Mass spectrometric evidence for the modification of small molecules in a cobalt-60 irradiated rodent diet

Jeevan K Prasain; Landon S Wilson; Alireza Arabshahi; Clinton Grubbs; Stephen Barnes

doi:10.1002/jms.3950

. Author manuscript; available in PMC: 2018 Aug 1.

Published in final edited form as: J Mass Spectrom. 2017 Aug;52(8):507–516. doi: 10.1002/jms.3950

Mass spectrometric evidence for the modification of small molecules in a cobalt-60 irradiated rodent diet^&

Jeevan K Prasain ^1,^*, Landon S Wilson ², Alireza Arabshahi ², Clinton Grubbs ³, Stephen Barnes ^1,²

PMCID: PMC5544575 NIHMSID: NIHMS877965 PMID: 28544323

Abstract

The purpose of this study was to investigate the effect of radiation on the content of animal diet constituents using global metabolomics. Aqueous methanolic extracts of control and cobalt-60 irradiated NIH 7001 diets were comprehensively analyzed using nanoLC-MS/MS. Among the over two thousand ions revealed by XCMS followed by data preprocessing, 94 positive and 143 negative metabolite ions had greater than 1.5 fold changes and p-values <0.01. Use of Metaboanalyst statistical software demonstrated complete separation of the irradiated and non-radiated diets in unsupervised principal components analysis and supervised partial least squares discriminant analysis. Irradiation led to an increase in the content of phytochemicals such as glucosinolates and oxidized lipids in the diet. Twenty-eight metabolites that were significantly changed in the irradiated samples were putatively identified at the level of molecular formulae by MS/MS. MS/MS^ALL analysis of chloroform-methanol extracts of the irradiated diet showed increased levels of a number of unique linoleic acid-derived branched fatty acid esters of hydroxy fatty acids. These data imply that Gamma-irradiation of animal diets causes chemical changes to dietary components which in turn may influence the risk of mammary cancer.

Keywords: animal diet, Cobalt-60 irradiation, mass spectrometry, oxidized lipids, metabolomics

Introduction

The importance of variability in the diet in animal models of disease is well appreciated and is an important consideration regarding the reproducibility of research funded by NIH. It’s been 40 years since the American Institute of Nutrition (AIN) developed criteria for diets used in experimental research. This involved a shift from an undefined laboratory chow diet consisting of mixtures of wheat, fishmeal and soy proteins to defined diets based on casein, oils and vitamin mixtures. The first of these, the AIN76A diet was developed in the late 1970s (1) and was reformulated in the 1990s as AIN93G and AIN93M (2), optimized for growth and maintenance, respectively. Despite this improvement in the nutrition of the rodents, it nonetheless also affected on the biological outcomes of experiments. When the transition from the lab chow diet to AIN76A diet occurred, the number of mammary tumors induced by N-methyl-N-nitrosourea (MNU) in female Sprague-Dawley rats rose sharply without any changes in growth rate and final body weight (3). This suggested that a component of the lab chow diet had chemopreventive properties and led to close examination of the role of soy and its isoflavones in breast cancer prevention (3). However, surprisingly, exposure of rats to the isoflavone genistein, once they were established on AIN-76A diet and had reached 40 days of age, had no effect on numbers of mammary tumors induced by MNU (4,5). This suggested that there were (early) windows of exposure that were relevant to diet-induced mammary cancer prevention.

In related studies, attempts by the NCI chemoprevention program in the 1990s to test combinations of formerly effective chemopreventive agents failed. Examination of experimental records revealed that these previous chemopreventive effects were obtained using a laboratory chow diet. Accordingly, the NCI chemoprevention program reverted to use of laboratory chow diet (containing soy, wheat and fishmeal). Importantly, the chemopreventive effect of soy isoflavones given in the young adult period of life became evident again when they were added to laboratory chow diet (5). This suggested that exposure to soy prior to and after weaning is critical to enable isoflavones to prevent formation of mammary tumors. It confirmed previous observation that addition of genistein alone to the semi-purified AIN76A diet in early life is chemopreventive (4).

As an appreciation of the role of the microbiome in biomedical research models has evolved (plus a separate concern about the stability of stored diets), commercial diet manufacturers have introduced Gamma-irradiation of diets with cobalt-60 to reduce microbial contamination. Many institutional animal resource programs have changed to using these irradiated diets. One of us (CJG) using these irradiated diets in the studies of breast cancer chemoprevention in rats, noted that the number of mammary tumors were 40% less than when using non-irradiated diets. Since there was a risk that the irradiated and non-irradiated diets were not strictly comparable since they had come from different batches of diets over several years, a single batch of NIH 7001 diet was prepared by Envigo and divided into two parts, one not receiving radiation (0 kGy) and the other exposed to 20–50 kGy of cobalt-60 Gamma-radiation. This experiment confirmed the earlier result in that there were 38% less tumors in animals consuming the irradiated diet (C. J. Grubbs, unpublished results).

The goal of this survey study, therefore, was to determine the effects of γ-radiation on the small molecule metabolome of the diet using a global LC-mass spectrometry approach. Compounds detected on the basis of their LC retention times and precursor ion masses were subjected to univariate and multivariate statistical analysis. In addition, the hydrophobic lipid components of the diet were examined using a comprehensive MSMS^ALL approach. Ions shown to be significantly changed by the γ-irradiation were analyzed by inspection of their accurate mass MS/MS spectra to obtain preliminary identification.

Materials and methods

Reagents

Solvents (acetonitrile, chloroform and methanol) were the best grades available. The lipid internal standard, C17 ceramide, was purchased from Avanti, Alabaster, AL.

Diets

Envigo (Indianapolis, IN) prepared a batch of NIH 7001 diet (composition provided in supplemental Table 1). Both non-irradiated (control) and irradiated diets were transferred to UAB for use in the chemoprevention experiments. The diets were stored at 4°C during the feeding experiments. During this time, aliquots for biochemical analysis were removed and stored at −20°C under N₂.

Diet processing for metabolomics analysis

LC-MS analysis

Ground, pelleted diet samples (1 g) were extracted in triplicate with ice-cold methanol-water (4:1, by volume) by shaking for 2 h under N₂. The mixtures were centrifuged at 3,000 × g for 10 min at 4°C. The supernatants were carefully decanted and taken to dryness under N₂. The dried extracts were re-suspended in 1 ml double-distilled H₂O and centrifuged at 14,000 × g for 10 min at 4°C.

