An inverse association between plasma benzoxazinoid metabolites and PSA after rye intake in men with prostate cancer revealed with a new method

Prostate cancer (PC) is a common cancer among men, and preventive strategies are warranted. Benzoxazinoids (BXs) in rye have shown potential against PC in vitro but human studies are lacking. The aim was to establish a quantitative method for analysis of BXs and investigate their plasma levels after a whole grain/bran rye vs refined wheat intervention, as well as exploring their association with PSA, in men with PC. A quantitative method for analysis of 22 BXs, including novel metabolites identified by mass spectrometry and NMR, was established, and applied to plasma samples from a randomized crossover study where patients with indolent PC (n = 17) consumed 485 g whole grain rye/rye bran or fiber supplemented refined wheat daily for 6 wk. Most BXs were significantly higher in plasma after rye (0.3–19.4 nmol/L in plasma) vs. refined wheat (0.05–2.9 nmol/L) intake. HBOA-glc, 2-HHPAA, HBOA-glcA, 2-HPAA-glcA were inversely correlated to PSA in plasma (p < 0.04). To conclude, BXs in plasma, including metabolites not previously analyzed, were quantified. BX metabolites were significantly higher after rye vs refined wheat consumption. Four BX-related metabolites were inversely associated with PSA, which merits further investigation.


Supplementary Text 1.
Discovery of BX-related metabolites from urine samples with high BX concentration using Data Dependent Acquisition Mass Spectrometric (DDA-MS) The LC-MS system consisted of an Agilent 1260 infinity HPLC system with a quaternary pump (Santa Clara, CA, USA) coupled to an AB Sciex 4500 triple-quadrupole trap mass spectrometer (QTRAP/MS) (AB Sciex, Framingham, USA) with electrospray ionization (ESI). The autosampler was fitted with a cooler set at 10 °C. The column oven was set at 30 °C and equipped with a Synergi Polar column and pre-column (4 µm, 2.1*250 mm, Phenomenex). The gradient was run as a binary gradient using a weak eluent (A) consisting of 7% acetonitrile and 20mM acetic acid in water; and a strong eluent (B) consisting of 78% acetonitrile and 20 mM acetic acid in water.
Nitrogen gas was used as a collision gas to generate MS/MS fragmentations. Instrumentdependent parameters for the mass spectrometer were as follows: Curtain gas, 35 psi; ion spray voltage, -4300 V; source temperature, 450 °C; gas 1, 90 psi; gas 2, 90 psi; the collision gas (CAD) was set to -2. The chromatographic method was run at 300 µl/min for 23 min with an injection volume of 10.0 µL. The gradient was as follows: At 0 min A/B was 100/0; at 1 min A/B: 92/8; at 3 mins A/B: 90/10; at 13 min A/B:30/70; at 14 min A/B:10/90; at 16 min A/B:10/90; at 17 min A/B:100/0; and end at 23 min in A/B:100/0. Two DDA-MS methods were set up, an untargeted and a targeted method. The untargeted DDA-MS method (EMS-EPI) was set up combining a full scan (EMS, m/z 200-550 Da) as survey scan, with a product ion scan (EPI, m/z 50-500 Da) as dependent scan, both utilizing the enhanced sensitivity of the linear ion trap. Similarly, a targeted DDA-MS method (MRM-EPI) was set up utilizing known fragmentation patterns of BX-standards and non-BX commercially available phase 2 glucuronide and sulfate metabolites (Group 4 compounds). The method took advantage of the high sensitivity of the triple quadrupole using targeted ion reaction mass transitions (Table 1, manuscript) as survey scan, combined with the EPI scan (m/z 50-500 Da) as dependent scan. Both the targeted and untargeted DDA-MS methods were set up with the purpose of obtaining spectral data as good as possible (cycle time 2.5 sec and general peak width at half height 15 sec). Further settings were based on the work published by Bhattarai et al (2021) 1 .

