Characterization of Nonenzymatic Glycation on a ... - ACS Publications

Nov 7, 2007 - Lowell J. Brady, Theresa Martinez, and Alain Balland*. Department .... Qibin Zhang , Jennifer M. Ames , Richard D. Smith , John W. Bayne...
0 downloads 0 Views 397KB Size
Anal. Chem. 2007, 79, 9403-9413

Characterization of Nonenzymatic Glycation on a Monoclonal Antibody Lowell J. Brady, Theresa Martinez, and Alain Balland*

Department of Analytical Sciences, Amgen, Incorporated, 1201 Amgen Court West, Seattle, Washington 98119

We present here an improved analytical method for the analysis of glycation events in proteins. Nonenzymatic glycation of an IgG2 monoclonal antibody was studied using affinity chromatography, mass spectrometry, and chemical derivatization. Analysis of both forced-degraded and bulk-drug substance (BDS) samples showed the presence of glycated protein. A new peptide mapping procedure, incorporating derivatization using sodium borohydride, allowed the development of a sensitive method for detecting and identifying the sites of modification. When combined with tandem mass spectrometry, peptides glycated by glucose showed dramatically improved MS/MS spectra as compared to underivatized controls. Using these methods we were able to map a number of glycation sites in both forced-degraded and BDS samples that were distributed across both light and heavy chain subdomains. The combination of affinity chromatography, high-resolution mass spectrometry, and a simple derivatization procedure should allow the facile analysis of glycation for other antibody and protein samples. Glycation, originally described by Maillard,1 refers to the nonenzymatic reaction of sugars with proteins. This process involves a series of stages starting with the reaction between the reducing carbonyl function of a carbohydrate and an amino group present on proteins, usually at the epsilon amine of lysine residues or at the N-terminus of the molecule. This early step forms a Schiff base that undergoes an Amadori rearrangement to yield an aminomethyl ketone. Oxidation of the Amadori compounds by reactive oxygen species leads to reactive 1,2 dicarbonyl derivatives that further react with other amino functions on proteins to generate a complex pattern of protein modifications collectively known as advanced glycation end-products (AGE).2-4 Nonenzymatic glycation of proteins is an active field of research due to the important implications of these post-translational modifications on various pathological conditions such as diabetes, osteoarthritis, Alzheimer’s, and other diseases.5,6 Development of sensitive and * Corresponding author. E-mail: [email protected]. Phone: 206-265-8603. Fax: 206-217-4692. (1) Maillard, L. C. Compt. Rend. 1912, 154, 66-68. (2) Sell, D. R.; Monnier, V. M. J. Biol. Chem. 1989, 264, 21597-21602. (3) Thorpe, S. R.; Baynes, J. W. Amino Acids 2003, 25, 275-281. (4) DeGroot, J. Curr. Opin. Pharmacol. 2004, 4, 301-305. (5) Goldin, A.; Beckman, J. A.; Schmidt, A. M.; Creager, M. D. Circulation 2006, 114, 597-605. (6) Valcourt, U.; Merle, B.; Gineyts, E.; Viguet-Carrin, S.; Delmas, P. D.; Garnero, P. J. Biol. Chem. 2007, 282, 5691-5703. 10.1021/ac7017469 CCC: $37.00 Published on Web 11/07/2007

