Targeted quantification of N-1-(carboxymethyl) valine and N-1-(carboxyethyl) valine peptides of β-hemoglobin for better diagnostics in diabetes

N-1-(Deoxyfructosyl) valine (DFV) β-hemoglobin (β-Hb), commonly referred as HbA1c, is widely used diagnostic marker in diabetes, believed to provide glycemic status of preceding 90–120 days. However, the turnover of hemoglobin is about 120 days, the DFV-β-Hb, an early and reversible glycation product eventually may undergo irreversible advanced glycation modifications such as carboxymethylation or carboxyethylation. Hence quantification of N-1-(carboxymethyl) valine (CMV) and N-1-(carboxyethyl) valine (CEV) peptides of β-Hb would be useful in assessing actual glycemic status. Fragment ion library for synthetically glycated peptides of hemoglobin was generated by using high resolution–accurate mass spectrometry (HR/AM). Using parallel reaction monitoring, deoxyfructosylated, carboxymethylated and carboxyethylated peptides of hemoglobin were quantified in clinical samples from healthy control, pre-diabetes, diabetes and poorly controlled diabetes. For the first time, we report N-1-β-valine undergoes carboxyethylation and mass spectrometric quantification of CMV and CEV peptides of β-hemoglobin. Carboxymethylation was found to be the most abundant modification of N-1-β-valine. Both CMV-β-Hb and CEV-β-Hb peptides showed better correlation with severity of diabetes in terms of fasting glucose, postprandial glucose and microalbuminuria. This study reports carboxymethylation as a predominant modification of N-1-β-valine of Hb, and quantification of CMV-β-Hb and CEV-β-Hb could be useful parameter for assessing the severity of diabetes.


Background
Poorly controlled diabetes leads to variety of complications including blindness, amputations and, kidney failure [1]. Thus, glycemic control is crucial in management of diabetes [2]. Diagnosis and management of diabetes heavily relies on detection of levels of HbA1c [N-1-(deoxyfructosyl) valine (DFV) β-hemoglobin], which is considered as a gold standard for assessing the glycemic status over preceding 90-120 days [3,4]. In normal individuals, HbA1c ranges from 3 to 6.5 % and up to 15 % in poorly controlled diabetes [5]. However, previous studies have suggested that the HbA1c is slowly reversible, and for a given glucose concentration it eventually reaches equilibrium [6]. Furthermore precise quantification of HbA1c using available methods in routine diagnostics, such as ion exchange [7], phenylboronate affinity

Open Access
Clinical Proteomics *Correspondence: mj.kulkarni@ncl.res.in 1 Proteomics Facility, Division of Biochemical Sciences, CSIR-National Chemical Laboratory, Pune 411008, India Full list of author information is available at the end of the article [8] or antibody based chromatography [9], are technically challenging. For example, ion exchange based measurement of HbA1c is influenced by other chemical modifications such as glutathionylation [10]. Whereas boronate affinity chromatography invariably detects the cis-diol groups of glycated hemoglobin not limited to DFV but also includes other amino acids like lysine and arginine, which may result in overestimation of HbA1c [11]. In case of antibody based affinity chromatography method, lack of specificity against DFV peptide poses serious problems.
Limitations in the light of current methods to measure accurate HbA1c, we explored the chemistry of HbA1c. HbA1c is chemically N-1-(deoxyfructosyl) valine (DFV) β-hemoglobin, an early and reversible glycated product formed by non-enzymatic reaction with glucose, can possibly undergo relatively stable advanced glycation modifications such as CMV and CEV during the lifespan of erythrocytes. Therefore, the assumption that HbA1c provides glycemic status over preceding 90-120 days is debatable. In fact, elevated levels of CMV-Hb were observed in diabetic nephropathic patients by immunosassy in 2001 [12]. However, its usefulness in diagnostics has been overlooked. Advanced mass spectrometry based quantitative approaches like MRM/PRM may provide better view of glycation status of hemoglobin. In MRM, a specific precursor and fragment ions are monitored for quantitation using triple quadrupole mass spectrometers. MRM based quantification is becoming quite popular in the area of proteomics research, as this approach is able to replace expensive antibodybased quantifications like western blotting and ELISA [13]. MRM performed on high resolution mass spectrometer is called parallel reaction monitoring (PRM) or HR-MRM. Unlike MRM, in PRM it is not possible to monitor the specific fragment ion during acquisition, as third quadrupole is replaced with high resolution mass analyzers such as Orbitraps and Time of flight (TOF). In PRM, post mass spectral acquisition, extracted ion chromatograms (XICs) for selected fragment ions are used for quantitation [14]. Hence, we explored the existence of CMV and CEV modifications for possibility of discovering alternative diagnostics in diabetes using PRM. In this pursuit, for the first time, we have comprehensively characterized and quantified DFV/DFL, CMV/CML and CEV/CEL peptides of hemoglobin. The major findings of this study are N-1-β-valine undergoes carboxyethylation; and carboxymethylation is the predominant modification of N-1-β-val-Hb. Furthermore, CMV and CEV peptides correlate well with fasting glucose, postprandial glucose, and microalbuminuria. Thus, we propose that CMVβ-Hb and CEV-β-Hb provides better insight of glycemic status in diabetes.

