Comprehensive quantification of carboxymethyllysine modified peptides in human plasma

Abstract

A prolonged hyperglycemic condition in diabetes mellitus results in glycation of plasma proteins. N(ε)-Carboxymethyllysine (CML) is a well-known protein advanced glycation end product, and one of its mechanisms of formation is through further oxidation of Amadori compound modified lysine (AML). Unlike enrichment of AML peptides using boronate affinity, biochemical enrichment methods are scarce for comprehensive profiling of CML-modified peptides. To address this problem, we used AML peptide sequence and site of modification as template library to identify and quantify CML peptides. In this study, a parallel reaction monitoring workflow was developed to comprehensively quantify CML modified peptides in Type 1 diabetic subjects’ plasma with good and poor glycemic control (n = 20 each). A total of 58 CML modified peptides were quantified, which represented 57 CML modification sites in 19 different proteins. Out of the 58 peptides, five peptides were significantly higher in poor glycemic control samples with the area under the receiver operating characteristic curve ≥ 0.83. These peptides could serve as promising indicators of glycemic control in Type 1 diabetes management.

Graphical Abstract

Introduction

Type 1 diabetes (T1D) is a chronic disease affecting millions of people worldwide, nearly 1.6 million in the US alone in 2018 according to the American Diabetes Association. T1D is an autoimmune disease that occurs in genetically susceptible individuals, where immune dysregulated pathogenic T cells destroy the insulin producing β cells in the islets of Langerhans [1]. The resulting chronic hyperglycemia leads to the development of diabetic microvascular pathology and eventually to blindness, renal failure, and nerve damage [2]. Good glycemic control is implicated in persistent reduction of progressive diabetic complications [3]. Consequently, optimal glycemic control is essential for the management of diabetes.

The hyperglycemic condition of diabetes results in chemical modification of plasma proteins by a nonenzymatic reaction known as glycation [4]. The reaction between the carbonyl group of the reducing sugar and primary amine of proteins first forms unstable Schiff’s base and then forms a reversible fructosamine/Amadori product; further rearrangement and dehydration reactions lead to the formation of advanced glycation end products (AGEs) [4]. Irreversible glycation of the N-terminal amino group of the hemoglobin β chain called HbA1c reflects the average concentration of glucose in the blood over the past 8 to 10 weeks and is considered as gold standard to measure glycemic control [5]. However, factors like glucose gradients across the red blood cell membrane, intracellular pH, glycation rate, anemia, blood loss, splenomegaly, and iron deficiency may affect HbA1c levels [6, 7]. Subsequently, Amadori-modified glycated albumin (GA) or its association with HbA1c are being assessed as alternative to HbA1c [8–10]. Further, apart from GA, other plasma proteins were evaluated for sensitivity toward glycemic control [10]. For the early glycation adducts, Schiff’s base and fructosamine/Amadori product, upon reaching equilibrium between glucose over a period, the measured levels attain steady-state value that does not increase as a function of time [4]. Meanwhile, Amadori product tends to undergo rearrangement to form stable AGEs. Carboxymethyl lysine (CML) and carboxyethyl lysine (CEL) are two major AGEs and serve as biomarkers of oxidative stress resulting from both carbohydrate and lipid oxidation reactions [11]. Indeed, CML and CEL constitute up to 80% of total AGEs [12]. Given the heterogeneous nature of AGEs, it is necessary to explore CML modified plasma proteins that could serve as stable and sensitive markers to glycemic control index.

CML being one of the most abundant AGEs has been extensively studied as it is implicated in various pathophysiology and its formation by different mechanisms. CML formation is found to occur in three different ways. First, the further oxidation of Amadori product modified lysine (AML) [13]. Second by glucose oxidation resulted glyoxal reaction with lysine [14] and third by glyoxal derived from Schiff base decomposition reaction with lysine [15]. Endogenously formed CML is more clinically relevant and has been studied mainly by GC—MS [16], enzyme-linked immunosorbent assay [17], fluorescence [18], multiple reaction monitoring (MRM) [19] and isotope dilution methods [20]. However, all of the above-mentioned studies measure free CML in enzymatic hydrolysates of either plasma, tissues, or cellular and extracellular proteins but fail in providing information about modified protein and the modification site, which are critical to claim CML as endogenous and understand the pathological roles of these proteins. Previously, CML and CEL modifications are reported in isolated lens protein crystallin and collagen from connective tissue [21, 22]. Immunochemical approaches have been adopted to characterize the structural aspects of AGEs [23, 24]. Monoclonal antibodies were generated against CML and used to quantify the CML adducts in small intramyocardial arteries in the heart tissue of diabetic subjects [17]. Similarly, CML modified cytokeratin 10 was identified from stratum corneum lysates from epidermis by using two-dimensional gel electrophoresis and amino acid sequence analysis [25]. Further, characteristic fragmentation pattern and modification specific reporter ions for CML (m/z 142.1 and 187.1) and CEL (m/z 156.1 and 201.1) were established and by using nano-UPLC-Orbitrap-MS, 21 CML sites in 17 proteins in pooled plasma of Type 2 diabetes (T2D) were identified [26]. Similarly, by using targeted sequential window acquisition of all theoretical mass spectra (SWATH) analysis, 4 AML, 7 CML, and 2 CEL modified peptides representing nine sites on albumin were quantified in pooled plasma samples of control, prediabetes, T2D, and T2D-microalbuminuria [27].

