- Email: [email protected]
Varietal discrimination of hop pellets by near and mid infrared spectroscopy
Author’s Accepted Manuscript VARIETAL DISCRIMINATION OF HOP PELLETS BY NEAR AND MID INFRARED SPECTROSCOPY Julio C. Machado, Miguel A. Faria, Isabel M.P.L.V.O. Ferreira, Ricardo N.M.J. Páscoa, João A. Lopes www.elsevier.com/locate/talanta
PII: DOI: Reference:
S0039-9140(17)31232-8 https://doi.org/10.1016/j.talanta.2017.12.030 TAL18166
To appear in: Talanta Received date: 15 September 2017 Revised date: 6 December 2017 Accepted date: 11 December 2017 Cite this article as: Julio C. Machado, Miguel A. Faria, Isabel M.P.L.V.O. Ferreira, Ricardo N.M.J. Páscoa and João A. Lopes, VARIETAL DISCRIMINATION OF HOP PELLETS BY NEAR AND MID INFRARED SPECTROSCOPY, Talanta, https://doi.org/10.1016/j.talanta.2017.12.030 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
VARIETAL DISCRIMINATION OF HOP PELLETS BY NEAR AND MID INFRARED SPECTROSCOPY
Julio C. Machado Jr.a, Miguel A. Fariaa, Isabel M. P. L. V. O. Ferreiraa, Ricardo N. M. J. Páscoab*, João A. Lopesc
a
LAQV/REQUIMTE/ Departamento de Ciências Químicas, Laboratório de Bromatologia e
Hidrologia, Faculdade de Farmácia, Universidade do Porto, Rua de Jorge Viterbo Ferreira, 228, 4050-313, Porto, Portugal b
LAQV/REQUIMTE, Laboratório de Química Aplicada, Departamento de Ciências
Químicas, Faculdade de Farmácia, Universidade do Porto, Rua de Jorge Viterbo Ferreira, 228, 4050-313, Porto, Portugal c
Research Institute for Medicines (iMed.ULisboa), Faculdade de Farmácia, Universidade de
Lisboa, Av. Prof. Gama Pinto, 1649-003, Lisbon, Portugal
*Corresponding author: [email protected]
Abstract
Hop is one of the most important ingredients of beer production and several varieties are commercialized. Therefore, it is important to find an eco-real-time-friendly-low-cost technique to distinguish and discriminate hop varieties. This paper describes the development of a method based on vibrational spectroscopy techniques, namely near- and mid-infrared spectroscopy, for the discrimination of 33 commercial hop varieties. A total of 165 samples (five for each hop variety) were analysed by both techniques. Principal component analysis, hierarchical cluster analysis and partial least squares discrimination analysis were the chemometric tools used to discriminate positively the hop varieties. After optimizing the spectral regions and pre-processing methods a total of 94.2% and 96.6% correct hop varieties discrimination were obtained for near- and mid-infrared spectroscopy, respectively. The results obtained demonstrate the suitability of these vibrational spectroscopy techniques to discriminate different hop varieties and consequently their potential to be used as an authenticity tool. Compared with the reference procedures normally used for hops variety discrimination these techniques are quicker, cost-effective, non-destructive and eco-friendly. Graphical abstract
Keywords: Hop, near infrared spectroscopy, mid infrared spectroscopy, discrimination, chemometrics;
1. Introduction Hop (Humulus lupulus L.) is one of the principal ingredients of beer, responsible for different intensities of bitterness and a wide range of specific flavours and aromas with a strong impact in beer organoleptic properties. The amount of specific flavours and aromas present in beer depends mainly on hops variety but also how it is used during the brewing process. For this reason the selection of hop’s variety is becoming the main target of interest for brewers aiming to produce different and exquisite beers styles. There is thus an increase on the research concerning the development and study of new varieties over the last years. Furthermore, several high valued hops had been launched in the market with much differentiated prices making its authentication crucial for brewers. In this context, fraudulent declarations or labelling of hop products must not be excluded. Total or partial replacement of expensive hops with less expensive ones results in economic benefit to dealers as it was reported recently [1]. Besides providing bitterness and aroma compounds to the beer, hop presents a wide range of bioactive properties, in particular antioxidant, anti-inflammatory, anticancer, immunomodulatory and antimicrobial activities associated to its composition [2-4]. These bioactive properties are related with the plant variety (each variety presents different chemical compounds profiles) and have impact also on its commercial value. Due to the abovementioned reasons more than two hundred hop varieties are presently cultivated worldwide. These differ qualitatively and quantitatively in their fraction of essential oils (characterized by terpenes, alcohols, acids, esters, ketones and aldehydes) and, in their non-volatile fraction of hard resins (composed by several polyphenols, such as xanthohumol, prenylnaringenin compounds and derivatives) and soft resins (containing prenylated phloroglucinol derivatives, α- and β- acids) [5]. The high number and phenotypic variability of hop fostered the development of several analytical methods to evaluate the chemical compounds present in each crop, which has been a common practice, as well as methodologies for the varietal authentication of hops in its different commercialized forms, being pellets of compressed dry flowers the most used nowadays. The
identification of hop varieties is generally performed using several strategies such as the evaluation of the morphological characteristics [6] - difficult to achieve when hop cones are processed - and the chemical analysis of hop compounds, namely bitter acids, flavonoids, essential oils, or proteins, individually or using the combination of different chemical groups [7-15]. As chemical markers are strongly influenced by environmental conditions during plant growth (region, soil, harvest time, agricultural practices, environmental factors), cones processing and storage, several DNA-based methods were developed targeting the plant phylogenetic studies and hops molecular identification, to circumvent the environmental variation of chemical markers. Several methods were thus proposed for hop varietal authenticity control as the AFLP - Amplified Fragment Length Polymosphism [16]; RAPD - Random Amplified Polymorphic DNA [17], SSR – Simple Sequence Repeats [1, 18-21] and DArT - diversity array technology markers [22]. Although several methods are available the analysis of SNPs (Single Nucleotide Polymorphisms), differences of single nucleotide in some homologous DNA among varieties, is presently considered the more reproducible tool for the identification of varieties of several plants [23]. Next Generation Sequencing (NGS) technologies applied to the study of the hop genome showed the presence, in average, of a Single Nucleotide Polymorphism (SNP) in every 346 bp of the hop genome [24]. These authors identified 17128 SNPs in the hop genome and concluded that a group of 3068 SNP retrieved the same level of discrimination amongst varieties. However the minimum number of markers to differentiate among all genotypes was not identified. Notwithstanding, Henning and collaborators (2015) identified 7 SNPs from a group of 374829 markers that effectively differentiated all 121 varieties and accessions (representing a broad spectrum of hop lines from around the world). Therefore these authors proposed the use of highresolution melting (HRM) curve analyses as a simple, rapid and more economically viable means to perform genetic fingerprinting on hop genotypes [25]. As SNP markers provide a greater likelihood of differentiating among varieties, the SNP/HRM approach was used in this study as an effective way of samples discrimination to validate their differentiation. All the above referred techniques usually require a considerable amount of sample and are destructive. To avoid such drawbacks, vibrational spectroscopic techniques, namely near infrared
(NIR) and mid-infrared (MIR) spectroscopy, emerge as suitable alternatives for hops discrimination. The NIR spectroscopy results from overtones and combination of fundamental vibrational bands, namely C-H, N-H, O-H and S-H bonds, in the spectral range of 14,000 to 4,000 cm-1. MIR spectroscopy is characterized for fundamental, bending and rotating vibrations in the spectral range of 4,000 to 400 cm-1. The MIR spectrum has a higher specificity than NIR spectrum and for this reason is considered more appropriate for identification and characterization purposes. Both these techniques are rapid, non-destructive with a low-cost per analysis [26]. There are already some examples in the literature exploring vibrational spectroscopic techniques for discrimination purposes, namely in genetically and non-genetically modified maize plants [27], grapevine varieties [28], bacteria species discrimination [29], and even hops [14] although focusing only a small sample size usually within the same country of origin. The aim of this study was thus to evaluate if NIR and MIR spectroscopy can be used as fast and nondestructive technique for the differentiation and identification of hop varieties targeting authenticity purposes. A total of 33 commercial varieties (representing about 75% of the total volume commercialized worldwide) were used. DNA-based SNP characterization was performed as a way to warrant samples differentiation. To the best of our knowledge this represents the first work exclusively based on NIR and MIR spectroscopy techniques to hop cultivars fast discrimination within a highly diverse group of varieties.
2. Experimental 2.1. Samples Commercial hops were acquired as pellets in local suppliers of raw materials for brewers. The varieties list is presented in Table 1 together with the content of α-acids and country of origin, as indicated by the manufacturer. The choice of the varieties to include in the study was done by selecting the most relevant in terms of market share accounting for 75% of the worldwide market volume in 2016 [30]. A total of 165 samples (five for each hop variety) were acquired including the
harvest years of 2015 and 2016. Samples were grinded to a homogeneous powder in a Retsch GM 200 mill (Retsch GmbH, Haan, Germany), stored in the dark and refrigerated until the moment of analysis.