© 2007 American Chemical Society

specific methods to monitor early glycation products is important in the clinic, especially to monitor the effectiveness of treatments proposed to diabetic patients.7 Analysis of early glycation products is also becoming increasingly important as their role in damaging nonstructural proteins such as enzymes and serum proteins becomes more appreciated.7 Improved and more specific assays are also needed to assess the impact of novel protection strategies aimed at designing glycation inhibitors for therapeutic applications.8,9 We investigated early glycation events in drug manufacturing. Addition of hexose sugars to proteins has the potential to increase the heterogeneity of protein biopharmaceuticals, which could impact product quality and impair regulatory approval.10 In addition to its impact on manufacturing processes, recent reports have shown that glycation could affect drug formulation and half-life, due to the addition of glucose or fructose moieties derived from hydrolysis of sucrose present as a buffer excipient.11 Boronate affinity chromatography has been used for the analysis of carbohydrates12 and, to a more limited extent, for the analysis of intact proteins13 and antibodies.14 This approach takes advantage of the known affinity of cis-diol bearing sugar moieties, such as those present in glycated protein products, for phenyl boronate chromatography resin. We applied this method to both quantify the degree of glycation and enrich low levels of glycated product for subsequent analysis. Other common approaches for analysis of glycated proteins have included mass spectrometry (MS) of the intact molecule to estimate degree of modification11 and peptide mapping with tandem mass spectrometry (MS/MS) analysis to determine the location of the modification sites. However, both of these approaches have limitations. Intact mass spectrometry of large proteins, especially antibodies, does not always yield sufficiently resolved data to detect low-level components, and the presence of additional heterogeneity due to N-linked carbohydrates can (7) Lapolla, A.; Traldi, P.; Fedele, D. Clin. Biochem. 2005, 38, 103-115. (8) Reddy, V. P.; Beyaz, A. Drug Discovery Today 2006 11, 2006, 646-654. (9) Harding, J. J.; Ganea, E. Biochim. Biophys. Acta 2006, 1764, 1436-1446. (10) Kozlowski, S.; Swann, P. Adv. Drug Delivery Rev. 2006, 58, 707-722. (11) Gadgil, H.; Bondarenko, P; Pipes, G.; Rehder, D.; Mcauley, A.; Perico, N.; Dillon, T.; Ricci, M.; Treuheit, M. J. Pharm. Sci. [Online] 2007, 96, JPS 20966. (12) Yan, J.; Springsteen, G.; Deeter, S.; Wang, B. Chem. Eur. J. 2003, 9, 21932199. (13) Brena, B.; Batista-Viera, F; Ryden, L.; Porath, J. J. Chromatogr. 1992, 604, 109-l 15. (14) Harris, R. In State of the Art Analytical Methods for the Characterization of Biological Products and Assessment of Comparability; Mire-Sluis, A. R., Ed.; Developments in Biologicals; Karger: Basel, Switzerland, 2005; Vol. 122, pp 117-127.

Analytical Chemistry, Vol. 79, No. 24, December 15, 2007 9403

obscure the presence of glycation, since both show the same +162 Da mass shift. At the peptide level, glycation sites are frequently distributed across several lysine residues, reducing the intensity of signal and resulting in only subtle changes between glycated and nonglycated proteins analyzed by peptide mapping. Furthermore, MS/MS analysis of glycated peptides results in a commonly observed neutral loss phenomena. Although this behavior can be exploited to aid in discrimination between glucose and fructose moieties15 or used for precursor ion experiments,16 it also limits the ability to use MS/MS for direct site identification of the modification. Where signals are low, the poor tandem MS data obtained for glycated peptides may reduce the confidence of species assignment and require additional confirmatory experiments or cause modified peptides to be missed. We developed a set of techniques to address these issues and demonstrated their effectiveness in the analysis of a monoclonal antibody early in the development stage. To quantify the degree of glycation, we used boronate affinity chromatography combined with off-line mass spectrometry to measure and confirm low levels of glycation in antibody bulk-drug substance (BDS). To improve the detection and analysis of glycated peptides, we incorporated a derivatization step using sodium borohydride followed by trypsin cleavage and peptide map analysis with tandem MS detection. By reducing the glycated sugar moiety, the bond between the peptide and the carbohydrate is stabilized and the resulting MS/MS spectra are of high quality and do not show neutral loss behavior. Furthermore, by comparing derivatized to underivatized samples, glycated peptides can be identified by a shift in retention time. These methods provide sensitive and accurate techniques that should prove valuable to characterize glycation in proteins, even where levels are relatively low. EXPERIMENTAL SECTION Monoclonal Antibody. The CHO-expressed monoclonal antibody described in this study was manufactured at Amgen, Bothell, WA. Forced Glucosylation of an IgG2 Antibody. A purified IgG2 antibody was diluted into 100 mM Tris, pH 7.5 containing 0.1 M glucose (Sigma) and incubated at 37 °C in the dark at 10 mg/ mL. Samples were taken at various time points and exhaustively buffer exchanged into 100 mM Tris, pH 7.5 to remove the glucose. The final protein concentration was 10 mg/mL. Aliquots were stored at -70 °C until analysis. Sodium Borohydride Reduction. Sodium borohydride was made immediately prior to use at 100 mM in 0.02N NaOH. An amount of 10 µL of this solution was added to 40 µL of water and immediately mixed 1:1 with buffer-exchanged glucosylated antibody such that the final protein concentration was 5 mg/mL. The sample was mixed and incubated at room temperature for 1 h. The reaction was stopped by addition of 8 µL of 1.0 N HCl and incubated at room temperature for 5 min. The pH was adjusted by 1:1 dilution with 100 mM Tris, pH 7.5 for subsequent analysis. Intact and Reduced Mass Analysis. To evaluate the completeness of glucosylation, the glycated antibody was digested overnight with the endoglycosidase PNGaseF (New England (15) Frolov, A.; Hoffmann, P.; Hoffmann, R. J. Mass Spectrom. 2006, 41, 14591469. (16) Gadgil, H.; Bondarenko, P.; Treuheit, M.; Ren, D. Anal. Chem. 2007, 79, 5991-5999.

