Tandem

ARTICLE pubs.acs.org/ac

Multistep Ultrahigh Performance Liquid Chromatography/Tandem Mass Spectrometry Analysis for Untargeted Quantification of Glycating Activity and Identification of Most Relevant Glycation Products Stefan Mittelmaier and Monika Pischetsrieder* Department of Chemistry and Pharmacy, Food Chemistry, Emil Fischer Center, University of Erlangen-Nuremberg, Schuhstrasse 19, 91052 Erlangen, Germany

bS Supporting Information ABSTRACT: The use of advanced glycation end-products (AGEs) as biomarkers for diagnosis and clinical studies is still hampered by insufficient knowledge on clinically relevant structures formed from precursors associated with defined disease states. The present study conducted untargeted analysis of the glycating activity of AGEprecursors by ultrahigh performance liquid chromatography/tandem mass spectrometry multiple reaction monitoring (UHPLC/MSMSMRM), monitoring the loss of a nonapeptide as the glycation target. Thus, the glycating activities of seven important AGE-precursors were determined (glucose 13% and the reactive carbonyl compounds glucosone 39%, 3-deoxyglucosone 15%, 3-deoxygalactosone 26%, 3,4-dideoxyglucosone-3-ene 79%, methylglyoxal 94%, and glyoxal 97% peptide loss; 12 h/37 °C). Furthermore, UHPLC/ MSMS with simultaneous precursor ion scan and information-dependent acquisition of enhanced resolution spectra and subsequent product ion scan was applied for untargeted analysis of the major AGE-structures derived from various AGEprecursors. The 20 most important modifications could be assigned to 8 AGE-structures previously reported in the literature. Seven loosely bound AGEs not yet covered by conventional methods were detected and assigned to hemiaminals. Five AGE structures did not match any known products. The method can be applied to analyze glycating activity and AGE-structures formed from various other precursors under defined reaction conditions, supporting the selection and evaluation of diagnostic AGE-markers for clinical studies.

N

onenzymatic posttranslational protein modifications (nePTMs), which are mainly caused by oxidation and glycation processes, are associated with healthy metabolism and seem to be readily handled by the human organism. Under conditions of carbonyl or oxidative stress, however, protein modification rates may tremendously increase. The advanced glycation endproduct (AGE) pentosidine, for example, is 2.5-fold increased in plasma from diabetic patients and 21.5-fold increased in plasma from uremic patients compared to healthy controls.1 Additionally, uremia and diabetes lead to protein modifications, for example, by pyrraline, Nε-carboxymethyllysine, Nε-carboxyethyllysine, imidazolone derivatives, or fructosyllysine.2 4 Some nePTMs are closely related to major diabetic complications, such as retinopathy and nephropathy5 7 as well as to mortality and cardiovascular death in uremia.8 Recent studies applied proteomic tools and other mass spectrometric techniques to analyze the distribution of early glycation products in the plasma proteome of diabetic patients and controls as well as in human hemolysate9,10 and to quantify glycated hemoglobin in different blood samples.11 Early glycation products of hemoglobin and other proteins are established clinical markers for medium-term monitoring of hyperglycemia in diabetes. Early glycation products, however, are considered as r 2011 American Chemical Society

physiologically relatively inactive, whereas diabetic and uremic complications are rather caused by AGEs. Furthermore, early protein glycation products are solely derived from glucose. There is strong evidence, though, that reactive carbonyl compounds such as methylglyoxal or 3-deoxyglucosone are more important AGE-precursors than glucose itself due to their high reactivity toward proteins, which can exceed the reactivity of glucose by several orders of magnitude.12,13 Reactive carbonyl compounds are formed in vivo by spontaneous glucose degradation or by metabolic processes, particularly in the state of diabetes and uremia.14 16 Alternatively, they may derive from exogenous sources such as nutrition, smoking, or medicinal products.17 20 Because of their high biological relevance and the origin from reactive carbonyl compounds, AGE-biomarkers are more conclusive in clinical studies predicting clinical complications in diabetes and uremia than early glycation products. Despite their biological relevance, the use of AGE-markers for clinical studies is still limited for several reasons:21 (i) Although several AGEs have Received: September 28, 2011 Accepted: November 13, 2011 Published: November 13, 2011 9660