Total Lipid analysis

Lipid extraction of the diets was carried out using the Bligh and Dyer method (6). Briefly, each ground diet sample (6 mg) was homogenized in water (1 mL). Methanol (2.5 ml), chloroform (1.25 ml) and internal standard (C₁₇ ceramide, 0.1 μg/mL, 10 μl) were added to the homogenate and sonicated for 10 s. Then an additional 1.0 ml water and 1.25 ml chloroform were added and samples shaken vigorously. Each sample was centrifuged and the lower chloroform layer was removed carefully. The upper layer was re-extracted following the same procedure. The chloroform-soluble fraction was evaporated to dryness under N₂ for analysis.

Metabolomics analysis

A clear aqueous solution of diet extract (5 μl) was injected onto an Eksigent reverse-phase C₁₈ pre-column cartridge (0.5 cm × 200 μm ID) in 0.1% formic acid. Metabolites bound to the cartridge were eluted with a 20 min linear gradient of 0–80% acetonitrile in 0.1% formic acid onto an Eksigent ChipLC C₁₈ column (15 cm × 200 μm ID) at a flow rate of 1 μl/min. The ChipLC column was placed in an Eksigent Nanoflex (SCIEX, Concord, ON, Canada) operating at 45°C. At the end of each analytical run, the column was washed with acetonitrile:0.1% formic acid for 1 min followed by re-equilibration with 0.1% formic acid for 4 min. Eluate from the ChipLC column was passed into the nanoelectrospray ionization interface of a Triple TOF™ 5600 Mass Spectrometer System operating in the negative and positive ion modes. The collision energy was set to 35 V, curtain gas to 20, GS1 and GS2 to 15, spray voltage to 5500 V (positive ion mode)/4500 V (negative ion mode), and temperature to 400°C. In each duty cycle lasting 1.25 s, high mass resolution MS spectra were collected for 250 ms followed by 50 ms MS/MS spectra of the top 20 most intense molecular ions. Ions producing successful MS/MS spectra were put onto an exclusion list for the next 90 s.

MS/MS^ALL analysis

Dried chloroform-methanol extracts were reconstituted with methanol:chloroform (2:1 v/v) with 5 mM ammonium acetate and directly infused at 5 μl/min into a TurboIon electrospray ionization interface and analyzed to record MS and MS/MS positive and negative ion spectra on a Triple TOF™ 5600 System. During MS/MS^ALL acquisition, precursor ion isolation windows of 1.2 Da width selected in Q1 are fragmented in the Q2 collision cell and the generated product ions are monitored at high mass resolution by the TOF analyzer. The MS/MS^ALL data included a 250 ms survey scan of TOF-MS data from m/z 200–1200, followed by collection of 1000 MS/MS spectra of the selected mass windows (7). The collision energy was set to 35 V, curtain gas to 20, GS1 and GS2 to 15, spray voltage to 5500 V (positive ion mode)/4500 V (negative ion mode), and temperature to 400°C. The total time to carry out the entire experiment was ~8 min.

Data analysis

Collected MS data were inspected using Mzmine 2.21 (8). Putative identities were obtained by submitting the ions to Metabosearch (http://omics.georgetown.edu/metabosearch.html). Raw MS data were uploaded as .wiff files to XCMSonline (https://xcmsonline.scripps.edu) in order to align metabolite ions across all of the diet extracts. A new feature of XCMSonline is Activity Network (Connections), based on the software program mummichog, which provides information on the pathways involved as well as (non-pathway) chemical groupings and calculates their statistical significance. XCMS processed data were downloaded from XCMSonline in .zip format. Once expanded, the data in the Excel report file were filtered to remove ions eluting before 5 min (not bound to the column) or after 25 min (only eluted by the high acetonitrile wash step at the end of the gradient) and ions with integrated peak areas less than 5,000. The data were divided into individual Excel .csv files for each sample containing the median m/z values, median retention times and peak areas, grouped into two folders (irradiated and non-irradiated samples) and uploaded as a .zip file to MetaboAnalyst 3.0 (http://www.metaboanalyst.ca/MetaboAnalyst/). The data were normalized using individual sample total ion currents, and subjected to mean centering and Pareto scaling. Accordingly, univariate (Volcano plots) and multivariate (Principal Components Analysis, PCA, and Partial Least Squares Discriminant Analysis, PLSDA) analysis of the data were carried out. For the latter, the metabolites contributing to the largest separation between the two groups were identified in principal component 1.

Results

Radiation-induced metabolomics changes in the diet

Using XCMSonline, a total of 317 features were observed in LC-MS data collected in the negative ion mode and 136 features in the positive ion mode that had undergone significant change (p<0.05 and fold change >1.5) between the irradiated and non-irradiated diets. After cleaning up the downloaded LC-MS data from XCMSonline to remove early (non-bound) metabolite ions and ions collected in the solvent wash period, a Volcano plot (Fig. 1) revealed 143 negatively charged ions and 94 positively charged ions with fold changes greater 1.5 and p-values <0.01. Principal Components Analysis (PCA) and Partial Least Square-Discriminant Analysis (PLS-DA) were performed to assess chemometric separation among the non-irradiated and irradiated diet samples. Metaboanalyst showed complete separation of the irradiated and non-radiated diets in PCA analyses (Fig. 2A and B).

Fig. 1 — Volcano plot in negative ion mode showing the statistical significance (y axis) and fold change (x axis) for difference between irradiated and non-irradiated diets.

Fig. 2 — Score plots of principle component analysis from control (NR) and irradiated (IR) samples in positive [A] and negative ion [B] modes. These plots display a clear separation between irradiated and control diet samples.

The molecular formulae of twenty-eight metabolites observed by nanoLC-MS that were significantly changed (mostly up-regulated in the irradiated samples) were determined based on their measured accurate mass and MS/MS product ion interpretation (Table 1).

Table 1.

