nm.
L. Smeesters et al. / Food Control 51 (2015) 408e416
415
Fig. 8. Contrast between the fluorescence spectra of the healthy and contaminated maize samples, expressed by the intensity ratio of the integrated fluorescence spectrum from 475 nm until 550 nm to the integrated fluorescence spectrum from 400 nm until 475 nm, after excitation with (a) 365 nm; (b) 730 nm; (c) 750 nm; (d) 810 nm.
an ultra-fast and non-destructive technique, since it does not use chemicals and does not requires grinding of the maize. Moreover, laser-based sorting systems are able to scan the sample, such that each maize grain can be individually evaluated and the possible local aflatoxin-contamination of the sample can be identified. 4. Conclusion We demonstrate the use of one- and two-photon induced fluorescence spectroscopy for the non-destructive detection of carcinogenic aflatoxins in maize. We measured the fluorescence spectrum of different healthy and contaminated maize batches, when excited with 365 nm, 405 nm, 730 nm, 750 nm, 780 nm and 810 nm. Both the healthy and contaminated maize grains show an intrinsic fluorescence signal. However in the contaminated samples (with 72.1 ppb and 78.9 ppb of aflatoxins) the intrinsic fluorescence of the healthy maize grains is influenced by the presence of the aflatoxin. In comparison to the healthy maize samples, the contaminated samples show lower fluorescence intensities and a wavelength shift of the fluorescence maximum. Both the measured one- and two-photon induced fluorescence spectra show a wavelength shift of approximately 50 nm. The largest intensity differences are observed after excitation with 730 nm, where the mean fluorescence intensity of the healthy maize is 4 times stronger than the fluorescence intensity of the contaminated maize. Moreover, to
consider both the shape and wavelength region of the fluorescence spectrum, we investigate the intensity ratio of the integrated fluorescence spectrum from 475 nm until 550 nm to the integrated fluorescence spectrum from 400 nm until 475 nm. The mean fluorescence ratio shows the largest contrast for excitation with 365 nm and 780 nm. Because the two-photon induced fluorescence spectra show a larger variance than the one-photon induced fluorescent ones, the largest class difference is obtained after excitation with 365 nm. In comparison to one-photon induced fluorescence, two-photon induced fluorescence requires higher excitation powers, smaller excitation spot sizes and a more sensitive detector. The measured one-photon induced fluorescence intensity is 500 times larger than the measured two-photon induced fluorescent one. However, the near-infrared excitation wavelengths for two-photon induced fluorescence are more widely commercially available than the required UV excitation wavelengths for one-photon induced fluorescence. Furthermore, when using UV laser light in an optical setup, the optical mirrors and lenses need to be coated or fabricated in fused silica, resulting in a more expensive configuration. The investigation of the optical differences between the fluorescence spectra of the healthy and contaminated products, together with the determination of the illumination and detector requirements, paves the way to real-time, industrial, non-destructive detection of the carcinogenic aflatoxins in food products.
416
L. Smeesters et al. / Food Control 51 (2015) 408e416
Acknowledgments This work was supported in part by FWO (G008413N), IWT (IWT120528), the COST Action (MP1205), the Methusalem and Hercules Foundations and the OZR of the Vrije Universiteit Brussel (VUB). The authors would also like to thank dr. ir. Alfons Callebaut and dr. ir. Bart Huybrechts of the CODA-CERVA, the National Reference Laboratory for Mycotoxins, for their valuable feedback and know-how about mycotoxins. We are very grateful to them for the execution of the liquid chromatography-tandem mass spectroscopy analytical technique. References Downie, N. M., & Heath, R. W. (1970). Basic statistical methods. New York: Harper. Held, P. (April 2013). Quantitation of peptides and amino acids with synergy™ HT using UV fluorescence. http://www.biotek.com/resources/docs/Synergy_HT_ Quantitation_of_Peptides_and_Amino_Acids.pdf. Hruska, Z., Yao, H., Kincaid, R., Darlington, D., Brown, R. L., Bhatnagar, D., et al. (2013). Fluorescence imaging spectroscopy (FIS) for comparing spectra from corn ears naturally and artificially infected with aflatoxin producing fungus. Journal of Food Science, 7(8), 1313e1320. Krska, R., & Molinelli, A. (2007). Mycotoxin analysis: state-of-the-art and future trends. Analytical and Bioanalytical Chemistry, 387(1), 145e148.
Lakowicz, J. R. (1999). Principles of fluorescence spectroscopy. New York: Kluwer Academic/Plenum Publishers. Meulebroeck, W., & Thienpont, H. (2006). Optical detection techniques for laser sorting machines. Proceedings of SPIE, 6189, 337e348. Meulebroeck, W., & Thienpont, H. (2012). The study of vegetation indices for the monitoring of differences in chlorophyll and carotenoid composition in green vegetables. Proceedings of SPIE, 8439, 843920. Petterson, H., & Aberg, L. (2003). Near infrared spectroscopy for determination of mycotoxins in cereals. Food Control, 14(4), 229e232. €ttcher, M., & Kumke, M. (2010). Determination of aflatoxin B1 in Rasch, C., Bo alcoholic beverages: comparison of one- and two-photon induced fluorescence. Analytical and Bioanalytical Chemistry, 397(1), 87e92. €hmannsro € ben, H. G. (2010). Sensing of mycotoxin proRasch, C., Kumke, M., & Lo ducing fungi in the processing of grains. Food and Bioprocess Technology, 3(6), 908e916. Romer Labs. (June 2014). Mycotoxins affect everyone: Regulations for food. http:// www.mycotoxins.info/myco_info/consum_regu.html#usa. Saranwong, S., Sornsrivichai, J., & Kawano, S. (2014). Prediction of ripe stage eating quality of mango fruit from its harvest quality measured nondestructively by near infrared spectroscopy. Postharvest Biology and Technology, 31(2), 137e145. Smeesters, L., Meulebroeck, W., Raeymaekers, S., & Thienpont, H. (2014). The use of one- and two-photon induced fluorescence spectroscopy for the optical characterization of carcinogenic aflatoxins. Proceedings of SPIE, 9198, 919803. Toximet. (June 2014). Our award-winning mycotoxin testing system. http://www. toximet.com/. Unnevehr, L., & Grace, D. (2013). Aflatoxins finding solutions for improved food safety. International food policy research institute.