Fluorescent Derivatization Method of Proteins for Characterization by

Jun 26, 2007 - Zac Yates , Kannan Gunasekaran , Hongxing Zhou , Zhonghua Hu , Zhi ... fluorescence via a bubble cell capillary and laser power adjustm...
0 downloads 0 Views 202KB Size
Anal. Chem. 2007, 79, 5963-5971

Fluorescent Derivatization Method of Proteins for Characterization by Capillary ElectrophoresisSodium Dodecyl Sulfate with Laser-Induced Fluorescence Detection David A. Michels, Lowell J. Brady, Amy Guo, and Alain Balland*

Department of Process and Analytical Science, Amgen Incorporated, 1201 Amgen Court West, Seattle, Washington 98119

A fast and improved sample preparation scheme was developed for protein analysis using capillary electrophoresis-sodium dodecyl sulfate (CE-SDS) with laserinduced fluorescence detection. This CE-SDS method was developed as a purity assay for recombinant monoclonal antibodies (rMAbs). In this assay, rMAbs are derivatized with the fluorogenic reagent 3-(2-furoyl)quinoline-2-carboxaldehyde (FQ) in the presence of a nucleophile (CN-), which fluoresces only upon covalent binding to the protein. Purification after labeling is therefore not necessary to remove unreacted reagents. Proteins are incubated at 75 °C for 5 min to facilitate denaturation and labeling. For nonreduced preparation, rMAbs are labeled at pH 6.5 with a dye-to-protein (D/P) molar ratio of 50:1, which forms conjugates having 6 ( 4 FQ labels. For reduced preparation, rMAbs are labeled at pH 9.3 with a D/P molar ratio of 10:1, which generates light chain conjugates incorporated with 3 ( 2 FQ labels. Labeling artifacts such as fragmentation or aggregation are absent with use of alkylation reagents. This efficient labeling scheme generates detection limits for FQ-labeled rMAbs as low as 10 ng/mL. In comparison to other labeling strategies, labeling proteins with FQ has the advantage of speed, ease of use, and robust quantification. The therapeutic importance of recombinant monoclonal antibodies (rMAbs) has fueled advances in the area of separation methodologies for size-variant characterization and accurate quantification.1-3 High-performance liquid chromatography (HPLC) methods, like size exclusion chromatography (SEC) and denatured SEC (dSEC), are commonly used for characterization of rMAb size variants but suffer limitations in sensitivity and resolution power for low molecular weight species. Alternatively, electrophoretic methods, such as sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), are well-established for purity analysis of complex protein mixtures.4 Such methods have superior resolution and provide 10 ng/mL detection limits * To whom correspondence should be addressed. E-mail: [email protected]. Phone: 260-265-8603. Fax: 206-217-4692. (1) Adams, G. P.; Weiner, L. M. Nat. Biotechnol. 2005, 23, 1147-1157. (2) Wurm, F. M. Nat. Biotechnol. 2004, 22, 1393-1398. (3) Lee, H. G.; Chang, S.; Fritsche, E. J. Chromatogr., A 2002, 947, 143-149. (4) Guttman, A.; Nolan, J. Anal. Biochem. 1994, 221, 285-289. 10.1021/ac0705521 CCC: $37.00 Published on Web 06/26/2007