Metabolite ions undergoing significant changes in irradiated diet

Ionization (−)
Fold change	Standard deviation	Retention time (min)	p-value	Experimental (m/z)	Empirical formula	Calculated (m/z)
3.00 up	0.136	14.20	0.0001	129.054	C₆H₁₀O₃	129.0557
2.73 up	0.078	7.24	0.00003	147.029	C₅H₈O₅	147.0299
1.31 up	0.019	11.95	0.00009	159.067	C₇H₁₂O₄	159.0657
1.59 up	0.073	14.27	0.0124	159.103	C₈H₁₆O₃	159.1021
6.9 up	0.980	20.87	0.0267	169.087	C₉H₁₄O₃	169.0870
1.5 down	0.051	19.87	0.0025	171.103	C₉H₁₆O₃	171.1027
1.8 up	0.023	14.93	0.00005	175.098	C₈H₁₆O₄	175.0976
2.11 up	0.065	16.02	0.00007	187.097	C₉H₁₆O₄	187.0976
3.74 up	0.223	17.81	0.00025	189.113	C₉H₁₈O₄	189.1132
2.49 up	0.036	18.93	0.000002	199.134	C₁₁H₂₀O₃	199.1340
4.22 up	0.150	14.19	0.00003	199.096	C₁₀H₁₆O₄	199.0970
3.46 up	0.159	20.70	0.00010	209.118	C₁₂H₁₈O₃	209.1183
4.56 up	0.128	14.33	0.00001	217.108	C₁₀H₁₈O₅	217.1081
3.0 up	0.039	15.70	0.00057	231.160	C₁₂H₂₄O₄	231.1602
27.4 up	0.652	17.67	0.000002	241.107	C₁₂H₁₈O₅	241.1081
2.83 up	0.091	14.85	0.000036	259.118	C₁₂H₂₀O₆	259.1187
2.03 up	0.064	10.75	0.00009	299.080	C₁₃H₁₆O₈	299.0772
2.49 up	0.048	12.62	0.000005	357.082	C₁₅H₁₈O₁₀	357.0827
2.79 up	0.055	10.84	0.000006	417.106	C₁₇H₂₂O₁₂	417.1038
13.0 up	0.393	8.07	0.000007	424.038	C₁₄H₁₉NO₁₀S₂	424.0378
3.27 up	0.264	20.93	0.0099	497.281	C₂₅H₄₂O₉	497.2751
2.80 up	0.196	12.33	0.00079	508.103	C₁₆H₃₁NO₁₁S₃	508.0986
6.00 up	0.092	13.82	0.000007	522.111	C₁₇H₃₂O₁₁NS₃	552.1143
2.36 up	0.193	15.48	0.0022	536.130	C₁₈H₃₅NO₁₁S₃	536.1299
2.89 up	0.060	14.92	0.000006	591.136	C₂₇H₂₈O₁₅	591.1350
Ionization (+)
5.74 up	0.187	17.75	0.0026	155.106	C₉H₁₄O₂	155.1067
5.90 up	0.748	17.61	0.0028	173.117	C₉H₁₈O₃	173.1172
3.62 up	0.505	14.29	0.0065	545.130	C₂₆H₂₄O₁₃	545.1290

Open in a new tab

The standard deviation was calculated from the pooled variance of the irradiated and non-irradiated groups which was then divided by the mean of the non-irradiated group. The p-value was generated by using Student’s T-test.

Effect of radiation on phytochemicals

Aqueous methanolic extracts of irradiated diet showed a series of peaks at Rt 8–16 min in nanoLC-MS analysis that corresponded to glucosinolates. Compared to the control diet, contents of these metabolites were significantly increased in the irradiated samples (Table 1).

The highest increase (13-fold) was noted with deprotonated molecular ion m/z 424.038 (Rt 8.07 min) with product ions m/z 96.961, 195.033, 259.014, 274.989 and 290.997 in its MS/MS spectra (Supplemental Fig. 1A). These product ions are highly specific for glucosinolates and are derived from the cleavage on either side of the thioether bond (9). The presence in the high-resolution MS/MS spectrum of m/z 424.038 also showed a neutral loss of 79.961 Da (SO₃) and an intense product ion m/z 96.961 [HSO₄]⁻, indicating the presence of a sulfate group. The exact mass for this ion corresponded to the elemental composition C₁₄H₁₉NO₁₀S₂ (calc. m/z 424.038) and is tentatively identified as a benzyl glucosinolate, most likely to be sinalbin, based on comparison of its MS/MS product ions with those of reference (9).

Similarly, three other compounds m/z 508.100, 522.112, and 536.130 were identified as methylsulfonyl alkyl chain containing glucosinolates on the basis of their high resolution MS/MS spectra. The MS/MS spectra of all these ions showed diagnostic product ions (m/z 290.981, 259.014, 96.961) of glucosinolates (10) (Supplemental Fig. 1B–D). The ion m/z 522.112, possibly one of the isomers of methylsulfonyl nonylglucosinolate, showed 6-fold increase (p = 0.001) in peak area in the irradiated sample, compared to control sample (Table 1). These results indicate that ionizing radiation increases the content of glucosinolates in the diet possibly by making them more available since the diet was pelleted and these compounds may have been encapsulated in non-digested solid particles. This is analogous to differences in bioavailability observed for drugs when prepared in different formulations.

A relatively polar metabolite ion m/z 299.080 (Rt 10.75 min) in negative ion mode had a prominent product ion m/z 137.025 due to the loss of 162.055 (characteristic of a hexose sugar) which on subsequent loss of 43.989 (CO₂) yielded the product ion m/z 93.036. These data are in agreement with an elemental composition for this compound as C₁₃H₁₆O₈ (calc. m/z 299.077), potentially a glycoside of salicylic acid. Another glycoside of phenolic acid m/z 357.088 [M-H]⁻ was observed to be up-regulated in the irradiated diet. This precursor ion yielded two intense product ions m/z 153.019 and m/z 109.030 due to subsequent losses of 204.069 and 43.989 Da. The loss of 204.069 may correspond to acetyl hexose sugar. The presence of dihydroxybenzoic acid moiety can be rationalized by the product ion m/z 153.019 (calc. m/z 153.019). These data suggested this compound is an acetyl hexoside of dihydroxybenzoic acid with molecular formula C₁₅H₁₈O₁₀ (calc. m/z 357.082).

