Plants grown in green houses under controlled conditions were supplied by the dedicated facilities of the city of Toulouse (Southern France) at flowering time in early March over a period of 3 years (2017–2019). The flowers, leaves and roots were harvested separately, and either directly processed or lyophilized and stored at −20 °C until use.

Plant material was ground to a fine powder in liquid nitrogen with a mortar and pestle and then extracted following a protocol adapted from Craik et al. (1999). The powder was suspended in a solution of dichloromethane/methanol (v/v) at a ratio of 10 mL per gram fresh weight (FW) of the starting material. After stirring at room temperature for 4 h, the suspension was centrifuged (8000g, 10 min, 3 °C), and the pellet re-extracted for an additional 3 h using the same conditions. The pellet was discarded and the recovered supernatants were pooled, filtered on a sintered G5 glass funnel, then poured into a separating funnel and water added until a separation phase was visible. The lower phase which contains chlorophyll and lipophilic substances was discarded, and the upper methanolic aqueous phase was recovered, concentrated under reduced pressure (Büchi) and adjusted to a small volume with water, giving rise to the flower, leaf and root aqueous crude extracts (CEs).

The quantitative determination of the phenolic and flavonoid contents of the CEs was achieved using standardized colourimetric methods: the Folin–Ciocalteu assay for measuring the C6–C3 phenolic compounds at 765 nm (Hughes et al. 2010) and the AlCl₃ assay for measuring flavonoids at 430 nm (Bahorun et al. 1994). The total phenolic and flavonoid contents were calculated on the basis of the calibration curve of ferulic acid (y = 0.0168x + 0.0881, R² = 0.9913) and rutin (y = 0.0211x + 0.0003, R² = 0.9978), respectively. The anthocyanidin content was determined at 530 nm as reported by Ganjewala et al. (2008) with cyanidin chloride as the standard. The data were expressed as micrograms equivalent of ferulic acid, rutin or cyanidin chloride per gram of FW starting material.

Ultra-high-performance liquid chromatography–high-resolution mass spectrometry (UHPLC–HRMS) analysis of the phenolic compounds of flower, leaf or root extracts was performed as previously reported (Chervin et al. 2017) with a UHPLC-LTQ Orbitrap XL instrument (Ultimate 3000, Thermo Fisher Scientific, Hemel Hempstead, UK) and a Acquity UPLC BEH C18 column (100 × 2.1 mm i.d., 1.7 μm, Waters, MA, USA) equipped with a guard column. The mobile phase A (MPA) was water with 0.1 % formic acid (FA) and mobile phase B (MPB) was acetonitrile with 0.1 % FA. The solvent gradient was as follows: 0 min, 95 % MPA; 0.5 min, 95 % MPA; 12 min, 5 % MPA; 15 min, 5 % MPA; 15.5 min, 95 % MPA; and 19 min, 95 % MPA. The flow rate was 0.3 mL min⁻¹, the column temperature was set to 40 °C and injection volume fixed to 2 μL. Mass detection was performed using an electrospray source (electrospray ionization technique) in negative ionization modes. The mass scanning range was m/z 100–2000 Da. Each full MS scan was followed by data-dependent tandem mass spectrometry (MS/MS) on the three most intense peaks.

Data files were processed with MS-DIAL version 3.08 (Tsugawa et al. 2015) with optimized detection threshold set to 3 × 10⁵. Annotation of main features was processed with MS-FINDER version 3.04 (Tsugawa et al. 2016) using an in-house Viola database based on matches within the Dictionary of Natural Products (DNP, CRC Press, version 25:2). Results afforded several candidates and they were ranked according to their similarity score based on comparison between experimental MS/MS fragments and in silico spectra of candidates.

For optical microscopy, 150- to 200-µm sections of fresh samples were made using a vibratome (VT1000S, Leica, Rueil-Malmaison, France), mounted on a glass slide in a drop of distilled water and observed using either a fluorescent macroscope (AxioZoom V16, Zeiss, Germany) or an inverted microscope (DMIRBE, Leica, Germany) with a CCD camera (MC190HD, Leica, Germany). Plant fluorescent tissue imprints were observed with epifluorescence using different excitation ranges: UV (excitation filter BP 340–380 nm, emission filter LP 425 nm), blue (excitation filter BP 450–490 nm, emission filter LP 515 nm) and green (excitation filter BP 515–560 nm, emission filter LP 590 nm). Observations were made on 10 different leaves, roots and petals from two plants.