© 2007 American Chemical Society

when visualized by silver stain.5 However, limitations of SDSPAGE are related to time-consuming gel preparations, lack of automation, and labor-intensive/nonlinear staining techniques. More recently, capillary electrophoresis-sodium dodecyl sulfate (CE-SDS) has been recognized as a powerful separation method for proteins. In CE-SDS, as with SDS-PAGE, proteins denatured in SDS acquire constant mass-to-charge ratios resulting in SDSprotein complexes with mobilities proportional to the logarithm of their effective molecular weight.6 The advantages of CE-SDS include direct quantification with absorbance or fluorescence detection, high resolution, fast automated analyses, nanoliter sample injection volumes, and use of replaceable UV-transparent sieving gels.7-11 Characterization of rMAbs with CE-SDS has become widely used for lot-to-lot consistency of the manufacturing process, detecting low-level impurities, and measuring size-variant heterogeneity.12-15 However, poor signal-to-noise (S/N) and irregular baseline are often experienced with conventional UV detection. To improve sensitivity, strategies based on derivatization of proteins with fluorescent dyes followed by CE separation with laser-induced fluorescence detection (LIF) have proved to be very efficient.16,17 The pros and cons of various methodologies and dyes, developed for LIF detection of proteins in CE separations, have (5) Oakley, B. R.; Kirsch, D. R.; Morris, N. R. Anal. Biochem. 1980, 105, 361363. (6) Weber, K.; Osborn, M. J. Biol. Chem. 1969, 244, 4406-4412. (7) Hu, S.; Michels, D. A.; Fazal, M. A.; Ratisoontorn, C.; Cunningham, M. L.; Dovichi, N. J. Anal. Chem. 2004, 76, 4044-4049. (8) Dovichi, N. J.; Hu, S.; Michels, D.; Zhang, Z.; Krylov, S. N. Biotechnology 2001, 5b, 269-277. (9) Manabe, T. Electrophoresis 1999, 20, 3116-3121. (10) Benedek, K.; Guttman, A. J. Chromatogr., A 1994, 680, 375-381. (11) Guttman, A.; Horvath, J.; Cooke, N. Anal. Chem. 1993, 65, 199-203. (12) Ma, S. State of the Art Analytical Methods for the Characterization of Biological Products and Assessment of Comparability; Mire-Sluis, A. R., Ed.; Karger: Basel, 2005; Vol. 122, pp 49-68. (13) Tous, G. I.; Wei, Z.; Feng, J.; Bilbulian, S.; Bowen, S.; Smith, J.; Strouse, R.; McGeehan, P.; Casas-Finet, J.; Schenerman, M. A. Anal. Chem. 2005, 77, 2675-2682. (14) Patrick, J. S.; Lagu, A. L. Electrophoresis 2001, 22, 4179-4196. (15) Krull, I. S.; Liu, X.; Dai, J.; Gendreau, C.; Li, G. J. Pharm. Biomed. Anal. 1997, 16, 377-393. (16) Liu, H.; Cho, B. Y.; Krull, I. S.; Cohen, S. A. J. Chromatogr., A 2001, 927, 77-89. (17) Quigley, W. C.; Dovichi, N. J. Anal. Chem. 2004, 76, 4645-4658.

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5963

recently been reviewed.18 In the case of rMAbs, Hunt and Nashabeh demonstrated that fluorescent labeling of rMAbs can be used for CE-SDS with LIF detection as a replacement for silver-stained SDS-PAGE.19 In Hunt and Nashabeh’s procedure, preparation artifacts in the form of thermal-induced fragmentation and aggregation were generated for rMAbs labeled with 5-carboxytetramethylrhodamine succinimidyl ester (5-TAMRA). The procedure was later revised to incorporate a low-pH alkylation step as a means for stabilizing the intact antibody.20 Although product stability improved, additional purification (via buffer exchange column) was needed to alleviate protein aggregation through exposure to excess dye. In this article, we describe an improved fluorescent derivatization method for proteins. In our scheme, labeling reagents are mixed with protein in a single sample tube and heated at 75 °C for 5 min; upon dilution, samples are ready for analysis. We employ fluorogenic 3-(2-furoyl)-quinoline-2-carboxaldehyde (FQ) as the labeling reagent since it becomes fluorescent only upon reaction with a primary amine, has very few fluorescent impurities, and generates an extremely low background; therefore, purification of labeled product is not necessary. We also demonstrate that FQ-labeled rMAb samples containing unreacted dye did not affect aggregation levels even at high FQ concentrations. Furthermore, detection sensitivity similar to those obtained with 5-TAMRAlabeled samples was also achieved for FQ-labeled rMAbs. Ultimately, we offer a new strategy for quantitative purity characterization of rMAbs using CE-SDS as the enhanced analytical technique used during product development and quality control. EXPERIMENTAL SECTION Reagents. rMAbs were manufactured in-house (Amgen, Inc., Seattle, WA). Sieving gel and 0.1 M NaOH basic and 0.1 M HCl acidic washes were purchased from Beckman Coulter, Inc. (Fullerton, CA). All chemicals and reagents were analytical grade. Buffers were prepared using 0.2 µm filtered Milli-Q deionized water. Sodium borate, sodium bicarbonate, sodium phosphate (dibasic), citrate-phosphate concentrate, dithiotreitol (DTT), iodoacetamide (IAM), N-ethylmaleimide (NEM), sodium iodoacetate (IAA), and 10% (w/v) SDS stock solution were purchased from Sigma-Aldrich (St. Louis, MO). Derivatizing reagents, FQ and potassium cyanide (KCN), were purchased from InvitrogenMolecular Probes (Eugene, OR). A 10 mM FQ stock solution was prepared by dissolving 2.5 mg into 1.0 mL of methanol. Daily working solutions of 1 or 5 mM FQ were made using appropriate dilutions with water. Solutions of FQ were stored in the dark at -20 °C in amber glass vials. Instrumentation. CE-SDS was performed using a Beckman ProteomeLab PA800 CE system equipped with LIF detection (Fullerton, CA). The LIF detector uses a 3.5 mW argon-ion laser having an excitation at 488 nm; emission was collected through a 600 ( 15 nm band-pass filter (Edmund Optics, Barrington, NJ). Separations were performed with 30.2 cm fused-silica capillaries having 50 µm i.d. × 375 µm o.d. (20 cm effective length). Data (18) Garcia-Campana, A. M.; Taverna, M.; Fabre, H. Electrophoresis 2007, 28, 208-232. (19) Hunt, G.; Nashabeh, W. Anal. Chem. 1999, 71, 2390-2397. (20) Salas-Solano, O.; Tomlinson, B.; Du, S.; Parker, M.; Strahan, A.; Ma, S. Anal. Chem. 2006, 78, 6583-6594.

