Plant propagation and harvesting
We used Dionysia tapetodes accession 20,140,435 that was received from the Royal Botanic Garden Edinburgh (accession 19,822,508) comprising wild material collected by Prof. T.F. Hewer (number 1164) between 1969 and 1971. D. tapetodes was grown at Cambridge University Botanic Garden (Cambridge, UK) in clay pots plunged in sand in an alpine house to keep the roots cooler and at a more stable temperature. The potting compost comprised 50% loam-based compost, 30% 1-9 mm grit, 10% sharp sand, 10% seramis and a small amount of slow release fertiliser. The compost is top-dressed with grit which is carefully worked under the collar of the plant to reduce the risks of basal rot. Given that Dionysia are prone to rosette burning from overexposure to sunlight, a temporary shading screen was placed over the collection from mid-March and removed later in the season after sun intensity decreased. Old flowers were carefully removed to prevent any botrytis infection moving from dead flowers into living material. Watering adhered to a strict regime: Young plants were watered from underneath via a water bath and once they passed one year old were then watered overhead without getting the cushion surface wet. Water was applied sparingly in the winter months and increased once the plants were in active growth. Primula marginata (accession 195,844,928) leaf samples were taken from plants grown in the mountain glasshouse at Cambridge University Botanic Garden. Primula bullata var bullata winter leaves were a gift from Prof. David Rankin (University of Edinburgh, UK).
Cryo-Scanning Electron Microscopy (cryoSEM) and cryo-fracture
Rosettes of 5–8 leaves were mounted, frozen in nitrogen slush, platinum coated and fractured as previously described [30]. To accommodate better fractures without dislodging trichomes, some samples were dipped in 70% v/v ethanol to remove the wool and then air-dried prior to freezing. Cryo prepared samples were viewed using a Zeiss EVO HD SEM fitted with a backscattered electron detector and 25 kV acceleration voltage.
SEM imaging of wool fibres
For high magnification, low kV imaging, clumps of uncoated wool fibres were placed on a sticky carbon tab and mounted on an SEM stub in the Zeiss EVO HD SEM at high vacuum and 1 kV accelerating voltage using SE detector fast scanning with frame averaging to prevent wool movement. Backscattered electron imaging of wool was carried out using the BSD detector and 25 kV accelerating voltage.
Embedding, sectioning and imaging of trichomes by transmission electron microscopy (TEM) and light (fluorescence) microscopy
The tip of the leaf (or leaf buds) was removed to ensure efficient fixative penetration, and then immediately treated as described in [31] for fixation, dehydration, and embedding steps. For fluorescence microscopy, semi-thin Sects. (1 μm) of the whole leaf bud were then obtained with a Leica EM UC7 ultramicrotome using a Histo Jumbo 8 mm diamond knife (DiATOME) and laid on droplets of sterile water on (uncoated) glass microscopy slides. In order to avoid folds on sections during the water drying process, slides were dried on a hot plate set at 55 °C (Leica). Slides were finally mounted in a 1:1 solution of AF1 Citifluor antifadent with PBS, containing calcofluor as a cell wall counterstain, and imaged with a confocal microscope (Zeiss LSM700).
For TEM observations, 100 nm thin sections were obtained with a Leica EM UC7 ultramicrotome using an ultra 45° diamond knife (DiATOME) and deposited on 200-300 μm mesh formvar-coated nickel grids.
Samples were post-stained in Reynold’s lead citrate (3 min) plus uranyl acetate (3 min). Thin sections were viewed in a Tecnai G20 electron microscope at 200 keV, 20 μm objective aperture. Images were taken with an AMT camera controlled by DEBEN software.