2.2 MIR spectral acquisition A Fourier transform PerkinElmer Spectrum BX FTIR System spectrophotometer (Waltham, USA) equipped with a DTGS detector and a PIKE Technologies Gladi ATR accessory was used to collect the hop MIR spectra in diffuse reflectance mode. The spectra were acquired in the spectral range of 4,000 to 400 cm−1, with a resolution of 4 cm−1 and 32 scan co-additions. The hop samples were transferred to the ATR crystal and compressed until constant signal. For each hop cultivar, one small portion was transferred and a spectrum was collected for each portion. Therefore, a total of 5 spectra were acquired for each hop variety. Between each hop variety, the ATR crystal was cleaned and a background was acquired.
2.3 NIR spectral acquisition A Fourier transform near infrared spectrometer, FTLA 2000 from ABB (Québec, Canada) equipped with an indium-gallium-arsenide (InGaAs) detector was used to collect the hop NIR spectra in diffuse reflectance mode. Each spectrum was recorded as the average of 64 scans in the spectral range of 10,000 cm−1 and 4000 cm−1, with 8 cm−1 resolution. The hop samples were transferred into borosilicate flasks in order to perform spectra acquisition. The background was measured at the beginning of each analysis using a reference substance (Teflon). A total of 5 spectra were acquired for each hop variety.
2.4 NIR and MIR data analysis
The NIR and MIR data analysis was performed using principal component analysis (PCA) [31] and partial least squares discriminant analysis (PLSDA) [32]. PCA was used to detect outliers and find common patterns while the later was used to develop discrimination models. For PLSDA models, samples were classified according to the respective variety. The available data were divided in calibration (70%) and validation (30%) sets in a random way, but maintaining the same proportion of each variety in both sets (to avoid unbalanced hops variety classes in both sets) [33]. The assessment of the optimal number of latent variables (LV) was done using the leave-one-sample-out crossvalidation procedure using only the calibration set. The NIR and MIR spectra were divided in five spectral regions. The NIR regions were: R1 from 10,000 to 7,472 cm-1, R2 from 7,468 to 6,083 cm-1, R3 from 6,079 to 5,389 cm-1, R4 from 5,385 to 4,964 cm-1 and R5 from 4,961 to 4,035 cm-1. The MIR regions were: R1 from 4,000 to 2,782 cm-1, R2 from 2,780 to 1,882 cm-1, R3 from 1,880 to 1,492 cm-1, R4 from 1,490 to 874 cm-1 and 872 to 600 cm-1. These regions were tested individually and in combination for the selection of the best spectral regions. The selection of best processing technique involved testing standard normal variate (SNV) and Savitzky-Golay filter (with first and second derivatives, different filter widths and polynomial orders) individually and in combination. After the optimization of the best spectral regions and pre-processing methods, the test set was projected in the optimized PLSDA calibration model to assess the percentage of correct predictions. Model predictions were expressed as confusion matrices [33]. The total percentage of correct predictions was obtained adding the diagonal elements of the confusion matrices. One PLSDA model was developed for each spectroscopic technique (NIR and MIR). Loadings of the first four LV of each developed PLSDA were also analysed to understand which specific wavenumbers were more important. Spectral data were mean centred before application of PCA and PLSDA. All chemometric analyses were performed with Matlab version 8.6 (MathWorks, Natick, USA) and PLS Toolbox version 8.2.1 (Eigenvector Research Inc., Wenatchee, USA).
2.5 DNA analysis
DNA was extracted from 50 mg powdered samples using the E.Z.N.A. HP Plant DNA Mini Kit from Omega Bio-Tek (Norcross, USA) following the manufacturer instruction, with a slight modification consisting in a previous sample homogenisation using a Precellys Evolution tissue homogeniser (Bertin Instruments, Montigny-le-Bretonneux, France), in 2 ml tubes containing 1.4 mm zirconium oxide beads. Extracts were quantified and evaluated for purity by UV absorption at 260 and 280 nm in a 4 µL sample aliquot using a BioTek Synergy HT plate reader equipped with the Take3 MicroVolume Plate adapter. After quantification, all samples were diluted to 10 ng/µL with water for molecular biology (W4502, Sigma Aldrich, St. Louis, USA). HRMA was carried out using the primers proposed by Henning and collaborators [25]. Amplification reactions were conducted in a Bio-Rad CFX Connect thermalcycler using the SsoFast™ EvaGreen® Supermix (Bio-Rad, CA, USA), in 10 µL volumes. PCR conditions were optimized to reach the shortest time of analysis and defined as: an initial denaturation of 98 ºC for 2 min, followed by 35 cycles of 98 ºC for 5 s and the annealing temperature (from 57 to 63 ºC) for 5 s. Immediately after PCR, melting analysis was performed from 60-95 ºC in 0.5 ºC increments. Melting curve clustering was performed using the Precision Melt AnalysisTM v 1.0 Software from Bio- Rad (Hercules, CA, USA) with a confidence of at least 98% in group assignment.
3. Results and discussion 3.1 Validation of commercial samples by SNP discrimination The varietal genotyping was assessed using 7 SNP markers (TP137094, TP15403, TP245055, TP295074, TP400349, TP411590 and TP437202) required to differentiate all the commercial hop accessions, as proposed by the study of Henning and collaborators (2015) in which 121 were
discriminated [25]. Amplicons of each marker originate three different HRM curves corresponding each one to one of the allelic variants (two homozygotes and one heterozygote). SNP calling was performed by assigning each sample melting curve to one of the three allelic combinations for each SNP site. It was possible to achieve the complete discrimination of all the 33 varieties in study and to confirm its phylogenetic relatedness, assuring this way that samples acquired are genetically different eliminating putative problems of cross mislabelling.
3.2 Exploratory spectral analysis The NIR and MIR spectra of all hop samples were depicted in Figure 1. As abovementioned, a PCA was performed with both NIR and MIR spectral data. The data was just mean-centred before application of PCA and the entire spectral regions were used. The results suggested that no outliers were identified in both data sets. Then, with the objective of understanding if the NIR and MIR spectra gather specific information regarding hop variety, the first two principal components scores were plotted (Fig. 2). In the plots of both NIR and MIR scores it was visible some cluster formation tendency regarding hop’s variety, suggesting that both NIR and MIR spectra contain information related with hop variety. The differentiation was not clear.
3.2.1 Spectral regions and pre-processing methods optimisation The NIR and MIR data were divided in five spectral regions as mentioned in the material and methods section with the objective of identifying spectral regions more informative in terms of the hop’s variety. Several pre-processing techniques were tested as well. The rationale for selecting spectral windows and pre-processing methods was to minimise the distance within varieties while maximising the distance between varieties. The PLSDA results obtained with the calibration set in terms of correct prediction percentages were used to determine the best spectral regions and preprocessing methods. The best pre-processing technique for both NIR and MIR spectra was SNV
followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative. The best spectral regions were: 6,079 to 5,389 cm-1 and 1,880 to 600 cm-1 for NIR and MIR technique, respectively.
3.2.2 Optimised models After defining the best spectral region and pre-processing methods the first two principal components were plotted again (Fig. 3). It is clear that after this optimization step, most samples are now clustered according to the variety, with some exceptions, for both NIR and MIR data. Looking with more detail for the NIR data (Fig. 3a), Nelson Sauvin clustered very close to Amarillo, Citra with Summit and Hallertau Tradition with Hallertau Mittelfrüh The hop varieties Ella, East Kent Golding (EKG), Fuggles and Goldings clustered far apart from the rest. Regarding MIR data (Fig. 3b), the varieties that clustered very close were: Nelson Sauvin with Amarillo, Hallertau Mittelfrüh with Saaz and Challenger with Hallertau Tradition. Hersbrucker and Crystal clustered far apart from the others. These findings indicates that both NIR and MIR spectral data contain information related to hop’s variety and suggest the application of a supervised classification method.
3.3 Unsupervised classification Hierarchical cluster analysis (HCA) was applied for both NIR (Figure 4) and MIR (Figure 5) data considering the average of the five spectra of each hop variety and the best wavenumber region and pre-processing techniques as identified before. The dendrogram obtained using NIR data (Figure 4) grouped the hop varieties in two distinct clusters (C1 and C2). This distribution can be associated with the α-acids amount, because most of the hop varieties that contain a high amount of α-acids are clustered in C2 while varieties with α-acids content lower than 6% are aggregated in C1 cluster. The most similar hop varieties were Citra and Summit, and this is in agreement with the PCA findings. Regarding the dendrogram obtained using MIR data (Figure 5), the hop varieties were also grouped
in two different clusters (C1 and C2). This distribution can also be correlated with α-acids amount considering that most of the hop varieties that contain a high amount of α-acids are aggregated in one cluster (in this case C1) but not so clearly as in the case of NIR spectroscopy.
