Identification of Plant-Pathogenic Fungi Using Fourier Transform Infrared Spectroscopy Combined with Chemometric Analyses

CHAI A-li, WANG Yi-kai, ZHU Fa-di, SHI Yan-xia, XIE Xue-wen, LI Bao-ju*

1. Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences, Beijing 100081, China 2. Beijing Entry-Exit Inspection and Quarantine Bureau, Beijing 100026, China

Identification of Plant-Pathogenic Fungi Using Fourier Transform Infrared Spectroscopy Combined with Chemometric Analyses

CHAI A-li1, WANG Yi-kai2, ZHU Fa-di1, SHI Yan-xia1, XIE Xue-wen1, LI Bao-ju1*

1. Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences, Beijing 100081, China 2. Beijing Entry-Exit Inspection and Quarantine Bureau, Beijing 100026, China

Identification of plant-pathogenic fungi is time-consuming due to cultivation and microscopic examination and can be influenced by the interpretation of the micro-morphological characters observed. The present investigation aimed to create a simple but sophisticated method for the identification of plant-pathogenic fungi by Fourier transform infrared (FTIR) spectroscopy. In this study, FTIR-attenuated total reflectance (ATR) spectroscopy was used in combination with chemometric analysis for identification of important pathogenic fungi of horticultural plants. Mixtures of mycelia and spores from 27 fungal strains belonging to nine different families were collected from liquid PD or solid PDA media cultures and subjected to FTIR-ATR spectroscopy measurements. The FTIR-ATR spectra ranging from 4 000 to 400 cm-1were obtained. To classify the FTIR-ATR spectra, cluster analysis was compared with canonical vitiate analysis (CVA) in the spectral regions of 3 050～2 800 and 1 800～900 cm-1. Results showed that the identification accuracies achieved 97.53% and 99.18% for the cluster analysis and CVA analysis, respectively, demonstrating the high potential of this technique for fungal strain identification.

Fourier transform infrared spectroscopy (FTIR); Plant-pathogenic fungi; Identification; Cluster analysis; Canonical vitiate analysis

Introduction

Fungi are considered as serious pathogens to many plants and can cause a severe economic damage. Early detection and identification of these pathogens is crucial for timely control. The current methods available for identification of fungi are time consuming and not always very specific. Typically, fungi cultures, DNA-based methods, and antibody-based detection schemes are used to identify fungi[1-3]. Cultural methodology is time-consuming, in some cases even for specialists, since strains must be grown in pure culture and need to be examined by microscopy[4]. Molecular biology method has been reported as one of the most rapid and sensitive methods for the identification of pathogens, but they are not yet in large-scale use due to their high costs and high requirements for skills[5]. Serological methods based on specificity of the antigen-antibody reaction in immune systems are not always highly specific[6]. Therefore, there is a need for the development of more direct, rapid, and sensitive techniques for identification of plant-pathogenic fungi.

Fourier transform infrared (FTIR) spectroscopy has been proved to be a perfect tool for identifying microorganisms[7]. Biological molecules such as lipids, proteins, nucleic acids, polysaccharides, and phosphate-carrying compounds all generate specific FTIR spectra, which provide biochemical information regarding the molecular composition, structure, and interaction in cells. The molecular compositions and cell structures differ among different microorganisms[8]. Therefore, based on the FTIR spectra of the whole cell, single microorganisms could be identified. In the last decades, FTIR spectroscopy has successfully been applied in various fields for the identification of single cell microorganisms[9-10], such as bacteria[11-13], yeasts[14-16]and actinomycetes[17]. However, few studies have focused on the identification of multicellular fungal mycelia and spores—the important growth forms of the vast majority of fungi—based on FTIR spectroscopy. In order to obtain representative reproducible spectra of mycelia and spores mixture, the procedure established for single cells should be modified for sample preparation and measurement techniques.