Supplementary Text 2.
Purification of BX-related metabolites from urine rich in BX from a human pilot study Some unidentified signals present only after the rye intervention and with BX-like MS fragmentation patterns, determined by the data dependent acquisition-based MS-methods (DDA-MS), were marked as interesting for further investigation. Available BX-related standards (See Table 1, group 1 and 2 (manuscript)) were used to build the methods and interpret the data. Most of the interesting signals could be recognized as probably being a known aglucone coupled to a glucuronic acid or sulfate i.e. common phase II metabolites. The combined urine taken after the intervention was subjected to subsequent solid phase extractions and DDA-MS was used for fraction selection with the purpose to isolate the compounds of interest sufficiently for identification by NMR. Fractions with a content of BX related metabolites as determined by the DDA-MS analysis, where mass spectra of BX-standards were compared to spectra of unknown peaks, were subjected to further SPE separation this time using weak anion exchange SPE (WAX), utilizing the inherent anionic properties of glucuronic acid and sulphate derivatives. Prior to passing the reconstituted sample through the WAX, the sample was passed through a strong cation exchange cartridge (MCX). This allowed for the removal of basic and twitter ionic compounds by binding to the strong cationic exchange before selective and reversable binding of anionic compounds to the weak anion exchange column. Meanwhile neutral, polar compounds were washed away due to poor interaction with the stationary phases. Hippuric acid, endogenic to urine, was a major obstacle in the cleaning process as it was present in very high concentrations and had acidic properties like the glucuronic acid derivates. Hence, method development also focused on conditions that would separate the BX-derivative from hippuric acid as much as possible. This was obtained to a large extent by eluding the weak acid metabolites first with consecutive small fractions of 2% hydrochloric acid and 20% methanol in water and afterwards 2% hydrochloric acid and 40% methanol in water leading to the BXglucuronides eluding in the first fractions and hippuric acid in the later. The BX-sulfates were retained until eluted by base. In this way a good separation into metabolite categories was obtained prior to moving on to semi-preparative HPLC. However, HPLC-UV revealed that the fractions still contained a lot of endogenic urine metabolites and that the BX-metabolites were only minor constituents in the fractions. Fractions were collected by 30 seconds time windows and all fractions were analyzed using the DDA-MS methods. In this way, the small peaks corresponding to interesting BX-metabolites were identified and collected. A second round of semi-preparative HPLC-UV using a column with a different selectivity afforded the purified standards. Their structures were primarily assigned by mass spectrometry comparing spectra to that of standard compounds of similar structures.

NMR Experiments
NMR spectra were acquired in DMSO-d6 at 300 K using a Bruker Avance III NMR spectrometer ( 1 H resonance frequency of 600.13 MHz), and 1 H and 13 C NMR chemical shifts were referenced to the residual solvent signal of DMSO-d6 (δH = 2.50 ppm, δC = 39.52 ppm). 1 H NMR spectra were recorded with a spectral width of 20 ppm using 30° pulses and 65 k data points. For the 2D NMR experiments, phase-sensitive DQF-COSY and ROESY spectra were recorded using gradient-based pulse sequences with 9 ppm spectral width and 2 k × 512 data points (processed with forward linear prediction to 1 k data points in F1); multiplicity-edited HSQC spectra were acquired with the following parameters: 1JC,H = 145 Hz, spectral width 12 ppm for 1 H and 140 ppm for 13 C, 1730 × 192 data points (processed with forward linear prediction to 512 data points and zero-filling to 1 k data points in F1), and 1.0 s relaxation delay; low-pass filtered HMBC experiments were optimized for n JC,H = 8.0 Hz (long-range), 1 JC,H min = 125 Hz, and 1 JC,H max = 160 Hz and recorded using a spectral width of 12 ppm for 1 H and 180 ppm for 13 C , 2k × 256 data points (processed with forward linear prediction to 512 data points and zero-filling to 1 k data points in F1), and 1.0 s relaxation delay.

Determination of recovery and limit of detection (LOD) of BX compounds in plasma
To obtain a plasma matrix that contained no BX compounds for spiking experiments, blood from pigs that had only been fed with refined wheat for several weeks was used. Four standard mixtures (A, C, L, and HH, see Table 1 in manuscript) were spiked into aliquoted pig plasma samples at a concentration of 5 ng/ml plasma in six replicates giving a total of twenty-four recovery samples. The spiked samples were stored together with the original samples at -80 ºC and subsequently thawed, prepared, and analyzed one set of four recovery samples together with each of the six batches of study plasma samples. The mean and the standard deviation of the recovery samples of all six batches were calculated and LOD was determined as 3 x STDDEV of six replicates. For the semi-quantified compounds, no recovery or LOD were determined).