3.4 Supervised classification The selected supervised classification method used was PLSDA. The test set of NIR and MIR data was then projected using the best spectral region and pre-processing technique. Models considering the NIR and MIR data obtained a total of 96.6% (8 LV) and 94.2% (10 LV) of correct predictions, respectively. The results demonstrate the applicability of both NIR and MIR techniques to discriminate the 33 hop varieties used in the present study in a non-destructive way which configures a useful tool for authenticity purposes. These results were obtained with only 5 samples per hop variety. Further studies should be made increasing the amount of different samples per variety and also using more varieties representing the cultivated plant full diversity. The confusion matrices (one for each vibrational technique) revealed that there was no hop variety that could not be identified or misclassified with another hop (see supplementary data Table 1 and 2). In both confusion matrices the elements that are zero were not shown. The confusion matrix for NIR spectroscopy (see supplementary data Table 1) revealed that the worst hop variety prediction involved Perle and Goldings with 74% and 79% of correct predictions, respectively. The hop variety Perle was incorrectly predicted as Northern Brewer and Saaz while Goldings was incorrectly predicted as Fuggles. Regarding MIR spectroscopy, the worst hop variety prediction involved Simcoe and Perle with 79% and 72% of correct predictions, respectively. Simcoe was incorrectly predicted with Mosaic while Perle was incorrectly predicted as Northern Brewer and Bobek. The verified similarity between Perle and Northern Brewer varieties (corresponding to the lowest number of correct predictions for both NIR and FTIR data) is also supported by the genetic data previously published using RAPD analysis [34] where authors describe the varieties as very close genetically, as well as the breeding history of the variety Perle which has the Northern Brewer as the female parent
[35]. This similarity was verified as well in our data concerning the SNP analysis performed for the varietal authenticity check. Globally, the results obtained show the ability and accuracy of both NIR and MIR spectroscopic techniques to discriminate between different hop varieties.
3.4.1 Models analysis The first four latent variables loadings obtained in the PLSDA models using NIR and MIR data were depicted in Figure 6. These variables of the PLSDA model using NIR data (encompassing approximately 96% of the total variance) revealed that the most important wavenumbers region was located between 6,000 and 5,600 cm-1. This region corresponds to CH3, CH2, CH, SH and Ar-CH absorption bands (first overtone region). There is no published work that connects these absorption bands with the chemical compounds present in hop. Therefore, further studies should be performed in order to understand which chemical compounds are responsible for these absorption bands. However, once hops are rich in a complex mixture of polyphenolic compounds including, prenylphloroglucinols (α- and β- acids) and flavonoids (proanthocyanidins, flavanol glycosides, prenylchalcones and prenylflavanones), the authors believe that these can be related with the referred absorptions bands, together with a large fraction of volatile compounds, including terpenes (as myrcene, humulene, caryophyllene and farnesene among others) and molecules containing sulphur (as thiols, namely 4-mercapto-4-methyl-pentan-2-one, 3 mercaptohexan-1-ol,among others) [36, 37]. This is supposed once the total of α-acids and β-acids can vary from low values as 5% to values over 20% of total dry weight depending on the hop variety [38]. Moreover, the flavonoids fraction can vary between 3% and 6% (w/w), highlighting that the amount of the prenylchalcone xanthohumol, the major constituent of this group, can also vary from 0.1% to 1% (w/w) [38, 39]. The quantity of total oil is also known to vary considerably from 0.5% to 3.0% (v/w) among the different varieties of hops [40].
The first four latent variables of the PLSDA model using MIR data (encompassing approximately 75% of the total variance) showed that the most important wavenumbers regions were located between 1,800 and 1,600 cm-1 and 1,400 and 1,000 cm-1. The first region can be associated with proteins while the second can be connected with carbohydrates and DNA/RNA/phospholipids [29]. In fact, another important parameter that substantially changes between hop varieties is the amount of proteins referenced to be between 12 to 22% dry matter [13] which the authors believe that can contribute to hops discrimination using MIR.
4. Conclusion This work assesses the suitability of NIR and MIR spectroscopy to discriminate hop varieties in its most used form: pellets. The 33 varieties analysed represent the big majority of the hops world market (about 75%) which assures the applicability of the present method as a rapid and nondestructive tool that can be used for authenticity purposes in the hops supply chains. The increasingly high value of hop elite varieties can promote fraudulent practices in the market which sharpens the need of better and more expedite authenticity tools like the one herein proposed. Unsupervised methods applied to spectra provided indications that these techniques can be able to discriminate between hop varieties and suggested the application of a supervised method. After optimising the wavenumbers region and pre-processing techniques, a total of 96.6% (8 LV) and 94.2% (10 LV) of correct hop varieties prediction were obtained for NIR and MIR spectroscopy, respectively. The results obtained in this work are very promising and demonstrate the suitability of both vibrational techniques. Moreover, these methods are rapid, cost-effective, non-destructive and environmentally friendly (do not generate residues). Further studies including more samples per variety and reflecting different geographical origins, when possible, and years of harvest are needed to attest the robustness of these techniques.
Acknowledgements. Authors are grateful to the projects NORTE-01-0145-FEDER-000011 - Qualidade e Segurança Alimentar - uma abordagem (nano) tecnológica and UID/QUI/50006/2013POCI/01/0145/FEDER/007265 with support from FCT/MEC (Fundação para a Ciência e Tecnologia and Ministério da Educação e Ciência) through national funds and co-financed by FEDER, under the Partnership Agreement PT2020. This work received financial support from the European Union (FEDER funds through COMPETE POCI-01-0145-FEDER-016735) and National Funds (FCT, Fundação para a Ciência e Tecnologia) through project PTDC/AGR-PRO/6817/2014. Ricardo N.M.J. Páscoa thanks FCT (Fundação para a Ciência e Tecnologia) and POPH (Programa Operacional Potencial Humano) for Post-Doc Grant SFRH/BPD/81384/2011. To all financing sources the authors are greatly indebted. The authors thank as well “Os Três Cervejeiros, Lda” for providing the samples.
References [1] G. Korbecka-Glinka, U. Skomra, H. Olszak-Przybys, Cultivar identification in dry hop cones and pellets using microsatellite loci, European Food Research and Technology, 242 (2016) 1599-1605. [2] T. Hudcova, J. Bryndova, K. Fialova, J. Fiala, M. Karabin, L. Jelinek, P. Dostalek, Antiproliferative effects of prenylflavonoids from hops on human colon cancer cell lines, J. Inst. Brew., 120 (2014) 225-230. [3] D. De Keukeleire, S. Milligan, J. Kalita, V. Pocock, L. De Cooman, A. Heyerick, H. Rong, Prenylated hop flavonoids are key agents in relation to health-beneficial properties of beer, Proceedings of the 28th International Congress of the European Brewery Convention, Budapest, Hungary, 12-17 May 2001 CD-ROM, Contribution 8, 10 pp. Fachverlag Hans Carl, Nürnberg, Germany, 2001, 2001. [4] T. Henley, L. Reddivari, C.D. Broeckling, M. Bunning, J. Miller, J.S. Avens, M. Stone, J.E. Prenni, J. Vanamala, American India Pale Ale matrix rich in xanthohumol is potent in suppressing proliferation and elevating apoptosis of human colon cancer cells, International journal of food science & technology, 49 (2014) 2464-2471. [5] B. Steenackers, L. De Cooman, D. De Vos, Chemical transformations of characteristic hop secondary metabolites in relation to beer properties and the brewing process: A review, Food chemistry, 172 (2015) 742-756. [6] A. Mongelli, M. Rodolfi, T. Ganino, M. Marieschi, C. Dall’Asta, R. Bruni, Italian hop germplasm: Characterization of wild Humulus lupulus L. genotypes from Northern Italy by means of phytochemical, morphological traits and multivariate data analysis, Industrial Crops and Products, 70 (2015) 16-27. [7] G.B. Nickerson, P.A. Williams, A. Haunold, Varietal differences in the proportions of cohumulone, adhumulone, and humulone in hops, J. Am. Soc. Brew. Chem, 44 (1986) 91-94.
