High-throughput analysis of anthocyanins in horticultural crops using probe electrospray ionization tandem mass spectrometry (PESI/MS/MS)

Abstract Plant secondary metabolites exhibit various horticultural traits. Simple and rapid analysis methods for evaluating these metabolites are in demand in breeding and consumer markets dealing with horticultural crops. We applied probe electrospray ionization (PESI) to evaluate secondary metabolite levels in horticultural crops. PESI does not require pre-treatment and separation of samples, which makes it suitable for high-throughput analysis. In this study, we targeted anthocyanins, one of the primary pigments in horticultural crops. Eighty-one anthocyanins were detected in approximately 3 minutes in the selected reaction-monitoring mode. Tandem mass spectrometry (MS/MS) could adequately distinguish between the fragments of anthocyanins and flavonols. Probe sampling, an intuitive method of sticking a probe directly to the sample, could detect anthocyanins qualitatively on a micro-area scale, such as achenes and receptacles in strawberry fruit. Our results suggest that PESI/MS/MS can be a powerful tool to characterize the profile of anthocyanins and compare their content among cultivars.


Introduction
Horticultural crops contain various secondary metabolites representing unique traits such as color, aroma, taste, and functionality [1][2][3]. Anthocyanins are important metabolites that ref lect unique colors [2]. The chemical structures of anthocyanins are glycosides of anthocyanidins, and six major basic structures are observed [4]. These compounds show various colors, except green, based on substituents at the B-ring, local pH, or both [2]. Anthocyanins act as part of the stress defense function of crops, and also contribute to human health by protecting against cardiovascular diseases and cancer [2,3,5,6]. It is important to understand anthocyanin profiles rapidly and spatially to enhance anthocyanin accumulation while breeding horticultural crops or anthocyanin intake while selecting horticultural products for consumption.
High-performance liquid chromatography (HPLC) and mass spectrometry (MS) were the main techniques used in previous anthocyanin analyses [4,7,8]. These are highly accurate but require many complicated steps and considerable time for preliminary treatment of plant samples and separation of anthocyanins. Ambient MS approach is one of simple analytical methods to solve these problems. Ambient MS create and analyze ions in various samples from plants and humans under ambient conditions [9]. For example, ambient MS techniques, such as desorption electrospray ionization (DESI) or direct analysis in real time (DART), were implemented for high-throughput analysis in plant and horticultural research [10][11][12]. Low-temperature plasma, electrospray laser desorption ionization, and dielectric barrier discharge ionization are also utilized in plant sciences and are suitable for imaging or single-cell analysis at the two-or three-dimensional stage [13][14][15].
Probe electrospray ionization tandem mass spectrometry (PESI/MS/MS), used in this study, is an ambient MS approach. PESI was developed by Hiraoka et al. [16,17]. The ionization mechanism of PESI involves electrospray ionization (ESI) using a probe needle with a tip diameter of approximately 700 nm. A high voltage is applied to the probe, and the adherent compounds at the probe tip become ionizable. A small quantity of compounds is sufficient as the technique prevents loss during pre-treatment step [18]. As ionization is manageable even in a high-salt environment [19], it can also be applied to plant samples. Probes are available in various shapes and materials [17,20,21], and the probe used in this study is based on the design of an acupuncture needle and has reasonable portability and moderate robustness. Additionally, tandem mass spectrometry (MS/MS) improved the selectivity of the target compounds in the presence of matrix [18]. PESI/MS/MS has been used in biomedical engineering and forensic medicine research [22][23][24][25]. This approach may be applicable to horticulture crops, having complicated shapes and metabolites with similar structures such as anthocyanins.
In this study, we considered the application of PESI/MS/MS to horticultural crops, for the first time, and detected 81 anthocyanins associated with the structures of aglycones and glycosides. These anthocyanins have been assessed widely in horticultural fruits, vegetables, and f lowers with little difference compared to using previous instruments like HPLC-based MS [4,7,26]. Thus, PESI/MS/MS could be used for high-throughput screening of anthocyanins in 16 species or cultivars and to develop a straightforward method for detecting molecular species qualitatively in micro-area organs.

