Construction of Raman spectroscopic fingerprints for the detection of Fusarium wilt of banana in Taiwan

Banana (Musa sp.) is cultivated worldwide and is one of the most popular fruits. The soil-borne fungal disease Fusarium wilt of banana (FWB), commonly known as Panama disease, is caused by Fusarium oxysporum f. sp. cubense (Foc) and is a highly lethal vascular fungal disease in banana plants. Raman spectroscopy, an emerging laser-based technology based on Raman scattering, has been used for the qualitative characterization of biological tissues such as foodborne pathogens, cancer cells, and melamine. In this study, we describe a Raman spectroscopic technique that could potentially be used as a method for diagnosing FWB. To that end, the Raman fingerprints of Foc (including mycelia and conidia) and Foc-infected banana pseudostems with varying levels of symptoms were determined. Our results showed that eight, eleven, and eleven characteristic surface-enhanced Raman spectroscopy peaks were observed in the mycelia, microconidia, and macroconidia of Foc, respectively. In addition, we constructed the Raman spectroscopic fingerprints of banana pseudostem samples with varying levels of symptoms in order to be able to differentiate Foc-infected bananas from healthy bananas. The rate at which FWB was detected in asymptomatic Foc-infected samples by using the spectral method was 76.2%, which was comparable to the rates previously reported for other FWB detection methods based on real-time PCR assays, suggesting that the spectral method described herein could potentially serve as an alternative tool for detecting FWB in fields. As such, we hope that the developed spectral method will open up new possibilities for the on-site diagnosis of FWB.


Introduction
Fusarium wilt of banana (FWB), which is commonly known as Panama disease and is caused by the soil-borne fungal pathogen Fusarium oxysporum f. sp. cubense (Foc), is a highly lethal vascular fungal disease in banana plants, and is thus reported to be one of the major limiting factors for banana production worldwide [1]. applications [26,27]. SERS, with its high sensitivity, little-to-no sample processing, and rapid detection times, was shown to be a potentially useful tool in clinical medicine for the earlier diagnosis of neurological, cardiovascular, and viral diseases, as well as cancers [28]. To the best of our knowledge, however, there have been no reports regarding the use of Raman spectroscopy or SERS for the detection of FWB. The purpose of this study, therefore, was to construct specialized Raman spectroscopic fingerprints for further detection of FWB based on SERS. In the study, we successfully determined Raman fingerprints related to the macroconidia, microconidia, and mycelia of Foc. We also used a field-detection assay to evaluate the feasibility of the SERS method. Here, we present the results of the first prospective vibrational spectroscopic study in which FWB could be detected without DNA extraction by using the SERS method.

Pathogens and growth conditions
Foc isolates and other fungal pathogens collected from banana including Colletotrichum musae (Colm, which causes Anthracnose of banana), Deightoniella torulosa (Dt, which causes Deightoniella leaf spot), Alternaria alternata (Aa, which causes Alternaria speckle of banana), Botryosphaeria dothidea (Bd, which causes Crown rot of banana), and Cordana musae (CorM, which causes Cordana leaf spot of banana) were used in this study. A single spore culture of each tested Foc and fungal isolate was grown on a potato dextrose agar (PDA) plate (200 g/l of potato extracts, 1% glucose, and 2% agar).

Sample preparation for SERS measurements
Fresh fungal mycelia (100 mg) and banana pseudostems (1 g) were homogenized using a taco™Prep Bead Beater (GeneReach Biotechnology Corp., Taichung, Taiwan) according to the manufacturer's instructions. The number of conidia was counted with a glass haemacytometer under a microscope (Carl Zeiss, Axioskop 2 plus, Germany). Aliquots of the homogenized fungal mycelia, banana pseudostems, and quantified conidia were loaded onto a Raman SERSchip (Labguide Co., Ltd., Taipei, Taiwan) illuminated with laser light and detected by a QE65 Pro spectrometer (Ocean optics, Inc., Dunedin, USA) for SERS measurements.