Confocal images were acquired with a confocal laser scanning microscope (CLSM; TCS SP8, Leica, Germany) using 405 and 561 nm diode lasers, with the emitted autofluorescence collected in the range of 420–600 and 650–720 nm, respectively. Images were acquired in the z dimension and 3D reconstructions were carried out using 3D Leica software. Emission spectra of intact cells and extracts were collected using the spectral module of the Leica SP2 scanning head. The 405-diode laser was used for excitation and spectra were collected in the range of 415–780 nm with a 10-nm band pass. Emission spectra (N = 20–30) were acquired for 10–15 cells per tissue types from two plants or from extracts made at least from three biological repeats.

In order to characterize the Parma violet Viola alba subsp. dehnhardtii, intact and/or transverse sections of the leaf blade, petal and root were examined (Fig. 2A–Q). Upon excitation in the UV range (340–380 nm), the fluorescence emitted (425 nm) by the adaxial (upper) surface of the leaf blade exhibited an intense fluorescence in the red range interspersed with the deep blue fluorescence of the midrib and the veins (Fig. 2A). This blue fluorescence was also detected at the distal part of the midrib (open arrow) and under the form of small outgrowths at each tooth of the dentate blade (solid arrows). Examined at higher magnification, these outgrowths resemble glandular structures in bright field (Fig. 2B), which emit blue (Fig. 2C) or yellow (Fig. 2D) fluorescence upon excitation in the UV (340–380 nm) or blue (450–490 nm) range.

Compared to the adaxial face, the abaxial (lower) surface exhibited a more intense fluorescence in the blue range than in red (Fig. 2E). The blue fluorescence was distributed in patches throughout the blade which holds numerous stomata, each being surrounded by highly fluorescent epidermal cells arranged in a typical jigsaw puzzle-type pattern (Fig. 2F); an intense fluorescence was also observed at the distal part of the midrib (Fig. 2G).

Observations of free-hand transverse sections of a leaf blade were made by conventional wide-field microscopy (Fig. 2H) and CLSM (Fig. 2I). Upon excitation in the UV range, the emitted fluorescence was observed mainly in the blue and the red channels, which roughly correspond to the fluorescence of phenolic compounds and chlorophyll, respectively. Blue fluorescence was associated primarily with the large vacuoles of palisade parenchyma cells (deep blue), large pockets within the spongy mesophyll (bright blue) and epidermal cells at the abaxial (lower) face of the leaf blade. Such a blue gradient might reflect the presence of distinct phenolic families. Cells of the adaxial epidermal layer covered by a thin fluorescent cuticle exhibited a distinct absence of fluorescence (Fig. 2H).

The red fluorescence clearly depicted the numerous chloroplasts dispersed in the cytosol around the vacuoles in leaf tissue. For a further examination, confocal images were acquired in the z dimension and used to generate 3D reconstructions focused on the mesophyll (Fig. 2I). As indicated above, the former range of emitted fluorescence (420–600 nm, pink pseudo colour) was mainly collected from the large vacuoles of both types of cells, and the latter (650–720 nm, green pseudo colour) from the numerous chloroplasts.

The general features of the flower petal are shown in Fig. 2J–P. Examination of intact petals and transverse sections showed that the violin colour was exclusively located in the epidermal layers, both on the adaxial and abaxial sides of the petal (Fig. 2O), typical of anthocyanidin pigments. This is consistent with the fluorescence properties of these two cell layers, which was mainly in the red range (Fig. 2P). A closer examination of the adaxial surface showed the jigsaw puzzle-like disposition of the cells (Fig. 2J and K), with a few cells exhibiting a blue autofluorescence (Fig. 2L and M). The cells of the abaxial surface exhibited different patterns, with the cell walls appearing to expand into the cellular space (Fig. 2N).

A transverse section of a Viola root is shown in Fig. 2Q. The emitted fluorescence was mainly collected in the blue range within cortical tissues, vascular tissues and the pith, again exhibiting a gradient from deep blue (two to three cell layers of the rhizoderm), to the bright turquoise blue of the xylem vessels and pith. This is indicative of the presence of different phenolics in these tissues, i.e. soluble phenolics in the vacuoles of the rhizoderm, versus insoluble wall-bound polymers in the vessels. Due to the high fluorescence intensity, the endodermis and pericycle tissues were not clearly delimited. Highly fluorescent patches, presumably secretory cells, were also visible at the outer cell layer of the rhizoderm.