Field Emission Scanning Electron Microscopy (FE-SEM) of leaf sections
Whole rosettes were dissected with razor blades to remove leaf tips and subsequently processed as previously described in [32]. Following dissection, samples were immediately submerged in fixative (2% formaldehyde, 2% glutaraldehyde, 2.0 mM calcium chloride in 0.05 M sodium cacodylate buffer at pH 7.40) under vacuum overnight at room temperature. After washing 5 times in deionised water (DIW), samples were osmicated (1% OsO4, 1.5% K3Fe(CN)6 in 0.05 M sodium cacodylate buffer, pH7.4) for 3 days, 4 °C. Then samples were washed 5 × with DIW and treated with 0.1% (w/v) thiocarbohydrazide in DIW for 20 min in the dark at room temperature. After washing 5 × in DIW, osmication was repeated for 1 h at RT with 2% OsO4/DIW then 5 washing steps in DIW. Samples were then block-stained with uranyl acetate block stain (2% uranyl acetate in 0.05 M maleate buffer, pH 5.5) for 3 days, 4 °C. Samples underwent a further 5 washes in DIW and then dehydrated in a graded series of ethanol (50% > 70% > 95% > 100% > 100% dry), then 100% dry acetone and then 100% dry acetonitrile, 3 × in each for minimum 5 min. Samples underwent infiltration with a 50/50 mixture of 100% dry acetonitrile/Quetol resin, minus BDMA, overnight, followed by 3 days in 100% Quetol (minus BDMA) and then 5 days in 100% Quetol resin plus BDMA, exchanging the resin every 24 hr. The Quetol resin mixture consists 12 g Quetol 651, 5.7 g MNA, 15.7 g NSA, 0.5 g BDMA. Samples were placed within embedding moulds and then cured at 60 °C, 3 days. Semi-thin Sects. (1 μm) of the whole leaf bud were then obtained with a Leica EM UC7 ultramicrotome using a Histo-Jumbo 8 mm diamond knife (DiATOME) and then laid on droplets of sterile water on (uncoated) glass microscopy slides. Slides were dried on a hot plate at 55 °C (Leica) in order to avoid folds appearing during drying. Glass slides were then trimmed with a glass knifemaker to be mounted on aluminium SEM stubs using conductive carbon tabs, and the edges of the slides were painted with conductive silver paint. Samples were coated with 30 nm carbon using a Quorum Q150 TE carbon sputter coater. Samples were imaged in a FEI/Thermofisher Verios 460 SEM at 4 kV accelerating voltage, 0.2 nA probe current using a concentric backscatter detector (CBS) in immersion mode with a working distance of 3.5–4 mm; 1536 × 1024 pixel resolution, 3 μs dwell time, 4 line integrated. Stitching of adjacent image areas was carried out using the FEI MAPS software and default stitching settings and a 10% overlap.
Measurements of wool fibre diameter
Images of wool fibres, attached to leaves of an isolated D. tapetodes rosette, were taken with a Keyence VHX-7000 microscope at 2500 × magnification and illuminated with full field coaxial light. 2D depth-up mode was used for in-focus acquisitions. Fibre width measurements were carried out using the point-to-point measuring tool in the Keyence software.
Raman microscopy of farina
Raman microscopy was carried out on a Renishaw InVia instrument fitted with a 785 nm laser. D. tapetodes farina wool or P. marginata powder was carefully placed on a quartz slide and brought in to focus under a 50 × dry objective lens. Raman acquisitions used a 1200 l/mm grating, 1200 cm−1 centre, 785 nm laser at 10% power, regular confocal mode and 4 s exposure with 3 accumulations. At least 3 spectra per sample were averaged in order to improve signal-to-noise. To find close matches with reference Raman spectra, the experimental spectra were used as a search input against the Raman databases, that include some flavone derivatives, in the KnowItAll software (Bio-Rad Inc.) using the “SearchIT” tool and then candidate spectra were visualised by eye to remove false positives. Both default and Euclidean distance search settings were used. Matches are ranked according to their hit quality index (out of a maximum of 100). P. marginata farina gave a close match (97/100) with flavone. D tapetodes gave no close matches but yielded good correlation (70–80/100) to reference spectra of hydroxy- and methoxy- flavone derivatives that included 7, 2′-dimethoxyflavone; 3,7-dimethoxyflavone and 6-, 7-, or 8- hydroxy-derivatives. For fastFLIM-Raman correlative imaging of leaves submerged in water (Additional file 4) a confocal-Raman microscope, described in [33], used the following settings: 25 × 0.95 NA water dipping objective lens, FLIM 440 nm pulsed laser (at 20 MHz) with detector window set between 448 and 511 nm and 80 iterations. Raman: 1200 l/mm grating, 1200 cm−1 centre, 785 nm laser, 50% power, 15 s exposure with 2 accumulations used in line scan mode that intersected a glandular head cell.