Methods
All the chemicals were procured from Sigma-Aldrich (Sigma-Aldrich, MO, USA). MS grade solvents (Acetonitrile and Water) were obtained from J T. Baker (J T. Baker, PA, USA). RapiGest SF was procured from Waters (Waters Corporation, MA, USA). Membrane filters of 3 kDa cut off were procured from Millipore (Millipore, MA, USA).

Study design and clinical details
The current study deals with the characterization and quantification of DFV/DFL, CMV/CML and CEV/CEL modified peptides of hemoglobin from 39 clinical samples. The overview of the study design is depicted in Fig. 1a, b. The study was approved by Chellaram Diabetes Institute Ethics Committee (CDIEC), Pune, India. The written consent was obtained from all subjects prior to blood collection. Blood samples were collected from Chellaram Diabetes Institute (CDI), Pune and grouped into healthy controls (n = 9), pre-diabetes (n = 10), diabetes (n = 10) and poorly controlled diabetes (n = 10) based on their fasting, postprandial blood glucose levels and HbA1c. The descriptive characters and biochemical parameters of the study subjects are provided in Additional file 1: Table S1.

Synthesis of AGE modified hemoglobin Chemical synthesis of carboxymethyl and carboxyethyl modified hemoglobin
Hemoglobin (10 mg/mL) and sodium cyanoborohydride (0.05 M) were dissolved in 0.2 M sodium phosphate buffer (pH 7.4), to which either glyoxylic acid (0.045 M) or methylglyoxal (0.05 M) were added [15,16], the reaction mixture was incubated at 37 °C for 24 h under sterile and dark conditions to obtain carboxymethyl and carboxyethyl modified hemoglobin respectively.

Synthesis of AGE modified hemoglobin
Hemoglobin (10 mg/mL) and glucose (0.5 M) in 0.2 M sodium phosphate buffer (pH 7.4) were incubated at 37 °C for 15 days under sterile and dark conditions [15,16].

Hemoglobin extraction from clinical samples
Blood was collected from subjects in heparin-coated capillary tubes containing aprotinin (BD Bioscience). The blood samples were kept at 4 °C for 15 min, and centrifuged at 3000 rpm for 15 min to remove plasma. The pellet containing erythrocytes was washed twice with saline buffer and stored at −80 °C until use. Erythrocytes were lysed in ammonium bicarbonate buffer (50 mM) containing 0.1 % RapiGest SF with vigorous vortexing for 30 min at 4 °C [17]. Further, cell lysate was centrifuged at 20,000 rpm for 30 min and protein in the supernatant was collected.

Sample preparation and LC-MS/MS acquisition
In vitro AGE modified hemoglobin and hemoglobin isolated from clinical samples, were diluted with 50 mM ammonium bicarbonate buffer containing 0.1 % Rapi-Gest SF followed by reduction (Dithiothritol), alkylation (Iodoacetamide) and trypsin digestion. Tryptic peptides were loaded onto reverse-phase column (150 × 2.1 mm, 1.9 µm) on an UPLC Accela 1250 (Thermo Fisher Scientific) coupled to a Q-Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific) at a flow rate of 350 µL/ min over a period of 45 min gradient, using 6 gradient segments (held at 2 % solvent A over 2 min, 2-40 % A over 35 min, 40-98 % A over 2 min, held at 98 % A over 2 min, 98-2 % A over 2 min held at 2 % A for 2 min). Solvent A was ACN with 0.1 % formic acid and solvent B was aqueous with 0.1 % formic acid.

Identification of glycation modifications
Mass spectra were analyzed by using Proteome Discover software (PD 1.4.0.288, Thermo Fisher Scientific). Sequest-HT was used as a search engine for peptide identification against protein database consisting of α and β Hb (UniProt IDs: P68871 and P69905). Peptide and fragment mass tolerance was set at 10 ppm and 0.5 Da respectively. DFV/DFL (+162.0211 Da), CMV/CML (+58.0055 Da) and CEV/CEL (+72.0211 Da) were given as variable modifications at N-terminal valine and lysine residues [15,16,18]. The false positive rate was set to 1 % in the PD workflow. Glycated peptides were manually validated for accurate mass shift in precursor ion due to modification, and presence of fragments retaining modification. Subsequently, fragment ion library was generated for DFV/DFL, CMV/CML, and CEV/CEL modified hemoglobin peptides (Additional file 2: Table S2).