In this study, we report a unique approach for the identification and comprehensive quantification of CML modified peptides by utilizing an in-house generated AML peptide sequence as templates. Construction of the library relies on our previously developed online 2D-LC—MS/MS platform for accurate and comprehensive identification of AML sequence specific modification sites. Parallel reaction monitoring (PRM) based method was then used to validate and quantify these in silico generated CML peptides. A total of 58 CML-modified peptides representing 57 CML modification sites in 19 different proteins were quantified in good and poor glycemic control plasma samples from T1D subjects, out of which 5 CML modified peptides were found to be significantly higher in the poor glycemic control than in the good control group with Receiver Operator Characteristic (ROC) area under curve ≥ 0.83 and showed positive correlation with HbA1c levels.

Experimental Section

Study design

Overview of the complete study design is shown in Figure 1. In-house AML modified peptide sequence library (310 peptides) was generated from identified AML peptides in different human plasma cohorts using online 2D LC MS/MS platform [10]. The AML sequence library served as a surrogate template in generation of CML-modified peptides. Based on the library sequences as template, accurate m/z and predicted retention time were established for CML-modified peptides. Representative synthetic CML peptides were used for RT calibration and for assessment of their resistance to trypsin. A PRM-based method was developed to quantify CML-modified peptide to identify plausible candidate markers of glycemic control.

Figure 1.

Overview of the overall study design. In-house AML peptide library was constructed from identified AML peptides in different human plasma cohorts using online 2D LC—MS/MS platform and served as surrogate sequence template for CML peptides. Synthetic CML peptides were used for RT calibration and to evaluate protease resistance. Scheduled LC—PRM assay was optimized to quantify CML peptides and their correlation to glycemic control was studied.

Chemicals and Materials

All chemicals and reagents were purchased from Sigma-Aldrich (St. Louis, MO) unless mentioned otherwise. Custom synthesized CML modified, stable isotopically labeled peptides and MS-grade modified trypsin were purchased from ThermoFisher Scientific. Human plasma samples were procured through T1D Exchange Living Biobank, collected from T1D patients with good and poor glycemic control based on HBA1c levels (n = 20 in each group). Descriptive characters including age, gender, and race are provided in Table S1, Supplemental File 1. Upon receiving shipment in dry ice, the samples were stored at − 80 °C until further use. This study was approved by the University of North Carolina Greensboro Institutional Review Board.

Determination of the integrity of CML modified peptides to trypsinization

Custom synthesized CML modified and stable isotopically labelled peptides K^#VPQVSTPTLVEVSR, K^#QTALVELVK and DSSLC*K^#LC*MGSGLNLC*EPNNK (K^#-CML modified lysine, C*-Carbamidomethylated cysteine, C-terminal R or K was heavy isotope labeled with ¹³C and ¹⁵N) were incubated with or without MS-grade trypsin (1: 50, trypsin to peptide ratio) overnight at 37°C, without reduction with dithiothreitol (DTT) and alkylation with iodoacetamide (IAA). The samples were analyzed using UltiMate 3000 RSLC nano system online coupled to a QExactive HF Orbitrap MS (ThermoFisher Scientific) to assess the integrity of CML modified peptides to trypsin proteolysis.