The protocol that was retained to solubilize small metabolites was adapted from Craik et al. (1999). Indeed, colourimetric assays showed that the aqueous CEs contain flavonoids and/or simple phenolics, whose amounts vary with the organs (Fig. 3). Over several extractions, the highest levels of flavonoids and phenolics characterized the flowers (mean values 1.34 ± 0.17 and 1.76 ± 0.11 mg g⁻¹ FW, respectively), as compared to leaves (0.22 ± 0.03 and 1.26 ± 0.08 mg g⁻¹ FW, respectively) and roots (0.2 ± 0.02 mg g⁻¹ FW phenolics only). A further experiment conducted on a narrow (3–5 mm) band of tissue excised from the margin of the leaf blade revealed that the small outgrowths observed by microscopy (Fig. 2A–D) contributed significantly (up to 60–70 % per gram of FW) to the phenolic C6–C3 and flavonoid content of the leaf, being more concentrated in this part than in the rest of the blade.

Among phenolic compounds, preliminary UHPLC–HRMS analysis [seeSupporting Information—Fig. S1 and Table S3] detected varying amounts of coumarins (peaks # 4, 8, 9) and phenolic acids, notably salicylic acid (peak # 44) in the aerial parts and root extracts of the violet, being higher in leaves than in other organs, on a FW basis. It also showed the richness of flower extracts in several flavonoids, most notably kaempferol (peaks # 16, 24, 26) and rutin glycosides (peaks # 46, 48). Anthocyanidins were restricted to flower extracts (Fig. 3), whose measured amounts (111.4 ± 3.2 µg g⁻¹ FW) accounted for about 8 % of the total flavonoid content.

To determine whether the compounds detected in the spectrophotometric assays (Fig. 3) could account for the autofluorescence observed in these tissues (Fig. 2), the fluorescent spectra of living tissues were determined using the lambda scan module of the CLSM. Using the 405-nm diode laser, spectra were collected from 415 to 780 nm, comparable to the conditions used for imaging, above. Figure 4 and Supporting Information—Table S2 show the spectra acquired for 15–20 cells per tissue types from two plants (Fig. 4A–C), the spectra of the aqueous extracts prepared from the same plant parts (Fig. 4D–F) and the spectra of standards (Fig. 4G–I) used for the chemical measurements above.

Based on the emission spectra, three distinct patterns were observed. The first of these (peak 1, A_max = 470 nm) was present in both intact tissues (Fig. 4A–C) and extracts (Fig. 4D–F) of flowers, leaves and roots. Because flavonoids were not detected in roots, it follows that peak 1 corresponds to the C6–C3 phenolic acid and/or coumarin compounds, in agreement with the spectra of the authentic standards (ferulic acid, 4-methyl umbelliferone). A closer examination of living leaf tissues showed that the fluorescence spectra displayed the same pattern across the section, being higher in the glandular-like structure and their extracts than in parenchyma cells (Fig. 4B and E), consistent with their respective concentration in C6–C3 compounds.

The second pattern, in the blue-green to yellow region of the visible spectrum (520–620 nm), was solely present in the leaf and flower tissues in the form of a peak (# 2) which slightly overlaps peak 1 in living tissues (Fig. 4A and B) and in the corresponding extracts (Fig. 4D and E). The presence of flavonoids in the extracts (Fig. 3B), including rutin [seeSupporting Information—Fig. S1 and Table S1], and the region of the visible spectrum where these fluoresce, indicate that peak 2 corresponds to the flavonol group of flavonoids, consistent with the similar range of fluorescence of the flavonoid standards, rutin and quercetin (Fig. 3H).

A third emission maximum (peak 3, 600–700 nm) was specifically associated with the flower petal epidermis, both in the adaxial cells of intact tissues (Fig. 4A) and in extracts (Fig. 4D), typical of the presence of anthocyanidin flavonoids and in agreement with the spectrum of the authentic standard, cyanidin. While only observed in trace amounts in leaf extracts (Fig. 4E), its presence was likely also masked by the intense overlapping fluorescence of chlorophyll in these tissues (Fig. 4B).

Data used for the chemical analysis and the fluorescence emission spectra are available as Supporting Information.