[8] L. De Cooman, E. Everaert, D. De Keukeleire, Quantitative analysis of hop acids, essential oils and flavonoids as a clue to the identification of hop varieties, Phytochem. Anal., 9 (1998) 145-150. [9] S. Kenny, Identification of US-grown hop cultivars by hop acid and essential oil analyses, J. Am. Soc. Brew. Chem., 48 (1989) 3-8. [10] M. Kovačevič, M. Kač, Solid-phase microextraction of hop volatiles: Potential use for determination and verification of hop varieties, Journal of Chromatography A, 918 (2001) 159-167. [11] M. Kovačevič, M. Kač, Determination and verification of hop varieties by analysis of essential oils, Food chemistry, 77 (2002) 489-494. [12] A. Čerenak, M. Pavlovič, M.Oset Luskar, I.J. Košir, Characterisation of Slovenian hop (Humulus lupulus L.) varieties by analysis of essential oil, Hop Bulletin, 18 (2011) 27-32. [13] C. Neugrodda, M. Gastl, T. Becker, Protein Profile Characterization of Hop (Humulus lupulus L.) Varieties, J. Am. Soc. Brew. Chem., 72 (2014) 184-191. [14] M. Ocvirk, J. Grdadolnik, I.J. Kosir, Determination of the botanical origin of hops (Humulus lupulus L.) using different analytical techniques in combination with statistical methods, J. Inst. Brew., 122 (2016) 452-461. [15] R.A. Shellie, S.D.H. Poynter, J.F. Li, J.L. Gathercole, S.P. Whittock, A. Koutoulis, Varietal characterization of hop (Humulus lupulus L.) by GC-MS analysis of hop cone extracts, J. Sep. Sci., 32 (2009) 3720-3725. [16] F. Gu, W. Zhang, H. Li, Identification of hop varieties with AFLP technique, Hop Bulletin, 13(2006) 53-59. [17] M.S. Abbott, M.J. Fedele, A DNA-based varietal identification procedure for hop leaf tissue, J. Inst. Brew., 100 (1994) 283-285. [18] A. Cerenak, J. Jakse, B. Javornik, Identification and differentiation of hop varieties using simple sequence repeat markers, J. Am. Soc. Brew. Chem., 62 (2004) 1-7. [19] N. Stajner, J. Jakse, P. Kozjak, B. Javornik, The isolation and characterisation of microsatellites in hop (Humulus lupulus L.), Plant science, 168 (2005) 213-221. [20] J. Jakse, Z. Luthar, B. Javornik, New polymorphic dinucleotide and trinucleotide microsatellite loci for hop Humulus lupulus L, Mol. Ecol. Resour., 8 (2008) 769-772. [21] J. Koelling, M.C. Coles, P.D. Matthews, A. Schwekendiek, Development of new microsatellite markers (SSRs) for Humulus lupulus, Molecular breeding, 30 (2012) 479-484. [22] E.L. Howard, S.P. Whittock, J. Jakše, J. Carling, P.D. Matthews, G. Probasco, J.A. Henning, P. Darby, A. Cerenak, B. Javornik, A. Kilian, A. Koutoulis, High-throughput genotyping of hop (Humulus lupulus L.) utilising diversity arrays technology (DArT), Theoretical and Applied Genetics, 122 (2011) 1265-1280. [23] H. Yamauchi, Y. Mukouzaka, T. Taniguchi, K. Nakashima, S. Furukubo, M. Harada, Newly developed SNP-based identification method of hop varieties, J. Am. Soc. Brew. Chem., 72 (2014) 239-245. [24] P. Matthews, M. Coles, N. Pitra, Next generation sequencing for a plant of great tradition: application of NGS to SNP detection and validation in hops (Humulus lupulus L.), Monatsschrift für Brauwissenschaft, 66 (2013) 186. [25] J.A. Henning, J. Coggins, M. Peterson, Simple SNP-based minimal marker genotyping for Humulus lupulus L. identification and variety validation, BMC research notes, 8 (2015) 542. [26] C.A.T. dos Santos, R.N. Páscoa, J.A. Lopes, A review on the application of vibrational spectroscopy in the wine industry: From soil to bottle, TrAC Trends in Analytical Chemistry, (2017). [27] K. Emura, S. Yamanaka, H. Isoda, K.N. Watanabe, Estimation for different genotypes of plants based on DNA analysis using near-infrared (NIR) and Fourier-transform infrared (FT-IR) spectroscopy, Breed. Sci., 56 (2006) 399-403. [28] S. Gutierrez, J. Tardaguila, J. Fernandez-Novales, M.P. Diago, Data Mining and NIR Spectroscopy in Viticulture: Applications for Plant Phenotyping under Field Conditions, Sensors, 16 (2016).
[29] M. Vaz, L. Meirinhos-Soares, C. Sousa, M. Ramirez, J. Melo-Cristino, J. Lopes, Serotype discrimination of encapsulated Streptococcus pneumoniae strains by Fourier-transform infrared spectroscopy and chemometrics, Journal of microbiological methods, 93 (2013) 102-107. [30] S. Barth, H. Meier, The Barth report, Barth-Haas Group, Nuremberg, (2017) 15-28. [31] T. Naes, T. Isaksson, T. Fearn, T. Davies, Interpreting PCR and PLS solutions, A User-Friendly Guide to Multivariate Calibration and Classification, 1 (2004) 39-54. [32] M. Barker, W. Rayens, Partial least squares for discrimination, J. Chemometr., 17 (2003) 166173. [33] R. Páscoa, M. Lopo, C.T. dos Santos, A. Graça, J. Lopes, Exploratory study on vineyards soil mapping by visible/near-infrared spectroscopy of grapevine leaves, Computers and Electronics in Agriculture, 127 (2016) 15-25. [34] A. Murakami, Hop variety classification using the genetic distance based on RAPD, J. Inst. Brew., 106 (2000) 157-161. [35] J.A. Henning, M.S. Townsend, S. Kenny, Potential heterotic crosses in hops as estimated by AFLP-based genetic diversity and coefficient of coancestry, J. Am. Soc. Brew. Chem., 62 (2004) 6370. [36] P. Zanoli, M. Zavatti, Pharmacognostic and pharmacological profile of Humulus lupulus L, Journal of Ethnopharmacology, 116 (2008) 383-396. [37] S. Hieronymus, For The Love of Hops: The Practical Guide to Aroma, Bitterness and the Culture of Hops, Brewers Publications2012. [38] C. Almaguer, C. Schonberger, M. Gastl, E.K. Arendt, T. Becker, Humulus lupulus - a story that begs to be told. A review, J. Inst. Brew., 120 (2014) 289-314. [39] J.F. Stevens, J.E. Page, Xanthohumol and related prenylflavonoids from hops and beer: to your good health!, Phytochemistry, 65 (2004) 1317-1330. [40] G. Eyres, J.-P. Dufour, 22 - Hop Essential Oil: Analysis, Chemical Composition and Odor Characteristics A2 - Preedy, Victor R, Beer in Health and Disease Prevention, Academic Press, San Diego, 2009, pp. 239-254.
Figures Figure 1. Hops raw spectra: NIR (a) and MIR (b).
a)
b)
Figure 2. Scores of the first two components using all NIR (a) and MIR (b) spectra (mean centred).
a)
b)
Legend:a) NIR scores; b) MIR scores; Fuggles; Ella; Nugget; Equinox; Hallertau Traditional; Cascade; Summit; Nelson Sauvin; Mount Hood; Citra; Cluster; Crystal; Tomahawk; Bobek; Bramling; Brewers Gold; Challenger; Goldings; Mosaic; Hallertau Magnum; Willamette; Target; Northern Brewer; Simcoe; Spalt Select; Saaz; Hersbrucker; Perle; Amarillo; Hallertau Mittelfrüh; Chinook; EKG; Tettnanger;
Figure 3. Scores of the first two components using the spectra pre-processed with SNV followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative and best spectral regions for NIR (a) and MIR (b) spectra.
a)
b)
Legend: a) NIR scores; b) MIR scores; Fuggles; Ella; Nugget; Equinox; Hallertau Traditional; Cascade; Summit; Nelson Sauvin; Mount Hood; Citra; Cluster; Crystal; Tomahawk; Bobek; Bramling; Brewers Gold; Challenger; Goldings; Mosaic; Hallertau Magnum; Willamette; Target; Northern Brewer; Simcoe; Spalt Select; Saaz; Hersbrucker; Perle; Amarillo; Hallertau Mittelfrüh; Chinook; EKG; Tettnanger;
Figure 4. Dendrogram obtained with hierarchical cluster analysis (Ward's method) using NIR data between 6,079 and 5,389 cm-1 and pre-processed with SNV followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative.
C1
C2
Figure 5. Dendrogram obtained with hierarchical cluster analysis (Ward's method) using MIR data between 1,880 to 600 cm-1 and with the best wavenumber region and pre-processed with SNV followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative.
C1
C2
Figure 6. Representation of the loadings of the first four latent variables of the PLSDA models using NIR (a) and MIR data (b).
a)
b)
Table 1. Variety designation, α-acids content (%) and country of origin of the 33 samples studied.
