The LC-MS grade methanol, ethanol, acetonitrile, formic acid, 0.1% (w/v) poly(L-lysine) hydrobromide solution, and gelatin were obtained from Sigma-Aldrich (Shanghai, China). Analytical grade ammonium acetate, chloral hydrate, phloroglucinol, iodine, sulfuric acid, and ethanol were obtained from Sinopharm Chemical Reagent Co., Ltd (Beijing, China). The optimum cutting temperature (OCT) compound was obtained from Leica (Nussloch, Germany). Water was purified by employing a Milli-Q filtration system (Millipore, Bedford, USA). Information for the reference standards could be found in Table 1. Sixty-four batches of Isatidis Radix were purchased from herb markets in the provinces of Hebei, Gansu, Sichuan, and Heilongjiang, China. The origin of each sample was authenticated as the dried root of Isatis tinctoria L. (Brassicaceae) by Associate Professor Shuai Kang in accordance with the Chinese Pharmacopeia, edition 2020. For future reference, the voucher specimens (S1–S64) were deposited in NIFDC, Beijing, China.

To investigate the quality characters of Isatidis Radix, macroscopic and microscopic features of the herb were observed. For macroscopic characterization, features including the texture, the color of surface, the depth of longitudinal wrinkle, and the existence of roothead were described and compared. For each batch of sample, the lengths of the longest, medium, and shortest individuals, and the diameters at the middle of the thickest, medium, and thinnest individuals were measured and recorded. Then, the root with the medium size of each batch was cut from the middle, and the character of the transverse section was observed and photographed using a digital single-lens reflex (DSLR) camera (Canon 5D4, Tokyo, Japan). For microscopic characterization, a 1-cm section was cut axially from the middle of the root with the medium size of each batch, which was embedded in the OCT compound before freezing at −13°C. Next, the frozen root was sectioned into a 30-μm slice with the tape at −13°C by employing a cryomicrotome (Leica CM 1950, Nussloch, Germany). The microscopic features of the cross-section of Isatidis Radix were observed via a microscope (Olympus BX51, Tokyo, Japan) after permeabilizing with chloral hydrate solution. To examine the lignified cells and the starches, the section was stained by phloroglucinol solution and sulfuric acid, and iodine test solution, respectively. An automatic digital slide scanning imaging system (Zeizz Axiocan.z1, Jena, Germany) was used to collect the images. Finally, the quality characters for Isatidis Radix were proposed based on morphological investigation combined with literature research.

The dried root was cut from the middle to obtain a 0.5-cm-long piece, which was embedded with 0.1 g/ml gelatin solution and frozen at −20°C. For cryosectioning, the frozen sample was axially fixed on the sample holder of a cryomicrotome (Leica CM1950, Nussloch, Germany), using OCT compound as the adhesive. Cross-sections of 100 μm thickness were obtained at−18°C and adhered on a glass slide with the double-sided conductive tape (3M, St. Paul, MN, USA).

An Ultimate 3000 ultra-high performance liquid chromatography system coupled with Quadruple-Enactive Orbitrap (UHPLC-Q-Orbitrap) high-definition tandem mass spectrometer (Thermo Fisher Scientific, Waltham, USA) was adopted to identify the chemical constituents in extracts of Isatidis Radix. Xcalibur software (Thermo Fisher Scientific, Waltham, USA) was used for instrument control and parameter setting. The sample was pulverized and shifted through a 60-mesh sieve, and the powder was extracted with mixtures of methanol and water at different ratios. Chromatographic separation was performed on a Waters CORTECS UPLC T₃ column (2.1 mm × 100 mm, 1.6 μm) of 35°C. The following binary gradient consisting of methanol (A) and water (B), both with 10 mM ammonium acetate at a flow rate of 0.2 ml/min was applied as follows: 0 min 2% A, 40 min 40% A, 45 min 68% A, 55 min 95% A, 65 min 95% A, 70 min 2% A, and 80 min 2% A. The sample inject volume was 1 μl. The full-MS and data-dependent scan (dd-MS²) under both ESI⁺ and ESI⁻ modes was used for Q-Orbitrap MS/MS data acquisition with the mass spectrometer parameters as follows: The data acquisition range was m/z 50–1,500 Da. The spray voltage was 3.5 kV for ESI⁺ mode and 3.0 kV for ESI⁻ mode. The sheath gas and auxiliary gas were operated at flow rates of 35 arb and 10 arb, respectively. The temperature of the capillary was 350°C and that of the auxiliary gas heater was 350°C, and the S-lens RF level was 50.