SERS measurements
A portable QE65 Pro Raman Spectrometer System (Ocean optics, Inc., Dunedin, USA) equipped with a 785-nm near-infrared diode laser source was used in this study. The samples of fungal mycelia, fungal spores, and banana pseudostems were detected with the aforementioned QE65 Pro Raman spectrometer at a laser power of 20 mW, and the integration time for the three types of samples were 15, 10, and 3 s, respectively. The strong [29] and reproducible [30] Raman peaks among the measured SERS spectra were selected as the fingerprinting/characteristics of the Raman spectra.

Sampling criteria of symptomatic banana samples for SERS measurements and molecular detections
Foc-infected banana pseudostems with varying symptoms were used for SERS measurement and molecular detection assays. The sampling criteria used for these infected banana pseudostems were previously described in detail elsewhere [1]. Specifically, we collected a total of 73 banana pseudostem samples (among them, there were 21 asymptomatic pseudostems and 52 symptomatic pseudostems, including 22 mildly symptomatic, 11 moderately symptomatic, and 19 severely symptomatic pseudostems) and 15 samples of banana infected with fungal pathogens other than Foc (3 samples for each disease) from 12 different fields that had been strongly affected by FWB. Necrosis covering less than 1/3 of the total area of a pseudostem, less than 2/3 but equal to or more than 1/3 of the total area, and equal to or more than 2/3 of the total area were classified as mild, moderate, and severe symptoms, respectively. The fieldinfected banana pseudostems showing varying symptoms were washed, surface-sterilized with 1% sodium hypochlorite (NaHClO), rinsed in sterile water, and dried under a laminar flow hood to eliminate epiphytic microbes. The surface-sterilized banana pseudostems were then cut into 1 cm 2 sections and put onto a Nash PCNB agar medium (1.5% peptone, 2% agar, 0.1% KH 2 PO 4 , 0.05% MgSO 4 . 7H 2 O, 0.1% pentachloronitrobenzene, 0.03% streptomycin, and 0.1% neomycin) for a plate-out assay [31]. Simultaneously, a piece of the banana pseudostems surrounding each section was used for further SERS measurements and molecular detections. In addition, the 21 asymptomatic pseudostems were also incubated at 28˚C for 2 months in a growth chamber to make sure the samples were Foc-infected.

Molecular detections of FWB
Three previously published molecular detection methods, namely MDIP (molecular detection of isolated pathogen) [5], IPDP (in-planta detection with PCR) [4], and IPDQP (in-planta detection with real-time PCR) [8], were also performed in this study in order to compare their results with the SERS measurements. The pathogen isolation and DNA extraction for the MDIP were previously described in detail elsewhere [5]. Specifically, the surface-sterilized field-infected banana pseudostems showing varying symptoms were used for the plate-out assay. After the plate-out assay, the pathogens grown on the Nash-PCNB agar medium were used for DNA extraction as previously described in detail elsewhere [5]. The DNA samples (50 ng) of the pathogens were then used for further PCR identification (specifically, using MDIP) according to the procedures described in Lin et al. [5]. In addition, a piece of each of the banana pseudostems (0.3 g) showing varying symptoms was directly used for DNA extraction and further PCR (specifically, using IPDP) or real-time PCR (specifically, using IPDQP) assays. The further procedures of PCR for IPDP and real-time PCR for IPDQP were performed as previously described in [5] and [8], respectively.