FTIR spectroscopy equipped with attenuated total reflection (ATR) is considered as one of simple, direct, flexible and sensitive technique, it offers possibility for directly investigating the chemical compositions of various materials on smooth surfaces[18-19]. This technique has been employed for discrimination of microorganisms through a very simple sample preparation[20-21]. However, fungal identification by FTIR-ATR spectroscopy is still a challenging task as vast information is made available from the microbial structure without the extraction of cell materials. The questions that originate from microfungal spectroscopic studies are as follows. (1) How to obtain representative reproducible spectra of fungal mycelia and spores mixture? (2) What information should be used for fungal spectra identification?

Cluster analysis and canonical vitiate analysis (CVA) are two other chemometric analyses that can be used to group data objects on the basis of similarities among the samples. Cluster analysis and CVA analysis are widely recognized as powerful tools in obtaining information about relations within a dataset[22-23].

The aim of present study was to use FTIR-ATR spectroscopy for identification of various pathogenic fungi strains of horticultural plants. Twenty seven important plants-pathogenic fungi from vegetables, flowers and fruits represent different hierarchic fungal phyla, orders and families were selected in the study in order to obtain a representative overview of the classification potential. Cluster analysis and canonical vitiate analysis (CVA) were used to classify the FTIR-ATR spectra of fungal strains.

1 Materials and Methods

1.1 Fungal Strains

A total of 27 fungal strains that cause severe damages to horticultural plants were selected for this study. The 27 strains belong to nine phylogenetic families from five classes and three phyla. OnlySclerotiniasclerotiorumis a member of the phylum Ascomycota,Phytophthoracapsici,Phytophthoracactorum,PythiumaphanidermatumandPythiumdebaryanumare members of phylum Chromista, whereas the remaining 22 strains are Anamorphic fungi. The hierarchy of the fungal genera was derived from CABI databases and is given in Table 1. Strains tested were deposited in the collection of the Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences.

1.2 Mycelia Harvest

Selected fungal strains were grown on PDA (Potato Dextrose Agar) plates at 26 ℃. A pre-culture aged 5～10 days was used for inoculation. Fungal mycelia and spores were produced using two different culture methods.

(1) Liquid PD (Potato Dextrose) culture. A small agar plug from precultured plate was inoculated in 150 mL liquid PD medium in a 250 mL flask, and kept shaking at 200 rpm for 7～10 days after inoculation. Fungal mycelia and spores were harvested in the late log growth phase, washed thrice with sterile 0.9% saline solution to remove medium, then fungal were desiccated under vacuum (0.1 bar) for several hours, and finally pulverized in 2 mL test tubes by a ball mill.

(2) Cellophane-overlay PDA solid culture. A small agar plug was inoculated in cellophane-overlay PDA in a 9 cm petri dish and kept static. After sufficient mycelia had developed for subsequent measurements (5～15 days), mycelia and spores were scraped off the cellophane. For each strain, mixtures of mycelia and spores were always harvested after the same number of days, frozen in liquid nitrogen, stored at -80 ℃ until freeze drying and pulverizing by a ball mill in 2 mL test tubes.

The experiment was repeated three times, which means that mixtures of mycelia and spores from the same strains were produced in three independent experimental sets in August 2010 (PD culture), November 2011 (PD culture) and January 2012 (PDA culture), respectively. This would give information on possible effects of annual development and growth patterns on fungal spectra discrimination. Moreover, three replicates were inoculated per strain in each sets (for a total of 9 replicates per strain) in order to check for the reproducibility of fungal FTIR spectra. Samples of experiments in 2010, 2011 and 2012 were indicated by adding “a”, “b” and “c” to the strain names, whereas replicates of the same sets were indicated by adding “1”, “2” and “3” (e.g.Alternariasolania 1).

1.3 Spectral Measurement

Fourier transform infrared spectra were recorded using a spectrometer (Spectrum 100，Perkin-Elmer Corporation, USA) equipped with deuterated tryiglycine sulphate (DTGS) detector. The sampling station was equipped with an attenuated total reflection (ATR) accessory (PIKE technology, USA), which consisted of transfer optics within the chamber, through which infrared radiation was directed to a detachable ATR zinc selenide crystal mounted in a shallow trough for sample containment. The spectrometer system was controlled by a computer running Spectrum Version 6.0 software.