Mass imaging of the tissue was carried out on a Q-TOF mass spectrometer equipped with the DESI source (Waters Xevo G2-XS, Milford, USA). The parameters were set as follows: mass range, m/z 50–1,200; ion source temperature, 150°C; capillary voltage, 4.0 kV; spray solvent, MeOH-H₂O (95:5, containing 0.1% formic acid); spray solvent speed, 2 μl/min; spray gas (N₂) pressure, 0.45 MPa; incident spray angle, 75°; collection angle, 10°; X and Y pixel sizes, 80 μm; and raster speed, 200 μm/s. A calibration solution of 0.2 mg/L leucine encephalin in methanol (Waters, Milford, MA, USA) was used as the lockmass for high-resolution mass spectra. To measure the positive and negative ions separately, two slices were prepared for each sample. To validate the identification results of metabolites in Isatidis Radix, all the reference standards were prepared at a concentration of 10 μg/ml with 70% (v/v) methanol. Then, 0.5 μl of each of the reference solutions was applied on a glass slide and dried in the air. Ultimately, the spots of the reference standards were scanned by the same parameters of the sample.

All images were reconstructed by linear smoothing and displayed in absolute intensity after total ion current (TIC) normalization. Screening of the ions was carried out by the aid of the database including more than 400 components in Isatidis Radix, which was constructed by the authors using UNIFI software (Waters, Milford, MA, USA) and other commercial databases. Further identification was based on the accurate mass to charge ratio with reference to the reference standards and/or the literature and databases and MS/MS information obtained by UHPLC-Q-Orbitrap MS analysis. The region of interest (ROI) was selected from the whole scanned sector. Each sample had 1 ROI that was converted into a data matrix of mass m/z and signal intensity. Then, orthogonal partial least squares discrimination analysis (OPLS-DA) was performed by EZinfo (Waters, Milford, MA, USA) for the differentiation of samples with different quality characters and for the discovery of the potential quality-associated markers.

Macroscopic examination on 64 batches of Isatidis Radix from different markets and habitats was performed. The results showed that the surfaces of major samples were gray, while those of the minor samples were black. Rootheads were observed in most samples except samples S22, S33, and S43. Regarding the lengths and diameters of Isatidis Radix, large spans were measured in different roots from the same batch. Likewise, the depth of longitudinal wrinkle varied without regularity. In contrast to the above features, the texture and characteristics of the transverse section of 64 samples could be categorized into 6 types named with quality character code 1–6 (see Table 2 and Figure 1). To take a closer look at the microscopic features of Isatidis Radix, axial slices of the samples were stained and observed. Normally, the lignified cell wall would be stained red in varying degrees depending on the extent of lignification. On the contrary, starch granules in the cell would be stained blue, and their distribution could be observed clearly. Just as expected, macroscopic and microscopic characteristics of Isatidis Radix were highly related. As shown in Table 2 and Figure 1, when the root was lax and the transverse section was cleft, the bundles of wood fibers and the starch granules were rarely seen under microscopy. Conversely, when the root was compact and the transverse section was oily or starchy, the bundles of wood fibers could be observed easily, and the starch granules were abundant. Nevertheless, exception was found in samples with a quality character code of 5. Although their textures were compact, the transverse sections were horny, and the bundles of wood fibers and the starch granules were scarce, which might be owing to the improper processing and storage of the herbs.