In-solution tryptic digestion of human plasma samples

Protein concentration of the human plasma samples was determined by BCA assay before being subjected to in-solution tryptic digestion. Then 1 mg of human plasma proteins was dissolved in 8 M urea. The proteome was reduced with 10 mM DTT for 1 h at 37°C and alkylated with 40 mM IAA for 1h at 37°C in dark. Samples were diluted with 50 mM ammonium bicarbonate to bring down the urea concentration below 1 M. Calcium chloride (1 mM) was added to the solution before addition of MS-grade trypsin (1: 50, trypsin to protein ratio), and the proteome was hydrolyzed overnight at 37°C. The samples were cleaned up by passing through C18 SPE cartridges (Waters), and the eluted peptides were concentrated by either using stream of nitrogen gas or vacuum concentrator. The peptide concentration was again determined by BCA assay before LC—MS/MS analysis.

Online 2D-LC—MS/MS platform to enrich and quantify AML modified peptides

The online 2D-LC—MS/MS platform was setup as described in previous study [28] on Ultimate3000 RSLC nano system along with a VWD-3400 UV detector. An in-house packed boronate affinity column (1 mm × 5 cm) with Glycogel II boronate affinity gel was used as the first dimension to enrich AML peptides and a C18 analytical column (2 μm, 100 Å particles, 75 μm × 50 cm) to separate enriched peptides in the second dimension. The RSLC nano system was online coupled to a QExactive HF Orbitrap MS with an EASY Spray interface. Buffers (A, 50 mM ammonium acetate, pH 8.1, in water; B, 0.1 M acetic acid in water) and a stepwise gradient (0% B for 7 min; 0–100% B in 0.1 min; 100% B for 13 min; 100–0% B in 0.1 min; 0% B for 20 min) at flow rate 30 μL/min were used in the first dimension LC to enrich AML peptides. The unbound peptides were first eluted to waste and AML peptides were diverted to trap column at 15 min and released at 19 min to separate on C18 analytical column. Mobile phases for the second-dimension peptide separation (A, 0.1% formic acid (FA) in water; B, 0.1% FA in acetonitrile) were delivered by the RSLC nano system with flow rate 0.3 μL/min. The gradient time was 90 min (0–19 min, 4% B; 19–19.1 min, 5% B; 19.1–63 min, 35% B; 63– 63.1 min, 90% B; 63.1–77 min, 90% B; 77–77.1 min, 4% B; 77.1–90 min, 4% B). Eluted peptides were ionized in positive ion mode and analyzed by QExactive HF MS with HCD fragmentation. The MS data set was acquired in data dependent mode with one full MS scan (resolution 60000 at m/z 200) followed by 15 MS/MS scans (resolution 15000 at m/z 200). Other parameters for acquisition included: full MS AGC target of 1 × 10⁶, MS/MS AGC target of 1 × 10⁵, dynamic exclusion of 20 s, and mass isolation window of 1.4 m/z, normalized collision energy at 28.

For the global plasma proteome profiling, a 1D-LC—MS/MS method was performed on the same QExactive HF Orbitrap instrument without boronate affinity enrichment. The mobile phases for peptide separation were the same as for the second dimension of the online 2D platform. The gradient time was 115 min (0–3 min, 4% B: 3–115 min 75% B). MS/MS data were acquired using the same parameter as the online 2D method.

LC-PRM targeted analysis of CML peptides

The PRM workflow was adopted to validate and quantify the candidate CML modified peptides. Human plasma tryptic digests spiked in with synthetic isotopically labeled CML peptides were separated and quantified using an inclusion list of candidate CML peptides, which had the same peptide sequence as the AML modified peptides identified using the online 2D LC-MS/MS platform. The spike concentration of peptide K^#VPQVSTPTLVEVSR and K^#QTALVELVK was 36 and 56 fM, respectively. The LC-PRM data set was acquired using Ultimate3000 RSLC nano and Orbitrap Exploris 240 (ThermoFisher Scientific). Plasma tryptic digest were resolved on a C18 analytical column (2 μm, 100 Å particles, 75 μm × 50 cm). Mobile phases for peptide separation (A, 0.1% FA in water; B, 0.1% FA in acetonitrile) were delivered by the RSLC nano system with flow rate 0.25 μL/min. The gradient time was 110 min (0–110 min, 45% B; 110.1–115 min, 75% B; 115.1–135 min, 4% B). Eluted peptides were ionized in positive mode and analyzed by Orbitrap Exploris 240. MS data set was acquired in full scan MS followed by PRM. MS parameters of full scan MS: resolution 60 000, custom AGC target and auto injection time; of PRM: resolution 60 000, standard AGC target and auto injection time and isolation window of 1.4 m/z.