Table 1 Hierarchy of the fungal strains included in this study

The FTIR-ATR spectra were acquired from 4 000 to 800 cm-1with a spectra resolution of 4 cm-1. For each FTIR-ATR spectrum, a number of 64 scans were collected in transmission mode and co-added to improve the signal-to-noise ratio of the spectrum. A background measurement (air) was recorded before each measurement.

1.4 Data Analysis

All fungal FTIR-ATR spectra were evaluated by cluster analysis and canonical vitiate analysis (CVA). Data analysis was carried out using the SAS version 9.0 (SAS Institute Inc., USA). Firstly, the spectra were normalized. Then first derivatives with 9 smoothing points were independently vector-normalized.

Cluster analysis was used to display similarities of fungal spectra graphically in a dendrogram. The dendrogram was constructed using Pearson’s correlation coefficient and Ward’s algorithm. The spectral distance, also calleddvalue, is a measure of the similarity of the spectra of two strains. Thedvalues between all spectra were calculated. The resulting distance matrix provided the basis for cluster analysis.

For CVA analysis, the principal component analysis (PCA) data compression method was firstly applied to transform the data set, which consisted of a large number of inter-correlated variates (wavenumbers), into a reduced new set of variates. Then CVA analysis was performed on the resulting compressed set of spectral variates to classify the fungal spectra.

2 Results and discussion

2.1 Selection of Spectral Windows

The classification of fungal spectra depends on inter- versus inter-strain spectral differences. To visualize these differences, mean FTIR-ATR spectra of fungal mycelia and spores mixtures were shown in Fig.1. As an example, one strain of the nine phylogenetic families included in this study was selected. The spectra were screened for those spectral windows that showed the most pronounced differences between the strains tested. The spectral windows were known to characteristic for certain chemical structures: [W1] 3 700～3 050 cm-1, O—H and N—H bands regions of various micromolecules; [W2] 3 050～2 800 cm-1, the fatty acid region; [W3] 1 800～1 485 cm-1, the amide Ⅰ and amide Ⅱ bands from proteins and peptides; [W4] 1 485～1 185 cm-1, the mixed regions with bands from proteins, fatty acid and phosphate compounds; [W5] 1 185～900 cm-1, the polysaccharides region; [W6] 900～800 cm-1, the true fingerprint region where most of the bands were unassigned to specific cellular compounds or functional groups. Spectral differences between the strains were especially visible in the [W2], [W3], [W4] and [W5] according to the first derivative spectra (Fig.2). Therefore, the spectra regions of 3 050～2 800 and 1 800～900 cm-1corresponding to various characteristic peaks of fungal strains were selected for cluster analysis and CVA analysis.

Fig.1 FTIR-ATR spectra of mycelia and spores mixtures from nine fungal strains

Mean spectra of 9 replicates (6 from liquid PD, 3 from solid PDA media cultures) per strain from 3 independent experimental sets. Subranges W2—W5 were used for classification of fungal spectra

Fig.2 First derivatives of FTIR-ATR spectra from nine fungal strains

3.2 Cluster Analysis of Fungal FTIR-ATR Spectra

Cluster analysis was performed for grouping all spectra according to their spectral similarity. Strikingly, uhe cluster analysis grouped all the nine replicate spectra of each strain in separate groups, apart from a few exceptions, as will be discussed bellowed.