5964

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

integration was performed with Beckman 32Karat software, version 7.0. Preparation of FQ-Labeled rMAbs. Labeling reactions were incubated for 5 min at 75 °C using a heating block. Samples were cooled with ice and diluted 6-fold with an SDS solution to quench the reaction. For nonreduced labeling, 75 µg of rMAb (0.5 nmol) was mixed in 35 µL of 0.1 M citrate-phosphate (pH 6.5), 2% SDS, 10 mM NEM, 1 mM KCN, and 25 nmol FQ; the reaction was quenched with 4% SDS. For reduced labeling, samples were heated for 3 min at 75 °C in 25 µL of sodium borate (pH 9.3), 3% SDS, and 20 mM DTT; upon cooling, labeling was initiated with 5 µL of 0.5 M IAM, 5 µL of 32 mM KCN, and 5 nmol of FQ; the reaction was quenched with 1% SDS. Preparation of Deglycosylated rMAbs. Approximately 300 µg of rMAb was mixed with 5 µL of peptidyl N-glycosidase (PNGase F) from New England BioLabs (Ipswich, MA) in a 100 mM Tris buffer, pH 8.0 and digested for 20 h at 37 °C. CE-SDS Analysis. All CE reagents were pressurized from the capillary inlet at 70 psi. Prior to injection, the capillary was washed with 0.1 M NaOH, 0.1 M HCl, water, and sieving gel for 4, 1, 1, and 10 min, respectively. Samples were injected at -10 kV for 10 s. Separations were performed using constant current (Ohm’s plot correlation coefficient ) 0.999; -23 µA resulted in a voltage of -15 kV) instead of traditional constant voltage to maintain run-to-run conductivity and better migration time reproducibility. Sample tray and capillary cartridge temperatures were set to 20 °C. Mass Spectrometry Analysis. To reduce mass complexity of rMAb structures, labeled samples were prepared using deglycosylated material. SDS was substituted with 4 M guanidinehydrochloride salt (Roche, Alameda, CA). On-Line LC/ESI-MS. Intact whole mass analyses were performed on an Agilent 1100 HPLC system connected in-line to an Agilent ESI-TOF mass spectrometer (Agilent Technologies, Santa Clara, CA). An amount of 2 µg of FQ-labeled rMAb was injected onto a polyhydroxethyl aspartamide column (PolyLC, 300 Å × 5 µm particles, 2.1 mm × 20 cm); the column was operated at 0.1 mL/min under isocratic conditions with 0.1% (v/v) formic acid. Under these conditions, large proteins elute prior to small molecules and buffer components. Acetonitrile with 1.0% (v/v) formic acid was added postcolumn at 25 µL/min to aid both protein desolvation and ionization in the ESI source. A divert valve in the mass spectrometer was configured to direct column eluent to waste before and after protein elution to avoid contaminating the instrument. The mass spectrometer was calibrated according to manufacturer recommendations with a resolution of 13 500 obtained prior to instrument operation. Optimized ESI source conditions for these samples were 4.0 kV for the capillary, 415 V for the fragmentor, and 9 L/min for source nitrogen. Raw spectra were summed and analyzed using deconvolution software available from Agilent. Mass accuracy values were below 100 ppm for all major species. MALDI-MS. Reduced samples were analyzed with a Bruker Daltonics Ultraflex MALDI-TOF/TOF mass spectrometer (Bremen, Germany). The instrument was controlled with Bruker FlexControl software, version 2.4. The total spectra were summed across 10 laser shots (100 spectra/shot). A standard 192-well stainless steel MALDI sample plate was used. Subsequently, a 1