Database search

The acquired 2D-LC—MS/MS and 1D-LC—MS/MS data sets were analyzed by using SEQUEST HT in Proteome Discoverer 2.2 (ThermoFisher Scientific) with a UniProt human database for protein identification. The search parameters included were variable modification- oxidation of methionine (15.99 Da), AML modification (162.0528 Da) (AML modification was not included while performing 1D-LC—MS/MS global proteomics data set) and fixed modification of cysteine carbamidomethylation (57.021464 Da). Peptide identification was performed using a 10-ppm precursor ion tolerance; the product ion tolerance was 0.05 Da. Peptide-spectrum matches were adjusted to 1% FDR. Skyline (64-bit, 20.1.0.155) [29] was used to analyze the PRM data set, and only the commonly matched high intense transitions in all samples were considered for quantification, preferably y ions were used. The peak area of each peptide reported by Skyline was further normalized with spiked in standard isotopically labeled CML peptides before performing the statistical analysis.

Statistical analysis

For the PRM results, ROC curve and Pearson correlation analysis were performed for each peptide using GraphPad Prism (64-bit, 8.3.1) (GraphPad Software). For ROC curve analysis, the 95% confidence interval was calculated using the Wilson/Brown method and area under the curve was reported as a percentage. An unpaired t test was performed with two tailed P values. The global proteome profiling data set obtained from Proteome Discoverer was analyzed by using Perseus software (Version 1.6.14.0) as described in a previous study (10). The Gene Ontology (GO) analyses for biological pathway, cellular component, molecular function, disease, and pathway enrichment of proteins were analyzed online by using DAVID Bioinformatics Resources 6.8. [30]

Results and Discussion

CML modification is resistant to trypsin proteolysis

Since the rationale is to use the AML modified peptide as template sequence for CML modified peptide, it is important to make sure that the CML modification is resistant to trypsin cleavage. Stable isotopically labeled CML peptides (i) K^#VPQVSTPTLVEVSR (m/z 854.4775, charge state 2+), (ii) K^#QTALVELVK (m/z-597.8618, 2+), and (iii) DSSLC*K^#LC*MGSGLNLC*EPNNK (m/z 822.0304, 3+) were synthesized according to the AML peptide sequence and incubated with and without trypsin; precursor ion intensities were measured to evaluate the integrity of these peptides. The precursor ion peak area was determined by extracted ion chromatogram (XIC), and ratios of peak area with and without trypsin equal to 1, suggesting no effect of trypsin on the integrity of CML-modified peptides [Table S2, Supplemental File 1]. Thus, it supports our novel approach of using AML modified peptide sequence to predict CML modified peptide harboring same site of modification.

Construction of CML peptide library from AML peptides

In each pathophysiological condition in vivo, the ratio of glycated to unmodified protein is low. Thus, the enrichment strategy using boronate affinity chromatography was utilized for AML-modified peptides [31], whereas no such enrichment strategies are available for CML modified peptides. Prediction algorithms are available for a few post-translational modifications (PTMs) which originate from the action of sequence specific enzymes [32]. However, such a prediction algorithm is unlikely to be feasible for a spontaneous chemical modification such as CML. To this end, and as demonstrated above for resistance to trypsin proteolysis, we used AML peptide as a template to predict CML peptide as investigation of CML modification on the same site of AML in peptide not only provides direct evidence of its formation from Amadori product but also underlines the clinical relevance of modification site and the protein bearing it. The predicted CML peptides are then validated and quantified, similar strategy has been used in targeted quantification workflows including RPM, MRM, and SWATH [33, 34].

For the AML-modified peptide library, the AML peptides identified from previous human plasma studies in our laboratory, including those using online 2D LC—MS/MS platform were merged [10, 28], which resulted in 310 AML peptides as putative peptide sequences harboring same modification site. Further, the retention times (RT) of the representative AML peptides (acquired using 2D LC—MS/MS platform) were correlated to that of the corresponding CML modified peptides (acquired using 1D LC—MS/MS platform), and a good correlation was achieved as depicted in Figure 2, consequently RTs were predicted for the shortlisted 270 corresponding CML modified peptides as a base for RT scheduling window in the PRM method, and then the exact m/z was calculated and verified by using Skyline software.

Figure 2.

Retention time correlation between AML and CML peptides identified in 2D and 1D LC—MS/MS platform.