Validation of a quantitative method for determination of BXs compounds/metabolites in plasma
Guidelines for validation of analytical methods aimed at being used for quantification of natural bioactive compounds in food or in consumer's plasma do not exist. Validation guidelines are generally dedicated to methods that are used for analysis of potential toxic contamination (pesticide residues in food) 2 . or for analysis that are meant to assure that then content of a pharmaceutical drug is accurate 3 . It is however a general recommendation that recovery and limit of detection are relevant parameters to include when an analytical ad hoc method is used 4 . The main goal of a method validation is to assure that the method is fit for the purpose. Recoveries between 70 and 120% are often mentioned as acceptable 2 . For seven of the compounds the recovery values are outside this range; however, for six of those compounds the relative standard deviation is <20%, which makes the deviation from recommended recovery range acceptable.
The BX-phase II metabolites semi-purified from urine were semi-quantified in the plasma samples against standard curves of the standards 4-HPAA-glcA and 4-HPAA-sulfate respectively. However, between-sample statistical comparison still provides meaningful interpretations because the inter-sample variation in matrix effect can be disregarded as was found in the work of Steffensen et al. 5 Likewise, a LOD could not be established for the BXphase II metabolites. In general, LOD varies considerably between compounds and is affected by many factors both of chromatographic properties and mass spectrometric properties. The relatively low and insignificant concentrations found in the plasma samples of some of the BXphase II metabolites might simply reflect a poor LOD. The recovery estimated from spiked controls, analyzed with all batches are shown below In Table S1 together with the LOD.

Structure elucidation of BX metabolites from urine samples
The compound (Figure S2), which by MS was suggested to be HBOA-glcA, was identified in an inseparable mixture. The HBOA skeleton was identified from the signals for H-5 at δ 6.94  The compound annotated as 2-HPAA-glcA ( Figure S3) was obtained as an inseparable mixture, evident from the complex 1 H NMR spectrum with signals of varying intensity (Spectrum 5). Starting with the broad unresolved signal for H-3 at δ 8.11, which is down field shifted compared to H-3 in N- (2-hydroxyphenyl)acetamid  where the carbonyl group is hydrogen-bonded with the phenol OH, the COSY spectrum (Spectrum 6) revealed correlations between H-3, the overlapping H-4 and H-5 at δ 7.02, and H-6 at δ 7.10. The characteristic singlet signal at δ 9.08 did not display any one-bond CH correlations in the HSQC spectrum (Spectrum 7), and was attributed to the amide N-H (similar signal observed for 2-HPAA). Similarly, a methyl singlet was observed at δ 2.09, which together with the abovementioned signals showed the presence of the 2-HPAA moiety. For the glucuronic acid moiety, doublet signals for H-1' (δ 4.70, d JH1',H2' = 7.7 Hz) and H-5' (δ 3.86, d JH4',H5' = 9.5 Hz) showed COSY cross peaks to H2' at δ 3.39 and H4' at δ 3.44, respectively, the two latter masked by a large water signal. However, also the HSQC spectrum clearly showed H-2' at δ 3.39 (δC2' 72.7) and H-4' at δ 3.44 (δC4' 71.1) in addition to H-3' at δ 3.33 (δC3' 74.8). The connection of the glucuronic acid at C-1 of the 2-HPAA moiety was confirmed by correlations observed between the amide proton H-3 and H-1', H-3, and the methyl group in the ROESY spectrum (Spectrum 8) as well as between H-6 of the 2-HPAA moiety and H-1' and H-5' of the glucuronic acid (see Figure S3). Finally, a HMBC correlation from H-5' to a signal at δC 169.7 confirmed C-6' to be a carboxylic acid (Spectrum 9). The compound annotated as 2-HPAA-glcA is thus 2-(acetylamino)phenyl β-D-glucopyranosiduronic acid and thereby the proposed structure is confirmed.
The compound (Figure S4), which by MS was suggested to be 2-HHPAA-glcA, was also identified in an inseparable mixture. The 1 H NMR spectrum of 2-HHPAA-glcA (Spectrum 10) showed the same overall pattern of a 1,2-disubstituted benzene ring as observed for 2-HPAA-glcA, with H-3 at  However, instead of the methyl singlet observed at δ 2.09 in the spectrum of 2-HPAA-glcA, a downfield shifted methylene singlet for H-9 was observed at δ 3.99 (δC 61.5), showing a 2-HHPAA skeleton rather than the 2-HPAA skeleton. The HSQC spectrum (Spectrum 11) revealed the same pattern for a glucuronic acid for 2-HHPAA-glcA as seen for 2-HPAA-glcA, and as seen in Figure S4, the same ROESY correlations (Spectrum 12) as seen for 2-HPAA-glcA was