A phenolic compound m/z 545.130 (Rt 14.29 min) in the positive ion mode showed a neutral loss of 248.053 Da generating the intense product ion m/z 297.077 in its MS/MS spectrum (Supplemental Fig. 2). The neutral loss of 248.053 Da indicated the presence of a malonyl hexosyl residue conjugated with a phenolic compound (11) with elemental composition C₁₇H₁₂O₅ (calc. m/z 297.076). Further loss of 42.011 Da (presumably an acetyl group) from the product ion m/z 297.077 produced m/z 255.066. These data suggested this compound is a malonyl conjugate of a polyphenol with molecular formula C₂₆H₂₄O₁₃ (calc. m/z 545.129). The content of this compound was significantly higher in the irradiated samples compared to the control diet.

Another ion m/z 591.136 [M-H]⁻ significantly increased in the irradiated samples compared to the control samples. It showed characteristic neutral losses of 144.042, 102.032 and 62.000 Da indicating the presence of a hydroxyl methyl-glutaryl (HMG) moiety (12). The neutral loss of 144.042 Da provided an abundant product ion m/z 447.093 which on further loss of a hexose sugar unit (162.052 Da) gave rise to the product ion m/z 285.042 (Fig. 3). Based on these data, the ion m/z 591.136 was tentatively identified as a flavonoid glycoside-HMG-conjugate with elemental formula C₂₇H₂₈O₁₅ (calc. m/z 591.1350).

Fig. 3 — nano-LC-MS/MS spectrum of *m/z* 591.136 [M-H]⁻ (Rt 14.92 min).

A relatively polar metabolite (Rt 7.24 min) with m/z 147.029 showed product ions m/z 129.020, 103.040 due to losses of H₂O and CO₂ in its MS/MS spectrum. Neutral losses of CH₃COOH and CO₂ plus H₂O resulted in the prominent product ions m/z 87.011 and 85.031, respectively. These data indicated this compound is most likely arabino-1,4-lactone with molecular formula C₅H₈O₅ (calc. m/z 147.029).

Lipid oxidation

Table 1 contains a number of oxidized lipids identified by in nanoLC-MS that were significantly altered by the irradiation. Enhancement of metabolite ion m/z 129.054 was observed in the irradiated diet, compared to the control diet (Table 1). Although precise identity of this ion remains unknown, loss of C₂H₄O (44.06 Da) from the precursor ion m/z 129.054 gave rise to m/z 84.985. These MS/MS data together with deprotonated molecular high resolution mass suggested this compound is probably α-ketoisocaproate (C₆H₁₀O_3; calc. m/z 129.056), arising from branched chain amino acids leucine or isoleucine oxidation.

A number of medium chain length oxidized fatty acids were detected with increased ion intensities in the irradiated samples. For example, isobaric ions m/z 159.067 (Rt 11.95 min) and 159.103 (Rt 14.27 min) had product ions due to neutral losses of H₂O, CO₂ and HCOOH (Supplemental Fig. 3A and B). Thus, these compounds were putatively identified as pimelic acid with molecular formula C₇H₁₂O₄ (calc. m/z 159.0657) and C₈H₁₆O₃ (calc. m/z 159.1021).

An ion m/z 169.087 (with Rt 20.8 min) had losses of CO₂ and HCOOH producing intense product ions m/z 125.098 and 123.082, respectively, in its MS/MS spectrum (Supplemental Fig. 4A). These characteristic losses together with high resolution mass of products and precursor ions suggested that this compound is an oxidized fatty acid with molecular formula C₉H₁₄O_3, (calc. m/z 169.087). The level of this compound increased 6.9 fold in the irradiated diet, compared to the control diet.

An analog of this compound with a difference of two mass units, i.e., m/z 171.103 (Rt 19.8 min) had product ions m/z 153.092, 127.113 and 125.098 due to losses of H₂O, CO₂ and HCOOH, respectively (Supplemental Fig. 4B). Inspection of its high resolution MS/MS spectra led to its assignment as a monohydroxynonenoic acid (C₉H₁₆O₃) (Table 1)

There were increased levels in the irradiated diet of a m/z 175.098 [M-H]⁻ precursor ion which had the product ions m/z 130.965 and 129.092 due to neutral losses of CO₂ and HCOOH, respectively, in its MS/MS spectrum. These along with additional, abundant product ions m/z 75.011 and 57.001 suggested that this compound is a dihydroxy-octanoic acid (C₈H₁₆O₄) (calc. m/z 175.098).

Two nonanoic acid derivatives m/z 187.097 (Rt 16.0 min) and 189.113 (Rt 17.81 min) were identified with molecular formulae C₉H₁₆O₄ and C₉H₁₈O₄, respectively, based on interpretation of their MS/MS spectra. The ion m/z 187.097 [M-H]⁻ yielded product ions m/z 169.087, 143.108 and 125.097 due to neutral losses of H₂O, CO₂ and H₂O plus CO₂, respectively. The MS/MS fragmentation pattern of the ion m/z 187.097 matched well with the Metlin database MS/MS record for nonanedionic acid (calc. m/z 187.098).

The other ion m/z 189.113 [M-H]⁻ showed losses of two molecules of H₂O (m/z 153.0930) and H₂O plus CO₂ losses giving rise to m/z 127.114. These data enabled us to propose the structure of this compound as either hydroxyl oxononenoic acid or dihydroxy-nonanoic acid (calc. m/z 189.113).

Two isobaric ions m/z 199.134 (Rt 18.93 min) and m/z 199.096 (Rt 14.19 min) both had losses of H₂O, HCOOH, and 72.021 Da, but with different ion intensities in their MS/MS spectra. The ion m/z 199.134 had weak product ions due to these losses, whereas m/z 199.096 showed significant product ions at m/z 181.086, 153.090, 127.113 and 125.063. In addition, it showed a prominent product ion m/z 171.102 [M-HCO(CH₂)₇COO⁻], indicating the presence a hydroxyl group at the C-9 position. Based on high resolution MS and MS/MS analyses, the ion m/z 199.134 was tentatively assigned as 9-hydroxy undecenoic acid C₁₁H₁₉O₃⁻ (calc. m/z 199.133). While the structure of the ion m/z 199.096 could not be proposed, it was an oxylipid with molecular formula C₁₀H₁₆O₄ (calc. m/z 199.097) (Table 1).

Another ion, m/z 209.118, whose intensity increased significantly on irradiation, had characteristic losses of oxidized fatty acids (H₂O, CO₂ and HCOOH), generating product ions m/z 191.106, 165.128 and 163.112, respectively. These data suggested it is a monohydroxy C₁₂ unsaturated fatty acid with molecular formula C₁₂H₁₈O₃ (calc. m/z 209.118).