Results and discussion
Detection of anthocyanins and discrimination from flavonols using PESI/MS/MS First, we considered the conditions for measuring anthocyanins using PESI/MS/MS. Specifically, we evaluated and determined the cleavage conditions of the targeted compounds using chemical standards in product ion scan mode. In this study, we targeted major anthocyanins based on six basic structures, i.e. cyanidin, delphinidin, malvidin, pelargonidin, peonidin, and petunidin [2]. Each precursor ion was searched against ESI-QQ spectral data from the MassBank spectral database [27].
Anthocyanin has a positive charge in solution ([M]+). Therefore, we measured all samples in positive ion mode during this study. Generally, f lavonol can be ionized efficiently in [M-H] − state [8]. However, PESI/MS/MS also detect f lavonols as [M + H] + with the same m/z sizes to anthocyanins, because of the lack of separators, such as those used in LC-MS. We challenged the limited cleavage of fragments by increasing the collision energy in the range of −50 to −70 V for distinguishing between anthocyanins and f lavonols. Specific product ions were identified in anthocyanins and f lavonols respectively, e.g. m/z 125 only in delphinidin 3hexoside (Del-3-Hex) and m/z 137 only in quercetin 3-hexoside (Que-3-Hex) (yellow arrows in Supplementary Fig. S1). In a few cases, the ion intensity of specific product ions was lower than that of aglycon ions, e.g. m/z 125 compared to 303 in Del-3-Hex ( Supplementary Fig. S1). However, we prioritized specificity in distinguishing between major anthocyanins and f lavonols.
Similarly, we also compared four other anthocyanins (cyanidin, pelargonidin, peonidin, and petunidin) and coordinated f lavonols (data not shown). A similar pattern of cleavage was observed in DESI-MS [12], where pelargonidin 3-glucoside was cleaved into tiny pieces while passing it through MS 3 . Malvidin did not correspond to any f lavonol in the MassBank database search results.
As another limitation, we could not distinguish several sugar structures using this method, e.g. the sugar structures of glucoside and galactoside. In this study, we made the notation for four main glycosides, i.e. hexoside (Hex), pentoside (Pen), rutinoside (Rut), and di-hexoside. Hexoside was selected as the main component of glycoside [4]. Rutinoside was selected because it is observed often in tree fruits and berries [28,29]. The other glycosides were based on previous studies using HPLC-ESI-MS/MS [4,7,[30][31][32][33][34][35][36][37]. The specific product ions in previous paragraphs were assigned to anthocyanins with four main glycosides. In contrast, other large glycosides were designated aglycone-sized ions based on detection sensitivity. In total, we could discriminate the five types of major anthocyanins from coordinated f lavonols, and determined the selected reaction monitoring (SRM) transition for 81 anthocyanins and 5 f lavonols (Table 1).
Next, we checked for the suitability of the SRM condition for plant samples (blueberry, apple, and onion). Specifically, we compared the intensities of product ions equivalent to Del-3-Hex or Que-3-Hex as the ion to be checked among differently colored crops. We used crude extracts to boost the intensity of the positive ions using a water/ethanol solvent, including formic acid. The skins of blueberries, apples, and onion bulbs were pared using a razor, dipped in the extraction solvent, and vortexed for 30 s. In this method of detecting anthocyanins in plants, the time and labor consumed in the pre-treatment steps are saved, unlike previous approaches, such as HPLC, which involve complicated purification steps.