Variety Amarillo Bobek (Styrian Golding B) Bramling Cross Brewers Gold Cascade Challenger Chinook
α-acids (%) 9 5 4 4.7 6.9 6.5 12.5
Origin USA Slovenia UK UK USA UK USA
Citra Cluster Crystal East Kent Golding Ella Equinox Fuggle Goldings Hallertauer Magnum Hallertauer Mittelfrüh Hallertauer Tradition Hersbrucker Mosaic Mount Hood Nelson Sauvin Northen Brewer Nugget Perle Saaz Simcoe Spalter Select Summit Target Tettnanger Tomahawk Willamette
13 7.5 6 5.2 14 13.4 5.2 5.1 10.5 3.6 5 2.3 12.8 6.3 11.9 6 11 5 3 13.1 5.2 15 8 2.1 16.5 4.7
USA USA USA UK Australia USA UK UK Germany Germany Germany Germany USA USA New Zeland UK USA Germany Czech Republic USA Germany USA UK Germany USA USA
Highlights
Discrimination of the most important commercial hop varieties through near and mid infrared spectroscopy;
Use of chemometric tools, optimization of spectral regions and pre-processing methods;
94.2% correct hop varieties discrimination with near-infrared spectroscopy;
96.6% correct hop varieties discrimination with mid-infrared spectroscopy;
The method proved to be accurate, cost-effective, non-destructive and ecofriendly;
PII: DOI: Reference:
S0039-9140(17)31232-8 https://doi.org/10.1016/j.talanta.2017.12.030 TAL18166
To appear in: Talanta Received date: 15 September 2017 Revised date: 6 December 2017 Accepted date: 11 December 2017 Cite this article as: Julio C. Machado, Miguel A. Faria, Isabel M.P.L.V.O. Ferreira, Ricardo N.M.J. Páscoa and João A. Lopes, VARIETAL DISCRIMINATION OF HOP PELLETS BY NEAR AND MID INFRARED SPECTROSCOPY, Talanta, https://doi.org/10.1016/j.talanta.2017.12.030 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
VARIETAL DISCRIMINATION OF HOP PELLETS BY NEAR AND MID INFRARED SPECTROSCOPY
Julio C. Machado Jr.a, Miguel A. Fariaa, Isabel M. P. L. V. O. Ferreiraa, Ricardo N. M. J. Páscoab*, João A. Lopesc
a
LAQV/REQUIMTE/ Departamento de Ciências Químicas, Laboratório de Bromatologia e
Hidrologia, Faculdade de Farmácia, Universidade do Porto, Rua de Jorge Viterbo Ferreira, 228, 4050-313, Porto, Portugal b
LAQV/REQUIMTE, Laboratório de Química Aplicada, Departamento de Ciências
Químicas, Faculdade de Farmácia, Universidade do Porto, Rua de Jorge Viterbo Ferreira, 228, 4050-313, Porto, Portugal c
Research Institute for Medicines (iMed.ULisboa), Faculdade de Farmácia, Universidade de
Lisboa, Av. Prof. Gama Pinto, 1649-003, Lisbon, Portugal
*Corresponding author: [email protected]
Abstract
Hop is one of the most important ingredients of beer production and several varieties are commercialized. Therefore, it is important to find an eco-real-time-friendly-low-cost technique to distinguish and discriminate hop varieties. This paper describes the development of a method based on vibrational spectroscopy techniques, namely near- and mid-infrared spectroscopy, for the discrimination of 33 commercial hop varieties. A total of 165 samples (five for each hop variety) were analysed by both techniques. Principal component analysis, hierarchical cluster analysis and partial least squares discrimination analysis were the chemometric tools used to discriminate positively the hop varieties. After optimizing the spectral regions and pre-processing methods a total of 94.2% and 96.6% correct hop varieties discrimination were obtained for near- and mid-infrared spectroscopy, respectively. The results obtained demonstrate the suitability of these vibrational spectroscopy techniques to discriminate different hop varieties and consequently their potential to be used as an authenticity tool. Compared with the reference procedures normally used for hops variety discrimination these techniques are quicker, cost-effective, non-destructive and eco-friendly. Graphical abstract
Keywords: Hop, near infrared spectroscopy, mid infrared spectroscopy, discrimination, chemometrics;
1. Introduction Hop (Humulus lupulus L.) is one of the principal ingredients of beer, responsible for different intensities of bitterness and a wide range of specific flavours and aromas with a strong impact in beer organoleptic properties. The amount of specific flavours and aromas present in beer depends mainly on hops variety but also how it is used during the brewing process. For this reason the selection of hop’s variety is becoming the main target of interest for brewers aiming to produce different and exquisite beers styles. There is thus an increase on the research concerning the development and study of new varieties over the last years. Furthermore, several high valued hops had been launched in the market with much differentiated prices making its authentication crucial for brewers. In this context, fraudulent declarations or labelling of hop products must not be excluded. Total or partial replacement of expensive hops with less expensive ones results in economic benefit to dealers as it was reported recently [1]. Besides providing bitterness and aroma compounds to the beer, hop presents a wide range of bioactive properties, in particular antioxidant, anti-inflammatory, anticancer, immunomodulatory and antimicrobial activities associated to its composition [2-4]. These bioactive properties are related with the plant variety (each variety presents different chemical compounds profiles) and have impact also on its commercial value. Due to the abovementioned reasons more than two hundred hop varieties are presently cultivated worldwide. These differ qualitatively and quantitatively in their fraction of essential oils (characterized by terpenes, alcohols, acids, esters, ketones and aldehydes) and, in their non-volatile fraction of hard resins (composed by several polyphenols, such as xanthohumol, prenylnaringenin compounds and derivatives) and soft resins (containing prenylated phloroglucinol derivatives, α- and β- acids) [5]. The high number and phenotypic variability of hop fostered the development of several analytical methods to evaluate the chemical compounds present in each crop, which has been a common practice, as well as methodologies for the varietal authentication of hops in its different commercialized forms, being pellets of compressed dry flowers the most used nowadays. The
identification of hop varieties is generally performed using several strategies such as the evaluation of the morphological characteristics [6] - difficult to achieve when hop cones are processed - and the chemical analysis of hop compounds, namely bitter acids, flavonoids, essential oils, or proteins, individually or using the combination of different chemical groups [7-15]. As chemical markers are strongly influenced by environmental conditions during plant growth (region, soil, harvest time, agricultural practices, environmental factors), cones processing and storage, several DNA-based methods were developed targeting the plant phylogenetic studies and hops molecular identification, to circumvent the environmental variation of chemical markers. Several methods were thus proposed for hop varietal authenticity control as the AFLP - Amplified Fragment Length Polymosphism [16]; RAPD - Random Amplified Polymorphic DNA [17], SSR – Simple Sequence Repeats [1, 18-21] and DArT - diversity array technology markers [22]. Although several methods are available the analysis of SNPs (Single Nucleotide Polymorphisms), differences of single nucleotide in some homologous DNA among varieties, is presently considered the more reproducible tool for the identification of varieties of several plants [23]. Next Generation Sequencing (NGS) technologies applied to the study of the hop genome showed the presence, in average, of a Single Nucleotide Polymorphism (SNP) in every 346 bp of the hop genome [24]. These authors identified 17128 SNPs in the hop genome and concluded that a group of 3068 SNP retrieved the same level of discrimination amongst varieties. However the minimum number of markers to differentiate among all genotypes was not identified. Notwithstanding, Henning and collaborators (2015) identified 7 SNPs from a group of 374829 markers that effectively differentiated all 121 varieties and accessions (representing a broad spectrum of hop lines from around the world). Therefore these authors proposed the use of highresolution melting (HRM) curve analyses as a simple, rapid and more economically viable means to perform genetic fingerprinting on hop genotypes [25]. As SNP markers provide a greater likelihood of differentiating among varieties, the SNP/HRM approach was used in this study as an effective way of samples discrimination to validate their differentiation. All the above referred techniques usually require a considerable amount of sample and are destructive. To avoid such drawbacks, vibrational spectroscopic techniques, namely near infrared
(NIR) and mid-infrared (MIR) spectroscopy, emerge as suitable alternatives for hops discrimination. The NIR spectroscopy results from overtones and combination of fundamental vibrational bands, namely C-H, N-H, O-H and S-H bonds, in the spectral range of 14,000 to 4,000 cm-1. MIR spectroscopy is characterized for fundamental, bending and rotating vibrations in the spectral range of 4,000 to 400 cm-1. The MIR spectrum has a higher specificity than NIR spectrum and for this reason is considered more appropriate for identification and characterization purposes. Both these techniques are rapid, non-destructive with a low-cost per analysis [26]. There are already some examples in the literature exploring vibrational spectroscopic techniques for discrimination purposes, namely in genetically and non-genetically modified maize plants [27], grapevine varieties [28], bacteria species discrimination [29], and even hops [14] although focusing only a small sample size usually within the same country of origin. The aim of this study was thus to evaluate if NIR and MIR spectroscopy can be used as fast and nondestructive technique for the differentiation and identification of hop varieties targeting authenticity purposes. A total of 33 commercial varieties (representing about 75% of the total volume commercialized worldwide) were used. DNA-based SNP characterization was performed as a way to warrant samples differentiation. To the best of our knowledge this represents the first work exclusively based on NIR and MIR spectroscopy techniques to hop cultivars fast discrimination within a highly diverse group of varieties.
2. Experimental 2.1. Samples Commercial hops were acquired as pellets in local suppliers of raw materials for brewers. The varieties list is presented in Table 1 together with the content of α-acids and country of origin, as indicated by the manufacturer. The choice of the varieties to include in the study was done by selecting the most relevant in terms of market share accounting for 75% of the worldwide market volume in 2016 [30]. A total of 165 samples (five for each hop variety) were acquired including the
harvest years of 2015 and 2016. Samples were grinded to a homogeneous powder in a Retsch GM 200 mill (Retsch GmbH, Haan, Germany), stored in the dark and refrigerated until the moment of analysis.
