Oxidised Met147 of human serum albumin is a biomarker of oxidative stress, reflecting glycaemic fluctuations and hypoglycaemia in diabetes

Oxidative stress has been linked to a number of chronic diseases, and this has aroused interest in the identification of clinical biomarkers that can accurately assess its severity. We used liquid chromatography-high resolution mass spectrometry (LC-MS) to show that oxidised and non-oxidised Met residues at position 147 of human serum albumin (Met147) can be accurately and reproducibly quantified with stable isotope-labelled peptides. Met147 oxidation was significantly higher in patients with diabetes than in controls. Least square multivariate analysis revealed that glycated haemoglobin (HbA1c) and glycated albumin (GA) did not significantly influence Met147 oxidation, but the GA/HbA1c ratio, which reflects glycaemic excursions, independently affected Met147 oxidation status. Continuous glucose monitoring revealed that Met147 oxidation strongly correlates with the standard deviation of sensor glucose concentrations and the time spent with hypoglycaemia or hyperglycaemia each day. Thus, glycaemic variability and hypoglycaemia in diabetes may be associated with greater oxidation of Met147. Renal function, high-density lipoprotein-cholesterol and serum bilirubin were also associated with the oxidation status of Met147. In conclusion, the quantification of oxidised and non-oxidised Met147 in serum albumin using our LC-MS methodology could be used to assess the degree of intravascular oxidative stress induced by hypoglycaemia and glycaemic fluctuations in diabetes.

(LC-MS), irrespective of the time the blood sample is left to clot at room temperature before centrifugation or repeated freeze/thaw cycles 28 . This may be because of the absence of MetO reductase activity in human blood 29 , although these have not been well characterised to date.
In the present study, we have quantified the levels of oxidised and non-oxidised Met at position 147 of human serum albumin (Met 147 ), to determine whether the oxidation status of this residue reflects the oxidative stress induced during disease pathophysiology, and ultimately whether this might represent a useful biomarker of oxidative stress. To this end, we have improved our mass spectrometric methodology for the accurate and stable quantification of such residues in clinically-derived samples.

Results
Improved methodology for the quantification of oxidised and non-oxidised Met-containing serum tryptic peptides. We first sought to improve the accuracy and reproducibility of our previous methodology for the quantification of the oxidation of Met residues in serum tryptic proteins 28 . We previously found that the ratio of trypsin-digested serum albumin fragments containing oxidised and a non-oxidised Met residues at position 147 of human serum albumin, Alb(Met 147 O) and Alb(Met 147 ), is one of the most promising potential clinical biomarker of intravascular redox status among the Met-containing tryptic serum proteins identified using a proteomic strategy 28 . Because the use of stable isotope-labelled peptides is known to enable the accurate quantification of peptide concentrations in biological samples, we synthesized two stable isotope-labelled peptides, SI-Alb(Met 147 ) and SI-Alb(Met 147 O), corresponding to the tryptic peptides, Alb(Met 147 ) and Alb(Met 147 O), respectively.