An ion m/z 231.160 that was three-fold lower in the irradiated diet had product ions in its MS/MS spectrum due to the losses of H₂O and HCOOH. Based on the accurate masses of its precursor and product ions, it corresponds to a molecular formula of C₁₂H₂₄O₄ (calc. m/z 231.160).

We also observed increased levels of oxidized long-chain fatty acids in nanoLC-MS analysis of the irradiated samples. An oxidized C₁₈ fatty acid m/z 497.282 (Rt 20.93 min) was significantly increased in the irradiated diet and had an abundant product ion m/z 327.219, along with other product ions m/z 171.103, 155.108 and 137.097, in its MS/MS spectrum. Although precise structure of this compound is yet to be identified, C18:2–3O (m/z 327.2191) may be conjugated with C8:1–3O, consistent with the molecular formula C₂₆H₄₂O₉ (calc. m/z 497.275).

An oxidized fatty acid m/z 217.108 (Rt 14.33 min) had product ions m/z 199.097, 171.103, and 155.108 due to neutral losses of H₂O, HCOOH, and CO₂ plus H₂O. Because there are two consecutive losses of CO₂ (one from m/z 199.097 and other from m/z 155.108) (Supplemental Fig. 5), it is proposed to be an oxidized dicarboxylic acid with molecular formula C₁₀H₁₈O₅ (calc. m/z 217.108) (Table 1).

Similarly, a deprotonated ion m/z 241.107 (Rt 17.67 min) also showed two consecutive losses of CO₂ yielding the product ion m/z 153.129. A combined loss of H₂O and CO₂ resulted in the product ion m/z 179.106. These data suggested that this compound is a C₁₂ unsaturated fatty acid. The ionization of only one of the two acidic groups in C₃-C₁₂ dicarboxylic acids has been reported (13) and therefore, its formula was tentatively assigned as C₁₂H₁₈O₅ (calc. m/z 241.108) (Fig. 4A). The content of this compound was significantly higher (27-fold, p = 0.0004) in the irradiated diet, compared to the control diet (Table 1).

Fig. 4 — ESI-MS/MS spectra of oxidized lipids [A] *m/z* 241.107; [B] *m/z* 259.118 in negative ion mode.

The ion m/z 259.118 showed the similar fragmentation pattern (the loss of two CO₂ and two H₂O molecules) suggesting this compound is an oxidized dicarboxylic fatty acid with molecular formula C₁₂H₂₀O₆ (calc. m/z 259.119) (Fig 4B).

Reactive aldehydes

Two closely eluting peaks (Rt 17.75 and 17.61 min) observed in positive ion data, m/z 155.106 [M+H]⁺ and 173.117 [M+H]⁺, respectively, were significantly up regulated in the irradiated diet, compared to the control diet (Table 1). The ion m/z 155.106 had a product ion m/z 137.095, due to the loss of H₂O. Further loss of CO (−27.994 Da) from this ion gave rise to m/z 109.102 (Fig. 5A). A prominent product ion m/z 67.056 (C₅H₇⁺) was observed in the MS/MS spectrum of the ion m/z 155.106. Interpretation of these data enabled putative identification of this compound as C₉H₁₄O₂, most likely 4-oxo-2-nonenal (calc. m/z 155.107) (Table 1).

Fig. 5 — ESI-MS/MS spectra oxidized lipids [A] *m/z* 155.106; [B] *m/z* 173.117 in positive ion mode.

The other ion m/z 173.117 showed similar fragmentation pattern as seen in the ion m/z 155.106 with mass difference due to loss of H₂O, possibly a hydroxylated oxononanal derivative with molecular formula C₉H₁₆O₃ (calc. m/z 173.118) (Fig. 5B).

Lipid analysis by MS/MS^ALL

MS/MS^ALL analysis was performed to evaluate changes in relatively hydrophobic lipid composition in diets after irradiation. Contents of linoleic and linolenic acids in foods changed significantly after irradiation. A marked decrease (~55%) in linoleic acid (the major precursor for the formation of reactive aldehyde compounds and oxylipins) was observed in the irradiated samples, based on their ion intensities (Supplemental Fig. 6). With the decrease in linoleic acid content, levels of oxylipins such as oxygenated products of linoleic acid increased in the irradiated samples.

An ion m/z 295.229 had a prominent product ion m/z 277.217 due to the loss of a water molecule in its MS/MS spectrum (Fig. 6A). Two diagnostic product ions of the epoxide of linoleic acid, m/z 155.144 and 171.102, were observed along with m/z 183.139 and 195.138 (14). Based on product ion information, the precursor ion m/z 295.229 was considered as a mono-oxygenated product of linoleic acid, probably a mixture of epoxides of linoleic acid – leukotoxin A (9,10 epoxy-12-octadecaenoic acid) and leukotoxin B (12,13 epoxy-9-octadecaenoic acid) or 9-and 13-hydroxy octadecadienoic acid (HODE) (elemental composition C₁₈H₃₂O₃, calc. m/z 295.227).

An increased level of another oxylipin m/z 293.211 was also observed in the lipid extract of the irradiated diet. The MS/MS spectrum of this ion contained the product ions m/z 275.203 and 249.223 with losses of H₂O and CO₂, respectively (Fig. 6B). In addition to these ions, diagnostic product ions m/z 185.115 and 113.098 derived from 9-keto-octadecadienoic and 13-keto-octadecadienoic acids, respectively, were observed (15). These data indicated that the ion m/z 293.211 is a mixture of 9- and 13-keto-octadecadienoic acids (calc. m/z 293.212).

An oxygenation product of oleic acid (m/z 297.243) was significantly increased in the irradiated diet. The MS/MS spectrum of the ion m/z 297.243 contained the product ion m/z 279.231 along with other ions m/z 171.102, 155.106, 141.128, 183.011, and 127.112. These product ions are characteristic of oxygenation product of oleic acid (14). Based on these data, the ion m/z 297.243 could be a monooxygenation product with molecular formula C₁₈H₃₄O₃ (calc. m/z 297.244).