9404 Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

Figure 1. Comparative deconvolution of de-N-glycosylated antibody samples. The figure shows a comparison of two de-N-glycosylated samples analyzed using two different deconvolution parameters. For each sample, peak widths of 0.5 and 1.0 Da were used for deconvolution as indicated. The top panel shows a comparison of deconvolution obtained for a control sample, and the bottom panel shows that for 10 day glucosylated antibody. For each deconvoluted profile, the mass error in ppm relative to the de-N-glycosylated antibody mass of 146332.38 Da is indicated.

Biolabs) to remove the single N-linked carbohydrate. An amount of 1 µL of enzyme (500 000 U/mL) per 100 µg of antibody was used. For reduced samples, the sample was diluted to 2 mg/mL, mixed 1:1 with 20 mM DTT in 6.0 M guanidine, 100 mM Tris, pH 8.0 and incubated at 75 °C for 5 min. Both intact and reduced antibodies were introduced into the mass spectrometer by separation using a 2.1 mm × 150 mm, 300 Å polyhydroxyethyl aspartamide column (Poly LC) operated isocratically at 0.1 mL/ min on 0.1% formic acid. Under these conditions, this column isocratically separates large molecules such as proteins from salts and buffer constituents. The eluate from the column was mixed 1:1 with acetonitrile containing 2.0% formic acid using an HPLC pump and a mixing tee and directed into the inlet of an Agilent time-of-flight (TOF) mass spectrometer. For intact mass analysis data, the multiply charged mass spectra were exported to a text file, imported into MassLynx (Waters), and deconvoluted using MaxEnt1. For reduced samples, we used the Agilent deconvolution algorithm incorporated into MassHunter software. Trypsin Peptide Mapping of Antibody Samples with and without Sodium Borohydride Treatment. Glucosylated anti-

Figure 2. Comparison of raw spectra for control and forced-glucosylated de-N-glycosylated antibody. The figure shows a comparison of raw spectra for de-N-glycosylated control (A) and 10 day forced-glucosylated (B) antibody samples. The inset (C) shows a comparison of the 48+ charge state ion for each sample (overlay) with 0-4 glc indicating the degree of glycation determined for each species in the forced-glucosylated sample. The arrow indicates the region of the raw spectra that corresponds to a potential glycation in the untreated (BDS) sample.

body treated with sodium borohydride and untreated glucosylated control was reduced and alkylated using DTT and iodoacetic acid, buffer exchanged into 100 mM Tris with 0.1 M guanidine hydrochloride, and then digested with trypsin (Roche) (1:10 w/w) for 4 h at 37 °C. The peptides were separated using a 0.3 mm × 150 mm PLRP-S column (Varian, Inc.) at 9 µL/min with a water/ acetonitrile/2-propanol (10:90:10) solvent system containing 0.2% v/v TFA in the aqueous phase and 0.019% TFA in the organic phase. The column eluate was directed to the electrospray ionization (ESI) source of a ThermoFinnigan LTQ mass spectrometer and analyzed by MS and MSn. Boronate Affinity Chromatography. Affinity chromatography was used to quantify glycation degree. A phenyl-boronate affinity column (7.5 mm × 75 mm, TosoHaas) with a 7.5 mm × 3.5 mm guard column was equilibrated in 0.1 M sodium phosphate, pH 8.6, containing 400 mM sodium chloride. Deglycosylated BDS was loaded onto the column and eluted with a 0.5 M sorbitol gradient over 5 min in equilibration buffer. Both unbound and bound peaks were collected and analyzed by intact and reduced mass analysis. Peak areas at A214 nm were used to estimate the overall degree of glycation on the antibody. RESULTS AND DISCUSSION Intact and reduced mass analysis by electrospray ionization is a useful and standard tool to obtain an image of sample heterogeneity. For monoclonal antibodies, intact mass analysis using a TOF instrument provides sufficient resolution to profile the distribution of N-linked carbohydrate forms typically present in the Fc domain. However, due to the natural heterogeneity observed for these sugars, the presence of other, lower level species may be obscured. Further complication may arise due to the potential for artifacts during the deconvolution procedure used to analyze the data. This is particularly true in the case of lowlevel species, where baseline effects may conceal the presence of modifications or generate peaks that are artifactual. In our experience, these effects are augmented with increasing mass of the analyte and decreasing resolution of species in the raw data.