dx.doi.org/10.1021/ac2025706 | Anal. Chem. 2011, 83, 9660–9668

Analytical Chemistry been determined so far, systematic and untargeted studies for the identification of novel AGEs are missing. Thus, it is still under debate whether the most relevant AGEs have been identified so far or if the known AGEs represent only “the tip of an ice berg or the top of an ice cube”.22 (ii) The specificity of most commonly used AGE-biomarkers is often relatively low, since they can be formed from different precursors.23 Therefore, systematic studies are required to identify novel AGE biomarkers, which are formed from specific disease-associated precursors. The first step toward novel AGE biomarkers is a comprehensive analysis of the major AGE structures. Second, it is important to determine AGE structures specifically formed from AGE-precursors that are predominantly present under certain states of disease. For this purpose, the present work developed a method that allows untargeted analysis of the most important AGE-modifications using ultrahigh performance liquid chromatography-electrospray ionization-quadrupole linear ion trap-tandem mass spectrometry (UHPLC-ESI-qLIT-MSMS) with simultaneous precursor ion scan and information-dependent acquisition (IDA) of enhanced resolution (ER) spectra and subsequent product ion scan. The method was then applied to identify the most important AGE structures derived from the six major AGE-precursors. Additionally, it was possible to determine the glycating activity of the AGEprecursors independent from the products formed.

’ EXPERIMENTAL PROCEDURES Reagents and Samples. All experiments were carried out using LCMS-grade methanol and acetonitrile from Sigma Aldrich (Taufkirchen, Germany) as well as purified water from a Synergi-185 labwater-system (Millipore, Schwalbach, Germany). The target peptide used in this study (sequence GWGKGCGRG, acetylated N-terminus, purity >95%) was synthesized by JPT Peptide Technologies (Berlin, Germany). Glucosone, 3-deoxygalactosone (3-DGal) and 3,4-dideoxyglucosone-3-ene (3,4DGE) were prepared as described before with the only exception that mixed-bed ion exchanger Serdolit MB-2 (Serva, Heidelberg, Germany) was used for the initial 3-DGal purification.18,24 26 Glucose (>99.5%), glyoxal (GO), and methylglyoxal (MGO) (40% aqueous solutions) were obtained from Sigma-Aldrich and 3-deoxyglucosone (3-DG, >95%) from Chemos (Regenstauf, Germany). All other chemicals were, unless otherwise noted, purchased from Sigma-Aldrich or Acros (Geel, Belgium) and were at least of analytical grade. Incubation of Target Peptide with Dicarbonyls and Dicarbonyl Quantification. In total, 80 μL of the test compound solutions (glucose, glucosone, 3-DG, 3-DGal, 3,4-DGE, MGO, or GO) in water (2.5 mM) was mixed with 10 μL of an aqueous solution of target peptide (5 mM) and 10 μL of 1 M sodium phosphate buffer (pH 7.2). The resulting mixtures (final concentration, 0.5 mM peptide, 2 mM test compound, 100 mM buffer) were incubated at 37 °C for 4 or 12 h in a dry block shaker (500 rpm, Eppendorf, Hamburg, Germany). Additionally, samples incubated with 80 μL of pure water instead of the test compounds (heated control) and unheated samples diluted with 80 μL of water (unheated control) were prepared. To study the peptide-independent effect of heating on the test compounds, all incubations were also carried out with 10 μL of water instead of the peptide solution. All solutions were stored at 20 °C directly after incubation and thawed immediately before analysis. All samples were incubated in three independent experiments.