In all three experiments, 25 out of 27 strains were clustered cosrectly in separate groups (Fig.3). The only two strains for which not all nine replicates were placed in one groups werePhytophthoracactorumandPythiumaphanidermatum. Specifically, five out of nineP.cactorumspectra (Fig.3:P.cactorumb3, a2, c3, b2 and a1, marked by ←1) were separated from the otherP.cactorum. Furthermore, these fiveP.cactorumspectra were grouped out of the other members of Chromista (Fig. 3:←3), e.g.Phytophthoracapsici,PythiumaphanidermatumandPythiumdebaryanum. In contract, the spectrum b1 ofP.aphanidermatum(Fig.3:←2) clustered with the rest ofP.cactorumspectra (P.cactorumc2, b1, c1 and a3). Only these 6 out of the total of 243 spectra, i.e. 2.47% of all spectra, definitely grouped to the wrong strain sub-cluster at the strains level and have to be considered wrongly classified.

The ambiguous clustering of theP.cactorumspectra was probably caused by large differences between the nine replicate spectra, which resulted in the extraordinary high heterogeneity of 4.1 (Fig.3). This heterogeneity was much higher than that within all other strains. The variability between replicates of other strains ranged mostly below 0.2 heterogeneity in all nine replicates, only with the exception ofF.moniliforme(1.0) (Fig.3). The spectrum b1 ofP.aphanidermatumwas placed in one group with four spectra ofP.cactorum, this may be explained by the spectral distance between these two strains. The spectral distance betweenP.cactorumandP.aphanidermatumwas 0.6, which was the smallest inter-strains spectral distance of all tested strains (Table 2). The spectrum distances betweenP.cactorumand other Chromista fungal strains were relatively lower (Table 2: 0.6, 0.8, 1.1) than the rest of the distance metric, indicating that the strains within this group were compositionally more similar to one another than with the other strains.

2.3 Canonical Vitiate Analysis of Fungal FTIR-ATR Spectra

As an alternative classification method, CVA analysis was performed to evaluate the same set of fungal spectra. Briefly, FTIR-ATR spectra of fungal strains were subjected to PCA data compression, and followed by CVA analysis. PCA explained 95% of the total variability using the first eight principal components (PCs), whereas PC1 and PC2 were particularly representative of the spectral information and accounted for 78% of the total variance. Figure 4 shows the CVA plots for the 27 fungal strains using the first two PCs. The results showed that CVA analysis correctly assigned 99.18% of spectra to their respective groups with first two PCs, i.e. 241 of the 243 fungal spectra could be classified correctly. Only two out of nineP.cactorumspectra were misclassified toP.aphanidermatum. Moreover, the misclassified spectra ofP.cactorumandP.aphanidermatumin the CVA analysis were identical to the cluster analysis.

Fig.3 Cluster analysis of fungal FTIR spectral from all three experiments

(a)—(c)： Indicate samples of experiments 1～3, respectively; “1—3”： Indicate replicates of the same strain. Numbers at dendrogram backbones indicate heterogeneity. Cluster analysis was done using Ward’s algorithm; spectral ranges selected were 3 050～2 800 and 1 800～900 cm-1

Table 2 Distance measures between strains expressed as heterogeneity given for mean value of 9 replicates per strain from 3 independent experimental sets

Notes: 1：A.brassicicola; 2:A.solani, 3：A.cucumerina；4：C.cucumernum；5：C.cassiicola；6：C.apii；7：C.canescens, 8：S.solani；9：B.cinerea；10：T.roseum；11：V.dahliae；12：P.penicillium；13：F.oxysporum；14：F.thapsinum；15：F.solani；16：F.semitectum；17：M.roridum；18：P.guepinii；19：C.orbiculare；20：P.vexans；21：A.citrullina；22：R.solani；23：S.sclerotiorum；24：P.capsici；25：P.cactorum；26：P.aphanidermatum；27：P.debaryanum

Fig.4 CVA analysis of fungal FTIR-ATR spectra with PC1 and PC2 scores

3 Conclusions

In the present investigation, FTIR-ATR spectroscopy as a novel technique procedure in combination with cluster analysis and CVA analysis was evaluated for identification of important fungal pathogens that cause severe damage to horticultural plants. Few studies have successfully investigated on the identification of microfungi based on FTIR spectroscopy up till now, partly due to the preparation of living mycelia and spore cells extremely difficult. In this study, mixtures of mycelia and spores—the important growth form—from 27 fungal strains were collected from liquid PD and solid PDA media cultures and subjected to FTIR-ATR spectroscopy measurements. The classification accuracies achieved 97%～99% for both cluster and CVA analysis methods, demonstrating the high potential of this technique for fungal strains identification.