Quantification of CML modified peptides in T1D plasma by LC-PRM workflow

To comprehensively quantify low abundant CML modified peptides, a PRM workflow was adopted and optimized for higher sensitivity and selectivity. Among the 270 potential CML peptides derived from the AML peptide library, a total of 58 were quantified in good and poor glycemic control plasma samples. A representative MS2 spectrum is depicted in Figure 3 (the MS2 spectrum of each peptide is provided in Supplemental File 2). In the pooled plasma digest of 40 samples, the CML modified peptides quantified were more than 100. Due to the inconsistency of their quantification across all samples because of low abundance, a total of 58 CML modified peptides are reported in the final study. However, this number was found to be higher in in vitro glycated plasma (data not shown). These 58 peptides represented 57 CML modification sites in 19 different proteins which is higher than previously reported by using targeted nano-UPLC-Orbitrap-MS in T2D plasma sample (21 CML sites in 17 different proteins) [26]. Out of the 19 proteins, 24 CML-modified peptides were found in serum albumin, eight in serotransferrin, and five in apolipoprotein A-I which represent the highest level of modifications on proteins. The details of modification site, sequence, protein, fold changes are provided in Table S3, Supplemental file 1. Out of 58 quantified CML peptides, five peptides were found to have significantly higher fold changes in poor compared to good glycemic control samples with ROC AUC ≥ 0.83 (Figure 4, detailed information in Table 1). Moreover, these peptides also showed positive correlation with HbA1c levels where the abundance of CML-modified peptides are higher in poor glycemic control samples, similarly as higher HbA1c levels in these poor glycemic control samples (Figure 5), but the correlation overall is not very strong.

Figure 3.

Representative MS/MS spectrum and signature ions (m/z 142.1 and 187.1) of CML-modified peptide. Sequence: K^#VPQVSTPTLVEVSR, K1-Carboxymethyl (58.00548 Da), Charge: +3, Monoisotopic m/z: 566.65318 Da (+0.59 mmu/+1.04 ppm), MH+: 1697.94500 Da, XCorr:5.30, Percolator q-Value:0.0e0, Percolator PEP:4.4e-7, fragment match tolerance used for search: 0.05 Da.

Figure 4.

ROC curve and dot plot of CML peptides with ROC AUC ≥ 0.83 in differentiation of poor and good glycemic control samples. Green and red color represents good and poor glycemic control samples, respectively.

Table 1.

Significantly Upregulated CML-Modified Peptides in Poor Control Samples in Comparison with Samples with Good Glycemic Control (a)

Accession	Protein	CML Peptide	SOM	Log2 FC	T-tests	ROC AUC	Protein level change
P00738	Haptoglobin	AVGDK#LPEC*EAVC*GKPK	K82	1.03	9.65E-06	0.87	NC
P00739	Haptoglobin-related protein	VMPIC*LPSK#NYAEVGR	K212	0.98	8.28E-05	0.89	NC
P01834	Immunoglobulin kappa constant	VDNALQSGNSQESVTEQDSK#DSTYSLSSTLTLSK	K62	0.95	5.53E-08	0.96	NC
P01857	Immunoglobulin heavy constant gamma 1	GPSVFPLAPSS K#STSGGTAALG C*LVK	K16	0.51	2.81E-04	0.83	NC
P01871	Immunoglobulin heavy constant mu	EG K#QVGSGVTTDQVQAEAK	K153	0.48	3.71E-05	0.86	NC

^a(SOM- Site of modification; Log2 fold change; Student t-test P value, NC-No change)

Figure 5.

Pearson correlation between the abundances of the 5 significantly upregulated CML peptides and the levels of HbA1c in good and poor glycemic control samples. Green and red color represents good and poor glycemic control samples, respectively.

As the most abundant plasma protein serum albumin attracts attention because of its role in pathophysiology and candidature for alternative marker of glycemic index to HbA1c. Previously serum albumin is studied with great interest for its AGE modification. In current study 24 CML modified peptides of serum albumin were quantified representing 32 different sites of modification (K36, K75, K88, K117, K186, K198, K214, K229, K236, K249, K257, K300, K337, K375, K383, K396, K402, K426, K438, K489, K499, K549, K569, and K588), the CML modified sites reported in this study are higher than previous studies [Table S4, Supplemental File 1] [26, 27, 35–38]. Along with serum albumin, alpha-1B-glycoprotein, apolipoprotein A-I, apolipoprotein A-II, fibrinogen beta chain, haptoglobin, haptoglobin-related protein, hemopexin, immunoglobulin heavy constant alpha 1, immunoglobulin heavy constant gamma 1 & 2, immunoglobulin heavy constant mu, immunoglobulin kappa constant, immunoglobulin lambda constant 2 & 7, immunoglobulin lambda-1 light chain, immunoglobulin lambda-like polypeptide 5, and serotransferrin protein were found to be CML modified in contrast to the previous study where no abundant proteins were found to be CML modified [26]. Similarly, other CML-modified low abundant proteins identified by others were not identified in this study [26], which may be because such AML-modified peptides were not enriched in the boronate affinity column and therefore did not end up in the AML/CML library. As reported in the previous study, occurrence of the signature ions m/z 142.1 and 187.1 was evaluated for 58 individual peptides. However, the occurrence of these signature ions was found to be somewhat inconsistent [Table S5, Supplemental File 1] [26], although m/z 187.1 was much more prevalent than m/z 142.1. While the exact nature of missing signature ions is elusive, we suspect that peptide length, amino acid sequence, composition and charge state of the precursor ion may play a role.