ARTICLE

The dicarbonyl levels were quantified by UHPLC-diode array detection (DAD) after 1:25 dilution and derivatization with ophenylenediamine (OPD) following a validated procedure described previously.18 Results were related to the levels found in unheated control solutions. General UHPLC/MSMS Setup. For UHPLC/MSMS analysis, an Ultimate 3000 RS UHPLC (degasser, binary pump, autosampler, column oven, diode array detector; Dionex, Germering, Germany) was coupled to an API 4000 QTRAP mass spectrometer equipped with an ESI-source (AB Sciex, Foster City, CA). Analyst 1.5.1 with BioAnalyst extensions was applied for instrument control as well as data acquisition and processing. An ACQUITY UPLC BEH300 C18 column (100 mm  2.1 mm, 1.7 μm particle size; Waters) was used at 30 °C with the following gradient: A, formic acid (0.1%); B, acetonitrile; flow rate, 0.3 mL/ min; [time (min)/% B] 4.0/6, 0.0/6, 6.0/24, 6.1/90, 10.0/90. Unless otherwise noted, an aliquot of 10 μL was injected. All flow eluting before 2.7 or after 7.5 min was discarded by a two-position valve prior to mass analysis. The ESI-source was operated at 500 °C with a voltage of +5500 V and nitrogen as the drying gas. Product-Independent Quantification of the Glycating Activity by UHPLC-ESI-MSMS-MRM Analysis of Target Peptide Loss. To monitor the decrease of unmodified target peptide resulting from glycation, a multiple reaction monitoring (MRM) method was developed. First, the most abundant molecule ion under the ionization conditions described above was determined by Q1MS-scanning. Subsequently, a product ion experiment was performed with a collision energy of 30 V. The three most intense peaks from the product ion spectrum were used to set up the MRM method. Each sample was analyzed in triplicate by LC/MSMS. To check linearity of the method, standard solutions with different concentrations of the target peptide (1000, 250, 100, 25, 10 nM in 100 mM sodium phosphate buffer containing 5 mM dithiothreitol (DTT)) were analyzed prior to quantification of the samples. The experiment was independently repeated on two other days. The resulting calibration curves were evaluated by linear regression analysis with a minimally acceptable correlation coefficient of 0.990. For quantitative MRM analysis, incubated samples were diluted by the factor 1000 with 5 mM DTT in 100 mM sodium phosphate buffer. For relative quantification of peptide loss, the mean area of three independent samples (three LC/MSMS runs for each) was calculated for each experiment. The final result was then calculated as loss of the initial concentration of target peptide obtained from the mean area of three unheated control samples, which was set as 100%. Within the range described above, this analytical setup can monitor up to 98% of the decrease in peptide concentration. If necessary, a lower dilution factor allows the assessment of even higher losses. Untargeted Analysis of Peptide Modification by Precursor Ion Scanning and IDA of ER Spectra. Prior to qualitative analysis, samples were diluted 10-fold with 5 mM DTT in 100 mM sodium phosphate buffer. To scan for modified peptides, an LC/ MSMS method combining precursor ion scanning and IDA of ER spectra was used. For precursor ion scans, the fragment m/z 159.1 was chosen as the product ion (see Results and Discussion) applying a collision energy of 50 V. A mass range from 400 to 800 Da was scanned within 1 s. IDA of ER spectra was triggered for the five most intense mass peaks of each spectrum, provided that the mass exceeded an intensity of 500 cps and did not trigger an acquisition in the preceding 10 s. The ER scans were acquired at a 9661

dx.doi.org/10.1021/ac2025706 |Anal. Chem. 2011, 83, 9660–9668

Analytical Chemistry

ARTICLE

Identification of the Binding Site by Product Ion Scanning. To elucidate the binding site of the three most abundant modifications of each sample, the mother ions were used as precursor ions for product ion experiments. For this purpose, the 1:10 diluted samples previously used for the screening experiments were injected, and the QTRAP enhanced product ion mode was applied with a collision energy of 35 V. Statistical Analysis. Statistical significance was calculated by a one-sided t test after evaluating the equality of variances by an f-test.