The results obtained here can serve as a basis for the development of a database for species identification and strain characterization of fungal pathogens. Considering the simplicity, rapidity, and low cost of the method presented, FTIR-ATR spectroscopy could be developed into a routine analysis tool for identification of microfungi, as it is less time-consuming than the conventional culture method and less expensive compared to the molecular approaches.

[1] Myint M S, Johnson Y J, Tablante N L, et al. Food Microbiol., 2006, 23(6): 599.

[2] Hein G, Flekna G, Krassning M, et al. J. Microbiol. Methods, 2006, 66(3): 538.

[3] Fischer G, Dott W. Int. J. Hyg. Environ. Health, 2002, 205(6): 433.

[4] Jong S C, Davis E. Mycopathologia, 1978, 66(3): 153.

[5] Correa A A, Toso J, Albarnaz J D, et al. J. Food Quality, 2006, 29(5):458.

[6] Lau A, Chen S, Sleiman S, et al. Future Microbiol., 2009, 4(9): 1185.

[7] Gupta B S, Jelle B P, Gao T. Int. J. Spectrosc., 2015, Article ID 521938.

[8] Sivakesava S, Irudayaraj J, DebRoy C. Trans. ASAE, 2004, 73(3): 951.

[9] Naumann D, Helm D, Labischinski H. Nature, 1991, 351(6321): 81.

[10] Naumann A. Analyst, 2009, 134(6):1215.

[11] Baldauf N A, Rodriguez-Romo L A, Yousef A E, et al. Appl. Spectrosc., 2006, 60(6): 592.

[12] Li H, Tripp C P. Appl. Spectrosc., 2008, 62(9): 963.

[13] Samuels A C, Snyder A P, Emge D K, et al., Appl. Spectrosc., 2009, 63(1): 14.

[14] Rellini P, Roscini L, Fatichenti F, et al. FEMS Yeast Research, 2009, 9(3): 460.

[15] Corte L, Rellini P, Roscini L, et al. Anal. Chim. Acta, 2010, 659: 258.

[16] Wenning M, Seiler H, Scherer S. Appl. Environ. Microbiol., 2002, 68(10): 4717.

[17] Haag H, Gremlich H U, Bergmann R, et al. J. Microbiol. Methods, 1996, 27: 157.

[18] Kos G, Lohninger H, Krska R. Anal. Chem. 2003, 75(5): 1211.

[19] Schmitt J, Flemming H-C. Int. Biodeter. Biodegr, 1998, 41(1): 1.

[20] Santos C, Fraga M E, Kozakiewicz Z, et al. Res. Microbiol.，2010, 161(2): 168.

[21] Salman A, Tsror L, Pomerantz A, et al. Spectroscopy, 2010, 24: 261.

[22] Poulli K I, Mousdis G A, Georgiou C A. Anal. Chim. Acta, 2005, 542(2): 151.

[23] Forina M, Oliveri P, Lanteri S, et al. Chemometr. Intell. Lab, 2008, 93(2): 132.

O657.3

A

Foundation item： the National Natural Science Foundation of China (31201473), the Science and Technology Innovation Program of the Chinese Academy of Agricultural Sciences (CAAS-ASTIP-IVFCAAS), and was funded by the Key Laboratory of Biology and Genetic Improvement of Horticultural Crops, Ministry of Agriculture, P.R. China

10.3964/j.issn.1000-0593(2016)11-3764-08

Received： 2015-10-13; accepted： 2016-03-10

Biography： CHAI A-li, (1983—), female, associate investigator in Chinese Academy of Agricultural Sciences e-mmail: chaiali@163.com *Corresponding author e-mail: libaoju@caas.cn