During routine characterization of a monoclonal antibody in the early stages of development, we detected a minor species consistent with +162 Da addition of mass to the light chain after reduction and mass analysis. This finding, pointing to potential glycation events, was interesting since orthogonal methods such as peptide mapping had initially failed to detect peptides consistent with this modification on the antibody. To test our ability to detect glycated species using several orthogonal techniques, we generated a set of glycated antibody samples by incubating a sample of our bulk-drug substance (BDS) with glucose at 37 °C as described in the Experimental Section. We were interested in the ability of intact mass analysis to show the presence of glycation after removal of the N-linked carbohydrates by PNGaseF treatment. Figure 1 shows deconvoluted data for both control and 10 day forced-glucosylated antibody samples. Although these data were collected on an Agilent TOF instrument, we exported the data to MaxLynx and deconvoluted using MaxEnt1. Processing the data at two different peak width settings illustrates the difficulty of analyzing intact antibody samples, particularly with respect to minor components. Although the higher peak width value generated a simplified baseline, it also resulted in a mass error value for the major species greater than 10 times that observed when a narrower peak width setting was used. However, use of a narrower peak width setting resulted in a baseline that was not as smooth and mass peaks that were relatively wider and not Gaussian. For the forced-glucosylated sample, the addition of +162 Da mass units was clear and showed successful addition of the glucose moiety to the protein. However, for the BDS (control) sample, no mass could be confidently associated with glycated protein. To further examine this question, we carefully compared the raw spectra for untreated and forced-glucosylated samples. These data are shown in Figure 2. The overall quality of the spectra for both samples was high, with good baseline and species resolution obtained. Furthermore, comparison of the raw data at the individual charge state level (as shown for the 48+ ion) revealed Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

9405

Figure 3. Reduced mass analysis of forced-glucosylated antibody. The figure shows reduced mass results for control and forced-glucosylated antibody after de-N-glycosylation. Shifts in mass of ∼162.0 Da consistent with nonenzymatic glucose addition were observed and labeled +1 and +2 glc. Minor levels of these masses were also observed for untreated samples and are further discussed in the text. The experimental mass accuracy for both light and heavy chains was below 10 ppm for all major species.

a signal in the untreated antibody that overlapped with that corresponding to a single glycation in the forced-glucosylated sample. These results suggested that glycation could be detected on the intact antibody but not easily observed in the deconvoluted data due to the low level of abundance and possibility for deconvolution artifacts. Antibodies are tetrameric proteins with bivalent symmetry, composed of two light and two heavy chains. As noted above, we initially detected glycation by reduced mass analysis of the light chain of our BDS sample. To test for glycation on both chains, we analyzed PNGaseF-treated control and forced-glucosylated samples by reduced mass as shown in Figure 3. For the forcedglucosylated samples, a clear increase of +162 Da additions to the expected mass of the two constituent chains confirmed the increase of glycation over increasing incubation time with glucose and showed that the added sugar moieties were distributed on both chains. For the BDS, minor forms could be detected for both 9406

Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

chains consistent with glycation. It is interesting to note that quality of the baseline for the deconvoluted data for light and heavy chain was significantly better than that observed for the intact antibody shown in Figure 1. Our experience showed that the Agilent deconvolution algorithm produced results comparable or superior to MaxEnt1 for reduced samples, so we processed all reduced data using Agilent deconvolution. This observation is consistent with the idea that the deconvolution algorithm has difficulty producing accurate output data at higher mass values. The potential for deconvolution artifacts and difficulty analyzing intact antibodies has been discussed previously.17 Another important observation from the reduced mass data was that even for antibody incubated for 10 days in glucose at 37 °C, the major form for both light and heavy chains was unmodified protein. This is in contrast with the observation of (17) Gadgil, H.; Pipes, G.; Dillon, T.; Treuheit, M.; Bondarenko, P. J. Am. Soc. Mass Spectrom. 2006, 17, 867-872.