Nile red staining and imaging
Freshly harvested leaves of D. tapetodes were placed in tubes containing 0.1 μg/ml w/v of Nile red in 0.0001% acetone (prepared from a Nile red stock of 1 mg/ml w/v in 100% acetone). 3D Imaging was carried out on a Zeiss LSM700 confocal microscope using a 555 nm laser and 575–625 nm emission filter. Deconvolution and surface rendering of the Z-stack was carried out on Huygens software (Scientific Volume Imaging, Netherlands).
Reagents, solvents and sample preparation for chemical analysis
Pure Flavone was purchased from Alfa Aesar as a white solid with 99% purity (CAS no. 525–82-6, catalogue no. A13627) and used without further purification. All solvents were anhydrous and used as purchased without any further purification. Flavone wool from was picked from D. tapetodes leaf surfaces using fine tweezers and placed in a microcentrifuge tube. The wool sample (approximately 0.5 mg) was dissolved in 50 μL of acetonitrile/water (1:1) with a few drops of dimethylsulfoxide (DMSO) to aid solubility. Sample preparation was performed in this way for analytical HPLC, LCMS and HRMS analysis.
Analytical high-performance liquid chromatography (HPLC)
Analytical HPLC was run on an Agilent 1260 Infinity using a Supelcosil ABZ + PLUS column (150 mm × 4.6 mm, 3 μm) eluting with a linear gradient system (solvent A: 0.05% (v/v) trifluoroacetic acid (TFA) in H2O, solvent B: 0.05% (v/v) TFA in acetonitrile (MeCN)) over 15 min at a flow rate of 1 mL/min.
Liquid chromatography mass spectrometry (LCMS)
Chromatographs were recorded on a Waters ACQUITY H-Class UPLC with an ESCi Multi-Mode ionisation Waters SQ Detector 2 spectrometer (LC system: solvent A: 2 mM ammonium acetate in water/MeCN (95:5); solvent B: 100% MeCN; column: AQUITY UPLC CSH C18, 2.1*50 mm, 1.7 μm, 130 Å; gradient: 63 5–95% B over 3 min with constant 0.1% formic acid).
High resolution mass spectrometry (HRMS)
HRMS was carried out on a Waters LCT Premier Time of Flight mass spectrometer. ESI refers to the electrospray ionisation technique.
Nuclear magnetic resonance (NMR) spectroscopy
NMR spectroscopy was carried out as described in reference [34], with the following modifications: All pulse sequences are the default (with the exception of the DEPT135) from the Topspin 3.2pl7 software used to control the acquisition. The analysis required 1H, 13C, DEPT135, DFQ-COSY, Heteronuclear Single Quantum Coherence (HSQC, with DEPT 135 editing) and Heteronuclear Multiple Bond Correlation Spectroscopy (HMBC) spectra. All necessary shaped and decoupling pulses were calculated by the software, using defined 90 degree pulses.
1H NMR
Proton magnetic resonance spectra were recorded using an internal deuterium lock (at 298 K unless stated otherwise) on Bruker DPX (400 MHz; 1H-13C DUL probe), Bruker Avance III HD (400 MHz; Smart probe), Bruker Avance III HD (500 MHz; Smart probe) and Bruker Avance III HD 62 (500 MHz; DCH Cryoprobe) spectrometers. Pulse sequence used zg30 – PLW1 = 14 W, P1 = 10.5 μs, SW = 20 ppm, TD = 64 K, AQ = 3.28 s, D1 = 1 s, NS = 16. Proton assignments are supported by 1H-1H COSY, 1H-13C HSQC or 1H-13C HMBC spectra, or by analogy. Chemical shifts (δH) are quoted in ppm to the nearest 0.01 ppm and are referenced to the residual non-deuterated solvent peak. Discernible coupling constants for mutually coupled protons are reported as measured values in Hertz, rounded to the nearest 0.1 Hz. Data are reported as: chemical shift, multiplicity (br, broad; s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet; or a combination thereof), number of nuclei, coupling constants and assignment.