Parallel reaction monitoring (PRM) data acquisition
A total of 42 m/z corresponding to DFV/DFL, CMV/ CML and CEV/CEL modified peptides observed in vitro were used for PRM analysis in clinical samples  Table S2). PRM was performed at a resolution of 17,500, isolation window of 2 Da, target AGC value of 2e 5 , maximum injection time of 150 ms, MSX count 1, and isolation offset 0.5 m/z. Fragmentation of precursor peptides was performed by high energy collision induced dissociation (HCD) of 28 eV.

Quantification of glycated peptides
Raw files obtained from PRM acquisitions were analyzed by Pinpoint software (version 1.4.0 Thermo Fisher Scientific). Ion chromatograms were extracted with a mass tolerance of 10 ppm using all b and y ions. The composite MS/MS spectrum of each targeted peptide was reconstructed from the area under curves (AUCs) of all the transitions (Fig. 4b, Additional file 3). The AUC for each modified peptide was deduced by addition of corresponding common co-eluted fragment ions AUCs observed in all replication and groups [14].

Statistical analysis
Mass spectra for clinical samples were acquired in technical triplicates. Graphpad Prism 5 ver 5.01 was used to determine statistical significance. Two-way ANOVA followed by Tukey's test and Bonferroni posttests were performed for the AUC of quantified glycated peptides and their fold change in prediabetes, diabetes and poorly controlled diabetes. The AUCs of the modified peptides were correlated with clinical parameters blood glucose, HbA1c, microlabuminuria, etc. by Pearson's correlation method. p values less than 0.05 were considered as significant (Additional file 4).

Construction of fragment ion library for synthetically glycated hemoglobin peptides
Mass-spectrometry based targeted quantification approaches like MRM and PRM heavily relies on fragment ion or spectral library [14,16]. Therefore, fragment ion library for synthetically glycated hemoglobin peptides was constructed. DFV/DFL, CMV/CML and CEV/ CEL containing peptides of hemoglobin were studied by LC-MS/MS analysis. Initially the AGE modified peptides identified by Proteome Discoverer were analyzed for exact mass increment of 162.0211 Da (DFV/DFL), 58.0055 Da (CMV/CML) and 72.0211 Da (CEV/CEL) in the precursor ions. Subsequently, the MS/MS spectra were manually inspected for presence of modified fragment ions as described earlier [16,18]. For example, the N-terminal peptide of β-Hb "VHLTPEEK" was interrogated for deoxyfructosyl (DF), carboxymethyl (CM),  Table S2 and their MS/MS annotations are provided in Additional file 3.
A total of 42 glycated peptides (Additional file 2: Table  S2) were identified in DDA analysis. Glyoxylic acid modified hemoglobin had maximum number of glycated peptides (20), followed by glucose modified (13) and methylglyoxal (9). Irrespective of glycating agent used, 26 peptides were observed to be carboxymethylated. Together these glycated peptides involved 19 sites, 9 from α-Hb and 10 from β-Hb (Additional file 2: Table S2). In case of glucose modified hemoglobin, apart from deoxyfructosylation of N-1-β-Val, it was interesting to observe carboxymethylation and carboxyethylation of N-1-β-Val and other lysine residues.

Quantification of advanced glycated peptides of hemoglobin in clinical samples
Using precursor ion information, DFV/DFL, CMV/ CML, and CEV/CEL modified hemoglobin peptides from Table S2 were quantified by PRM followed by PIN-POINT analysis in clinical samples. Both b and y ions from all the transitions were used to generate the composite MS/MS spectrum for each targeted peptide. The co-eluted fragment ions of the glycated precursor were reconfirmed for the presence of fragment ions observed in the synthetically glycated hemoglobin. For relative quantification of modified peptides, the total AUCs of corresponding common co-eluted fragment ions observed in all replication and groups were used. The list of modified peptides and their corresponding fragment ions used for quantification is mentioned in Table 1. Total 26 glycated peptides (15 from α-Hb and 11 from β-Hb) were identified and quantified in clinical samples (Fig. 3a, b). Fold change in AUCs was calculated for all the modified peptides across different clinical conditions and is represented in Fig. 3a, b. A total of 13 peptides of α-Hb and 9 peptides of β-Hb were significantly elevated in poorly controlled diabetes as depicted in Fig. 3a In context of diagnostics of diabetes, DFV of β-Hb is quantified. Therefore, here we have emphatically discussed the quantification of advanced glycated peptides involving N-1-β-Val. Figure 4a displays the Log (10) values of average TIC, and average AUC of CMV, CEV and DFV peptides, indicating that there was no major variation in TIC across different samples, although the AUC of CMV, CEV and DFV increased with severity of diabetes. Subsequently, quantification of CMV, CEV and DFV peptides was performed by PINPOINT analysis. For the first time, we have observed that N-1-β-valine undergoes carboxyethylation; and carboxymethylation as a predominant modification of N-1-β-Val in all the subjects (Fig. 4b, c). Apart from V*(CM)HLTPEEK (m/z-505.75), an additional peptide with 2 missed cleavages V*(CM) HLTPEEK*(CM)SAV TALWGK*(CM)VNVDEVG-GEALGR (m/z-1112.56) involving CMV modification was found to be significantly elevated in all the diabetic conditions.    Although previous studies have reported that CMV-Hb was more prevalent in diabetic patients than in healthy subjects, and was proposed to be a valuable marker for the progression of diabetic nephropathy, and oxidative stress [13,19], this study reports that CMV is the predominant modification of β-N-1-Val in all the clinical conditions (Fig. 4c). Furthermore, CEV showed higher fold change in AUC followed by CMV and DFV, in all diabetic conditions compared to healthy controls (Fig. 4d, Additional file 5) suggesting its usefulness as marker in the early stages of diabetes. Thus, monitoring CMV and CEV, along with DFV could be useful to assess the glycemic status in diabetes.