Figure 4. Boronate affinity chromatography of untreated (BDS) and forced-glucosylated antibody samples. The figure shows the chromatograms obtained for control and forced-glucosylated samples of antibody obtained by boronate affinity chromatography in phosphate buffer (A). Also shown are the chromatographic results obtained for a series of standards made by diluting untreated and 21 day glucosylated antibody (both samples at the same concentration) at different ratios, where the highest absorbing unbound peak is BDS and the highest bound peak is 21 day glucosylated material, with 1:3, 1:1, and 3:1 mixtures of these samples representing the remaining three intermediate chromatograms (B). The results shown are for the 100 µg load level. Quantitative values for these data are shown in Supporting Information Figure S-2.

the 10 day sample by intact mass, where the major form was singly glycated protein. This observation demonstrates the difficulty in using reduced mass analysis to estimate levels of glycation for multichain proteins such as antibodies, where back-calculation for each chain would be required. For proteins of this type, a method is needed to evaluate the modification on the intact molecule level. To obtain an orthogonal determination of glycation degree for both control and forced-glucosylated antibody samples, each was analyzed by boronate affinity chromatography as shown in the top panel of Figure 4. This type of affinity chromatography is specific for solvent exposed sugars containing a cis-diol moiety18 and does not retain proteins through N-linked carbohydrates.19 As described in Harris,14 boronate affinity chromatography operated using HEPES based buffers has been previously used to estimate the level of glycation on antibodies. However, with our antibody HEPES was a poor buffer choice due to limited column capacity and the necessity to use absorbance at 280 nm for quantification. We tested several different buffer systems and found optimal performance in sodium phosphate buffer at pH 8.5. Under these conditions, nonglycated protein flows through the column, while glycated protein is retained on the column and eluted by inclusion of sorbitol in the elution buffer. By integrating the A214 nm traces and comparing the areas of the two peaks obtained for each sample, quantitative information about the degree of glycation can be obtained. We tested the linearity and (18) Springsteen, G.; Wang, B. Tetrahedron 2002, 58, 5291-5300. (19) Dowlut, M.; Hall, G. J. Am. Chem. Soc. 2006, 128, 4226-4227.

loading capacity of the column using two different approaches. Since known standards of our glycated antibody were not available, we mixed BDS with 21 day forced-glucosylated BDS (both at 10 mg/mL) in known ratios to generate a standard curve. To determine the maximum capacity of the column, we injected different mass amounts for samples with different levels of glycation. The loading capacity experiments showed that quantification values for injections of forced-glucosylated samples after day 7 were not linear at the 150 µg on column level (see Supporting Information Figure S-1), suggesting that the higher levels of glycated protein exceeded the capacity of the column. Based on this information, we tested linearity of the method at 75 and 100 µg by plotting experimental values obtained by mixing BDS and day 21 samples at 1:3, 1:1, and 3:1 ratios to obtain intermediate points. Plotting the values determined by integrating the chromatograms (see Supporting Information Figure S-2), we were able to fit linear equations to the data points for each load level with R2 values of 0.998 and 0.999 for the 100 and 75 µg load levels, respectively. These results showed that the method was linear for quantification of bound and unbound components. Representative data for the 100 µg load level for this experiment is shown in the lower panel of Figure 4. Using the linearity and load level parameters determined above, we quantified the level of bound material, corresponding to glycated antibody, for each sample as shown in Table 1. A clear increase in the degree of glycation was measured with increasing Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

9407

Figure 5. Intact mass analysis of untreated (BDS) boronate chromatography pools. The figure shows the intact mass data for the unbound and bound purified fractions from boronate affinity chromatography of antibody BDS. The mass for unbound (top panel) was consistent with that expected for PNGaseF-treated antibody within 2.6 ppm. The bound fraction (lower panel) showed a mass increase of +166.5 Da, consistent with glycated protein. See the text for further discussion. Table 1. Quantification of Antibody Glycation by Boronate Affinity Chromatography sample

degree of glycation (bound area %)

untreated (BDS) day 4 day 7 day 10 day 21

8.17 32.4 46.5 57.3 79.4

incubation time, consistent with the intact and reduced mass results shown in Figure 3. To provide further confidence in the specificity of this method, we purified the bound and unbound peaks from several successive runs of BDS separated by boronate affinity at the 100 µg load level. Since the retained component was only about 8% of the total peak area, we were only able obtain approximately 160 µg of this component of the sample from 20 successive chromatographic runs. After buffer exchange into 100 mM Tris, pH 7.5, both peaks were analyzed by intact mass analysis as shown in Figure 5. A clear difference in mass between the two peaks was observed, with the major form for the flow-through peak consistent with unmodified antibody within 3 ppm mass error (note that a 0.5 Da peak width was selected for both samples). The major species observed for the retained fraction showed a +166.5 Da shift in mass corresponding to glycated protein. As discussed above, differences in deconvolution settings influenced the mass accuracy of the major species with a tradeoff observed between a cleaner baseline and better accuracy (data not shown). The higher mass error for the glycated fraction is probably due to the apparent presence of adducts (a +44 Da 9408 Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