Results and discussion
To evaluate the potential of the Raman spectroscopy assay for the rapid differentiation of banana pathogens, the SERS spectra databases of the fungal pathogens on banana in Taiwan were built, including those for Foc, Colm (which causes Anthracnose of banana), Dt (which causes Deightoniella leaf spot), Aa (which causes Alternaria speckle of banana), Bd (which causes Botryosphaeria crown rot of banana), and Corm (which causes Cordana leaf spot of banana).
The mean SERS spectra of the tested mycelia and conidia covering the spectral range of 400-1800 nm are presented in Fig 1 and Fig 2, respectively. The characteristic band assignments of the SERS spectra are listed in Table 1 (mycelia) and  Fig 2E) characteristic SERS peaks were observed in the mycelia, microconidia, and macroconidia of Foc, respectively.
It is notable that the 652 cm −1 peak, which corresponds to guanine [32], was only present in the mycelia of Foc (Table 1). The Raman spectra reflected from these characteristic SERS peaks may be potentially useful for the identification of Foc. Several characteristic peaks were observed from most of the tested pathogens. Specifically, most of the tested conidia featured peaks around 802, 896, and 1416 cm -1 (Table 2), while most of the tested mycelia also featured SERS peaks around 733 (except for Bd), 896 (except for Corm), and 1332 (except for Dt) ( Table 1) cm -1 . The peaks around 802 and 896 cm −1 may have been from C-C ring breathing [32,33] and CH 2 groups of fatty acid chain [34], respectively. Meanwhile, the major assignments of the peaks around 733, 1332, and 1416 cm −1 were to saccharides (trehalose, dextrose, and chitin, respectively). The vibration mode and major assignment of the peaks around 1590 cm −1 , however, were not yet determined.
Comparisons of the SERS patterns showed high similarity between the conidia of Foc ( Fig  2D & 2E) and Colm ( Fig 2C). This was especially true for the peaks around 802, 896, 1003, 1113, 1255, 1289, 1346, 1416, and 1639 cm −1 , as they were observed in all of the conidia samples (that is, both macroconidia and microconidia) of Foc and Colm (Table 2). However, the SERS peak at 1189 cm -1 attributed to in-plane deformation vibrations of C-N was only observed in the macroconidia of Foc, while the peaks around 1576 cm -1 , which may correspond to nucleic acid [35,36], were only observed in the microconidia of Foc. The spectral region at 1003 and 1113 cm −1 can mainly be attributed to the C-C aromatic ring of phenylalanine and deformational vibration of tryptophan [32], respectively, and the bands at 1255 and 1289 cm −1 may be attributable to the amide III of collagen [30] and C-C stretching of unbranched saturated fatty acids [34], respectively. The peaks of 1346 and 1639 cm −1 were possibly related to the CH 2 of tryptophan [32] and C-C stretching ring or in-plane C-H bending, respectively [37]. In addition, we tested Bd (YJC-F004) and Dt (PM-YJL-F119) on potato dextrose agar, yeast peptone dextrose agar, and V8 agar. However, there was no conidia formation found in these conditions. Furthermore, the Raman spectroscopic fingerprinting databases for conidia of Bd and Dt are not currently available due to the technical difficulties.
A reliable method for the early monitoring of banana health is essential for formulating appropriate and timely disease management strategies to counteract the diseases affecting bananas. Traditional laboratory methods for diagnosing FWB involve lengthy assays including pathogen isolation, culturing, and morphological observation, as well as pathogenicity testing, which can take a few days to months [8]. One solution to this problem has been the use of advanced molecular techniques such as PCR [5,6], multiplex PCR [7], real time-PCR [8,9] LAMP [10], RealAmp [11,12], and iiPCR [1] assays, all of which provide a more rapid diagnosis of FWB. However, use of these approaches for on-site diagnosis of FWB has been limited by their requirement of appropriate DNA extraction protocols, because it is sometimes difficult to detect Foc in infected banana pseudostem samples when the DNA quality from the tested samples is poor. In this study, therefore, we constructed the Raman spectroscopic fingerprints of various symptomatic banana pseudostem samples infected by Foc (Fig 3). The characteristic peaks of pseudostems with varying levels of symptoms were located at 447 cm −1 (for Foc-free pseudostems, Fig 3A), 447, 802, 896, 1044, and 1400 cm −1 (for asymptomatic pseudostems, Fig 3B); 802, 896, 1044, 1113, and 1400 cm −1 (for mildly symptomatic pseudostems, Fig 3C); 447, 614, 896, 1044 and 1400 cm −1 (for moderately symptomatic pseudostems, Fig 3D); and 447, 614, 802, 830, 896, 1039, 1044, 1163, 1212, 1332, 1400, and 1572 cm −1 (for severely symptomatic pseudostems, Fig 3E), respectively. In contrast to the characteristics of

PLOS ONE
the samples with more severe symptoms, the Raman spectral features of asymptomatic and mildly symptomatic pseudostems were similar (but differentiable with those of Foc-free pseudostems). These data indicated that there seemed to be only slight changes in the overall molecular composition detected by Raman spectroscopy during the development of symptoms in pseudostems as they went from healthy to mildly symptomatic. In addition, the prominent SERS peaks at 896, 1044, and 1400 cm -1 , which mainly come from fatty acid chains [34] and proline [38], could be consistently observed in both the asymptomatic and symptomatic pseudostems. There were more and a greater variation in the Raman spectra of the severely symptomatic pseudostems than in the other pseudostems, indicating that there seemed to be great change in the overall molecular composition detected by Raman spectroscopy when the pseudostems were severely infected by Foc. Moreover, the Raman peaks (e.g. at 447, 830, 1044,