Irradiation also increased the levels of an ion m/z 311.223 which generated the product ions m/z 293.212, 267.192 and 275.201 due to neutral losses of H₂O, CO₂ and 2x H₂O, respectively, characteristic of oxidized lipids. The presence of the product ions m/z 171.101, 139.113 and 113.099 in its MS/MS spectrum is consistent with molecular formula C₁₈H₃₂O₄ (calc. m/z 311.223), most likely an epoxyalcohol of octadecenoic acid.

Similarly, ions m/z 327.216 and 329.232 were observed in higher intensities in the irradiated samples, compared to the controls. The ion m/z 327.216 showed a similar MS/MS fragmentation pattern (m/z 309.209, 291.194, 281.247 and 267.233) to those of m/z 297.243 (m/z 279.231, 261.223, 251.238 and 237.217), with characteristic mass difference by 30, indicating the addition of CHOH in the structure of m/z 297.243 (Supplemental Fig. 7A). Thus, the ions m/z 327.216 was tentatively identified as oxidized C18-fatty acids with molecular formulae C₁₈H₃₂O₅ (calc. m/z 327.218). MS/MS of the ion m/z 329.232 produced characteristic fragment patterns including ions m/z 229.143, 211.133, 171.102 and 139.112 (Supplemental Fig. 7B), indicating this compound is trihydroxy-octadecenoic acid (C₁₈H₃₄O₅ calc. m/z 329.233). The structure is also supported by the published reference (16).

When precursor ions containing a product ion m/z 295.224 was search on the extracted sample to detect precursors of C18:2-O (15), a series of ions appeared in the mass range of 500–900 Da (Supplemental Fig. 8A). For example, m/z 577.481 is one of the most abundant ions and its MS/MS spectrum contained the prominent product ions m/z 541.367, 295.228 (deprotonated monohydroxy C18:2, calc. m/z 295.228), 281.248 (deprotonated C18:1, calc. m/z 281.248) and 279.233 (deprotonated C18:2, calc. m/z 279.232) due to the neutral losses of 36.114, 282.225, 296.233 and 298.248 Da, respectively (Supplemental Fig. 8B). These data suggested that the ion m/z 577.482 is a mixture of oleic acid-dihydroxy oleic acid and linoleic acid-dihydroxy stearic acid with the molecular formula C₃₆H₆₆O₅ (calc. m/z 577.484). Structurally, they are likely to be similar to previously reported branched fatty acid esters of hydroxy fatty acids (FAHFAs) (17).

Further oxidation of this compound may give rise to another novel FAHFA with m/z 593.473 which had product ions m/z 281.247 (C18:1) and 311.223 (C18:2–2O) in its MS/MS spectrum, consistent with molecular formula C₃₆H₆₆O₆ (calc. m/z 593.479). We also detected m/z 625.469 showing a product ion m/z 589.451 due to the loss of 2xH₂O in its MS/MS spectrum. The presence of product ions m/z 327.217 (calc. m/z 327.218) and the ion m/z 297.244 indicated this compound to be an oxidized lipid with molecular formula C₃₆H₆₆O₈ (calc. m/z 625.469).

An ion m/z 487.328 [M-H]⁻ showed prominent product ions m/z 451.350, 329.232, 281.247, 279.233, 171.103 and 159.102 in its MS/MS spectrum (Supplemental Fig. 9). The presence of product ions m/z 329.232 and m/z 171.103 were indicative of an oxidized C₁₈ fatty acid. The observed accurate mass of the deprotonated molecular ion and MS/MS product ions, the molecular formula of this compound is proposed to be C₂₆H₄₈O₈ (calc. m/z 487.328). This is a novel lipid consisting of an oxidized long-chain fatty acid (C18:O-3O) esterified with medium chain fatty acid (C8:1–2O).

Discussion

Several changes in the profiles of dietary metabolites and lipids were observed in the irradiated samples. There was a significant difference in the content of phytochemicals between the irradiated and the control diet samples. Metabolomics data and pathway analyses are consistent with radiation-induced increases in oxidized forms of linoleate. A limitation of all these procedures was the formation of novel oxidized lipid metabolites not recognized by the databases.

We observed the marked increase in extractable glucosinolates in the irradiated diets. Glucosinolates are sulfur and nitrogen containing glycosides, widely found in cruciferous vegetables (10). Their hydrolyzed products, isothiocyanates, are well known chemopreventive agents against the development of cancers (18). Effect of radiation on glucosinolates of cabbage has previously been reported (19). The increased levels of glucosinolates in the irradiated diets may be responsible for breast cancer chemoprevention in rats, as the number of mammary tumors in rats were 40% less than when using non-irradiated diets (Clinton Grubbs, unpublished results).

Up-regulation of glucosinolates upon irradiation may be possible through inactivation of myrosinase enzyme, causing intact glucosinolates to remain in the diet. Other phytochemicals such as polyphenol glycosides containing a malonyl or hydroxyl methyl-glutaryl moieties were significantly increased in the irradiated diet. Gamma rays can also penetrate cells and may cause the breakdown of middle lamella in cell wall thereby releasing these compounds (20,21). This suggests that gamma irradiation may alter the pharmaceutical state of solid food.

We observed significant differences in the levels of PUFAs and their oxygenation products between the control and irradiated diet. Linoleic acid and α-linolenic acid are the main polyunsaturated fatty acids in plant-based diets. Irradiation of diet will lead to formation of reactive oxygen species that promote lipid oxidation in foods rich in these fatty acids (22). Oxylipins are a diverse family of oxygenated fatty acids that are produced by oxidative metabolism, either enzymatic or non-enzymatic of polyunsaturated fatty acids such as linoleic acid and linolenic acid (22). Chemically-derived, racemic mixtures of oxylipins are formed and may be esterified with other oxidized lipids.

In our study, irradiation increased the content of a number of oxidized lipids, particularly, linoleic acid-derived oxidation products containing multiple combination of hydroxyl, epoxide, and or ketone moieties. A mixture of 13- and 9- HODE is one of the possibilities which is synthesized in vivo from 18:2 released from the endogenous triacylglycerol pool (23). Effects of storage and Gamma-irradiation on metabolic pattern, including linoleic acid metabolism in red blood cells has previously been reported (24).