Figure 6. Reduced mass analysis of boronate-purified glycated antibody. The figure shows reduced mass results for the de-Nglycosylated boronate retained BDS compared to unfractionated BDS. Mass shifts of +162.0 Da consistent with glycation were observed for both light and heavy chains.

secondary species was observed) due to disodiation. This is consistent with the presence of 0.4 M sodium chloride in the boronate affinity mobile phase. Lower levels of species with these adducts were also observed for the flow-through fraction. Although the signal for the boronate retained fraction was lower, resulting in a noisier deconvolution profile, the intact mass data indicate that this fraction pool is composed of glycated protein. A very low level species could be consistent with nonglycated protein, but it was observed at the level of many other low-level species that could be noise. To confirm the mass shift of the glycated species and determined the distribution across light and heavy chains, reduced mass analysis was run in comparison to unseparated BDS as shown in Figure 6. A clear and significant enrichment of glycated species for both heavy and light chains for the boronate-purified fraction was observed relative to unfractionated BDS. Reduced mass analysis of the flow-through peak from boronate affinity showed the absence of glycated protein (data not shown). These results confirm that boronate affinity provides an accurate measure of antibody glycation and indicate that the glycation observed in our BDS is distributed on both light and heavy chains. The level of glycation determined by boronate affinity chromatography of BDS (8.17%) is difficult to detect in the deconvoluted intact mass data, though the species is apparent in the raw data. This observation, combined with the results of forcedglucosylated samples showing successful detection of glycated species suggests that there is a low level of glycated forms that may be difficult to detect by intact mass. Although reduced data may allow detection, it is difficult to use these data for quantification. Boronate affinity addresses this issue by providing an

Figure 7. Peptide map analysis of forced-glucosylated antibody. The figure shows a comparison of 17 day forced-glucosylated antibody treated with sodium borohydride (top panel) prior to tryptic peptide map analysis compared to the same sample without sodium borohydride treatment (lower panel). Arrows in the top panel indicate peptides that were observed to shift to earlier retention time after sodium borohydride treatment.

orthogonal technique to intact mass analysis that can quantify glycation even at low levels. However, it may be necessary to show the specificity of the column for each new antibody or protein sample, as others have noted specificity issues when attempting to resolve serum and other proteins on this resin.20,21 As noted earlier, initial peptide map analysis failed to detect glycated species during characterization of BDS material. However, careful inspection of peptide map data revealed that this was due to the low level of the species, coelution with other, more intense peaks, and poor quality of the mass spectra for the glycated species. To address these issues and determine the sites of modification for both forced-glucosylated and BDS samples, we employed a derivatization procedure that improved the mass spectra and increased the confidence of assignment for glycated peptides. Each sample was derivatized with sodium borohydride prior to reduction and alkylation of the cysteine residues and subsequent trypsin digestion. This treatment reduces the double bond present in the Amadori moiety to a single bond, stabilizing the linkage between the peptide and the sugar residue. Reduction of the double bond results in a mass shift of 2.0 Da for glycated peptides. For these experiments a sample of antibody forcedglucosylated for 17 days was used as a positive glycation control and compared to unfractionated BDS. Material limitations prevented using the boronate affinity retained pool described above for this work. Companion samples for each antibody were also prepared for peptide mapping using the same procedure as for derivatization except with omission of the sodium borohydride reducing agent. (20) Li, Y. C.; Larsson, E.; Jungvid, H.; Galaev, I.; Mattiasson, B. J. Chromatogr., A 2001, 909, 137-145. (21) Brena, B.; Batista-Viera, F; Ryden, L; Porath, J. J. Chromatogr. 1992, 604, 109-l 15.

Figure 8. Peptide map analysis of forced-glucosylated antibody (detailed view). The figure shows a detailed view of a region of the chromatogram from Figure 7. Three glycated peptides were observed to elute in this region as indicated by the arrows. For each peptide, a clear shift to earlier retention time was observed for the sodium borohydride treated relative to the untreated sample.