When comparing the profiles of product ions in this sample, the trend was found to be similar to that observed with the chemical standards. The product ion equivalent to Del-3-Hex at −60 V (m/z 125) was stably detected in blueberries (purple bar in Fig. 1). This is consistent with the results of previous studies, which the main anthocyanins in blueberry are delphinidin and malvidin [4,30]. In contrast, the product ion equivalent to Del-3-Hex could hardly be detected in either green or red apple skins. A product ion comparable to Que-3-Hex (m/z 137) was detected stably in both the skins (blue bars in Fig. 1). In a previous study of apple [38], cyanidin was detected as the major anthocyanin in red skin, and delphinidin was not detected. Quercetin glycosides were detected in both green and red skins. Similar to apple, the product ion of Del-3-Hex was barely detected in onion, whereas that of Que-3-Hex was stably detected in both cultivars (Fig. 1). Regardless of the species and skin color, the product ion (m/z 303) was stably detected (pink bars in Fig. 1). Thus, we determined that the detected ion (m/z 303 at −20 V) in apple and onion was not mainly Del-3-Hex but Que-3-Hex. The two compounds could be distinguished by increasing the collision energy (m/z 125 at −60 V and m/z 137 at −50 V).

Profiles of anthocyanins in horticultural crops using PESI/MS/MS
We utilized the SRM transition for 81 anthocyanins measurable for a total of 3 min (Table 1), and characterized the anthocyanin content profiles from the total area values calculated using absolute intensities. Crude extracts from edible parts in 16 species were used in the same manner as described in the previous section.
In principal component analysis (PCA), variance in PC1 score (37.9%) ref lected the difference between red lettuce and red onion against other crops (Fig. 2). Cyanidins added to malonyl-hexose are the main contributors to this difference. This trend is consistent with previous reports on LC-based analysis [36,37]. Variance in PC2 score (26.6%) ref lected the difference between fruits (especially blueberries and grapes) and strawberries. These profiles were similar to those obtained by HPLC-ESI-MS/MS analysis [4,7]. Most anthocyanins could not be detected in cultivars without a red color, including those of grape, apple, cabbage, lettuce, and onion (Supplementary Table S1). Thus, these cultivars did not vary in origin in PCA.
Most of the set product ions could be detected stably in cultivars with a red color during measurement time, which might be because the sample solution was stably supplied by repeatedly stirring the probe to a sufficient depth. Additionally, the SRM method was designed such that the scan rate per compound (1-2 ms) was faster than the ionization timing difference (150 ms), even when measuring compounds had different solvophobicities. This method provides exhaustive spectral intensities averaged for each anthocyanin over the entire analysis time. However, when detecting compounds with different polarities simultaneously, it is necessary to determine whether the ionization efficiencies of compounds are sequential and exhaustive [19].

Figure 1. Detection of delphinidin 3-hexoside (Del-3-Hex) and quercetin 3-hexoside (Que-3-Hex) product ions in horticultural fruits and vegetables
using PESI/MS/MS. Blueberry skin, apple skin, and onion bulb were pared using a razor and dipped in the extraction solvent. Each extract supernatant (10 μL) was placed on a plate for liquid and ionized using the PESI unit of DPiMS-8045. The spectral intensities of each product ion equivalent to anthocyanin or f lavonol fragments or both were measured in selected reaction monitoring (SRM) transition mode, as described in Table 1  PCA score and loading plots for anthocyanins in horticultural crops using PESI/MS/MS. Blueberry, grape, apple, cherry, strawberry, and radish skins; cabbage and lettuce leaves; and onion bulbs were pared and dipped in the extraction solvent. Grape, apple, cherry, cabbage, and onion cultivars had different colors. Lettuce has two colors in one leaf. Two strawberry cultivars were analyzed, i.e. "Yumenoka" (Yume) and "Sagahonoka" (Saga). Each extract supernatant (10 μL) was placed on a plate for liquid and ionized using the PESI unit of DPiMS-8045. The intensities of the product ions and the total area values were calculated as described in Table 1 and Fig. 1. Score scatterplots show the PCA model of anthocyanins in the edible parts of horticultural fruits (circles) and vegetables (triangles) based on Pareto distribution (n = 3-4). Loading scatterplots were drawn from PCA analysis (rhombus, n = 80, anthocyanin peaks). Additionally, various anthocyanins, not reported in earlier studies, could be detected in many cultivars with a red color using the PESI/MS/MS method (Supplementary Table S1). We could not identify these compounds as precise anthocyanins in this study because of a lack of chemical standards. However, the PESI/MS/MS method may be effective as the first easy and rapid screening method for anthocyanin detection.

