Method details

Selection of Food Items

Approximately 220 items of representative foods in Japan were selected among the foods listed in the Standard Tables of Food Composition in Japan -2015 (Seventh Revised Version) published by the Ministry of Education, Culture, Sports, Science and Technology, Japan. Approximately 2,000 food items, including variation of cooking states, are listed in the standard table. Among them, we selected ~220 items that are easily available in local markets and commonly consumed in Japan as an initial set of target foods.

Sample Preparation

Foods without 'Removed Portion' were ground into fine powder in liquid nitrogen. 'Removed Portion' described in the 'Remarks' in the standard table were removed. Solid samples were ground into fine powder in liquid nitrogen with mortal and pestle. When food size is large, a representative portion is cutout and ground. For example, in the case of a water melon, a portion of comb cut where the edge is along to the line between the petiole and the tip of the fruit is used to minimize the deviation of maturity in the fruit. Liquid samples are used without grinding.

Extraction

Compounds in foods were extracted by approximately 75%(v/v) methanol. In the case of liquid samples and powders from foods containing abundant water (such as vegetables), 3 times volume (v/v or w/v) of 100% methanol were added. In the case of samples with low water content, such as dried seaweed, 40 times volume (w/v) of 75% methanol were added. We did not measure the moisture content of each food. A 25 uM of 7-hydroxy-5-methylflavone is contained as internal standard (IS) in the added methanol. Strong peaks always detected around 73 min in our LC condition are derived from the IS.

The sample in methanol in 2 mL tubes was homogenized using Mixer Mill MM 300 (QIAGEN) at 25 Hz, for 2 min, twice. A homogenate was centrifuged under 17,400 x g, 5 min at 4 C. A supernatant was passed through a Polytetrafluoroethylene (PTFE) filter (pore size 0.2 um, Millipore), and a filtrate was applied to a C18 silica column (MonoSpin C18, GL Science) to remove highly-hydrophobic contaminants. The filtrate passed through the C18 column was used for LC-MS analyzes.

The same extraction procedure with 75% methanol were performed without sample to prepare mock samples and used as negative controls.

Liquid chromatography (LC)-mass spectrometry (MS) analysis

A LC separation with a longer elution time and a high-resolution mass spectrometry was conducted. Agilent 1100 system (Agilent) and Finnigan LTQ-FT (Thermo Fisher Scientific) were used. Aliquots (5 uL) of the methanol extract were applied to a TSK-gel ODS-100V column (4.6 x 250 mm, 5 um, TOSO), and separated by water containing 0.1%(v/v) formic acid (Solvent A) and acetonitrile containing 0.1%(v/v) formic acid (Solvent B). The gradient program was as follows: 3% B (0 min), 97% B (90 min), 97% B (100 min), 3% B (100.1 min), and 3% B (107 min). The flow rate was set at 0.25 mL/min for 0-100 min, and 0.5 mL/min for 100.1-107 min.

Each food extract was analyzed in three conditions as below.

Method Name	Description
Method 1 Positive	A ESI positive mode analysis for measurement of accurate mass and for obtaining MS/MS spectra. Full scan: resolution = 100,000 by FT MS2 scan: data dependent scan by IT for the most intense 5 ions in the full scan
Method 5 Positive	A ESI positive mode analysis for obtaining MS3 spectra. Full scan: resolution = 12,500 by FT MS2 scan: data dependent scan by IT for the most intense 5 ions in the full scan MS3 scan: data dependent scan by IT for the most intense 2 ions in the MS2 scan
Method 1 Negative	A ESI negative mode analysis for measurement of accurate mass and for obtaining MS/MS spectra. The mass conditions are same as those in the Method 1 Positive.
ESI: electrospray ionization, FT: Fourier transformation ion cyclotron resonance type mass spectrometry, IT: ion trap type mass spectrometry. m/z range for Full scan by FT: 100-1500 The dynamic exclusion conditions for MS2 scan: repeat count, 3; repeat duration, 30s; exclusion list size, 500; margin, 10 ppm, exclusion duration, 20s

The Method 5 Positive was omitted form analysis for the mock samples. Raw data were obtained by Xcalibur (Thermo Fisher Scientific).

Data processing

PoweGetBatch was used for peak detection and database searching. FlavonoidSearch was used for predicting flavonoid aglycones. The PowerGetBach software, which is good at to detect whole peaks including lower intensity ones and therefore good at to predict appropriate adducts, was used.

Peak Detection

Three sets of peak detection parameters were used for the Method 1 Positive/Negative data of foods and a single set of parameters was applied for each Method 5 Positive data of foods and Method 1 Positive/Negative data of mock samples. The peaks detected with these sets of parameters, namely Method 1 x 3 and Method 5 x 1 for a food, and Method 1 x several repetition of mock samples for positive mode, and Method 1 x 3 for a food and Method 1 x several repetition of mock samples for negative mode, were aligned for each mode using PowerGetBatch. Peaks in the following criteria were selected as 'valid' peaks: The peak is detected in all three sets of parameters for Method 1; The peak is never detected in mock samples. The most intense and major pattern of the MS2 spectrum and MS3 spectra for two different precursors were selected among the alignment results for each peak.

Database Searching

Compound database searching was performed using the accurate mass value and the estimated adduct. Following databases are used: KEGG, KNApSAcK, Human Metabolome Database (HMDB), LipidMAPS, flavonoid database in metabolomics.jp. The UC2 database in the MFSearcher web service was used for rapid cross-database searching and compiling the constitutional isomers in one record.

Please note that different stereoisomers are included in one record. The marks such as '[-1fr]' shown in the DB search results in the peak information page represents the states of the molecule in the original database as follows: The number, the charge of the molecule, 'f', a fragment in the molecule registered as multiple compartments (such as salts) was hit, 'r', the molecule was registered as a radical.

FlavonoidSearch

Prediction of flavonoid aglycones were conducted by FlavonoidSearch using MS2 and MS3 spectra in positive mode. A possibility of the peak being a flavonoid is high when a higher hit score, usually larger than 0.3, is observed. Candidates of known flavonoids are listed in the FlavonoidSearch results.

FlavonoidSearch calculates similarities of query spectra to those of virtual mass fragment database which is constructed based on knowledge and heuristics of CID. The virtual mass fragment data were constructed approximately 4500 flavonoids among approx. 7000 known flavonoids. No virtual mass fragment data were attached to the rest of approx. 2500 flavonoids. If the precursor mass was hit to these no virtual fragment ones, FlavonoidSearch returns score zero. Therefore, results with score zero remained in the result table to show the possibility of them.

More Details of the analytical methods

Visit Metabolonote website for further details. Mock analyses are shared among several food samples which are measured together in a same period. These samples are registered in the same sample set (SE) in Metabolonote record. If you need, please visit Metabolonote by clicking link button at the bottom of the peak list page of a food (Example is here).