The change of CML modification can be either from the change at the parent protein level or the change at the level of CML peptide. To rule out changes at the protein level, a global proteome profile of undepleted plasma was performed and levels of protein and peptide were compared. In 1D-LC—MS/MS based global proteome profiling, the quantified peptides were primarily unmodified peptides, and no significant fold change was observed in the proteins containing CML peptides listed in Table1 [detailed information in Table S6, Supplemental File 1].

To further understand the role of those 20 parent proteins bearing CML modification, GO annotation analysis and disease and pathway enrichment analyses were performed [Table S7, Supplemental File 1]. The GO analysis indicated that, these proteins are functionally involved in platelet degranulation, complement activation cascade and innate immune response. The enriched diseases involved T2D, cardiovascular diseases, and atherosclerosis. The results of enrichment analysis corroborate with the fact that, T1D is primarily an immune response dysregulation destroying β-cells and causing a higher occurrence rate of cardiovascular diseases and atherosclerosis in diabetic subjects compared to the regular population. Moreover, the complement system which is an effector for both adaptive and innate immunity has been shown to promote inflammation, proliferation, thrombosis and associated with pathogenic role in the development of diabetic complications including nephropathy [39,40].

HbA1c is considered as the gold standard to assess the glycemic control in diabetes. From a clinical point of view, many variables affect the glycemic control, level, and site of CML modifications. Along these lines, correlation between the HbA1c and CML modification becomes important. In a study with T2D subjects involving insulin treatment, HbA1c levels were improved with treatment and not the CML levels suggesting CML levels are influenced by other factors in addition to the overall glycemia [41]. Similarly, in a study with T2D subjects with and without ischemic heart disease, no correlation was observed between HbA1c and serum CML levels [42]. Interestingly, skin autofluorescence of AGEs values were positively correlated with HbA1c in children with T1D [43]. Further, the free and protein bound serum acid hydrolysates of CML levels were positively correlated with HbA1c in children with T1D [44]. Evidently, the correlation of CML levels and HbA1c in T1D is still debatable. In this respect, having selective assays to measure the free and protein bound CMLs separately very likely can help define the correlations between HbA1C and CML in diabetes.

Conclusions

In this study we report a unique approach for the identification and quantification of CML-modified peptides by utilizing an in-house curated AML peptide library as sequence template, based on CML as further oxidation product of AML at the same lysine site. Targeted profiling of CML peptides in T1D human plasma samples using an optimized LC-PRM workflow resulted in quantification of 58 peptides representing 57 CML modification sites in 19 different plasma proteins, with serum albumin showing highest number CML modified sites. Therefore, the use of template library coupled with targeted analysis proves to be a feasible strategy in comprehensive quantification of AML-derived CML modified peptides. In this respect, the workflow that we developed may be analogously extended to analysis of other difficult to enrich PTM peptides. However, this approach is limited by the AML peptides curated in the library which currently does not include peptides from very low abundant proteins. Depletion of high abundant proteins may help to reveal sensitive sites of CML in low abundant proteins including kinases, transcription factors that could play crucial role in the pathogenesis of diabetic complications. In addition, this approach is also limited to CML peptides sharing the same modification site with AML peptides; in this respect, separate identification of glyoxal-derived CML-modified peptides that may have unique modification sites other than the AML ones would be complementary to the approach implemented in this work. In this study, the peptide-bound CML data suggests a positive correlation with glycemic control. However, given the inconsistencies in literature, further studies with bigger cohorts are warranted. Nevertheless, the AML correlation with HbA1c was observed to be stronger in comparison with CML [10], suggesting AML are more sensitive and potentially better indicators of glycemic control.