2.2 MIR spectral acquisition A Fourier transform PerkinElmer Spectrum BX FTIR System spectrophotometer (Waltham, USA) equipped with a DTGS detector and a PIKE Technologies Gladi ATR accessory was used to collect the hop MIR spectra in diffuse reflectance mode. The spectra were acquired in the spectral range of 4,000 to 400 cm−1, with a resolution of 4 cm−1 and 32 scan co-additions. The hop samples were transferred to the ATR crystal and compressed until constant signal. For each hop cultivar, one small portion was transferred and a spectrum was collected for each portion. Therefore, a total of 5 spectra were acquired for each hop variety. Between each hop variety, the ATR crystal was cleaned and a background was acquired.
2.3 NIR spectral acquisition A Fourier transform near infrared spectrometer, FTLA 2000 from ABB (Québec, Canada) equipped with an indium-gallium-arsenide (InGaAs) detector was used to collect the hop NIR spectra in diffuse reflectance mode. Each spectrum was recorded as the average of 64 scans in the spectral range of 10,000 cm−1 and 4000 cm−1, with 8 cm−1 resolution. The hop samples were transferred into borosilicate flasks in order to perform spectra acquisition. The background was measured at the beginning of each analysis using a reference substance (Teflon). A total of 5 spectra were acquired for each hop variety.
2.4 NIR and MIR data analysis
The NIR and MIR data analysis was performed using principal component analysis (PCA) [31] and partial least squares discriminant analysis (PLSDA) [32]. PCA was used to detect outliers and find common patterns while the later was used to develop discrimination models. For PLSDA models, samples were classified according to the respective variety. The available data were divided in calibration (70%) and validation (30%) sets in a random way, but maintaining the same proportion of each variety in both sets (to avoid unbalanced hops variety classes in both sets) [33]. The assessment of the optimal number of latent variables (LV) was done using the leave-one-sample-out crossvalidation procedure using only the calibration set. The NIR and MIR spectra were divided in five spectral regions. The NIR regions were: R1 from 10,000 to 7,472 cm-1, R2 from 7,468 to 6,083 cm-1, R3 from 6,079 to 5,389 cm-1, R4 from 5,385 to 4,964 cm-1 and R5 from 4,961 to 4,035 cm-1. The MIR regions were: R1 from 4,000 to 2,782 cm-1, R2 from 2,780 to 1,882 cm-1, R3 from 1,880 to 1,492 cm-1, R4 from 1,490 to 874 cm-1 and 872 to 600 cm-1. These regions were tested individually and in combination for the selection of the best spectral regions. The selection of best processing technique involved testing standard normal variate (SNV) and Savitzky-Golay filter (with first and second derivatives, different filter widths and polynomial orders) individually and in combination. After the optimization of the best spectral regions and pre-processing methods, the test set was projected in the optimized PLSDA calibration model to assess the percentage of correct predictions. Model predictions were expressed as confusion matrices [33]. The total percentage of correct predictions was obtained adding the diagonal elements of the confusion matrices. One PLSDA model was developed for each spectroscopic technique (NIR and MIR). Loadings of the first four LV of each developed PLSDA were also analysed to understand which specific wavenumbers were more important. Spectral data were mean centred before application of PCA and PLSDA. All chemometric analyses were performed with Matlab version 8.6 (MathWorks, Natick, USA) and PLS Toolbox version 8.2.1 (Eigenvector Research Inc., Wenatchee, USA).
2.5 DNA analysis
DNA was extracted from 50 mg powdered samples using the E.Z.N.A. HP Plant DNA Mini Kit from Omega Bio-Tek (Norcross, USA) following the manufacturer instruction, with a slight modification consisting in a previous sample homogenisation using a Precellys Evolution tissue homogeniser (Bertin Instruments, Montigny-le-Bretonneux, France), in 2 ml tubes containing 1.4 mm zirconium oxide beads. Extracts were quantified and evaluated for purity by UV absorption at 260 and 280 nm in a 4 µL sample aliquot using a BioTek Synergy HT plate reader equipped with the Take3 MicroVolume Plate adapter. After quantification, all samples were diluted to 10 ng/µL with water for molecular biology (W4502, Sigma Aldrich, St. Louis, USA). HRMA was carried out using the primers proposed by Henning and collaborators [25]. Amplification reactions were conducted in a Bio-Rad CFX Connect thermalcycler using the SsoFast™ EvaGreen® Supermix (Bio-Rad, CA, USA), in 10 µL volumes. PCR conditions were optimized to reach the shortest time of analysis and defined as: an initial denaturation of 98 ºC for 2 min, followed by 35 cycles of 98 ºC for 5 s and the annealing temperature (from 57 to 63 ºC) for 5 s. Immediately after PCR, melting analysis was performed from 60-95 ºC in 0.5 ºC increments. Melting curve clustering was performed using the Precision Melt AnalysisTM v 1.0 Software from Bio- Rad (Hercules, CA, USA) with a confidence of at least 98% in group assignment.
3. Results and discussion 3.1 Validation of commercial samples by SNP discrimination The varietal genotyping was assessed using 7 SNP markers (TP137094, TP15403, TP245055, TP295074, TP400349, TP411590 and TP437202) required to differentiate all the commercial hop accessions, as proposed by the study of Henning and collaborators (2015) in which 121 were
discriminated [25]. Amplicons of each marker originate three different HRM curves corresponding each one to one of the allelic variants (two homozygotes and one heterozygote). SNP calling was performed by assigning each sample melting curve to one of the three allelic combinations for each SNP site. It was possible to achieve the complete discrimination of all the 33 varieties in study and to confirm its phylogenetic relatedness, assuring this way that samples acquired are genetically different eliminating putative problems of cross mislabelling.
3.2 Exploratory spectral analysis The NIR and MIR spectra of all hop samples were depicted in Figure 1. As abovementioned, a PCA was performed with both NIR and MIR spectral data. The data was just mean-centred before application of PCA and the entire spectral regions were used. The results suggested that no outliers were identified in both data sets. Then, with the objective of understanding if the NIR and MIR spectra gather specific information regarding hop variety, the first two principal components scores were plotted (Fig. 2). In the plots of both NIR and MIR scores it was visible some cluster formation tendency regarding hop’s variety, suggesting that both NIR and MIR spectra contain information related with hop variety. The differentiation was not clear.
3.2.1 Spectral regions and pre-processing methods optimisation The NIR and MIR data were divided in five spectral regions as mentioned in the material and methods section with the objective of identifying spectral regions more informative in terms of the hop’s variety. Several pre-processing techniques were tested as well. The rationale for selecting spectral windows and pre-processing methods was to minimise the distance within varieties while maximising the distance between varieties. The PLSDA results obtained with the calibration set in terms of correct prediction percentages were used to determine the best spectral regions and preprocessing methods. The best pre-processing technique for both NIR and MIR spectra was SNV
followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative. The best spectral regions were: 6,079 to 5,389 cm-1 and 1,880 to 600 cm-1 for NIR and MIR technique, respectively.
3.2.2 Optimised models After defining the best spectral region and pre-processing methods the first two principal components were plotted again (Fig. 3). It is clear that after this optimization step, most samples are now clustered according to the variety, with some exceptions, for both NIR and MIR data. Looking with more detail for the NIR data (Fig. 3a), Nelson Sauvin clustered very close to Amarillo, Citra with Summit and Hallertau Tradition with Hallertau Mittelfrüh The hop varieties Ella, East Kent Golding (EKG), Fuggles and Goldings clustered far apart from the rest. Regarding MIR data (Fig. 3b), the varieties that clustered very close were: Nelson Sauvin with Amarillo, Hallertau Mittelfrüh with Saaz and Challenger with Hallertau Tradition. Hersbrucker and Crystal clustered far apart from the others. These findings indicates that both NIR and MIR spectral data contain information related to hop’s variety and suggest the application of a supervised classification method.
3.3 Unsupervised classification Hierarchical cluster analysis (HCA) was applied for both NIR (Figure 4) and MIR (Figure 5) data considering the average of the five spectra of each hop variety and the best wavenumber region and pre-processing techniques as identified before. The dendrogram obtained using NIR data (Figure 4) grouped the hop varieties in two distinct clusters (C1 and C2). This distribution can be associated with the α-acids amount, because most of the hop varieties that contain a high amount of α-acids are clustered in C2 while varieties with α-acids content lower than 6% are aggregated in C1 cluster. The most similar hop varieties were Citra and Summit, and this is in agreement with the PCA findings. Regarding the dendrogram obtained using MIR data (Figure 5), the hop varieties were also grouped
in two different clusters (C1 and C2). This distribution can also be correlated with α-acids amount considering that most of the hop varieties that contain a high amount of α-acids are aggregated in one cluster (in this case C1) but not so clearly as in the case of NIR spectroscopy.