’ RESULTS AND DISCUSSION The purpose of the present study was to develop an untargeted method (i) to quantify glycating activity and reactivity of AGE-precursors and (ii) to determine the structure of the major glycation products derived from these precursors. Untargeted quantification of glycating activity of AGE-precursors was achieved by UHPLC/MSMS-MRM analysis of the target peptide loss, whereas precursor reactivity was determined by quantification of the precursor by UHPLC-ultraviolet visible spectroscopy (UV vis) analysis. Untargeted structural assigment of the major peptide modification was conducted by UHPLC-ESI-MSMS using precursor ion scan coupled to an IDA of ER scans and subsequent enhanced product ion scan analysis. The workflow is summarized in Figure 1. This setup allows performing all experiments with the same sample. A peptide was used as the glycation target. In contrast to free amino acids, all potential glycation sites can be combined in one single peptide molecule. Additionally, peptides model the reactivity of amino acid side chains in proteins more closely than free amino acids. In contrast to proteins, the use of peptides avoids the appearance of several charged species in ESI-MSMS spectra and the necessity to perform partial enzymatic protein hydrolysis prior to MS analysis, which both hinder untargeted adduct analysis. Therefore, a model peptide (GWGKGCGRG, acetylated N-terminus) was designed for the application as the glycation target. It contained lysine, cysteine, and arginine, the three amino acid residues that are the main targets for glycation in proteins. Tryptophan served as a combined UV- and mass-label (see below). The glycine residues were used as spacers between the reactive residues. Use of this model peptide allows direct comparison of the reactivity of the different amino acid side chains. If N-terminal activity shall be considered as well, the method can also be adopted applying the free peptide. Figure 1. Workflow for assessing quantitative (A, B) and qualitative (C, D) changes resulting from reactions between AGE-precursors and the target peptide.

scan speed of 250 Da/s within a 20 Da region around the triggering mass. Data analysis was carried out by Bioanalyst software from AB Sciex. First, the total ion chromatogram of the precursor ion scans from each sample was subjected to LC-MS Reconstruct. The resulting peak list for each sample was then compared with the corresponding heated control. All peaks that occurred in samples and control were discarded. Subsequently, ER spectra were analyzed for the remaining peaks to obtain better resolution and determine charge states. Finally, the molecules’ masses were calculated and the three most abundant mass shifts for each sample were further analyzed.

Untargeted Analysis of the Glycating Activity of Seven Different Precursors. In order to compare the glycating activity

of different precursors, the target peptide loss upon reaction with the test compounds was recorded. For this purpose, a UHPLC/ MSMS-MRM method was developed for the quantification of the unmodified peptide. MSMS-MRM instead of UV detection was applied, because the former distinguishes between the modified and unmodified target peptide even in the case of coelution. With this method, the target peptide eluted at 5.1 min and the mass spectrum revealed m/z 460.4 [M + 2H]2+ to be the predominant ion. Singly and multiply charged molecule ions were only found in negligible amounts. In MSMS-experiments, m/z 449.3 (y5), m/z 577.5 (y6), and m/z 634.5 (y7) were identified as the major fragments of m/z 460.4 (Figure S-1 in the Supporting Information). The most intense fragment y5 was assigned as quantifier for the MRM method, whereas y6 and y7 were used as qualifiers. The formation of cystine-dimers of the target peptide, 9662

dx.doi.org/10.1021/ac2025706 |Anal. Chem. 2011, 83, 9660–9668

Analytical Chemistry

Figure 2. Loss of target peptide after 4 or 12 h incubation (37 °C, pH 7.2) with water (heated control), glucose, or various dicarbonyls. The peak area of the target peptide measured in an unheated control sample was set as 100%. Significance of the decline was tested against the corresponding heated control (n.s. not significant, * p < 0.05, ** p < 0.01, *** p < 0.001).

which are not captured by the MRM method and may thus cause underestimation of the target peptide concentration, was avoided by adding DTT prior to mass analysis. To validate the method, a calibration curve (1000, 250, 100, 25, 10 nM of the target peptide) was recorded on three different days. The peak intensity increased proportionally to the concentration (coefficients of correlation 0.9972 0.9999). RSD calculated from three analyses for each sample was always