PLOS ONE
1212, 1400, and 1572 cm -1 ) of the field banana samples did not always perfectly match the Raman peaks of the Foc mycelia, microconidia, and macroconidia because it is easy for Raman peaks (300-1800 cm −1 ) to be interfered with by the Raman signals from endogenous biomolecules [39]. Before this study, Raman spectroscopy had not been used for the diagnosis of banana diseases. In this study, the most reproducible characteristic Raman spectra (asymptomatic: 802, 896, and 1044 cm −1 ; mildly symptomatic: 896, 1044, and 1113 cm −1 ; moderately symptomatic: 614, 896, and 1044 cm −1 ; severely symptomatic: 830, 896, 1039, 1400, and 1572 cm −1 ; and 447 cm −1 as healthy control) were collected to differentiate field-infected bananas with varying levels of symptoms. A field detection evaluation was further performed to determine whether the spectroscopic method based on the reproducible characteristic Raman spectra was suitable for the diagnosis of FWB. For this purpose, we collected a total of 73 banana pseudostem samples (among them, there were 21 asymptomatic pseudostems and 52 symptomatic pseudostems, including 22 mildly symptomatic, 11 moderately symptomatic, and 19 severely symptomatic pseudostems) and 15 samples of banana infected with fungal pathogens other than Foc (3 samples for each disease) from 12 different fields that had been strongly affected by FWB for the Raman spectroscopic fingerprinting analysis and the three molecular detections. We chose three published methods for comparison, namely PCR-based pathogen-identification (MDIP), Table 2. Raman shift and putative peak assignments of the SERS spectra from conidia of the common causative pathogens used in this study.

PLOS ONE
PCR-based in-planta detection (IPDP), and qPCR-based in-planta detection (IPDQP). As shown in Table 3, the detection rates of the Raman spectroscopic fingerprinting for the diagnosis of FWB in the asymptomatic, mildly symptomatic, moderately symptomatic, and severely symptomatic pseudostems were 16 0%) respectively, when the 21 asymptomatic pseudostems (each of the samples yielded at least two positive detection results within the four methods) were used as test samples. The detection rate for the asymptomatic samples when using the spectral method was 76.2%, which was comparable to that of the published molecular detection using real-time PCR (81.0%, IPDQP). In addition, after a 2-month incubation period, we were able recover Foc from those 21 asymptomatic pseudostems that progressed to being symptomatic by using the plate-out assay. It indicated that the field-detection results of this study were supported by those of traditional plate-out assays. The Raman spectroscopic fingerprinting assay is a simple and rapid method (requiring no DNA extraction procedures) for differentiating conidia or mycelia of Foc from those of other fungal pathogens collected from bananas, such as Colm, Dt, Aa, Bd, and Corm, as well as for the diagnosis of FWB on pseudostem samples with varying levels of symptoms. An acceptable field-detection rate for FWB detection by the spectroscopic fingerprinting assay was observed  Mild symptoms = less than 1/3 area of pseudostem necrosis; moderate symptoms = less than 2/3 but equal to or more than 1/3 area of pseudostem necrosis; severe symptoms = equal to or more than 2/3 area of pseudostem necrosis b SERS: surface-enhanced Raman spectroscopy; MDIP: molecular detection of isolated pathogen; IPDP: in-planta detection with PCR; IPDQP: in-planta detection with

PLOS ONE
among the results generated from the 73 random field-collected Foc-infected pseudostems with varying levels of symptoms, including asymptomatic pseudostems. In conclusion, to our knowledge, this is the first study to report the Raman spectroscopic fingerprinting databases of phytopathogenic fungi. The Raman spectroscopic fingerprinting assays developed in this study provide an alternative to conventional PCR and real-time PCR assays for the field-detection of FWB. The spectroscopic fingerprinting assays have the potential to serve as a rapid and simple tool for the routine diagnosis of FWB.