A simpler assay for detecting anthocyanins using PESI/MS/MS; probe sampling
We used a more intuitive method of sticking a probe directly to the sample without pre-extraction; thus, we named this method "probe sampling". We focused on strawberries as a measurement model. There are achenes and receptacles in strawberry fruit, and we visually distinguished the colors of these organs (Fig. 3). These shapes cannot be sampled using other ambient techniques, such as DART, which tests in an extensively open system environment. DESI can also perform fine sampling; however it requires imprinting on the TLC plate and is currently unsuitable for hard and small tissues such as achenes [12].
The degree of sample sticking to the probe tip was light only once, and the probe was placed immediately in the PESI unit. The extraction solvent was placed on a plate for liquid, and the adherent sample compounds at the tip were diluted and homogenized at the same time as ionization. The invasion depth into the solvent and the number of probe movements for repetitive samplings were the same as those in the crude extraction.
We used the same SRM transition in the crude extract samples, and 31 anthocyanins were detected in either strawberry organ. Of these, 29 overlapped with the results obtained using the crude strawberry extracts mentioned above. The signal intensities in the first segment were occasionally strong compared to those of the crude extraction, which then stabilized during the measurement period (data not shown), likely because the attached sample was initially over-ionized and then optimized upon dilution with the solution via vertical movement of the probe. Although these signal patterns may not be suitable for quantification, they are sufficiently stable for qualitative analysis.
We focused on the above-mentioned anthocyanins and compared the total area of the organs (Fig. 5). Cya-3-6"-Mn-Hex specifically accumulated in achenes compared to that in the receptacles. Cya-3-Hex and Pet-3-Pen mainly accumulated in the red-colored achene and on the receptacle surface. Contrastingly, Peo-3-Hex accumulated only on the surface of the receptacle. Pel-3-Hex, Pel-3-Rut, and Pel-3-6"-Ac-Hex were mainly present on the surface of the receptacle. These trends are generally consistent with those of earlier studies using MALDI-imaging MS and UPLC-qTOF-MS/MS, where Cya-3-Hex and Cya-3-6"-Mn-Hex were detected in achenes, and Cya-3-Hex, Pel-3-Hex, and Pel-3-Rut in the receptacle in "Tochiotome", "Seascape" and "Herut" [39][40][41]. The intensity of Pel-3-6"-Mn-Hex in achene and receptacle was not consistent in "Herut" [41], and accurate quantitative comparisons between different organs in the probe sampling are currently not feasible because each cell state is different. Additionally, in a few cases, the simplistic handling of probe sampling may be a factor in data variability (Fig. 5). Thus, this method can be regarded as a qualitative analytical method. . PCA score and loading plots for anthocyanins in strawberry organs using probe sampling PESI/MS/MS. Achene (Ach; triangle), surface (ReSur; circle), and inside (ReIn; square) of the receptacle in strawberry cultivars, i.e. "Yumenoka" (pink; Yu) and "Sagahonoka" (purple; Sa), in red (R) or non-colored (Y or W) areas were measured using the probes shown in Fig. 3. The pure extraction solvent was set on a plate for liquid, as shown in Fig. 1, and the probe was ionized using the PESI unit of DPiMS-8045. The intensities of the product ions and the total area values were calculated as described in Table 1 and Fig. 1. Score scatterplots show the PCA model of anthocyanins in strawberry organs based on Pareto distribution (n = 4-8). Loading scatterplots were drawn from PCA analysis (rhombus, n = 31, anthocyanin peaks).