13C NMR
Carbon magnetic resonance spectra were recorded using an internal deuterium lock (at 298 K unless stated otherwise) on Bruker DPX (101 MHz), Bruker Avance III HD (101 MHz) and Bruker Avance III HD (126 MHz) spectrometers with broadband proton decoupling. 1024 scans (NS) were acquired using pulse sequence 'zgpg30′, with waltz16 1H decoupling. 90 degree 13C pulse set to 21 W (PLW1) for 9.5 μs (P1). 209,786 points (TD) were digitised over 3.02 s (AQ), relaxation delay set to 2 s (D1). Sweep width was 276 ppm (SW), with an irradiation frequency of 110 PPM (O1P). Carbon spectra assignments are supported by DEPT editing, 1H-13C HSQC or 1H-13C HMBC spectra, or by analogy. Chemical shifts (δC) are quoted in ppm to the nearest 0.1 ppm and are referenced to the deuterated solvent peak. Data are reported as: chemical shift, number of nuclei, multiplicity, coupling constants and assignment. Magnetic resonance spectra were processed using TopSpin (Bruker). An aryl, quaternary, or two or more possible assignments were given when signals could not be distinguished by any means. Standard flavone numbering was followed.
DEPT 135
Pulse sequence dept135sp – This is a minor modification of the DEPT sequence to optimise the spectral baseline and uses an adiabatic shape for 180 degree carbon pulses. Carbon pulse powers as 13C experiment above, SW = 236.7 ppm, TD = 65,536, AQ = 1.10 s, D1 = 2 s, O1 = 100 ppm, NS = 64. Waltz16 decoupling.
DFQCOSY
This is a double-quantum filtered experiment; using gradient selection; pulse sequence cosygpmfqf. Non-uniform sampling; using a Poisson-gap weighted schedule was used to acquire 37.5% of 512 increments, each with 2 scans (SWF2 = 13.37 ppm, TD = 4 k, AQ = 0.31 s, D1 = 2 s). Proton pulse powers as above. Processed to 2 k x2k points using a sine function (SSB = 2.5).
HMBC
This experiment is phase sensitive; uses Echo/Antiecho gradient selection, with a three-fold low-pass J-filter to suppress one-bond correlations; pulse program 'hmbcetgpl3nd'. Acquired in phase sensitive mode using Echo/Antiecho-TPPI gradient selection, with a 3 step low pass j-filter to suppress 1 bond correlations. Long-range J-JCH parameters set to 10 Hz. Non-uniform sampling; with a Poisson–gap weighted schedule was used to acquire 37.5% of 768 increments; each with 2 scans (SWF1 = 250, SWF2 = 12.02 ppm, TD = 4096, AQ = 0.34 s, D1 = 2 s). Processed to 2048 × 2048 using a sine function (SSB = 4 & 2 for F2 and F1), then converted to magnitude mode in F2. (Topspin command 'xf2m').
HSQC
The HSQC was acquired using a Bruker Avance III HD 500Mhz equipped with a dual 13C/1H cryoprobe; using Topspin 3.2pl7. It was acquired in 'non-uniform sampling' mode and samples 25% of 1024 increments, using a 'poisson-gap' schedule. The data was processed using the default compressed sensing (CS) method in Topspin 3.5pl7 (on your computer) to 2048 × 2048 data points. This experiment is configured to give -CH2 groups an opposite phase to the -CH and -CH3 groups, using the hsqcedetgpsp.3 pulse sequence. Acquired in phase sensitive mode using Echo/Antiecho-TPPI gradient selection, multiplicity edited during selection step with shaped adiabatic pulses. Non-uniform sampling; with a Poisson-gap weighted schedule was used to acquire 25% of 1024 increments; each with 2 scans (SWF1 = 190 ppm, SWF2 = 12.99 ppm, TD = 1816, AQ = 0.14 s, D1 = 0.8 s). Carbon and proton pulse powers as above. Processed to 2048 × 2028 points using a qsine function (SSB = 2).
Standard flavone numbering: