Plant samples. This study was conducted using cabbage and Brussels sprouts purchased from supermarkets and natural food cooperatives located in Minneapolis and Saint Paul, Minnesota, between August and October, 2019. To capture a wide range of variation in GBS concentrations, samples were selected for variation in size, pigment, and when possible, production location. To further increase the variability within each sample, cabbage heads were divided into wrapper, inner, and core tissue as we have observed variation among these tissues previously. Plants were used for analysis one to five days after purchase. Each sample was split into two parts, one part to be used for wet chemistry analysis and the other for NIRS. The sample was chopped to approximately 2 cm square pieces and mixed to ensure homogeneity. Half of the sample was flash frozen for later lyophilization, and the remaining half was immediately placed in boiling water to begin sample preparation for HPLC analysis.
Chemical quantification of GBS. GBS was quantified as described by Hecht et al. [36] using modifications from Rosen et al. [37]. Samples of 100 to 150 g fresh weight were boiled in water in a volume of water three-times their weight for five minutes to deactivate myrosinase. Samples were cooled at room temperature for ten minutes, then macerated in a blender for two minutes, then a 40-mL aliquot of blended sample was stored at -30˚ C until further analysis which occurred within 30 days. Later, samples were thawed and homogenized for two minutes at 12,000 rpm with a Polytron PT 1300 D homogenizer (Kinematica AG, Lucerne, Switzerland), and then 2-mL sample of the homogenate was centrifuged for 4 minutes at 8,000 g at 4˚ C.
Desulfonated GSLs were extracted from the supernatant with solid phase strong anion exchange (SAX) columns (Sigma-Aldrich, St. Louis, MO). In a vacuum manifold, SAX columns were washed with 2 mL of 0.50 M sodium acetate buffer (pH 4.6), followed by 2 mL of deionized water. 500 µL of supernatant from centrifuged samples were filtered through columns followed by 1 mL of 0.02 M sodium acetate (pH 4.0) buffer. Finally, 1 mL of 0.2 mg·mL-1 sulfatase solution (aryl-sulfate sulfohydrolase from Helix pomatia - Type H-1; Sigma-Aldrich, St. Louis, MO) was vacuum-infiltrated through the columns. Columns were incubated for ~ 15 hours at room temperature before elution with 3 mL of water (elution occurred in two steps, first with 2 mL and then with 1 mL) and collected volumes were determined by weight. Eluents were stored at -30˚ C until HPLC analysis which occurred within 14 days.
HPLC analysis was carried out using an Agilent 1200 Series Quaternary system (Agilent Technologies, Inc., Santa Clara, CA), with the diode array detector set at λ = 229 nm, using a Luna C18, 5 µm, 250 × 4.6 mm guard column (Phenomenex, Torrence, CA) set at 30 °C. 50 µL of eluent was injected and separated on the column with the following flow rates and gradients: 0–2 min. gradient 5–15% acetonitrile, 1 mL/min.; 2–20 min. gradient 15–47% acetonitrile, 1 mL/min.; 20–22 min. gradient 47–100% acetonitrile, 1 to 1.15 mL/min.; 22–26 min., 100% acetonitrile, gradient 1.15 to 1.3 mL / min.; 26–28 min., 100% acetonitrile, gradient 1.3 to 1.5 mL/min.; 28–35 min., 5% acetonitrile, 1 mL/min. GSL peaks were viewed in OpenLAB Chromatography Data System with rev. C.01.06 software and GBS was identified based on retention time. Concentration was determined using sinigrin as an external standard and previously published response factors [38].
Near-infrared reflectance spectroscopy (NIRS) analysis. Plant samples of 80 to 150 g were wrapped in aluminum foil, flash frozen in liquid nitrogen, placed on dry ice, and stored at -80° C (S-7805, Uline, Pleasant Prairie, WI) until lyophilization. Frozen samples were placed inside a 35 L VirTis 24Dx48 general purpose freeze dryer (SP Scientific, Stone Ridge, NY, USA) which was kept at -30° C for seven days, then after one week the temperature was raised to -20° C, 24 hours later it was raised again to -10° C, and 24 hours later the temperature was then allowed to slowly raise over 3 days until it achieved room temperature and samples were then removed. Freeze-dried samples of ~ 10 g were pre-ground in a 12-cup electric coffee grinder and further ground using a Retsch ZM200 (Retsch group, Haan, Germany) grinder with 35 mesh (0.5-mm particle size). Between grinding different samples, the mesh was removed and cleaned with a brush and pressurized air to avoid cross contamination. Samples were stored in small coin envelopes (S-14719, Uline, Pleasant Prairie, WI), packed with indicating silica gel desiccant packets (MiniPax absorbent packets, Sigma-Aldrich, Saint Louis, MO), and stored inside an airtight container with for up to two weeks prior to NIRS analysis.
Diffuse reflectance NIRS was applied using a diode array NIRS instrument (DA 7250 NIR Analyzer; Perten Instruments, Hägersten, Sweden). For each NIRS run, a 10-g sample of freeze-dried cabbage or Brussels sprout leaf tissue was placed in a 22-ml volume NIRS sample dish by pouring the powdered sample into the dish and using a straight edge to scrap off any extra sample so that the surface was consistent across all scans. Each sample was scanned twice, and the calibration model was developed using the average of the two. Scans were taken in the range from 950 to 1650 nm at an interval of 5 nm. Testing and training data sets were split using a stratified random sampling procedure to split the data into calibration (n = 68) and validation (n = 24) data sets. Calibration and validation were performed with R Statistical software version 3.2.4 [39] using the pls package [40]. An additional data file generated for chemometrics are included [Additional file 1].