Based on the difference of organs, we considered only the cultivar-specific differences within the same organ category for the two cultivars, i.e. "Yumenoka" and "Sagahonoka". The inside of the receptacle in "Yumenoka" was widely reddish, and that of the "Sagahonoka" receptacle was reddish only at the edge of the fruit (Figs. 3C and 3D). Pelargonidins were detected mainly on the surface of the receptacle, with the total area of Pel-3-6"-Ac-Hex in "Yumenoka" being significantly higher than that in "Sagahonoka" (Fig. 5). These trends were similar, even the inside of receptacle obtained by probe sampling and the extract of the pre-treated sample.
Several methods similar to probe sampling have been reported, including dipping PESI/MS (dPESI/MS) [42], remote sampling in sheath-f low PESI (sfPESI) and adjustable sfPESI (ad-sfPESI) [17,43]. dPESI/MS is a different control setting for probe ionization than in probe sampling. Compared to the shallow invasion depth (<0.5-1 mm) and several shots in dPESI/MS, PESI/MS/MS was set at a depth of 2 mm which was deeper, and more repetitive samplings were conducted. This condition stabilized the signal profile and resulted in more continuous ionization and detection in the SRM mode. In sfPESI and ad-sfPESI, the structure of the probes was more precise, but more complicated than that in PESI/MS/MS. Additionally, water-soluble compounds could not be detected using sfPESI in the remote sampling of apple skin. The m/z intensities corresponding to the anthocyanin fragments could not be observed. However, in probe sampling of strawberry skin (receptacle and achene) by PESI/MS/MS, hydrophilic anthocyanins were stably detected even with one probe stick and pure solvent alone. A general comparison was not feasible because the measurement conditions and samples differed, but the difference in solvents, i.e. water/methanol in sfPESI and water/ethanol in PESI/MS/MS, might have contributed to the ionization efficiencies at plant surfaces. Additionally, regarding the detectors, previous ambient approaches mainly used single quadrupole or Orbitrap mass spectrometer [11,12,19,43], and MS/MS spectra have not yet been exhibited. In other instruments [39][40][41], sample pre-treatment, such as matrix application in MALDI-imaging, was complicated and qualitative. LC-MS/MS requires large amount of samples for separation using HPLC. In that respect, PESI/MS/MS could be effective as the first screening to detect anthocyanins in horticulture crops at the micro-area scale, with the combination of simplified probe sampling and avoiding the use of overly large scale MS instruments.
As we confirmed that partially independent metabolic pathways are locally regulated in the fruit using probe sampling (Figs. 3 and 4), it is possible to clarify temporal and spatial metabolic changes that are difficult with conventional analysis of the whole fruit or complicated analysis operations. In strawberries, the differences observed between achene and receptacle would provide insights to accelerate the elucidation of the physiological mechanism and breeding of new cultivars with independent colors in these organs [44]. A combinatorial approach using microscopy and probes may help elucidate mechanisms underlying the differences in metabolite accumulation at the cellular level [15].
In conclusion, we established a high-throughput analysis method for measuring 81 anthocyanins in approximately 3 min using PESI/MS/MS. The number of measurable molecular species was increased by devising the method that enables specific and long-term measurements. In addition to extracting solutions, probe sampling can qualitatively detect anthocyanins at a microscale. Proper implementation of these methods in research using plant materials is crucial. Measurable compounds other than anthocyanins and measurable plant species, including crops and ornamental f lowers, need to be amplified for future research. For research on the metabolism of animals, such as mice, pipelines were constructed using PESI/MS/MS for high-throughput metabolome and statistical analysis [22,24]. PESI/MS/MS can be a powerful tool for breed selection and quality assessment in the field of horticultural science, which would help detect the indicators of physiological responses in cultivation field, such as stress response markers. The portable probes allow the sample to be remote from the front of the MS instrument. Figure 5. Comparisons of anthocyanins in strawberry organs using PESI/MS/MS. Strawberry organs of cultivars "Yumenoka" (Yume) and "Sagahonoka" (Saga) were collected using the probes described in Fig. 3. Each data point was extracted from Fig. 4 (achene, the inside, and surface of the receptacle) and Fig. 2 (extract, pre-treated sample). "n.d." indicates "not detected" (i.e. total area value was zero). Asterisks indicate significant differences between the two red cultivars in the same organ (Wilcoxon test, p < 0.05). "n.s." indicates "not significant". Vertical bars indicate standard errors (n = 4-8).
Consequently, it would be possible to understand crop characteristics of several remotely located fields simultaneously if the transportation conditions of the sampled probes are established.

Plant materials
We purchased horticultural fruits and vegetables from local markets in Nagoya City, including blueberry (Vaccinium corymbosum), grape (Vitis vinifera, green and red skin), apple (Malus × domestica, green and red skin), cherry (Prunus avium, red and black skin), strawberry (Fragaria × ananassa, "Yumenoka" and "Sagahonoka"), cabbage (Brassica oleracea var. capitata, green and red leaves), lettuce (Lactuca sativa var. crispa, green and red leaves within a plant), onion (Allium cepa, yellow and red bulbs), and radish (Raphanus sativus var. sativus). The edible parts (skin, leaf, and bulb) were pared using a razor and translated into micro-tubes. After measuring the fresh weights of these pieces, tubes were immediately frozen in liquid nitrogen and stored at −80 • C until use.

Analysis conditions using PESI/MS/MS
A DPiMS-8045 (LCMS-8045 tandem mass spectrometer with probe electrospray ion source; Shimadzu Corporation, Kyoto, Japan) and PESI MS solution (ver 2.0.8) were used for analysis. Frozen sample pieces, except cabbages, were dipped in a volume of the extraction solvent (0.1% formic acid and 49.95% ethanol (v/v)) ten times their weight, and vortexed for 30 s. The cabbage pieces were dipped a volume of the extraction solvent two times their weight. The supernatant (10 μL) was placed on a plate for liquid (Shimadzu Corporation), and ionized using a PESI unit of DPiMS-8045. We used a PESI probe with a tip diameter of approximately 700 nm and a body diameter of nearly 0.14 mm (Shimadzu Corporation). For probe sampling, the tip of the probe was attached to the organs of ripe raw strawberry fruits at room temperature (approximately 20 • C) for one second. Subsequently, 10 μL of pure extraction solvent was placed on a plate for liquid and ionized.
The product ion scan modes were set as follows: scan range from 100 to precursor ion +10 (m/z), collision energy to the maximum from −10 to −70 V, scan speed to 5000 u/s, and measurement time to 60 s. The details of SRM mode for each analysis, such as the SRM transition and collision energy, are listed in Table 1. The dwell and pause times for SRM transition were 1 ms each. We set up ten parallel segments in the SRM transition mode, with each measurement time set to 0.3 min. There were 92 total measuring segments, i.e. the measurement time per sample was 3 min. The frequency of the PESI probe was 11.47 Hz, and the ionization and sampling times were 150 and 30 ms, respectively. The probe was at an upper position during ionization, and at a lower position during sampling. Repetitive samplings were performed 60 times per 0.3 min. The invasion depth of the probe in the sample solution was approximately 2 mm. The applied voltages for the PESI positive modes were 2.5 kV.

Statistical analyses
The calculation software, LabSolutions (ver. 5.86, Shimadzu Corporation) was used to confirm each anthocyanin peak in the MS chromatogram and calculate the total area values. Principal component analysis (PCA) was performed using Statistics in Microsoft Excel [45]. Other statistical analyses were performed using JMP 13 (SAS Institute, Inc., Cary, NC, USA). The data represent mean values ± standard error of three to eight biological replicates.