Our previous methodology, which did not use stable isotope-labelled peptides, quantified the signal intensity of Alb(Met 147 O) and Alb(Met 147 ), and employed their ratio as an indicator of oxidised Met. In the current analysis, the serially diluted stable isotope-labelled peptides, SI-Alb(Met 147 ) and SI-Alb(Met 147 O), were spiked into serum samples from participants prior to tryptic digestion and LC-MS analysis to generate the extracted ion chromatogram (XIC) intensities for the two endogenous tryptic peptides, Alb(Met 147 ) and Alb(Met 147 O), and the corresponding stable-isotope-labelled peptides, SI-Alb(Met 147 ) and SI-Alb(Met 147 O) (Fig. 1). The serum concentrations of Alb(Met 147 ) (C Alb(Met ) 147 ) and Alb(Met 147 O) (C ) Alb(Met O) 147 were then extrapolated from the XICs generated using the respective endogenous peptides and the corresponding spiked stable isotope-labelled peptides. The oxidation ratio for Met 147 was obtained by dividing C Alb(Met O) 147 by C Alb(Met ) 147 . Standard calibration curves were generated using serially diluted serum samples vs. quantified Alb(Met 147 ) or Alb(Met 147 O), and resulting equations from the regression analyses demonstrated very high degrees of linearity ( Supplementary Fig. S1). This updated method, using stable isotope-labelled peptides, yielded a coefficient of variation (%CV) of 9.8%, while the original method, which used signal intensity ratio of Alb(Met 147 O) and Alb(Met 147 ), yielded a %CV of 19.7% Figure 1. Representative extracted ion chromatograms (XICs) of tryptic peptides containing oxidised and nonoxidised Met 147 residue and relevant stable isotope-labelled peptides. The stable isotope-labelled peptides, SI-Alb(Met 147 ) and SI-Alb(Met 147 O), were spiked to the serum prior to trypsin digestion. XICs with charge states of three (a,c) and four (b,d) of endogenous (Alb(Met 147 ) and Alb(Met 147 O)) (a,b) and stable isotope-labelled peptides (c,d) are presented. Oxidised peptides were magnified 100-fold and 10-fold, respectively, and are shown above the original peaks.
We next evaluated the benefits of using of L-Met and L-cysteine (L-Cys) to prevent the spontaneous oxidation of Met and the carbamidomethylation of N-terminal amino acid residues. The addition of L-Met after the tryptic digestion of serum samples suppressed the spontaneous oxidation of Alb(Met 147 ). C Alb(Met O) 147 /C Alb(Met ) 147 , measured in serum samples from six healthy participants without the addition of L-Met, increased after a week or a month of storage (Fig. 2a-c). In contrast, the addition of excess L-Met to the serum samples markedly suppressed this increase (Fig. 2a-c). We next compared the XICs of Alb(Met 147 ) and SI-Alb(Met 147 ), with and without the use of L-Cys. The addition of excess L-Cys prior to the trypsin digestion of serum samples inhibited carbamidomethylation at the N-terminus of these peptides (Fig. 2d,e), leading to higher XIC values for uncarbamidomethylated Alb(Met 147 ) and SI-Alb(Met 147 ) (Fig. 2f,g). Therefore, we added excess L-Cys and L-Met prior to and immediately after the enzymatic digestion of serum samples in subsequent experiments.
The effect of long-term storage on Met oxidation was evaluated using SI-Alb(Met 147 ) and SI-Alb(Met 147 O), and excess L-Cys and L-Met. C Alb(Met O) 147 /C Alb(Met ) 147 did not significantly change after 2 years of storage at −80 °C (0.001267 ± 0.0001391, compared with the value obtained immediately after blood withdrawal: 0.001447 ± 0.0002763; n = 6; p = 0.2188). We next determined whether the length of the clotting time prior to serum separation affected the spontaneous oxidation of Alb(Met 147 ). Blood samples were allowed to clot at room temperature, and the obtained sera were alkylated and trypsin-digested for subsequent LC-MS analysis. C Alb(Met O) 147 /C Alb(Met ) 147 , determined in samples centrifuged after periods of time on the bench of between 10 min and 6 h, did not show any appreciable differences (0.001331 ± 0.00008874, 10 min; 0.001325 ± 0.00006762, 30 min; 0.001221 ± 0.0001424, 1 h; 0.001264 ± 0.0001149, 2 h; 0.001267 ± 0.0001608, 3 h; 0.001312 ± 0.0001276, 6 h; p = 0.5783). Therefore, C Alb(Met O) 147 /C Alb(Met ) 147 is a robust and reproducible measurement that is not affected by clotting time. www.nature.com/scientificreports www.nature.com/scientificreports/ Oxidised Met ratio in diabetes. We next measured C Alb(Met O) 147 /C Alb(Met ) 147 in 40 healthy volunteers and 124 patients with diabetes (Table 1). C Alb(Met O) 147 /C Alb(Met ) 147 levels were significantly higher in the diabetic patients than in the healthy volunteers (Fig. 3). Single regression analysis revealed that age, glycated albumin (GA)/glycated haemoglobin (HbA 1c ), blood urea nitrogen (BUN) serum creatinine (Cr) and uric acid positively correlated with C Alb(Met O) 147 /C Alb(Met ) 147 level, and this negatively correlated with body mass index (BMI), estimated glomerular filtration rate (eGFR) and serum total bilirubin ( Table 2). Least square multivariate analysis was undertaken using these statistically significant parameters as explanatory variables, as well as those reported to have an antioxidant activity, and revealed that GA/HbA 1c , eGFR, high-density lipoprotein (HDL)-cholesterol and total bilirubin significantly and independently influenced C Alb(Met O) 147 /C Alb(Met ) 147 level (Table 3).   (Fig. 4e). Because glycaemic excursions appeared to be closely associated with higher C Alb(Met O) 147 /C Alb(Met ) 147 , we next determined whether the administration of a sodium glucose cotransporter 2 inhibitor, which suppresses glycaemic fluctuations, would reduce the oxidation of the Met residue. Indeed, the C Alb(Met O) 147 /C Alb(Met ) 147 of 18 diabetic participants (nine men and nine women; 54.3 ± 9.5 years; HbA 1c 9.3 ± 1.7%, BMI 32.1 ± 5.6 kg/m 2 ) was significantly lower after either canagliflozin, luseogliflozin or empagliflozin was administered for 28 days (Fig. 5).
Finally, we assessed the advantage of this updated methodology over our previous method that simply measured signal intensity ratios of Alb   only showed a positive correlation with %CV of glucose levels and length of time spent with hypoglycaemia, but did not reveal any significant correlation with standard deviation or average glucose concentration, nor the percentage of time spent with normal or high glucose levels ( Supplementary Fig. S2). [Alb(Met 147 O]/[Alb(Met 147 )] of 18 diabetic participants did not significantly decrease after the oral intake of a sodium glucose cotransporter 2 inhibitor for 28 days (Supplementary Fig. S3). These results demonstrate that the updated quantification approach using stable isotope-labelled peptides is far more sensitive in detecting redox status changes in disease pathophysiology over our previous techniques which only measured signal intensity ratios even when excess L-Met and L-Cys were used upon enzyme digestion of serum.

Discussion
In our previously published study, we determined whether the oxidation of Met residues of serum proteins could be used as a clinical marker of oxidative stress, and concluded that the mass spectral intensity ratio of the levels of oxidised and non-oxidised Met residues of some serum tryptic proteins could be quantified stably and reproducibly 28 . In the present study, we have successfully improved the accuracy and reproducibility of our method for the quantification of the ratio of Met 147 to Met 147 O-containing tryptic peptides.
First, we measured the concentrations of serum Alb(Met 147 ) and Alb(Met 147 O) using stable isotope-labelled peptides by extrapolating the respective XIC data obtained using LC-MS analysis. In our previous method, the mass spectral signal intensity ratio of the serum tryptic peptides with and without MetO was determined 28 . However, this method was limited in accuracy because of broadening of the elution peak, which likely occurs as the LC column condition deteriorates. The degree of this broadening may differ between oxidised and non-oxidised Met, which could lead to lower accuracy when many samples are analysed. The quantification of peptides containing both oxidised and non-oxidised Met that we used in the present study eliminates this drawback by reducing the effects of column condition on the analysis. Met is prone to spontaneous oxidation once the surfactant used for digesting the serum proteins is removed prior to LC-MS analysis. The Alb(Met 147 ) in the trypsinised serum samples becomes significantly oxidised from 1 week of storage, but the addition of excess L-Met prior to this prevents the oxidation of Alb(Met 147 ) for at least 1 month. Another difficulty associated with the use of LC-MS technology for the accurate quantification of serum tryptic peptides containing Cys residues is the undesirable carbamidomethylation of peptide N-termini that is associated with the reductive alkylation procedure that prevents disulfide formation. In the present study, we have shown that the addition of excess L-Cys greatly enhances the XIC intensity of non-carbamidomethylated Alb(Met 147 ) and SI-Alb(Met 147 ). Furthermore, the addition of excess L-Cys and L-Met did not significantly affect C Alb(Met O) 147 /C Alb(Met ) 147 levels, even after serum samples were stored for 2 years at −80 °C. Thus, the addition of excess L-Met and L-Cys eliminates undesirable spontaneous Met oxidation and carbamidomethylation of N-terminal residues of target peptides, which leads to a marked improvement in the accuracy of the quantification of Alb(Met 147 ), Alb(Met 147 O), SI-Alb(Met 147 ) and SI-Alb(Met 147 O) using LC-MS.
Having successfully established an accurate, stable and reproducible method for the quantification of the oxidation rate of Met residues in blood proteins using a single drop of human serum, we then determined whether this would reflect the redox status associated with conditions that are known to modulate oxidative stress status in experimental diabetes [30][31][32][33] . In humans, it remains controversial whether glucose fluctuation during diabetes activates oxidative stress [34][35][36] and it has yet to be established whether hypoglycaemia affects redox status 37 . In the current study, we found that C Alb(Met O) 147 /C Alb(Met ) 147 was significantly higher in diabetic than in healthy participants. It did not correlate with HbA 1c or GA, which reflect mean blood glucose over a period, but significantly correlated with GA/HbA 1c ratio, which reflects glycaemic variability 38 . Therefore, we used CGM to determine whether C Alb(Met O) 147 /C Alb(Met ) 147 reflects features of the pathophysiology of diabetes, such as glucose fluctuations and hypoglycaemia. C Alb(Met O) 147 /C Alb(Met ) 147 positively correlated with parameters that directly reflect blood glucose fluctuations, such as SD, %CV, duration of hypoglycaemia and duration of hyperglycaemia, and negatively correlated with the length of time the patients were normoglycaemic. In agreement with this, 4 weeks of treatment of 18 type 2 diabetes patients with a sodium glucose cotransporter 2 inhibitor, which reduces glycaemic fluctuation by enhancing urinary glucose excretion, significantly reduced C Alb(Met O) 147 /C Alb(Met ) 147 . Least square multiple regression analysis also showed that serum bilirubin, an endogenous antioxidant 39,40 , HDL-cholesterol and eGFR are independent variables that negatively influence C Alb(Met O) 147 /C Alb(Met ) 147 . Taken together, these findings suggest that a variety of pathophysiological factors that have been shown experimentally to affect oxidative stress, such as 147 /C Alb(Met ) 147 was determined in 18 type 2 diabetic participants before and after 28 days' oral administration of a sodium glucose cotransporter 2 inhibitor. **p < 0.01, calculated using the Wilcoxon signedrank test.
glucose fluctuation and hypoglycaemia in diabetes, serum bilirubin, and renal function, do indeed affect indicators of oxidative stress in human blood proteins in a clinical setting.
To assess the superiority of our updated methodology, we reanalyzed all clinical samples using previous methodology that measures signal intensity ratios of MetO-and Met-containing peptides, and compared the results with those of the new method. [Alb(Met 147 O]/[Alb(Met 147 )] levels were not associated with the GA/HbA 1c ratio or serum total bilirubin when regression analyses were performed using entire samples. In patients undergoing continuous glucose monitoring, [Alb(Met 147 O]/[Alb(Met 147 )] only showed a positive correlation with %CV of glucose levels and length of time spent with hypoglycaemia, but did not reveal any significant correlation with the standard deviation or average glucose concentration, nor the percentage of time spent with normal or high glucose levels (new Supplementary Fig. S2). In contrast, C Alb(Met O) 147 /C Alb(Met ) 147 were associated with all of the measured parameters reflecting glycemic excursion, hypoglycaemia, and hyperglycaemia (Fig. 4), and endogenous factors theoretically affecting oxidative stress status, such as serum bilirubin, HDL cholesterol, and eGFR (Table 3). These results demonstrate that the updated approach is far more sensitive in detecting redox status changes over our previous techniques which simply measured signal intensity ratios. To date, clinical biomarkers that can detect the intravascular oxidative stress status elicited by blood glucose fluctuations, hypoglycaemia, and hyperglycaemia in diabetes, as well as that affected by endogenous factors such as serum bilirubin and HDL cholesterol, have been unavailable. Our results demonstrate that our upgraded mass spectrometry approach shows a marked improvement in usefulness such that it can detect the intravascular redox status in human disease pathophysiology.
In our previous report, we identified 53 trypsin-digested serum peptides that contained MetO 28 . However, the XICs of many of their MS1 overlapped with other enzyme-digested peptides when measured by LC-MS with a short analysis time (<60 min), which prevented detection of their signal intensities. MetO-containing peptides derived from low serum concentration proteins elicit comparatively larger background signals, rendering accurate measurement difficult. As a consequence, we selected five methionine-containing peptides technically quantifiable with relatively high concentrations and little signal overlap 28 . Met residues already highly oxidised at baseline in healthy subjects, such as Met 1181 of complement C3, cannot reflect redox status changes elicited by disease pathophysiology 28 . Therefore, in the current study, we focused on Met 147 of serum albumin as a promising clinical biomarker candidate because Met 147 has lower baseline oxidation levels than most other Met residues, such as Met 332 or Met 363 . The oxidation of two Met residues in the rat brain, Met 288 and Met 572 , was also reported 41 . Interestingly, these Met residues with elevated baseline oxidation levels are located on the surface of rat and human serum albumin molecules. It is speculated that baseline oxidation and/or susceptibility to oxidation depends upon the position of each Met residue in the three-dimensional structure of the protein molecule. Met is highly susceptible to oxidation, and elevated Met(O)-protein levels have been demonstrated in a variety of oxidative stress-related diseases. Further, Met(O) reductase has not been detected in human blood samples 29 . Therefore, Met(O) in selected Met residues could be used as oxidative stress biomarkers as long as Met-and Met(O)-containing peptides are accurately quantified.
In conclusion, our LC/MS methodology appears to be sufficiently accurate and sensitive for the detection of the effects of intravascular redox status in human pathophysiology.

Participants.
The study population consisted of 40 healthy volunteers and 124 diabetic patients (94 with type 2 diabetes and 30 with type 1 diabetes) who visited the Kitasato University Hospital between May 2016 and July 2019. A diagnosis of type 2 diabetes was made on the basis of insulin-independence, according to the criteria of the Japan Diabetes Society (patients with anti-glutamic acid decarboxylase autoantibody >1.5 U/ml or serum C-peptide <0.5 ng/ml were excluded) 42 . Clinical records were reviewed for all the potential participants, and those with acute inflammatory diseases, malignancies, or recent episodes of cerebrovascular or cardiovascular accident were excluded from the study. Serum sample collection. Venous blood was collected into vacutainers containing pro-coagulant, allowed to clot at room temperature for approximately 1 h, unless otherwise indicated, and then centrifuged at 1,000 × g for 20 min at room temperature. The separated serum was stored at −80 °C until processing. Patients underwent routine evaluations for the systemic diseases covered by the universal health coverage system in Japan 42,43 using electrocardiography (ECG), radiography (chest and abdomen), ultrasonography (neck and/or abdomen), urinalysis, complete blood count, and the measurement of 15 serum biochemical analysis items. The diabetic participants also underwent ophthalmological and neurological testing, the anti-glutamic acid decarboxylase antibody test, and measurements of GA and/or HbA 1c , fasting serum insulin, and urinary albumin-to-creatinine ratio.
Synthesis of stable isotope-labelled peptides. The following two stable isotope-labelled peptides were synthesized by Scrum Inc. (Tokyo, Japan) using L-phenylalanine-N-9-fluorenylmethoxycarbonyl ( 13 C 9 , 98%; 15  www.nature.com/scientificreports www.nature.com/scientificreports/ (Sigma-Aldrich) and incubated at 50 °C for 30 min. The mixture was then incubated in a dark room at room temperature with 2 μl of 375 mM iodoacetamide (Nacalai Tesque, Kyoto, Japan) for 30 min and then with 2 μl of 400 mM L-Cys (Fujifilm Wako Pure Chemical, Japan) for 10 min, after which 2 μl of 100 ng/μL Lys-C and 2 μl of 100 ng/μL trypsin were added and digestion was carried out for 24 h at 37 °C. Eighteen microlitres of the digest was then mixed with 50 μL of 10% acetonitrile (ACN) and 50 μL of 5% trifluoroacetic acid, and the mixture was centrifuged at 19,000 g for 15 min. The supernatant was recovered and 2 μL of 100 mM L-Met (Tokyo Chemical Industry, Tokyo, Japan) was added, prior to LC-MS analysis.
Quantification of the oxidised and non-oxidised Met residue-containing tryptic peptides using LC-MS. Analysis of the tryptic digests of the serum samples was performed using LC-MS, essentially as described 28 . Tryptic digests of the serum samples were injected onto a 2.0 mm (inner diameter) × 50 mm CAPCELL PACK MGIII-H S3 column attached to a Nanospace SI-2 HPLC system (Shiseido Fine Chemicals, Tokyo, Japan). The column temperature was maintained at 45 °C. The flow rate of the mobile phase was 200 μL/ min, and mobile phase A consisted of 0.05% formic acid (FA) and mobile phase B consisted of 0.05% FA/90% ACN. The mobile phase gradient was programmed as follows: 0% B (0-3  Continuous glucose monitoring. Seven non-diabetic volunteers and 28 diabetic patients underwent continuous glucose monitoring (CGM) using the iPro TM 2 CGM system (Medtronic Minimed Inc. Northridge, CA) 46,47 for 4-7 days and provided an overnight-fasted blood sample during the period. The complete SGL data were used to assess each glucose profile. The SD and %CV were used to assess glycaemic fluctuations, and the maximum, minimum and mean SGL were used to evaluate glycaemic control. The amount of time over a 24-h period each participant spent in the hypoglycaemic range (SGL < 70 mg/dL), in the normoglycaemic range (70 < SGL < 140 mg/dL) and in the hyperglycaemic range (SGL > 140 mg/dL) were used to analyse their relationships with C Alb(Met O) 147 /C Alb(Met ) 147 levels.
Statistical analysis. Data are expressed as mean ± standard deviation unless stated otherwise. Differences in C Alb(Met O) 147 /C Alb(Met ) 147 levels according to clotting time were analysed using one-way ANOVA. The Mann-Whitney U test was used to compare C Alb(Met O) 147 /C Alb(Met ) 147 values between diabetic and non-diabetic groups. The remaining comparisons, evaluating the effects of treatment on C Alb(Met O) 147 /C Alb(Met ) 147 level, were performed using the Wilcoxon signed-rank test for paired data. Linear regression models were used to compare the measured values of C Alb(Met O) 147 /C Alb(Met ) 147 and to determine the correlations with age, sex, BMI, GA, HbA 1c GA/HbA 1c and biochemical parameters. Multivariate analyses were performed essentially as described 46,47 , except that age, sex, BMI, GA/HbA 1c, eGFR, uric acid, total bilirubin, HDL-cholesterol and the use of metformin, a statin or an angiotensin-converting enzyme inhibitor/angiotensin II receptor blocker were used as explanatory variables, and C Alb(Met O) 147 /C Alb(Met ) 147 was used as the objective variable. All analyses were performed using GraphPad Prism 5.02 software (GraphPad Software Inc. San Diego, CA, USA) and/or JMP ver. 5.0.1a (SAS, Cary, NC, USA). P < 0.05 was considered to represent statistical significance.

Ethics approval and consent to participate. The protocol was approved by the Kitasato University
Medical School/Hospital Ethics Committee (B17-040, B15-181) and written informed consent was obtained from all participants.
All study protocols were performed in accordance with the relevant guidelines and regulations of Kitasato University Medical School as well as the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan and under the Code of Ethics of the Helsinki Declaration.

Data availability
All data generated or analyzed during this study are included in this article.