Figure 7 shows an example of 17 day forced-glucosylated antibody analyzed by peptide map analysis with and without derivatization. For the derivatized sample, the glucosylated peptides showed a slightly earlier migration time than for the underivatized sample as indicated. Figure 8 shows a magnified region of the chromatogram containing several glycated peptides. Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

9409

Figure 9. MS/MS analysis of heavy chain glycated peptide H25/26. The figure shows the MS/MS spectra for glycated peptide H25/26 without (lower panel) and with (top panel) sodium borohydride pretreatment. Ions have been labeled as b- or y-types as indicated by the sequence inset shown in the lower panel. The glycated lysine (residue 325 in the heavy chain) is indicated by an asterisk. For the top panel, ions with the ‘#” symbol indicate a shift of +164.0 Da relative to the value predicted by amino acid sequence.

The difference in rpHPLC retention time between derivatized and untreated glycated peptides can be exploited to rapidly detect glycated peptides. This observation is especially useful given the low levels observed even for forced-modified samples. By using this retention time difference, we were able to rapidly and easily detect numerous glycated peptides in our forced-glucosylated samples. Confirmation of the sites of modification was obtained by tandem MS analysis using a linear ion trap mass spectrometer. For derivatized samples, a clear and dramatic improvement of MS/ MS spectral quality was obtained. Figure 9 shows an example for peptide H25/26 from the constant domain sequence of heavy chain. The upper panel shows the MS/MS spectra for the 2+ ion corresponding to the derivatized peptide. Both y-ion and b-ion series were observed, with a strong b-ion series showing mass shifts of +164 Da corresponding to derivatized, glycated peptide fragments. These fragment ions allowed positive confirmation of 9410

Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

the site of modification at heavy chain lysine 325 as indicated by the sequence inset in Figure 9. As with other glycated peptides observed in this experiment, trypsin failed to cleave C-terminal to glycated lysine residues. For the underivatized sample, the major species observed was consistent with the loss of three water molecules from the peptide, and the generation of informative fragment ions was virtually nonexistent. This lack of data quality is a major limitation of the current methods of analysis of glycated peptides, consistently reported in publications.11,22 This also explains the initial failure to detect glycation in the BDS samples by peptide mapping: since the low-level species also yielded very poor MS/MS spectra, they were not assigned during either automated or manual MS data analysis. This observation shows that one advantage of derivatization for analysis of glycated (22) Lapolla, A.; Fedele, D.; Reitano, R.; Arico, N. J. Am. Soc. Mass Spectrom. 2004, 15, 496-509.

Figure 10. MS/MS analysis of heavy chain glycated peptide H17/18. The figure shows the MS/MS spectra for glycated peptide H17/18 without (lower panel) and with (top panel) sodium borohydride pretreatment. The glycated lysine is indicated in the sequence by an asterisk; note that the cysteine residues have been reduced and alkylated to their carboxymethyl form. For the top panel, ions with the ‘#” symbol indicate a shift of +164.0 Da relative to the value predicted by amino acid sequence. Table 2. Glycation Sites Identified in Forced-Glucosylated and BDS Antibody (A214 absorbance) light chain residues (lysine position no.) 24 36 109 132 151 155 175 189 196 213 heavy chain residues (lysine position no.) 12 19 63 150 247 316 325 391

identified in 17 day forced-glucosylated antibody × × × × × × × × × × × × × × × × × ×

identified in BDS (untreated) antibody ×

× ×

peptides is the ability to collect high-quality MS/MS data that can confirm the peptide sequence. For most peptides, it may be possible to assume that the site of glycation is at the location of missed cleavage; however, if very poor MS/MS data are obtained, it may not be possible to confirm the identity of the peptide to make such an assignment.

The observation of improved MS/MS fragmentation data for derivatized as compared to untreated glycated peptides was universal and independent of parent ion charge state or sequence, although not always as dramatic as for peptide H25/26 shown above. Figure 10 shows an example for the constant domain peptide H17/18, which contains the hinge region sequence from the heavy chain. Furthermore, this peptide contains two internal lysine residues due to a trypsin-resistant lysine/proline sequence (see Figure 10, inset). Fragmentation of the 2+ ion corresponding to the glycated peptide resulted in a high-quality series of fragment ions shifted by +164 Da, confirming the presence of glycation. For the underivatized peptide, the MS/MS spectrum was dominated by an ion corresponding to the neutral loss of three water molecules. To obtain further fragmentation data for this large peptide, MS3 of the most abundant ion from each spectrum was performed (Figure 11). For the derivatized peptide, both b- and y-series shifted by +164 Da relative to unmodified predicted values were observed. Furthermore, the observation of ion y9 shifted by +164 Da allowed the site of modification at Lys247 to be confirmed, excluding Lys245. This is a particularly satisfying result given the close proximity of these two potential modification sites. Since Lys245 is followed by a proline residue, trypsin would not be expected to cleave at this location and using protease susceptibility as a determinant for glycation site location would be ambiguous. MS3 data obtained for the underivatized peptide showed the generation of several fragment ions shifted by +108 Da, corresponding to a rearrangement product of the attached sugar resulting from the initial neutral loss. Although the presence of these ions confirms the presence of a glycated residue, the Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