We also observed a series of novel oxidized lipids with a branched ester linkage between a fatty acid and a hydroxyl fatty acid or oxidized lipids (Supplemental Fig. 8 and 9). While biology of these lipids are unknown, branched fatty acid esters of hydroxyl fatty acids, such as palmitic-acid-9-hydroxy-stearic acid, are implicated in insulin sensitivity and are found in reduced levels in adipose tissue and serum of insulin-resistant humans (17). Although previously reported as endogenous lipids, our results demonstrate that irradiation increases levels of branched fatty acid esters of hydroxylated fatty acids in the diet as well. Emerging studies have shown that oxylipins derived from linoleic acid can affect a broad range of physiological effects such as inflammation, cell proliferation and vascular function (25) and these may be implicated in cancer risk.

While infection, wounding and stress are known to induce oxidized lipids by biochemical enzymatic processes (26), effects of radiation on oxylipin formation in diet is unknown. Dietary lipid oxidation is one of the major problems in maintaining food quality (27). While in vitro studies on biological effects of individual oxylipins have been reported, effects of dietary oxylipins and or other oxidized lipids which exist together with many other nutrients and phytochemicals on health and diseases such as breast cancer are yet to be evaluated in vivo.

Conclusions

Our studies clearly indicate that Cobal-60 irradiation induces modifications of animal diet components in terms of content and or chemical structures. Several peaks corresponding to phytochemicals and oxidized fatty acids of animal diet that had been modified (mostly upregulated) by irradiation and they were identified at the level of their molecular formulae. However, the biological significance of the food metabolite modification after irradiation is still unknown. Treatment of rodent diets with Gamma-irradiation is not benign. Despite reducing the microbiome content of the food, it leads to increased oxidation of lipids and altered component availability. Surprisingly, it led to a reduction in carcinogenesis in a rodent model of breast cancer. Future studies will examine which of the radiation generated-components contribute to the changes in cancer risk.

Supplementary Material

Supp FigS1-6

Fig. 1. ESI-MS/MS spectra of glucosinolates [A] m/z 424.0380; [B] m/z 508.1003; [C] m/z 522.112; [D] m/z 536.131 in negative ion mode.

Fig. 2. nano-LC-MS/MS spectrum of m/z 545.130 [M+H]⁺ (Rt 14.29 min).

Fig. 3. nano-LC-MS/MS spectra of m/z 159.067 [A]; m/z 159.103 [B] in the negative ion mode.

Fig. 4. ESI-MS/MS spectra of oxidized lipids [A] m/z 169.080; [B] m/z 171.102 in negative ion mode.

Fig. 5. nano-LC-MS/MS spectrum of m/z 217.108 [M-H]-(Rt 14.33 min).

Fig. 6. Effect radiation on linoleic acid (C18:2) and linolenic acid (C18:3) in the diet. Values are expressed as mean ± SD (n=3).

Fig. 7. ESI-MS/MS spectrum of m/z 327.216 [A]; m/z 329.233 [B] in the negative ion mode.

Fig. 8. An extracted ion chromatogram (XIC) generated from product ion m/z 295.224 (18:2-OH) from an extracted diet sample for oxidized lipids [A]; ESI-MS/MS spectrum of m/z 577.482 in the negative ion mode [B].

Fig. 9. ESI-MS/MS spectrum of m/z 487.328 in the negative ion mode.

NIHMS877965-supplement-Supp_FigS1-6.pptx^{(241.8KB, pptx)}

Supp TableS1

NIHMS877965-supplement-Supp_TableS1.docx^{(26.2KB, docx)}

Acknowledgments

The mass spectrometers used for the analysis were purchased from funds from a NIH Shared Instrumentation Grant (S10 RR027822, SB, PI) and from the UAB Health Systems Foundation General Endowment Fund (SB, PI). Support for the purchase of the Eksigent nanoLC system came from the Office of the Vice-President for Research and Economic Development, the Deans of the School of Medicine and the College of Arts and Sciences, the UAB-UCSD O’Brien Acute Kidney Injury Center (P30 DK0739037, Anupam Agarwal, PI), and the Gene Bartow Fund for Cancer Research.

Footnotes

This work was presented in part at the 64^th Annual Conference of the American Society for Mass Spectrometry in San Antonio, Texas, June 5–9, 2016 and appeared in abstract form in the Proceedings of the Conference.

Conflict of interest

The authors declare that there are no conflicts of interest.