Correlation of DFV, CMV and CEV modified peptides of β-Hb with severity of diabetes
Next we correlated the levels of the modified peptides with clinical parameters by using Pearson's correlation method ( Table 2, Additional file 4). However, in the context of HbA1c, glycated peptides involving β-N-1-Val are discussed here. All three DFV, CMV and CEV peptides showed significant positive correlation with each other. However, CMV and CEV peptides showed significant correlation with HbA1c (r = 0.729*** and r = 0.549***), fasting blood glucose (r = 0.515** and r = 0.304*) and postprandial blood glucose (r = 0.696*** and r = 0.673***). While DFV peptide showed significant positive correlation with postprandial glucose (r = 0.284*) and HbA1c (r = 0.283*). Interestingly, the CMV peptide also showed a relatively better correlation with microalbuminuria (r = 0.305) and negative correlation with HDL (r = −0.362), although both were statistically not significant. This study suggests that CMV and CEV are better correlated with severity of diabetes.
Proteins undergo heterogeneous modifications during dynamic advanced glycation reactions [20]. Generally lysine residues of protein reacts with glucose to form fructosyl lysine, which eventually undergo structural rearrangement involving fragmentation and glyoxidation to form AGEs like carboxymethyl lysine (CML) and carboxyethyl lysine (CEL). The occurrence of CML and CEL is accelerated in hyperglycemic conditions. Especially long lived proteins are known to have elevated levels of CML modifications [21]. Therefore, these AGE modifications may reflect the glycemic status over a long time. Despite this fact, HbA1c i.e. DFV-β-Hb, an early glycated product is used for diagnosis of diabetes. Furthermore, given the turnover of hemoglobin is 120 days, it is expected that DFV-β-Hb, the first and reversible modification of glycation, to undergo stable AGE modifications like carboxymethylation and carboxyethylation (Fig. 5). Moreover, elevated levels of dicarbonyl such as glyoxal and methylglyoxal can also cause carboxymethylation and carboxyethylation [20]. Our observation of carboxymethylation and carboxyethylation of N-1-β-valine supports the hypothesis that hemoglobin undergoes advanced glycation modifications.

Conclusion
This is the first study that describes detailed characterization and quantification of advanced glycated peptides of hemoglobin including CMV and CEV peptides of β-Hb. Since advanced glycation modifications are associated with diabetic complications [22], quantification of CMV and CEV peptides may be more meaningful in the evaluation of glycemic status in diabetes. In addition, CMV and CEV show better correlation with severity of Fig. 3 a Relative fold change in AUC for glycated peptides of a α-Hb and b β-Hb with respect to healthy control. Statistical analysis was performed by two-way ANOVA followed by Tukey's test. PD-prediabetes, D-diabetes and PCD-poorly controlled diabetes (*p < 0.05, **p < 0.005, ***p < 0.0005) Fig. 4 a Displays the Log (10) values of average of TIC and average of AUC of CMV, CEV and DFV peptides, indicating that there was no major variation in TIC across different samples, although the AUC of CMV, CEV and DFV increased with severity of diabetes. b Spectra depicting co-eluted fragment ions of DFV, CMV and CEV peptides of β-hemoglobin using PINPOINT software. c AUC of DFV, CMV and CEV peptides of β-hemoglobin depicting relative abundance. d Relative fold change in AUC for DFV, CMV and CEV peptides of β-hemoglobin by PRM. Statistical analysis was performed by two-way ANOVA followed by Tukey's test and Bonferonnis posttests. Clinical groups are represented as C control, PD prediabetes, D diabetes, PCD poorly controlled diabetes (*p < 0.05, **p < 0.005, ***p < 0.0005)