Isatidis Radix has a long history of clinical use and market sale, during which the empirical morphological indicators presenting the superior quality of the herb have been formed. As could be summarized from Table 3, “Isatidis Radix with compact texture and starchy or oily fracture possessing good quality” was the consensus of different classic monographs on herbalogy and pharmacognosy. On the contrary, in the classic monograph “Species Systematization and Quality Evaluation of Commonly Used Chinese Traditional Drugs (Northern Edition),” Isatidis Radix is recorded as follows: “The root must be harvested before flowering. If collected after seed bearing, the herb would have a dry surface and a starchy fracture, which is not suitable for medicinal use (Lou, 1995).” Based on the above analysis, samples possessing quality character codes 2, 4, and 6 were considered as superior products. On the contrary, those with quality character codes 1, 3, and 5 were considered as inferior products.

A comprehensive metabolite profile of Isatidis Radix was achieved by a high-resolution DESI-Q-TOF/MS. The representative overall average mass spectra under positive and negative ionization modes are shown in Figure 2. The positive ions were mostly detected in the range of m/z 100–400 and m/z 550–600, while the negative ions were observed in the range of m/z 50–800, but with much lower intensity. The putative assignment of the components was based on ions screening using the self-constructed Isatidis Radix UNIFI database and with reference to the isotopic peak, the reference standards, the literature, and other databases. Due to low ion abundances, MS/MS data were difficult to obtain by DESI MSI. Instead, UHPLC-Q-Orbitrap MS/MS analysis was conducted to provide stronger evidence for qualitative identification. As indicated in Supplementary Table 1, a great variety of metabolites including alkaloids, sulfur-containing compounds, phenylpropanoids, nucleosides, amino acids, organic acids, flavonoids, phenols, terpenes, saccharides, peptides, and sphingolipids were detected readily. The positive ions were mainly found as H⁺ adducts of all amino acids, majority of the alkaloids, parts of the phenylpropanoids, the nucleosides and the phenols, several sulfur-containing compounds, a few flavonoids and sphingolipids, as well as organic acids with basic groups. Besides, a small part of alkaloids, phenylpropanoids, nucleosides, phenols, terpenes, peptides, and sulfur-containing compounds were detected in the form of sodium or potassium adduct. At the same time, the negative ions were detected prominently as [M–H]⁻ ions and inconspicuously as [M-H+FA]⁻ ions, evolving the majority of the sulfur-containing compounds and the organic acids, some phenylpropanoids and phenols, a few saccharides, flavonoids, terpenes, as well as alkaloids with acid group. Several components held an isomeric relationship and could not be differentiated by their exact mass, which is listed together in Supplementary Table 1, such as (R, S)-goitrin and progoitrin/epiprogoitrin.

The optical image of the transverse section (Figure 3) showed the anatomical structure of Isatidis Radix: xylem, cambium, phloem, cortex, and cork from inside to outside. Thanks to the MSI technology, tissue distributions of different kinds of metabolites could be visualized in these structures. For example, Figure 3 represented the H⁺ adducts of 3-formyl-indole and indoxyl. Although they are both typical alkaloids from Isatidis Radix, a minor difference was observed in their distribution pattern. The former was rich in the cambium and the latter was more abundant in the outer area of phloem. Likewise, the accumulation sites of amino acids and organic acids were also diversified. The signals of [M+H]⁺ ions of histidine and lysine were more obvious in the xylem (Figure 3), while that of proline tended to disperse from cambium to both sides (Figure 3). At the same time, the protonated adduct of arginine, the most abundant amino acid in Isatidis Radix, was located with strong signal intensity in all regions except the cortex and cork (Figure 3). A similar localization pattern (Figure 3) was found for sucrose, another component with high concentration in the herb. Unlike amino acids, organic acids in Isatidis Radix were detected mostly in the form of [M-H]⁻ ions. The accumulation sites of malic acid and maleic acid bore some similarities, which were observed mainly in the xylem (Figure 3). Conversely, the high signal intensity of citric acid was detected in the outer areas of the tissue including phloem, cortex, and cork (Figure 3). Interestingly, the ion of linolenic acid was located in xylem, cortex, and cork, and in the innermost and outermost regions of the root (Figure 3).