3.4 Supervised classification The selected supervised classification method used was PLSDA. The test set of NIR and MIR data was then projected using the best spectral region and pre-processing technique. Models considering the NIR and MIR data obtained a total of 96.6% (8 LV) and 94.2% (10 LV) of correct predictions, respectively. The results demonstrate the applicability of both NIR and MIR techniques to discriminate the 33 hop varieties used in the present study in a non-destructive way which configures a useful tool for authenticity purposes. These results were obtained with only 5 samples per hop variety. Further studies should be made increasing the amount of different samples per variety and also using more varieties representing the cultivated plant full diversity. The confusion matrices (one for each vibrational technique) revealed that there was no hop variety that could not be identified or misclassified with another hop (see supplementary data Table 1 and 2). In both confusion matrices the elements that are zero were not shown. The confusion matrix for NIR spectroscopy (see supplementary data Table 1) revealed that the worst hop variety prediction involved Perle and Goldings with 74% and 79% of correct predictions, respectively. The hop variety Perle was incorrectly predicted as Northern Brewer and Saaz while Goldings was incorrectly predicted as Fuggles. Regarding MIR spectroscopy, the worst hop variety prediction involved Simcoe and Perle with 79% and 72% of correct predictions, respectively. Simcoe was incorrectly predicted with Mosaic while Perle was incorrectly predicted as Northern Brewer and Bobek. The verified similarity between Perle and Northern Brewer varieties (corresponding to the lowest number of correct predictions for both NIR and FTIR data) is also supported by the genetic data previously published using RAPD analysis [34] where authors describe the varieties as very close genetically, as well as the breeding history of the variety Perle which has the Northern Brewer as the female parent
[35]. This similarity was verified as well in our data concerning the SNP analysis performed for the varietal authenticity check. Globally, the results obtained show the ability and accuracy of both NIR and MIR spectroscopic techniques to discriminate between different hop varieties.
3.4.1 Models analysis The first four latent variables loadings obtained in the PLSDA models using NIR and MIR data were depicted in Figure 6. These variables of the PLSDA model using NIR data (encompassing approximately 96% of the total variance) revealed that the most important wavenumbers region was located between 6,000 and 5,600 cm-1. This region corresponds to CH3, CH2, CH, SH and Ar-CH absorption bands (first overtone region). There is no published work that connects these absorption bands with the chemical compounds present in hop. Therefore, further studies should be performed in order to understand which chemical compounds are responsible for these absorption bands. However, once hops are rich in a complex mixture of polyphenolic compounds including, prenylphloroglucinols (α- and β- acids) and flavonoids (proanthocyanidins, flavanol glycosides, prenylchalcones and prenylflavanones), the authors believe that these can be related with the referred absorptions bands, together with a large fraction of volatile compounds, including terpenes (as myrcene, humulene, caryophyllene and farnesene among others) and molecules containing sulphur (as thiols, namely 4-mercapto-4-methyl-pentan-2-one, 3 mercaptohexan-1-ol,among others) [36, 37]. This is supposed once the total of α-acids and β-acids can vary from low values as 5% to values over 20% of total dry weight depending on the hop variety [38]. Moreover, the flavonoids fraction can vary between 3% and 6% (w/w), highlighting that the amount of the prenylchalcone xanthohumol, the major constituent of this group, can also vary from 0.1% to 1% (w/w) [38, 39]. The quantity of total oil is also known to vary considerably from 0.5% to 3.0% (v/w) among the different varieties of hops [40].
The first four latent variables of the PLSDA model using MIR data (encompassing approximately 75% of the total variance) showed that the most important wavenumbers regions were located between 1,800 and 1,600 cm-1 and 1,400 and 1,000 cm-1. The first region can be associated with proteins while the second can be connected with carbohydrates and DNA/RNA/phospholipids [29]. In fact, another important parameter that substantially changes between hop varieties is the amount of proteins referenced to be between 12 to 22% dry matter [13] which the authors believe that can contribute to hops discrimination using MIR.
4. Conclusion This work assesses the suitability of NIR and MIR spectroscopy to discriminate hop varieties in its most used form: pellets. The 33 varieties analysed represent the big majority of the hops world market (about 75%) which assures the applicability of the present method as a rapid and nondestructive tool that can be used for authenticity purposes in the hops supply chains. The increasingly high value of hop elite varieties can promote fraudulent practices in the market which sharpens the need of better and more expedite authenticity tools like the one herein proposed. Unsupervised methods applied to spectra provided indications that these techniques can be able to discriminate between hop varieties and suggested the application of a supervised method. After optimising the wavenumbers region and pre-processing techniques, a total of 96.6% (8 LV) and 94.2% (10 LV) of correct hop varieties prediction were obtained for NIR and MIR spectroscopy, respectively. The results obtained in this work are very promising and demonstrate the suitability of both vibrational techniques. Moreover, these methods are rapid, cost-effective, non-destructive and environmentally friendly (do not generate residues). Further studies including more samples per variety and reflecting different geographical origins, when possible, and years of harvest are needed to attest the robustness of these techniques.
Acknowledgements. Authors are grateful to the projects NORTE-01-0145-FEDER-000011 - Qualidade e Segurança Alimentar - uma abordagem (nano) tecnológica and UID/QUI/50006/2013POCI/01/0145/FEDER/007265 with support from FCT/MEC (Fundação para a Ciência e Tecnologia and Ministério da Educação e Ciência) through national funds and co-financed by FEDER, under the Partnership Agreement PT2020. This work received financial support from the European Union (FEDER funds through COMPETE POCI-01-0145-FEDER-016735) and National Funds (FCT, Fundação para a Ciência e Tecnologia) through project PTDC/AGR-PRO/6817/2014. Ricardo N.M.J. Páscoa thanks FCT (Fundação para a Ciência e Tecnologia) and POPH (Programa Operacional Potencial Humano) for Post-Doc Grant SFRH/BPD/81384/2011. To all financing sources the authors are greatly indebted. The authors thank as well “Os Três Cervejeiros, Lda” for providing the samples.
References [1] G. Korbecka-Glinka, U. Skomra, H. Olszak-Przybys, Cultivar identification in dry hop cones and pellets using microsatellite loci, European Food Research and Technology, 242 (2016) 1599-1605. [2] T. Hudcova, J. Bryndova, K. Fialova, J. Fiala, M. Karabin, L. Jelinek, P. Dostalek, Antiproliferative effects of prenylflavonoids from hops on human colon cancer cell lines, J. Inst. Brew., 120 (2014) 225-230. [3] D. De Keukeleire, S. Milligan, J. Kalita, V. Pocock, L. De Cooman, A. Heyerick, H. Rong, Prenylated hop flavonoids are key agents in relation to health-beneficial properties of beer, Proceedings of the 28th International Congress of the European Brewery Convention, Budapest, Hungary, 12-17 May 2001 CD-ROM, Contribution 8, 10 pp. Fachverlag Hans Carl, Nürnberg, Germany, 2001, 2001. [4] T. Henley, L. Reddivari, C.D. Broeckling, M. Bunning, J. Miller, J.S. Avens, M. Stone, J.E. Prenni, J. Vanamala, American India Pale Ale matrix rich in xanthohumol is potent in suppressing proliferation and elevating apoptosis of human colon cancer cells, International journal of food science & technology, 49 (2014) 2464-2471. [5] B. Steenackers, L. De Cooman, D. De Vos, Chemical transformations of characteristic hop secondary metabolites in relation to beer properties and the brewing process: A review, Food chemistry, 172 (2015) 742-756. [6] A. Mongelli, M. Rodolfi, T. Ganino, M. Marieschi, C. Dall’Asta, R. Bruni, Italian hop germplasm: Characterization of wild Humulus lupulus L. genotypes from Northern Italy by means of phytochemical, morphological traits and multivariate data analysis, Industrial Crops and Products, 70 (2015) 16-27. [7] G.B. Nickerson, P.A. Williams, A. Haunold, Varietal differences in the proportions of cohumulone, adhumulone, and humulone in hops, J. Am. Soc. Brew. Chem, 44 (1986) 91-94.