9411

Figure 11. MS3 analysis of heavy chain glycated peptide H17/18. The figure shows the MS3 spectra for glycated peptide H17/18 without (lower panel) and with (top panel) sodium borohydride pretreatment. The glycated lysine is indicated in the sequence by an asterisk; note that the cysteine residues have been reduced and alkylated to their carboxymethyl form. For the top panel, ions with the ‘#” symbol indicate a shift of +164.0 Da relative to the value predicted by amino acid sequence. For the lower panel, the symbol ‘&’ indicates ions that are shifted by +108.0 Da relative to the predicted value.

Figure 12. Comparison of forced-glucosylated and BDS-derivatized tryptic digests (detail). The figure shows a comparison of the A214 nm traces obtained for forced-glucosylated and BDS control trypsin-digested, sodium borohydride treated antibody. The arrow indicates the elution position of glycated peptide L2/3 from the light constant domain sequence in the forced-glucosylated sample and its absence (by UV detection) in the BDS samples. See Supporting Information Figure S-3 and the text for discussion of detection of this species below the UV detection limit.

limited sequence information that can be obtained from the fragmentation pattern does not allow the two lysine residues in the sequence to be distinguished. With the use of the identification and confirmation approach outlined above, 18 different glycation sites could be confirmed in our antibody after incubation with glucose for 17 days. Using the information from the forced-glucosylated samples, we searched both derivatized and underivatized peptide maps corresponding to BDS for these species. Three glycated peptides could be identified that were associated with minor A214 nm absorbing peaks. Table 2 shows the different sites that were identified for 9412 Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

the two samples that had UV absorbance. Since the species identified in the BDS samples were relatively low level and closely eluting with other, more abundant forms, it is not surprising that they were not detected in initial characterization efforts. In addition to the more abundant minor forms identified in BDS listed in Table 2, ions corresponding to glycated peptides could be observed where a UV peak was absent. Figure 12 shows a comparison of an expanded region of UV traces for forced-glucosylated and BDS trypsin digests. Both samples were treated with sodium borohydride. In the forced-glucosylated sample, a peak is present corresponding to glycated peptide L2/3 from light chain that is

absent in the BDS sample. However, close inspection of the MS and MSn data showed that an ion could be detected in the BDS sample corresponding to the glycated species and confirmed by MS/MS analysis (see Supporting Information Figure S-3). This observation indicates that the derivatization procedure allows the successful detection of glycated peptides, with confident MSn assignment, below the level of UV detection. Without derivatization, this ion would not have been assigned in the BDS sample since it had no UV absorbance and apparently poor MS/MS spectra. Since the sites of glycation in an antibody or protein may be commonly distributed across many very low-level peaks, quantification of glycation based on peptide mapping alone often may not be feasible. For reporting purposes, Table 2 shows only those glycated peptides that could be associated with UV absorbance.

simple procedure addresses a data quality problem well documented in the literature and allows less ambiguous assignment of location of glycation based on MS/MS data. We show in this paper that a combination of methods is best for quantification and identification of glycation sites in our antibody and that no one method reveals a comprehensive analysis of glycation by itself. This approach should be useful for other antibody and protein characterization efforts.

CONCLUSIONS We have applied affinity chromatography, mass spectrometry, and peptide mapping to analyze the glycated products of a monoclonal antibody present in both BDS and forced-glucosylated samples. By incorporating derivatization with sodium borohydride, the elution position of glycated peptides in peptide maps was shifted, facilitating their visualization. Derivatization also resulted in high-quality tandem MS spectra for clear site identification. This

SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

ACKNOWLEDGMENT The authors thank Sibylle Wilbert of ZymoGenetics, Inc. for her suggestions regarding the borohydride reduction method. We also thank our Amgen colleagues, Himanshu Gadgil and Gerd Kleemann, for helpful discussions on glycation. We thank Dean Pettit for supporting this work.

Received for review August 17, 2007. Accepted September 24, 2007. AC7017469

Analytical Chemistry, Vol. 79, No. 24, December 15, 2007

9413