The (R, S)-goitrin, the mixture of epigoitrin [(R)-goitrin] and goitrin [(S)-goitrin-] (Nie et al., 2016, 2017), was one of the most characteristic sulfur-containing compounds in Isatidis Radix and was acknowledged as the quality marker of this herb (Chinese Pharmacopoeia Commission, 2020). Epigoitrin and goitrin were verified as the breakdown products of their glucosinolate prototypes, epiprogoitrin and progoitrin, which also existed in Isatidis Radix (Nie et al., 2011). Epigoitrin, goitrin, epiprogoitrin, and progoitrin exhibited anti-influenza activity in vitro/ovo (Nie et al., 2020), and they were detected in the cross-section of Isatidis Radix by DESI MSI. As indicated in Figure 3, positive ion of (R, S)-goitrin and negative ion of epiprogoitrin/progoitrin all showed dominant abundance in the phloem. When investigated further, different preferential occurrences of (R, S)-goitrin and epiprogoitrin/progoitrin were observed in the cortex, cork, and the inner core of the xylem, respectively. Similar to epiprogoitrin/progoitrin, isatindigobisindoloside G and gluconapin, another two glucosinolates in Isatidis Radix, were mainly detected as the [M-H]⁻ ions that accumulated mostly in phloem (Figure 3). Conversely, the negative ion of clemastanin B, an impotent anti-influenza phenylpropanoid discovered from Isatidis Radix (Yang et al., 2013), presented the highest abundance in the cambium and around (Figure 3). Apart from the anti-viral effect, Isatidis Radix posed anti-inflammatory function as well. According to the literature, the major therapeutic reagents were nucleosides including guanine, adenine, and adenosine (Evaldsson et al., 2007; Clercq, 2019; Wu et al., 2019). As noticeable in Figure 3, the proton adducts of the above-mentioned nucleosides were located almost in the cambium and the outer area of the phloem. In particular, adenosine and its nucleobase, adenine, shared nearly the same tissue distribution.

To explore the potential of mass spectrometry imaging in quality illustration of herbal medicine, every 3 batches of representative Isatidis Radix possessing 6 categories of quality characters (indicated in Table 2) were analyzed by DESI-Q-TOF/MS. Although the metabolites' spatial distribution pattern bore some similarities in different categories of samples, their relative signal intensity in smaller tissues varied. The pattern recognition method OPLS-DA can distinguish among different groups and, simultaneously, identify the major contributing components. In our cases, OPLS-DA models based on MSI data of the cross-section within the whole spectral ranges (m/z 50–1,200) under positive and negative modes were further established to clarify the difference between superior and inferior samples. As analyzed in the “Quality-associated macroscopic and microscopic characters of Isatidis Radix” section, respective 3 batches of Isatidis Radix with quality characters codes 2, 4, and 6 were classified in group 1, representing good quality. Conversely, respective 3 batches of Isatidis Radix with quality character codes 1, 3, and 5 were classified in group 2, representing poor quality. As shown in Figures 4A[1],B[1], these two groups could be clearly clustered in the score plots of both positive and negative ion modes. Furthermore, the major differential components were acquired from the S-plots (Figures 4A[2],B[2]), in which each point represented an ion. The farther a data point was from the center point, the greater that ion contributed to the sample difference. Normally, compounds distributed at the two ends of the S-plot were regarded as the potential markers. Among them, arginine, progoitrin/epiprogoitrin, (R, S)-goitrin, 3-formyl-indole, syringin, adenosine, malic acid, adenine, isovitexin, clemastanin B, and uridine exerted higher ion intensities in samples from group 1 than in those from group 2. The very important in projection (VIP) values for their corresponding ions were 17.90, 18.77, 8.93, 7.08, 6.31, 6.10, 5.99, 5.50, 4.35, 4.31, and 4.08, respectively, with P-values all below 0.05. In all, these components could be used as potential quality-associated markers for Isatidis Radix.