[8] L. De Cooman, E. Everaert, D. De Keukeleire, Quantitative analysis of hop acids, essential oils and flavonoids as a clue to the identification of hop varieties, Phytochem. Anal., 9 (1998) 145-150. [9] S. Kenny, Identification of US-grown hop cultivars by hop acid and essential oil analyses, J. Am. Soc. Brew. Chem., 48 (1989) 3-8. [10] M. Kovačevič, M. Kač, Solid-phase microextraction of hop volatiles: Potential use for determination and verification of hop varieties, Journal of Chromatography A, 918 (2001) 159-167. [11] M. Kovačevič, M. Kač, Determination and verification of hop varieties by analysis of essential oils, Food chemistry, 77 (2002) 489-494. [12] A. Čerenak, M. Pavlovič, M.Oset Luskar, I.J. Košir, Characterisation of Slovenian hop (Humulus lupulus L.) varieties by analysis of essential oil, Hop Bulletin, 18 (2011) 27-32. [13] C. Neugrodda, M. Gastl, T. Becker, Protein Profile Characterization of Hop (Humulus lupulus L.) Varieties, J. Am. Soc. Brew. Chem., 72 (2014) 184-191. [14] M. Ocvirk, J. Grdadolnik, I.J. Kosir, Determination of the botanical origin of hops (Humulus lupulus L.) using different analytical techniques in combination with statistical methods, J. Inst. Brew., 122 (2016) 452-461. [15] R.A. Shellie, S.D.H. Poynter, J.F. Li, J.L. Gathercole, S.P. Whittock, A. Koutoulis, Varietal characterization of hop (Humulus lupulus L.) by GC-MS analysis of hop cone extracts, J. Sep. Sci., 32 (2009) 3720-3725. [16] F. Gu, W. Zhang, H. Li, Identification of hop varieties with AFLP technique, Hop Bulletin, 13(2006) 53-59. [17] M.S. Abbott, M.J. Fedele, A DNA-based varietal identification procedure for hop leaf tissue, J. Inst. Brew., 100 (1994) 283-285. [18] A. Cerenak, J. Jakse, B. Javornik, Identification and differentiation of hop varieties using simple sequence repeat markers, J. Am. Soc. Brew. Chem., 62 (2004) 1-7. [19] N. Stajner, J. Jakse, P. Kozjak, B. Javornik, The isolation and characterisation of microsatellites in hop (Humulus lupulus L.), Plant science, 168 (2005) 213-221. [20] J. Jakse, Z. Luthar, B. Javornik, New polymorphic dinucleotide and trinucleotide microsatellite loci for hop Humulus lupulus L, Mol. Ecol. Resour., 8 (2008) 769-772. [21] J. Koelling, M.C. Coles, P.D. Matthews, A. Schwekendiek, Development of new microsatellite markers (SSRs) for Humulus lupulus, Molecular breeding, 30 (2012) 479-484. [22] E.L. Howard, S.P. Whittock, J. Jakše, J. Carling, P.D. Matthews, G. Probasco, J.A. Henning, P. Darby, A. Cerenak, B. Javornik, A. Kilian, A. Koutoulis, High-throughput genotyping of hop (Humulus lupulus L.) utilising diversity arrays technology (DArT), Theoretical and Applied Genetics, 122 (2011) 1265-1280. [23] H. Yamauchi, Y. Mukouzaka, T. Taniguchi, K. Nakashima, S. Furukubo, M. Harada, Newly developed SNP-based identification method of hop varieties, J. Am. Soc. Brew. Chem., 72 (2014) 239-245. [24] P. Matthews, M. Coles, N. Pitra, Next generation sequencing for a plant of great tradition: application of NGS to SNP detection and validation in hops (Humulus lupulus L.), Monatsschrift für Brauwissenschaft, 66 (2013) 186. [25] J.A. Henning, J. Coggins, M. Peterson, Simple SNP-based minimal marker genotyping for Humulus lupulus L. identification and variety validation, BMC research notes, 8 (2015) 542. [26] C.A.T. dos Santos, R.N. Páscoa, J.A. Lopes, A review on the application of vibrational spectroscopy in the wine industry: From soil to bottle, TrAC Trends in Analytical Chemistry, (2017). [27] K. Emura, S. Yamanaka, H. Isoda, K.N. Watanabe, Estimation for different genotypes of plants based on DNA analysis using near-infrared (NIR) and Fourier-transform infrared (FT-IR) spectroscopy, Breed. Sci., 56 (2006) 399-403. [28] S. Gutierrez, J. Tardaguila, J. Fernandez-Novales, M.P. Diago, Data Mining and NIR Spectroscopy in Viticulture: Applications for Plant Phenotyping under Field Conditions, Sensors, 16 (2016).
[29] M. Vaz, L. Meirinhos-Soares, C. Sousa, M. Ramirez, J. Melo-Cristino, J. Lopes, Serotype discrimination of encapsulated Streptococcus pneumoniae strains by Fourier-transform infrared spectroscopy and chemometrics, Journal of microbiological methods, 93 (2013) 102-107. [30] S. Barth, H. Meier, The Barth report, Barth-Haas Group, Nuremberg, (2017) 15-28. [31] T. Naes, T. Isaksson, T. Fearn, T. Davies, Interpreting PCR and PLS solutions, A User-Friendly Guide to Multivariate Calibration and Classification, 1 (2004) 39-54. [32] M. Barker, W. Rayens, Partial least squares for discrimination, J. Chemometr., 17 (2003) 166173. [33] R. Páscoa, M. Lopo, C.T. dos Santos, A. Graça, J. Lopes, Exploratory study on vineyards soil mapping by visible/near-infrared spectroscopy of grapevine leaves, Computers and Electronics in Agriculture, 127 (2016) 15-25. [34] A. Murakami, Hop variety classification using the genetic distance based on RAPD, J. Inst. Brew., 106 (2000) 157-161. [35] J.A. Henning, M.S. Townsend, S. Kenny, Potential heterotic crosses in hops as estimated by AFLP-based genetic diversity and coefficient of coancestry, J. Am. Soc. Brew. Chem., 62 (2004) 6370. [36] P. Zanoli, M. Zavatti, Pharmacognostic and pharmacological profile of Humulus lupulus L, Journal of Ethnopharmacology, 116 (2008) 383-396. [37] S. Hieronymus, For The Love of Hops: The Practical Guide to Aroma, Bitterness and the Culture of Hops, Brewers Publications2012. [38] C. Almaguer, C. Schonberger, M. Gastl, E.K. Arendt, T. Becker, Humulus lupulus - a story that begs to be told. A review, J. Inst. Brew., 120 (2014) 289-314. [39] J.F. Stevens, J.E. Page, Xanthohumol and related prenylflavonoids from hops and beer: to your good health!, Phytochemistry, 65 (2004) 1317-1330. [40] G. Eyres, J.-P. Dufour, 22 - Hop Essential Oil: Analysis, Chemical Composition and Odor Characteristics A2 - Preedy, Victor R, Beer in Health and Disease Prevention, Academic Press, San Diego, 2009, pp. 239-254.
Figures Figure 1. Hops raw spectra: NIR (a) and MIR (b).
a)
b)
Figure 2. Scores of the first two components using all NIR (a) and MIR (b) spectra (mean centred).
a)
b)
Legend:a) NIR scores; b) MIR scores; Fuggles; Ella; Nugget; Equinox; Hallertau Traditional; Cascade; Summit; Nelson Sauvin; Mount Hood; Citra; Cluster; Crystal; Tomahawk; Bobek; Bramling; Brewers Gold; Challenger; Goldings; Mosaic; Hallertau Magnum; Willamette; Target; Northern Brewer; Simcoe; Spalt Select; Saaz; Hersbrucker; Perle; Amarillo; Hallertau Mittelfrüh; Chinook; EKG; Tettnanger;
Figure 3. Scores of the first two components using the spectra pre-processed with SNV followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative and best spectral regions for NIR (a) and MIR (b) spectra.
a)
b)
Legend: a) NIR scores; b) MIR scores; Fuggles; Ella; Nugget; Equinox; Hallertau Traditional; Cascade; Summit; Nelson Sauvin; Mount Hood; Citra; Cluster; Crystal; Tomahawk; Bobek; Bramling; Brewers Gold; Challenger; Goldings; Mosaic; Hallertau Magnum; Willamette; Target; Northern Brewer; Simcoe; Spalt Select; Saaz; Hersbrucker; Perle; Amarillo; Hallertau Mittelfrüh; Chinook; EKG; Tettnanger;
Figure 4. Dendrogram obtained with hierarchical cluster analysis (Ward's method) using NIR data between 6,079 and 5,389 cm-1 and pre-processed with SNV followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative.
C1
C2
Figure 5. Dendrogram obtained with hierarchical cluster analysis (Ward's method) using MIR data between 1,880 to 600 cm-1 and with the best wavenumber region and pre-processed with SNV followed by Savitzky-Golay with 15 points of filter width, 2nd polynomial order and 2nd derivative.
C1
C2
Figure 6. Representation of the loadings of the first four latent variables of the PLSDA models using NIR (a) and MIR data (b).
a)
b)
Table 1. Variety designation, α-acids content (%) and country of origin of the 33 samples studied.
Variety Amarillo Bobek (Styrian Golding B) Bramling Cross Brewers Gold Cascade Challenger Chinook
α-acids (%) 9 5 4 4.7 6.9 6.5 12.5
Origin USA Slovenia UK UK USA UK USA
Citra Cluster Crystal East Kent Golding Ella Equinox Fuggle Goldings Hallertauer Magnum Hallertauer Mittelfrüh Hallertauer Tradition Hersbrucker Mosaic Mount Hood Nelson Sauvin Northen Brewer Nugget Perle Saaz Simcoe Spalter Select Summit Target Tettnanger Tomahawk Willamette
13 7.5 6 5.2 14 13.4 5.2 5.1 10.5 3.6 5 2.3 12.8 6.3 11.9 6 11 5 3 13.1 5.2 15 8 2.1 16.5 4.7
USA USA USA UK Australia USA UK UK Germany Germany Germany Germany USA USA New Zeland UK USA Germany Czech Republic USA Germany USA UK Germany USA USA
Highlights
Discrimination of the most important commercial hop varieties through near and mid infrared spectroscopy;
Use of chemometric tools, optimization of spectral regions and pre-processing methods;
94.2% correct hop varieties discrimination with near-infrared spectroscopy;
96.6% correct hop varieties discrimination with mid-infrared spectroscopy;
The method proved to be accurate, cost-effective, non-destructive and ecofriendly;