References

1.American Institute of Nutrition. Report of the American Institute of Nutrition Ad Hoc Committee on Standards for Nutritional Studies. J Nutr. 1977;107:1340–1348. doi: 10.1093/jn/107.7.1340. [DOI] [PubMed] [Google Scholar]
2.Reeves PG, Nielsen FH, Fahey GC. AIN-93 purified diets for laboratory rodents: final report of the American Institute of Nutrition Ad Hoc Writing Committee on the Reformulation of the AIN76A Rodent Diet. J Nutr. 1993;123:1939–1951. doi: 10.1093/jn/123.11.1939. [DOI] [PubMed] [Google Scholar]
3.Barnes S, Grubbs C, Setchell KDR, Carlson J. Soybeans inhibit mammary tumors in models of breast cancer. Prog Clin Biol Res. 1990;347:239–253. [PubMed] [Google Scholar]
4.Lamartiniere CA, Cotroneo MS, Fritz WA, Wang J, Mentor-Marcel R, Elgavish A. Genistein chemoprevention: timing and mechanisms of action in murine mammary and prostate. J Nutr. 2002;132:552S–558S. doi: 10.1093/jn/132.3.552S. [DOI] [PubMed] [Google Scholar]
5.Kim H, Hall P, Smith M, Kirk M, Prasain JK, Barnes S, Grubbs C. Chemoprevention by grape seed extract and genistein in carcinogen-induced mammary cancer in rats is diet-dependent. J Nutr. 2004;134:3445S–3552S. doi: 10.1093/jn/134.12.3445S. [DOI] [PubMed] [Google Scholar]
6.Bligh EG, Dyer WJA. Rapid method of total lipid extraction and purification. Can J Biochem Physiol. 1959;37:911–917. doi: 10.1139/o59-099. [DOI] [PubMed] [Google Scholar]
7.Prasain JK, Wilson L, Hoang HD, Moore R, Miller MA. Comparative Lipidomics of Caenorhabditis elegans Metabolic Disease Models by SWATH Non-Targeted Tandem Mass Spectrometry. Metabolites. 2015;5:677–696. doi: 10.3390/metabo5040677. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Pluskal T, Castillo S, Villar-Briones A, Orešič M. MZmine 2: Modular framework for processing, visualizing, and analyzing mass spectrometry-based molecular profile data. BMC Bioinformatics. 2010;11:395. doi: 10.1186/1471-2105-11-395. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Fabre N1, Poinsot V, Debrauwer L, Vigor C, Tulliez J, Fourasté I, Moulis C. Characterisation of glucosinolates using electrospray ion trap and electrospray quadrupole time-of-flight mass spectrometry. Phytochem Anal. 2007;18:306–319. doi: 10.1002/pca.983. [DOI] [PubMed] [Google Scholar]
10.Lin LZ, Sun J, Chen P, Zhang RW, Fan XE, Li LW, Harnly JM. Profiling of glucosinolates and flavonoids in Rorippa indica (Linn.) Hiern. (Cruciferae) by UHPLC-PDA-ESI/HRMS(n) J Agric Food Chem. 2014;62:6118–6129. doi: 10.1021/jf405538d. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Sun J, Liu X, Yang T, Slovin J, Chen P. Profiling polyphenols of two diploid strawberry (Fragaria vesca) inbred lines using UHPLC-HRMS(n.) Food Chem. 2014;146:289–298. doi: 10.1016/j.foodchem.2013.08.089. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Rodríguez-Rivera MP, Lugo-Cervantes E, Winterhalter P, Jerz G. Metabolite profiling of polyphenols in peels of Citrus limetta Risso by combination of preparative high-speed countercurrent chromatography and LC-ESI-MS/MS. Food Chem. 2014;158:139–152. doi: 10.1016/j.foodchem.2014.02.077. [DOI] [PubMed] [Google Scholar]
13.Grossert JS, Fancy PD, White RL. Fragmentation pathways of negative ions produced by electrospray ionization of acyclic dicarboxylic acids and derivatives. Can J Chem. 2005;83:1878–1890. [Google Scholar]
14.Goyal S, Banerjee S, Mazumdar S. Oxygenation of monoenoic fatty acids by CYP175A1, an orphan cytochrome P450 from Thermus thermophilus HB27. Biochemistry. 2012;51:7880–7890. doi: 10.1021/bi300514j. [DOI] [PubMed] [Google Scholar]
15.Oliw EH, Garscha U, Nilsson T, Cristea M. Payne rearrangement during analysis of epoxyalcohols of linoleic and alpha-linolenic acids by normal phase liquid chromatography with tandem mass spectrometry. Anal Biochem. 2006;354:111–126. doi: 10.1016/j.ab.2006.04.010. [DOI] [PubMed] [Google Scholar]
16.Park SK, Ha JS, Kim JM, Kang JY, du Lee S, Guo TJ, Lee U, Kim DO, Heo HJ. Antiamnesic Effect of Broccoli (Brassica oleracea var. italica) Leaves on Amyloid Beta (Aβ)1–42-Induced Learning and Memory Impairment. J Agric Food Chem. 2016;64:3353–3361. doi: 10.1021/acs.jafc.6b00559. [DOI] [PubMed] [Google Scholar]
17.Yore MM, Syed I, Moraes-Vieira PM, Zhang T, Herman MA, Homan EA, Patel RT, Lee J, Chen S, Peroni OD, Dhaneshwar AS, Hammarstedt A, Smith U, McGraw TE, Saghatelian A, Kahn BB. Discovery of a class of endogenous mammalian lipids with anti-diabetic and anti-inflammatory effects. Cell. 2014;159:318–332. doi: 10.1016/j.cell.2014.09.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Conaway CC, Yang YM, Chung FL. Isothiocyanates as cancer chemopreventive agents: their biological activities and metabolism in rodents and humans. Curr Drug Metab. 2002;3:233–255. doi: 10.2174/1389200023337496. [DOI] [PubMed] [Google Scholar]
19.Banerjee A, Variyar PS, Chatterjee S, Sharma A. Effect of post-harvest radiation processing and storage on the volatile oil composition and glucosinolate profile of cabbage. Food Chem. 2014;151:22–30. doi: 10.1016/j.foodchem.2013.11.055. [DOI] [PubMed] [Google Scholar]
20.Choe E, Min DB. Chemistry and reactions of reactive oxygen species in foods. Crit Rev Food Sci Nutr. 2006;46:1–22. doi: 10.1080/10408390500455474. [DOI] [PubMed] [Google Scholar]
21.Kovács E, Keresztes A. Effect of gamma and UV-B/C radiation on plant cells. Micron. 2002;33:199–210. doi: 10.1016/s0968-4328(01)00012-9. [DOI] [PubMed] [Google Scholar]
22.Gabbs M, Leng S, Devassy JG, Monirujjaman M, Aukema HM. Advances in Our Understanding of Oxylipins Derived from Dietary PUFAs. Adv Nutr. 2015;6:513–540. doi: 10.3945/an.114.007732. [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Haas TA, Bertomeu MC, Bastida E, Buchanan MR. Cyclic AMP regulation of endothelial cell triacylglycerol turnover, 13-hydroxyoctadecadienoic acid (13-HODE) synthesis and endothelial cell thrombogenicity. Biochim Biophys Acta. 1990;1051:174–178. doi: 10.1016/0167-4889(90)90190-o. [DOI] [PubMed] [Google Scholar]
24.Patel RM, Roback JD, Uppal K, Yu T, Jones DP, Josephson CD. Metabolomics profile comparisons of irradiated and nonirradiated stored donor red blood cells. Transfusion. 2015;55:544–52. doi: 10.1111/trf.12884. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Vangaveti V, Baune BT, Kennedy RL. Hydroxyoctadecadienoic acids: novel regulators of macrophage differentiation and atherogenesis. Ther Adv Endocrinol Metab. 2010;1:51–60. doi: 10.1177/2042018810375656. [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Vu HS, Tamura P, Galeva NA, Chaturvedi R, Roth MR, Williams TD, Wang X, Shah J, Welti R. Direct infusion mass spectrometry of oxylipin-containing Arabidopsis membrane lipids reveals varied patterns in different stress responses. Plant Physiol. 2012;158:324–339. doi: 10.1104/pp.111.190280. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Kanner J. Dietary advanced lipid oxidation end products are risk factors to human health. Mol Nutr Food Res. 2007;51:1094–101. doi: 10.1002/mnfr.200600303. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials