Selective and sensitive UHPLC-ESI-Orbitrap MS method to quantify protein oxidation markers

A targeted UHPLC-MS/MS isotopic dilution method has been developed for the simultaneous quantification of 18 different free and protein-bound aromatic amino acid oxidation products in food and biological matrices. All analytes, including critical isomeric pairs of Tyr, o-Tyr, m-Tyr, and dioxyindolylalanine diastereomers were chromatographically resolved to obtain high selectivity, without the need for derivatizing or ion pairing agents. The results of method validation showed adequate retention time reproducibility [0.1–0.6% coefficient of variation (CV) for over 224 injections], accuracy (within ±1–20% of the nominal concentration), and precision (1–17% CV) for all target analytes. The lower limit of quantification was calculated in different matrices using both the Hubaux-Vos approach and accuracy and precision data showing values in the range of 0.2–15 ng/mL. Use of stable isotope-labelled internal standards compensated errors due to matrix effects and artefactual degradation of analytes. Both acid and enzymatic hydrolyses were tested to obtain the best possible results for the quantification of protein oxidation products, demonstrating the stability of target analytes under hydrolytic conditions. The method was successfully applied to quantify target analytes in serum, tissue, milk, infant formula, pork liver pâté, chicken meat and fish. The method was also applied to assess the role of Fenton’s reagent in oxidizing Trp, Phe and Tyr residues in different proteins, with results showing o-Tyr, dioxyindolylalanine diastereomers, kynurenine, dityrosine being the main oxidation products. The Fenton chemistry favored the formation of o-Tyr over m-Tyr from Phe with 2–36 folds higher yields. 3-Nitrotyrosine, a marker of protein nitration, was also detected in samples treated with Fenton’s reagent.


Introduction
Proteins undergo oxidation in the presence of reactive oxygen species (ROS) such as singlet oxygen ( 1 O 2 ), superoxide (O 2 •− ), peroxides (H 2 O 2 ), hydroxyl radical (•OH), and hypochlorous acid (HOCl). Protein oxidation leads to the formation of a diverse group of stable and unstable oxidation products, depending on the amino acid residue and the type of oxidant involved. Since proteins are an integral part of the human body and diet, the phenomenon of protein oxidation is investigated both in biological and food science disciplines. In vivo protein oxidation occurs due to various endogenous (redox enzymes and metal ions) and exogenous factors (UV light and ozone) [1]. Similar factors are responsible for protein oxidation in food systems, but are mostly investigated on the basis of food composition, processing and storage [2]. Protein oxidation in vivo is associated with maintaining cellular redox homeostasis [3] and various pathological conditions [4]. In foods, it is associated with change in texture, flavour and nutritional quality [5]. There is a growing interest among food, nutrition, and biological scientists in understanding the role of dietary protein oxidation products on human health [6], but conclusive evidence remains to be established. The catalytic disproportionation of H 2 O 2 by transition metal ions, known as Fenton chemistry, is one of the major reactions involved in the production of highly reactive •OH in vivo and in vitro. •OH can attack all amino acid residues, where greater rates for the attack was observed with Cys, Met, His, Phe, Tyr and Trp [1,7]. Among these amino acids, oxidation of aromatic amino acid residues (Trp, Tyr, and Phe) is a topic of great scientific interest in food and health science disciplines as they form numerous relatively stable oxidation products [4,8,9]. Oxidation of Trp residues leads to the formation of N ′ -formylkynurenine (NFK), kynurenine (Kyn), 3-hydroxykynurenine (3-OH-Kyn), dioxindolylalanine (DiOia) and 5-hydroxytryptophan (5-OH-Trp) as known major oxidation products [10,11]. Similarly, oxidation of Phe leads to the formation of o-Tyr and m-Tyr. 3,4-Dihydroxyphenylalanine (DOPA) and dityrosine (DiTyr) are two well-characterised oxidation products of Tyr [10]. Similar to ROS, reactive nitrogen species (RNS), such as ONOOH and •NO 2 , can react with amino acid residues, leading to protein nitration. One of the known protein nitration markers is 3-nitrotyrosine (3-NO 2 -Tyr). In meat, 3-NO 2 -Tyr is formed either due to endogenous RNS or exogenous nitrating compounds added during food processing (e.g. nitrate salts added during curing of meat) [12]. Furthermore, non-radical oxidants such as HOCl can be formed in vivo, which can chlorinate Tyr residues, leading to the formation of 3-chlorotyrosine (3-Cl-Tyr) [13]. 3-Cl-Tyr may also be formed in foods, provided HOCl is used as a disinfectant during food processing [14]. Free amino acids can also undergo oxidation similar to protein amino acid residues, forming free amino acid oxidation products, although they are less abundant [1]. Free forms of 5-OH-Trp, 3-OH-Kyn, NFK, tryptamine (Tra) and Kyn could also be formed from Trp catabolism in vivo [15,16].
Historically protein oxidation in food and biological matrices is assessed by quantifying generic markers of protein oxidation, mainly protein carbonyls by 2,4-dinitrophenylhydrazine assay [17]. Although such assays can estimate the extent of protein oxidation, it is difficult to conclude on the mechanism of protein oxidation as these methods have poor specificity and may produce biased results due to interfering Maillard reaction or lipid oxidation-derived carbonyl compounds present in the matrices [18]. Several HPLC methods have been developed eventually to quantify specific oxidation products. Aromatic amino acid oxidation products such as o-Tyr, m-Tyr, NFK, Kyn, and DiTyr exhibit absorption and emission properties, therefore, can be detected using HPLC systems coupled to UV and fluorescence detector (FLD) [19]. HPLC coupled with electrochemical detection (ECD) is often used to detect Kyn and its derivatives [20]. However, oxidation products are generally present in far lower levels compared to parent amino acids in most food and biological systems (nmol vs mmol), therefore, the sensitivity of HPLC-UV/FLD/ECD methods becomes insufficient to quantify these analyte. These methods also suffer from low selectivity due to co-eluting endogenous interfering compounds with similar structural and absorption/emission features [21]. In addition, comprehensive analysis of different oxidation products is not possible by UV and FL detectors. For instance, Phe cannot be detected using UV and FLD detection due to its low molar absorptivity, while 3-NO 2 -Tyr and 3-Cl-Tyr cannot be detected using FLD as they do not exhibit characteristic emission properties. Often a combination of multiple detector systems or independent analysis is used to quantify protein oxidation products [19,22].
UHPLC-MS/MS methods are inherently sensitive and selective compared to spectrophotometric or HPLC-UV/FLD/ECD based methods. Only a limited number of validated methods are available for the quantification of individual oxidation products using HPLC-MS, wherein only a selected number of oxidation products were targeted [23,24]. Considering the importance of aromatic amino acid residues and their oxidation products in the human biology and diet, we have developed a UHPLC-ESI-(Orbitrap)-MS/MS method for their analysis. The aim was to enable reliable quantification of Tyr, Trp and Phe and their oxidation products in the single chromatographic run using isotopic dilution assay. Met was also included in the method to assess loss of Met residues due to oxidation. A comprehensive validation covering sensitivity, selectivity, linearity, accuracy, precision, stability, and matrix effects is also provided. The developed method was applied to quantify target analytes in several food and biological matrices as well as to study the role of Fenton chemistry in oxidation of proteins. The method could be used to understand the mechanism of protein oxidation in foods and biological systems as well as to assess the metabolic fate of dietary oxidation products in future nutritional investigations.

Biological and food materials
Mice serum and mice small intestinal tissue were obtained from Department of Veterinary and Animal Sciences, University of Copenhagen, Denmark. Roasted chicken breast (30.5% w/w protein), baked pork liver pâté (33% w/w liver, 10% w/w protein), ultrahigh temperature treated (UHT) liquid infant formula (1.2% w/w protein, stored at room temperature for 3 months), and fresh trout fish (21% w/w protein) were purchased from a local supermarket in Copenhagen. Lactosehydrolysed UHT milk (3.4% w/w protein) was obtained from Arla Foods, Viby, Denmark. Blank, calibration and quality control (QC) standards: Water without SIL IS served as blank. Calibration standards (calibrators) were prepared by diluting known volumes of SD mix in water (see Table 1). Four different QCs were used during the method development, namely LLOQ QC, low QC, mid QC, and high QC, with concentrations equal to LLOQ, less than or equal to three times the LLOQ, near mid-range of the calibration curve, and near a high-range of the calibration curve, respectively (see Table 1). Freshly prepared standards and QCs were used in all experiments.

Preparation of samples
The method was validated to analyze both free and protein-bound forms of target analytes. Mice serum and small intestinal tissue were chosen to validate the analysis for free-form analytes, while chicken, pork liver pâté, milk, and infant formula were used for protein-bound forms after total hydrolysis of proteins. Accordingly, different sample preparation techniques were employed as described below.
Mice serum: Thirty microliters of serum was vortexed with 500 μL of ice-cold methanol and kept in a freezer (− 20 • C) overnight for protein precipitation. The sample was centrifuged at 15,000 g for 20 min at 4 • C.
The supernatant (460 μL) was collected and dried under vacuum using a centrifugal vacuum concentrator (Speedvac, Thermo Fisher Scientific, MA, USA) for 1.5 h at room temperature. The residue was dissolved in water (200 μL) and used for LC-MS analysis.
Mice small intestinal tissue: Thirty milligram of lyophilized tissue was vortexed with 800 μL of 25% ice-cold methanol containing 0.1% formic acid. The mixture was homogenized using an ultrasonic homogenizer for ca. 2 min and kept in a freezer for 1.5 h. Samples were centrifuged at 15,000 g for 30 min at 4 • C. The supernatant, without sampling the top fat layer, was used for the LC-MS analysis.
Roasted chicken breast, baked pork liver pâté, and fish: These samples were defatted before enzymatic hydrolysis of proteins. Briefly, 10-20 mg of sample (corresponding to 3-4 mg of protein) was vortexed with 2 mL of ice-cold defatting reagent (water: methyl tert-butyl ether: methanol, 1:5:4 v/v/v). The mixture was kept in a freezer (− 20 • C) for 2 h and centrifuged at 15,000 g for 30 min at 4-8 • C. The supernatant was discarded carefully and the pellet was dried under vacuum for 15 min at room temperature. The dry pellet was used for hydrolysis. For acid hydrolysis, samples were defatted using trichloroacetic acid (TCA) (see section Acid hydrolysis below and [19]).
LH UHT Milk and infant formula: Milk and infant formula samples were centrifuged at 15,000 g for 10 min to remove the fat layer and then subjected to enzymatic hydrolysis, while they were defatted using TCA before acid hydrolysis (see section Acid hydrolysis below and [19]).

Total enzymatic hydrolysis of proteins
Enzymatic hydrolysis of proteins was carried out using multiple proteases according a procedure described elsewhere with some modifications [25]. Samples equivalent to 3-4 mg of proteins were weighed into 2 mL Eppendorf tubes (n = 3) and 1000 μL of 0.02 M HCl was added.
The sample was allowed to solvate for 2 h at 4 • C and sonicated for 1 min using an ultrasonic homogenizer. Thereafter, 62 μL of pepsin solution (0.91 mg/mL, 200 U/sample) prepared in 0.02 M HCl was added to each tube and vortexed for 30 s. Samples were agitated for 24 h at 37 • C by Table 1 Concentrations of the calibration standards (C) and QCs in ng/mL. ULOQ is upper limit of quantification that was used to plot calibration curves and ULOQ' is the upper limit of quantification that could be used with acceptable accuracy and precision.

Acid hydrolysis of proteins
Acid hydrolysis of proteins (n = 3) was carried out in vacuo using 4 M MSA (with 0.2% w/v Tra) solution for 17 h at 110 • C using Pico Tag reaction vials according to the procedure described in Ref. [19]. The sample volume or weight corresponding to 0.5 mg protein was used for hydrolysis. Samples were deoxygenated with 3 alternative cycles of vacuum and ultrahigh purity grade nitrogen before incubation. The neutralized dry hydrolysates obtained were reconstituted in 333 μL of water and further diluted 15 folds in 0.1% v/v formic acid before LC-MS analysis.

Protein oxidation by Fenton's reagent
Pure proteins (BSA and β-LG), milk, infant formula, chicken breast, pork liver pâté, or fish were sequentially added with aqueous solutions of FeCl 2 ⋅4H 2 O (4.8 μL, 100 mM), EDTA (4.8 μL, 100 mM), and H 2 O 2 (9.8 μL, 9.8 M). The final concentrations of protein, Fe 2+ , EDTA, and H 2 O 2 in the reaction mixture were 10 mg/mL, 1.5 mM, 1.5 mM, and 300 mM, respectively. Samples were incubated at 55 • C for 30 min in a thermoshaker (Eppendorf, Hamburg, Germany) and cooled immediately thereafter by placing the reaction tubes on ice. After adding 1.6 mL of ice-cold methanol, samples were vortexed for 30 s and kept in a freezer (− 20 • C) for 1 h for protein precipitation. Thereafter, samples were centrifuged at 15,000 g for 15 min at 4 • C and the supernatant containing residual metal ions, EDTA, and H 2 O 2 was discarded. The pellets were then washed with ice-cold methanol (3 × 1.5 mL) including resuspension and centrifugation steps between washes. The pellet was dried under vacuum for 20 min at room temperature. The dry protein pellets were then subjected to acid and enzymatic hydrolysis as described above. All experiments were carried out as independent triplicate (n = 3).

UHPLC-ESI-MS analysis
Calibrators and diluted samples (470 μL) were mixed with 30 μL of IS mix (1000 ng/mL) and filtered through 0.22 μm regenerated cellulose syringe filters (Phenomenex, Vaerløse, Denmark) before LC-MS analysis. The LC system used in the study consisted of a Thermo Ultimate 3000 UHPLC equipped with a high pressure gradient binary pump (Thermo Scientific, MA, USA). Analytes were separated using an Acquity UPLC HSS T3 column (100 Å, 1.8 μm, 2.1 mm × 100 mm; Waters, Taastrup, Denmark) connected to a pre-column (100 Å, 1.8 μm, 2.1 mm × 5 mm). The column was equilibrated for 6 min with 100% A. The column oven and the auto-sampler temperature was set to 25 and 10 • C, respectively.
The injection volume was 10 μL. The needle was washed after drawing every sample using 100 μL of 10% v/v isopropanol.
Analytes were identified and quantified using an Orbitrap Q Exactive high resolution mass spectrometer (Thermo Scientific, MA, USA) equipped with a heated electrospray ionization (HESI) source. The following generic MS tune parameters were used: spray voltage, 3.5 kV (positive mode); capillary temperature 320 • C, sheath gas (nitrogen) flow, 35 arbitrary units; auxiliary gas (nitrogen) flow, 10 arbitrary units; probe heater temperature, 320 • C. A full MS scan was acquired for each run in the range of m/z 80-600 with a resolving power of 17,500 FWHM. The automatic gain control (AGC) target value and maximum injection time used for the full MS scan were 3e6 and 65 ms, respectively. Analytes were quantified through parallel reaction monitoring (PRM) mode with a resolving power of 17,500 FWHM. The parameters of PRM were as follows: AGC target value, 2e5; maximum injection time, 64 ms; isolation window, 2.0 m/z. Remaining parameters specific to target analytes are shown in Table 2. A programmable post-column diverter valve was used to divert the unwanted early and late eluting compounds into waste. The raw data obtained from UHPLC-ESI-MS were processed using Thermo Xcalibur software version 3.1. The peak area of a most abundant or unique product ion in the MS/MS spectrum was used in all quantitative calculations (see Table 2).

Method validation
The chromatographic method was validated based on the guidelines provided by the United States Food and Drug Administration (US-FDA) [26]. A brief protocol used for the validation experiments is given in the following sections. Since true analyte-free blank matrices were unavailable for any tested samples, sample (plasma) or protein hydrolysates (pork liver pâté, fish, and infant formula) were used as matrices in the validation experiments (referred to as pseudo-blank matrices).
Linear range: Appropriate calibration range to quantify target analytes was evaluated by using matrix-matched and solvent calibration curves. The matrix-matched calibration curves were prepared according to the following procedure. Initially, 19 different levels of calibrators were prepared by diluting the SD mix in water (n = 3). Each calibrator (440 μL) was mixed with 30 μL each of the pseudo-blank matrix and IS mix. The final concentration of each analyte in the calibrators were in the range of 0.018-4700 ng/mL, except for Trp, Tyr, Met and Phe, which were at 0.036-9400 ng/mL. Upon completing the LC-MS run, the peak area ratio of the analyte to the IS versus the nominal concentration of the analyte was plotted. A weighting factor of 1/x was applied during linear regression. Since pseudo-blank matrices were used, the peak area of the target analyte present in standard-spiked matrix was subtracted with the peak area of the target analyte present in the original pseudo-blank matrix before plotting matrix-matched calibration curves. The solvent calibration curves for each analyte was prepared following the same procedure as above, but 30 μL of water was used instead of a matrix. The final linear ranges were adjusted for each analyte once the lower (LLOQ) and upper limits of quantification (ULOQ) were established (see sections below).

Limit of detection (LOD):
The US-FDA does not explicitly provide guidelines to calculate LOD, but it states that LLOQ defines the sensitivity (see section LLOQ below). In the present study, LOD was therefore calculated according to Hubaux and Vos method (referred to as LOD H-V ) [27]. Initially, an estimate of the LOD was obtained by injecting a series of solvent calibrators at a lower concentration range (0.013-18.00 ng/mL). Once the preliminary estimate of the LOD is obtained, a series of solvent and matrix-matched calibrators having concentrations within the estimated range of the LOD were prepared (8 levels, n = 3). After LC-MS analysis, the peak area of each analyte was plotted against the nominal concentration and the corresponding LOD H-V for each matrix was calculated based on the Hubaux and Vos mathematical function [28]. Limit of quantification (LOQ H-V ) was calculated based on LOD H-V according to the following equation: LLOQ: Matrix-matched and solvent calibrators having concentrations in the range of 2-8 times the LOD H-V were analyzed (n = 6). The lowest tested concentration having accuracy within ±20% of the nominal concentration and precision within ±20% CV was chosen as LLOQ (referred to as LLOQ FDA ) [26,29]. The highest tested concentration having accuracy within ±15% of the nominal concentration and precision within ±15% CV was chosen as ULOQ. The LLOQ FDA values were used for the calibration curves and QCs, while LOD H-V and LOQ H-V were used only for comparison purpose.
Intraday and interday precision: Intra-day precision was determined by injecting 4 different QCs (LLOQ, low, mid and high, see Table 1) for 9 times a day. Inter-day precision was determined by injecting QCs for 3 times a day for 5 consecutive days. Precision is expressed as %CV of the calculated concentration.
Recovery: Standard spiked samples were prepared by spiking SD mix in four different concentration levels equivalent to LLOQ, low, mid and high QCs (n = 6). Similarly blank spiked-samples were prepared by spiking water instead of standards (n = 6). Standard-spiked and blankspiked samples were analyzed according to the LC-MS procedure described above. Recovery of a given analyte was calculated according to the following equation: where, Cs, Co and Ns are calculated concentration in standard spiked sample, calculated concentration in original sample and spiked concentration, respectively. Carryover: Carryover was determined by injecting a blank after running a ULOQ calibrator (4700 ng/mL, except for Trp, Tyr, Met and Phe, for which it was 9400 ng/mL). The injection cycle was repeated 20 times. The appearance of analytes in the blanks was monitored.
Stability: Stability of target analytes in the autosampler at 10 • C was tested by analyzing low, mid and high QCs stored in the autosampler over 5 consecutive days (n = 3 for each QC level for each day). Stability during freeze-thaw cycle was tested by analyzing QCs in 5 successive freeze-thaw cycles. In each cycle, QCs were thawed to room temperature and stored immediately at − 20 • C after injection. Long-term storage stability of target analytes could not be tested due to a significant variation in the absolute signal response of the MS over the time period of the work.
Matrix effect: A set of matrix-matched and solvent calibrators at 12 different concentration levels in the range of LLOQ-ULOQ were prepared according to the procedure described in the section 'Linear range'. Plots of peak area versus the nominal concentration were constructed. The following equation was used to calculate the matrix effect [30]: where Sm and Ss are slopes of matrix matched and solvent calibration curves, respectively. A value of 0% implied no matrix effect, while a negative and positive value implied ion suppression effect and ion enhancement effects, respectively. The role of IS in compensating the matrix effect was evaluated by calculating the matrix effect according to the above equation, but slopes of the calibration curves of area ratio (area of analyte/area of SIL IS) versus concentration were used in the calculation (referred to as ME IS %).

Run acceptance criteria
The column was conditioned by repeated injections of 2 blanks and 6 QCs before starting a sequence to ensure consistent chromatographic performance. Calibration standards were included at the beginning and end of the sequence. Blanks and QCs were included in regular intervals after every 10 samples in all validation sequences. Calibration curves were accepted only if the calculated concentrations of each calibrator were ±15% of nominal concentrations, except at the lower limit of quantification (LLOQ) where ± 20% was accepted. The final result was accepted only if the acquired data met the calibration acceptance criteria, and ≥50% of QCs per level were ±15% of their nominal concentrations, and ≥67% of total QCs were ±15% of the nominal concentrations [26]. Table 2 List of target analytes, stable isotope-labelled internal standards, and their chromatographic and mass spectrometric parameters.

Method optimization
The present method employed an Orbitrap Q Exactive MS, which is known for providing high-resolution accurate mass data for qualitative and quantitative analysis of small molecules [31]. While respecting its power to ensure exceptional mass selectivity, we attempted to obtain chromatographic separation of target analytes for the following three main reasons: (1) to improve sensitivity and selectivity by minimizing interference from co-eluting analytes, (2) to differentiate isobaric and isomeric target analytes (e.g. isomeric group of m-Tyr, o-Tyr and Tyr, isomeric group of DiOia-1 and DiOia-2, and isobaric group of DiOia-1, DiOia-2 and NFK), and (3) to increase the number of data points across chromatographic peaks as the Orbitrap analyzer acquires only limited number of scans when several co-eluting analytes are simultaneously targeted [31], while ≥12 scans/peak is desired for reliable quantification. In the present study 18 different analytes (Supplementary Figure S1) with diverse and comparable polarities were targeted, which made it challenging to obtain chromatographic separation of all target analytes. Different reversed-phase columns, including Zorbax Eclipse Plus C-18 (Agilent), Syncronis aQ (Thermo Scientific), and Waters Acquity UPLC HSS T3 (all with a dimension of 2.1 mm × 100 mm) and different eluents composed of formic acid, ammonium formate, acetonitrile, and methanol were tested to obtain optimal separation. Although it was possible to separate relatively non-polar target analytes in Zorbax Eclipse Plus C-18 column, it was difficult to resolve polar analytes as the column sought 5% organic modifier at the beginning of the run to avoid potential phase dewetting. On the other hand, Syncronis aQ and Acquity HSS T3 columns were compatible with 100% aqueous mobile phases, which ultimately provided the best retention and resolution for all analytes when 0.1% v/v formic acid in water and acetonitrile were used as a mobile phase. Acquity HSS T3 provided slightly better resolution compared to Syncronis aQ, and consequently chosen as the ideal column for the study. The oven temperature was varied between 20 and 45 • C to optimize the separation, but the best separation was obtained at 25 • C.
As illustrated in Fig. 1, all target analytes were chromatographically separated, except Kyn and DiTyr which had slightly overlapping peaks.
The method was also able to resolve diastereomers of DiOia, obtaining two distinct peaks with retention times of 5.44 and 7.21 min, referred to as DiOia-1 and DiOia-2, respectively. Symmetrical peak shapes were obtained for all the analytes which eased peak integration during data processing. Slightly broader peaks were obtained for 5-OH-Trp, Phe and DiTyr, which could not be improved any further. All analytes had a stable retention time as evidenced by a low CV in the range of 0.1-0.54% even after continuous operation for 6 days representing 224 injections ( Table 2). The retention time drift was <0.1 min for all analytes when tested with two different batches of the mobile phase.
The collision energy needed for fragmentation of precursor ions was optimized for each analyte by varying the normalized collision energy (NCE) values between 20 and 50 arbitrary units in independent runs. The optimal value was selected based on the NCE that provided superior sensitivity and a unique fragmentation pattern (see Table 2).
Apart from defatting and protein precipitation, samples were not purified before LC-MS analysis. The use of a programmable valve to divert unwanted early and late eluting compounds into waste avoided the need for sample desalting and left the ion source clean even after continuous operation for 5 consecutive days. If a diverter valve is unavailable, it is recommended to perform solid phase extraction, particularly for protein hydrolysates, as these samples contain salts that would otherwise deposit on the source.
Quantification of analytes were performed based on isotopic dilution method using SIL IS. Target analytes including Met, DOPA, 3-OH-Kyn, Tyr, Kyn, DiTyr, Phe, 3-Cl-Tyr, 3-NO 2 -Tyr, Trp, Tra, and KynA were quantified by using analogous SIL IS that co-eluted with the corresponding analyte. However, SIL IS for DiOia, m-Tyr, o-Tyr, 5-OH-Trp, and NFK could not be purchased or synthesized in their stable forms. Therefore, these analytes were quantified based on a surrogate SIL IS ( Table 2). A final concentration of 60 ng/mL of each SIL IS in the samples was found to be optimal for the quantitative analysis. At this concentration, the magnitude of isotopic signal contribution from analyte to IS and vice versa either did not exist or was negligible. Potential cross-talk between the product ions of target analyte and co-eluting SIL IS was monitored. It was found that 3-NO 2 -Tyr and its co-eluting SIL IS, 3-NO 2 -Tyr-13 C 6 , shared a common product ion of m/z 187.08, which was also the quantifier ion of the latter. The signal abundance of m/z 187.08 in the MS/MS spectra of 3-NO 2 -Tyr corresponded to ≤4.5% to that of 3- Fig. 1. Parallel reaction monitoring chromatograms of target analytes (A) and stable isotope-labelled internal standards (B). NO 2 -Tyr-13 C 6 . This should not affect the quantitative analysis as 3-NO 2 -Tyr and 3-NO 2 -Tyr-13 C 6 had distinct monoisotopic mass, which should essentially be resolved during PRM transitions. However, such situation could lead to cross interference (cross-talk) only if the MS fails to completely evacuate all ions between transitions [32]. A cross interference of 5% is generally considered tolerable, although specific guidelines do not exist at the moment [32][33][34].
The MS signal drift was assessed by monitoring the change in the absolute signal response of chemically stable SIL IS over time. The intraday drift was less than 3% (expressed as a mean CV of signal response from QCs injected for 20 times), while the inter-day drift was in the range of 2-4% (expressed as a mean CV of signal response from QCs injected over 5 days). Signal drift is a common phenomenon in LC-MS and an increased drift could be due to ion source contamination and fluctuation in the hardware parameters [35]. Normally, SIL IS-based quantification method compensates the signal drift and improves quantification accuracy [36,37].

Linearity
A linear relationship between area ratios (analyte peak area/internal standard peak area) and nominal concentrations was observed in both solvent calibration curve and matrix-matched calibration curves ( Table 3). The correct weighting factor for calibration curves was selected by monitoring the accuracy of the calculated concentration for each calibrator by choosing 1/x 0 , 1/x or 1/x 2 as weighting factors [38].
For most analytes, 1/x and 1/x 2 found to be an ideal weighting factor, while 1/x 0 (no weighting) found to be unsuitable as calculated concentrations showed high deviation from the nominal concentrations particularly at the low and high end of the calibration curve. The weighing factor of 1/x was chosen as it provided least error compared to nominal concentrations [38]. A previous report, however, demonstrated that a weighting factor of 1/x 2 should always be selected for all bioanalytical LC-MS methods [39]. As food and biological samples contain variable levels of parent amino acids (Tyr, Trp and Phe) and their oxidation products, suitable calibrators covering the range of expected analyte concentration were chosen to plot the calibration curves. More than 85% calibrators in the calibration curve had the calculated concentrations within ±15% of nominal concentrations, except at LLOQ where it was within ±20%, which met the validation recommendation [26]. The R 2 values for linear regression were in the range of 0.997-0.9999 for calibration curves plotted in the range of LLOQ-ULOQ (Table 3).

Sensitivity
Sensitivity parameters were determined using two different approaches: (1) Hubaux and Vos method, which is based on computing the prediction interval of the calibration model to estimate detection limits (LOD H-V ) [27], and (2) accuracy and precision method, which is based on calculating the accuracy of precision values of calculated concentrations of calibrators (LLOQ FDA ) [26]. As shown in Table 4, the LOD H-V of analytes ranged from 0.1 to 0.5 ng/mL in blank (water) indicating a high instrumental sensitivity, while slightly higher values were observed when analytes were present in a matrix (chicken, fish, infant formula and plasma), possibly due to the matrix effect. LOQ H-V of target analytes were in the range of 0.3-4.6 ng/mL, depending on the analyte and the matrix, with an exception of DiOia-2 that showed slightly higher LOQ H-V (14.9 ng/mL). In general, lower LOD H-V values were observed for late eluting analytes (3-Cl-Tyr, 3-NO 2 -Tyr, Trp, Tra and KynA) compared to early analytes. The difference is likely due to the appearance of taller and narrower peaks for late eluting compounds compared to peak broadening observed for early eluting ones. The LLOQ FDA of target analytes in blank (water) was in the range of 0.6-3.2 ng/mL, while it increased, reaching up to 13.8 ng/mL, when present in a matrix (e.g. Phe and 5-OH-Trp in roasted chicken breast matrix). It was evident that LOQ H-V was different than LLOQ FDA for most analytes, although only marginal difference was observed. Considering the performance of the method in quantifying trace level analytes in different matrices, we propose that the calculation of detection limits based on the precision and accuracy data provides a more realistic estimate of sensitivity. However, a systematic investigation in this regard is needed.

Selectivity
With an exception of having slightly overlapping peaks for Kyn and DiTyr, the separation of all the analytes including isomers offered a high degree of chromatographic selectivity (Fig. 1). Identification and quantification of Kyn and DiTyr were unaffected as they were resolved by the high detection selectivity of the Orbitrap MS in PRM transitions. The chromatographic separation of the analytes supports that the current chromatographic method could be readily translated into conventional HPLC-UV/FLD/ECD systems, at least to analyze target analytes in less complex samples. None of the target analytes simultaneously shared identical retention time, precursor ion and quantifier ion, which provided the method a high degree of selectivity. A high mass accuracy (1-4 ppm mass error) obtained for precursor and quantifier ions in the Orbitrap mass analyzer and use of co-eluting SIL IS provided another degree of selectivity for compound identification in complex samples. However, a slight difference in retention time between the analyte and SIL IS was observed, particularly for deuterated IS. This problem is inevitable in the reversed-phase chromatography as deuterated SIL IS are relatively polar compared to their protiated analogs [40].

Accuracy and recovery
The recovery values of analytes spiked in water were in the range of 94-112% for low, mid and high QC spiking levels, and 81-119% for LLOQ QC spiking level (Table 5), which indicated the analytical accuracy within ±15% of nominal concentrations, except within ±20% at LLOQ. Recovery of most of the analytes from chicken, fish, infant formula, and plasma matrices were typically in the range of 75-120%, but DiTyr and Kyn had a lower recovery (ca. 55%) from plasma samples at LLOQ spiking level. It is known that recovery of analytes varies based on the matrix composition [31].

Precision
The method was found to be precise for the quantification of all the target analytes as evidenced by acceptable CVs for calculated concentrations during intra-day and inter-day analyses (see Table 6). The intraday precision was in the range of 0.7-3% CV and inter-day precision was in the range of 0.8-3.3% CV for mid and high QCs. A slightly higher, but acceptable, CV% was observed for low and LLOQ QCs (≤8%), with only 5-OH-Trp being the exception, which had a CV of ca. 17% for the low QC. This result suggests that care must be taken while analyzing 5-OH-Trp. The US-FDA recommends a precision corresponding to ±15% CV between runs for all QCs, except ±20% CV for LLOQ QC.

Carryover
Carryover was not observed for any target analytes and SIL IS. It appeared that application of needle wash program constituting 10% isopropanol as wash solvent avoided any carryover issues. According to US-FDA guidelines, the carryover in the blank sample after running a ULOQ calibrator should not be greater than 20% of the LLOQ [26].

Matrix effect
The matrix effect from plasma, infant formula, chicken, and fish matrices was assessed by comparing the slopes of matrix-matched calibration curves with that of solvent calibration curves. The matrix effect without considering SIL IS in the calculation (see columns ME% in Table 7) was in the range of +13% to − 71%, depending on the matrix type. In general, ion suppression was observed for most of the analytes as indicated by negative ME% values. Clearly, the use of SIL IS compensated for the possible analytical error due to the matrix effect as the results obtained show a lower ME IS % value (+0.1% to − 9.7%, Table 7) than for those obtained without considering SIL IS (ME%).

Table 4
Limit of detection [LOD (H -V) ] and limit of quantification [LOQ (H -V) ] calculated based on Hubaux-Vos procedure, and lower limit of quantification calculated based on accuracy and precision data [LLOQ (FDA) ] in different matrices. All values are given in ng/mL.  However, the matrix effect for the quantification of 5-OH-Trp could not be compensated for effectively as evidenced by a higher ME IS % observed from chicken and fish matrices (11.8% and 23.6%, respectively). It is likely due to the absence of analogous (co-eluting) SIL IS for 5-OH-Trp in the method. Therefore, quantitative values of this analyte must be considered indicative only. In general, matrix effect of ±10% could be considered negligible as the deviations in the accuracy and precision of an LC-MS method will also usually be at this magnitude. As a conclusion, the isotopic dilution method found to be an efficient technique to compensate for the matrix effect and to improve analytical accuracy and precision during quantification of tested analytes. It should, however, be noted that the present method uses surrogate IS for the quantification of certain analytes, which may not adequately address the matrix effect from all types of samples. The method should be upgraded by incorporating analogous SIL IS for all analytes when they are commercially available or synthesis routes are established. Several other studies have also proven that this approach effectively compensate for the matrix affect during the LC-MS quantification of various analytes from food and biological matrices [41][42][43][44][45].

Stability
The chromatographic method was originally optimized including QA, a Trp metabolite, which had a retention time of 3.12 min. However, QA degraded rapidly in its aqueous solution in the autosampler at 10 • C, therefore, it was excluded from the current method. A previous study has shown that stock and working solutions can be prepared in dilute citric acid, which prevents the degradation of Trp metabolites in the auto-sampler [46]. As citric acid eluted as a broad peak with a retention time of 3.92 min in the current method, it was not used to avoid potential matrix effects. All analytes remained stable for 1 day in the autosampler as evidenced by the absolute signal responses being in the range of 91-101% compared to that of freshly prepared QCs (Table 8). A long term storage stability study indicated that all analytes, except DOPA, DiTyr, Tra, and their respective SIL IS, were found to be stable for 5 days showing 90-100% signal response and coefficient of absolute signal variation ≤15% at all tested concentration levels (Table 8). DOPA, DOPA-d 3 , DiTyr, DiTyr-13 C 2 , Tra and Tra-d 4 were found to be stable for 3 days, but showed significant loss in signal response afterwards. It was evident that both standards and analogous SIL IS degraded in equal magnitude, therefore, the absolute quantification was   unaffected as supported by the calculated concentrations within ±15% of the nominal concentration (Supplementary Table S1). Freeze-thaw stability of target analytes was tested over 5 cycles. All analytes were found to be stable against 5 freeze-thaw cycles, with only 3-NO 2 -Tyr and 3-NO 2 -Tyr-13 C 6 being the exceptions (Table 9); but the rate of degradation of 3-NO 2 -Tyr-13 C 6 was equivalent to that of 3-NO 2 -Tyr, therefore, the quantification of 3-NO 2 -Tyr was unaffected (Supplementary  Table S2).

Stability of analytes under acidic and enzymatic hydrolysis conditions
To analyze protein-bound oxidation products, it is essential to hydrolyze proteins completely to obtain free forms of amino acid oxidation products prior to LC-MS analysis. Acid-catalyzed hydrolysis using 4 M MSA (with 0.2% Tra) or 6 M HCl carried out at 110 • C for 16-24 h is the most commonly used hydrolysis methods to analyze individual oxidation products [18,19,24]. Enzymatic hydrolysis is referred to as a gentle method to hydrolyze proteins, but its efficiency to characterize protein oxidation products is rarely explored [11]. Nonetheless, these hydrolysis methods have several inherent limitations; and the instability of amino acid oxidation products under hydrolytic conditions is of major concern [11,18]. In the present study, both acid hydrolysis (using 4 M MSA with 0.2% Tra in vacuo) and enzymatic hydrolysis were carried out to obtain the best possible results. Initially, the stability of oxidation products under hydrolytic condition was tested by subjecting SD mix to acid and enzymatic hydrolysis in the absence of sample matrices. As illustrated in Fig. 2, the stability of target analytes varied depending on the applied hydrolysis method. Met, 3-OH-Kyn, Tyr, m-Tyr, o-Tyr, DiTyr, Phe, Trp, 3-Cl-Try, and KynA were stable under acid hydrolytic conditions (stability ≥ 95%). DOPA and 3-NO 2 -Tyr were found to be partially stable with stability values of 31 ± 8% and 80 ± 3%, while DiOia-1, DiOia-2, NFK, and 5-OH-Trp were found to be completely unstable. The calculated stability of Kyn was 207 ± 11%, which was higher than the expected value, likely due to hydrolysis of NFK into Kyn under acidic conditions [19]. The stability of Tra could not be determined as Tra was added externally in the hydrolysis reagent to minimize artefactual oxidation of target analytes. Under enzymatic hydrolytic condition, o-Tyr, 3-Cl-Tyr, 3-NO 2 -Tyr, Tra and KynA were found to be stable (stability ≥ 90%), while DiOia-1, m-Tyr, DiOia-2, NFK, DiTyr, and 5-OH-Trp were found to be partially stable with stability values of 22 ± 0, 31 ± 0, 43 ± 3, 3 ± 0, 2 ± 0 and 5 ± 0, respectively. DOPA and 5-OH-Trp were found to be completely unstable. The stability of parent amino acids, including Trp, Tyr, Met and Phe, could not be tested as enzymes added in the reaction mixture released significant amount of parent amino acids due to autolysis in the absence of protein substrates. Similar to what was observed after acidic hydrolysis, the calculated recovery of Kyn was higher than the expected value after enzymatic hydrolysis (165 ± 13%). It was clear that, unlike during acidic hydrolysis, NFK did not completely convert into Kyn during enzymatic hydrolysis as NFK was partially recovered after enzymatic hydrolysis. A previous study has reported 90, 58, <0.1 and < 1% stability of DiOia, Kyn, 5-OH-Trp and NFK, respectively, during enzymatic hydrolysis [11]. The stability may vary based on the traces of metal ions present in enzyme preparations and buffer salts used. It should, however, be noted that hydrolytic stability of free forms of analytes does not necessarily reflect the stability of protein-bound forms and their subsequent recovery after hydrolysis. Since a reference protein bearing known amounts of oxidation products is currently unavailable, the actual stability could not be calculated.
Based on the stability data, all protein-bound analytes, except NFK, DiOia-1 and DiOia-2, were quantified from acid hydrolysates of the samples, while DiOia-1, DiOia-2, and NFK were quantified from enzymatic hydrolysates. The data for protein bound Kyn is reported as the sum of Kyn and NFK.

Application of the method
Mice serum and small intestinal tissue samples were tested for the presence of target analytes, with results showing presence of diverse oxidation products or metabolites at different concentrations (Table 10). DOPA was not detected in serum, while 3-OH-Kyn was not detected in small intestinal tissue samples and o-Tyr, m-Tyr, DiTyr, Tra, 3-Cl-Tyr, and 3-NO 2 -Tyr were not detected in these two samples. Previous studies aiming at investigating the Kyn pathway in health and disease have reported comparable levels of 3-OH-Kyn, Kyn, NFK, and KynA in rat and mice plasma samples by using independent HPLC methods coupled with a diode array (for Kyn), fluorescence (for KynA) or an electrochemical (for 3-OH-Kyn) detection system [47,48]. Although not detected in the samples tested in the present study, it has previously been reported that free o-Tyr and m-Tyr could be formed in biological tissues from free Phe due to oxidative stress exerted by •OH [49]. Free m-Tyr can be misincorporated into proteins during cellular protein synthesis and induce cytotoxicity [50]. Tra and KynA are formed in vivo via Trp metabolic pathway; wherein tryptophan decarboxylase catalyzes conversion of Trp into Tra [51], and kynurenine aminotransferase catalyzes conversion of Kyn to KynA [15]. Tra and KynA play significant Table 7 Matrix effect [(ME(%)] and internal standard corrected matrix effect [(ME IS %)] from different matrices. Positive and negative values imply ion enhancement and ion suppression, respectively.

Table 8
Autosampler stability (10 • C for 12 h and over 5 days) of standards and stable isotopically labelled internal standards (SIL IS) used in the study. Stability of each analyte was calculated based on the average signal response of three replicate injections on each day, except for Day 1, where 12 injections spread over a day were considered. The signal from the first three injections soon after thawing the sample were considered equivalent to 100% stability. The nominal concentration of standards were equal to high QC, mid QC and calibration level C4 (see Table S1). Calibration level C4 instead of Low QC was used to study stability as the remaining concentration of analytes in Low QC was expected to be < LLOQ after storage. The concentration of each SIL IS was 60 ng/mL.  Table 9 Freeze-thaw stability of standards and stable isotopically labelled internal standards (SIL IS) over 5 cycles. The signal from the first injection soon after the first freeze-thaw cycle was considered equivalent to 100% stability. Three concentration levels were used in the experiment. The nominal concentration of each standards in concentration level 1, 2 and 3 were shown in Table S2. roles in neurophysiological processes [52]. Tra can be formed in vitro through photochemical reactions of Trp in aqueous solutions [53]. Free 3-NO 2 -Tyr is one of the biomarkers of oxidative stress induced by RNS and is measured in biological tissues using HPLC-ECD [54]. Obviously, simultaneous determination of these analytes as in the present method will not only save time and resources, but also provides a better picture of the metabolic pathway by minimizing the bias associated with sample preparation and analysis. The presence of protein-bound oxidation products were tested in roasted chicken breast, baked pork liver pâté, fresh fish, stored infant formula, and LH UHT milk, upon acid or enzymatic hydrolysis. The concentration of 5-OH-Trp was not reported as it was found to be completely destroyed under the hydrolytic conditions used. The concentration of parent amino acids and oxidation products in different samples is shown in Fig. 3 (see bars with solid fill). As expected, all samples contained higher concentrations of parent amino acids (Trp, Tyr, Met and Phe) compared to oxidation products. Pure proteins did not contain oxidation products, except trace levels of DiOia-1, DiOia-2, and Kyn (≤4 mg/100 g protein). Thermally processed samples, including LH UHT milk, infant formula, roasted chicken breast and baked pork liver pâté, contained relatively higher levels of oxidation products ( Fig. 3 bars with solid fill). These oxidation products could either be inherently present in samples or formed during thermal processing, which cannot be distinguished.

Role of fenton reaction in forming oxidation products
The role of Fenton reaction on protein oxidation was evaluated by incubating samples with Fenton's reagent before hydrolysis. As expected, the Fenton reaction led to a significant reduction in the concentrations of parent amino acids Met, Tyr, Phe and Trp, and concomitantly increased corresponding oxidation products irrespective of sample type (Fig. 3, comparison of bars with solid fill with pattern fill). Met was found to be almost completely destroyed in all samples after the treatment with Fenton's reagent. Previous kinetics studies have shown that Phe, Met, Tyr, and Trp can be rapidly oxidized by •OH with rate constants ranging from 6.5 × 10 9 to 1.3 × 10 10 dm 3 mol − 1 s − 1 , respectively, at pH ca. 7 [10]. It is known that oxidation of Met leads to the formation of methionine sulfoxide and methionine sulfone, but these analytes could not be quantified in the present study as they eluted with a retention time of <1.5 min where high level of interfering salts co-eluted. When all tested samples were compared, it was clear that DiOia-1, DiOia-2, Kyn and o-Tyr were the major oxidation products formed during the Fenton reaction with concentrations in the range of 20-120 mg/100 g protein, depending on the sample type. Interestingly, the concentrations of DiOia-1 and DiOia-2 were found to be similar in a given sample, irrespective of sample type, indicating these two isomers form in equimolar ratios during oxidation of Trp. The diastereomeric abundance of DiOia in food matrices is not reported in the literature. DiOia is formed due to oxidation of the pyrrole moiety of the Trp residues. Little is known about the physiological role of DiOia [55]. It was also clear that the Fenton chemistry favored the formation of o-Tyr over m-Tyr as 2-36 folds higher levels of o-Tyr was found in samples compared to m-Tyr (Fig. 3).
DiTyr was not detected in any control samples, but the treatment with Fenton's reagent led to its formation in all samples (1.2-9.0 mg/ 100 g protein), with only infant formula being the exception (Fig. 3). DiTyr is one of the major biomarkers of oxidative stress [56]. It is associated with protein cross-linking and pathogenesis of neurodegenerative diseases [57]. Previous studies have shown that DiTyr is produced in human neutrophils and macrophages in the presence of H 2 O 2 [58]. In food systems, the formation of DiTyr has been linked with structural and functional properties of foods [2,59]. A previous study has reported the presence of DiTyr in commercial infant formula samples using a UHPLC-FLD method, with concentrations in the range of 100-200 mg/100 g protein [60]. DiTyr was also quantified in wheat flour using LC-MS isotopic dilution assay, reporting a concentration of 24 μg/100 g sample [23]. Although NFK was detected in all enzymatically hydrolysed samples, no clear trend demonstrating the effect of the Fenton reaction was observed (Supplementary Figure S2).
Interestingly, 3-NO 2 -Tyr was detected in all samples treated with Fenton's reagent, with a concentration < LLOQ, but not in any control samples. It is known that superoxide (O 2˙− ) reacts with nitric oxide (•NO) and produce reactive peroxynitrite anion (ONOO − ), which can nitrate Tyr residues [61]. Perhaps traces of NOx present in the Fenton's reagent artefactually nitrated the proteins. Similarly, trace levels of KynA was detected in all Fenton reagent treated samples (1.22-2.45 mg/100 g proteins), but not in any control samples (Supplementary Figure S2). KynA could be formed from the free form of Kyn upon deamination and intramolecular cyclization. KynA can essentially not occur in protein-bound form. It has been reported that •OH can cleave peptides Fig. 2. Stability of different analytes under acidic and enzymatic hydrolysis condition. # Stability of Tra was not determined under acidic hydrolytic condition as the acid reagent used for the hydrolysis contained Tra, while stability of Trp, Tyr, Met and Phe could not be calculated under enzymatic hydrolysis condition due to artefactual release of these amino acids by enzymes due to autolysis. The stability is expressed as % of recovered analytes after subjecting corresponding standards into hydrolytic condition in three independent experiments. The nominal concentration for each analyte used in the stability study is 686 ng/mL, except for DiaOia-1 and DiaOia-2 (343 ng/mL each), and Met, Tyr, Trp and Phe (1372 ng/mL each).

Table 10
Concentration of free forms of target analytes (amino acids, amino acid oxidation products, and tryptophan metabolites) in mice serum (μg/mL) and mice small intestinal tissue (mg/100 g dry weight) a . The result is expressed as mean ± standard deviation (n = 3). and proteins and release free amino acids [62]. Therefore, we postulate that it could be formed from free Kyn released from proteins during incubation with Fenton's reagent due to protein backbone cleavage or is artefactually formed during hydrolysis.

Perspectives on the quantification of protein-bound oxidation products
In addition to an optimal LC-MS method, accurate quantification of protein-bound oxidation products is dependent on the efficiency of hydrolysis or, in other words, the recovery of free-form analytes from their protein-bound forms. Recovery is dependent on the interplay between the kinetics of the peptide bond hydrolysis and the degradation of the released free oxidation products under hydrolytic condition. The latter was evaluated to an extent by subjecting pure standards to hydrolytic condition (see Section 3.3 above). However, the absolute recovery of oxidation products could not be calculated due to lack of reference proteins containing known amounts of oxidation products. The bias in analyte recovery due to the hydrolysis step could be compensated for by spiking SIL IS before hydrolysis. This strategy provides reliable results only if the rate of degradation of protein-bond oxidation products is equal to that of the spiked free-form of SIL IS. Otherwise, this analytical approach could produce erroneous results due to unequal loss of target analytes and SIL IS during hydrolysis. Furthermore, oxidation products can be formed artefactually during the hydrolysis step. To test this, hydrolysates of pure BSA and β-LG were analyzed for the presence of oxidation products, assuming pure proteins does not contain any oxidation products. Trace levels of DiOia-1, DiOia-2, Kyn, and NFK were detected in these hydrolysates (≤4 mg/100 g proteins, Fig. 3 and Supplementary Figure S2), indicating either artefactual oxidation has taken place or the original protein itself has oxidation products, which cannot be distinguished. The presence of oxygen and metal ions in the reagents and enzyme preparation may trigger artefactual oxidation. Possible interconversion of oxidation products during hydrolysis may also not be Fig. 3. Concentrations of different analytes in samples tested. Bars with solid fill (black) represent samples analyzed without Fenton's reagent treatment (control samples) and bars with pattern fill (grey) represent samples treated by Fenton's reagent. Roasted chicken, fresh fish, baked pork liver pate, UHT treated 3 months stored infant formula and lactose hydrolysed UHT treated milk were used in the study. All experiments were carried out as independent triplicate (n = 3).

Conclusions
The combination of chromatographic resolution, mass resolution, high mass accuracy, and sensitivity enabled reliable simultaneous quantification of multiple aromatic amino acid oxidation products including parent amino acids in complex samples. The isotopic dilution method was found to be an efficient technique to compensate for the matrix effects and to improve analytical accuracy and precision during quantification of tested analytes. It should, however, be noted that the present method uses surrogate IS for the quantification of certain analytes, which may not adequately address the matrix effect from all types of samples. The method should be upgraded by incorporating analogous SIL IS for all analytes when they are commercially available or synthesis routes are established. The method also provided a quantitative data for the Fenton reaction-mediated formation of various protein oxidation products, identifying relevant markers for such reactions in food and biological matrices.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.