Figure 1. (A) CE-SDS separations of nonreduced FQ-labeled rMAb samples prepared at different dye-to-protein (D/P) molar ratios; the inset is an expanded view to illustrate detection of minor species. The asterisk points to the pre-main peak used for resolution assessment. (B) Deconvoluted whole mass spectra of deglycosylated labeled rMAb samples; mass peaks are assigned with the number of FQ conjugates per rMAb molecule. The dashed trace is the nonlabeled control material. The relative ion intensity (y-axis) was removed for clarity purposes.

µL aliquot of a 1:1 mixture containing FQ-labeled rMAb and MALDI matrix (50 mg/mL sinapinic acid in 50% (v/v) acetonitrile/ water and 0.1% (v/v) trifluoroacetic acid) was deposited onto the sample probe. Prior to sample spotting, 1 µL of the saturated sinapinic acid solution was applied to provide a sublayer. Each spot had ∼3 pmol of sample. Mass accuracy values were within 100 ppm. RESULTS AND DISCUSSION Selection of FQ Dye. Two categories of fluorescent reagents can be used to derivatize proteins: (i) highly fluorescent dyes

that react readily with proteins and (ii) nonfluorescent dyes that form highly fluorescent protein conjugates upon reacting to the target site. Ideally, a fluorescent dye for proteins should possess the following: (a) high quantum yields with low quenching, (b) fast reaction rates at low concentrations, (c) photostable dyeprotein conjugates, and (d) absence of labeling artifacts. One such dye, FQ, is a fluorogenic reagent that reacts with -amines of lysine in the presence of molar excess cyanide and forms a stable fluorescent isoindole derivative that generates FQ-protein conjugates having high quantum fluorescence efficiencies.21 Derivatization is rapid at high temperatures and moderate pH, and Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5965

conjugates have an absorption maximum which closely matches the 488 nm main radiation line of argon-ion lasers. In comparison to other commercial fluorophores, FQ is ideal because of its simple straightforward labeling procedure: reagents and protein are added to a single tube, incubated for 5 min, and diluted in SDS solutions to quench the reaction. Alternatively, labeling proteins with 5-TAMRA (AnaSpec, San Jose, CA) requires tedious sample handling: after 15-30 min of conjugation, excess dye requires removal to minimize background noise, system peaks, and protein aggregation; this procedure involves two NAP-5 column buffer exchanges that risk poor protein recovery. In contrast, labeling proteins with FQ does not require buffer exchange. Furthermore FQ labeling is highly specific to aminebased compounds, which was confirmed by the absence of system peaks from sample matrix and derivatization reagents, resulting in a flat baseline in the analysis of the sample blank. Characterization of Dye-to-Protein (D/P) Molar Ratio. Nonreduced rMAb Labeling Scheme. The D/P molar ratio, which we define as the ratio between the initial stoichiometric amounts of free dye to rMAb before conjugation, is essential for biomolecules containing a large number of reactive sites. Derivatization of proteins can alter their properties such as charge, molecular weight, hydrophobicity or can affect stability by rupture of disulfide linkages. Furthermore, labeling reactions result in a distribution of labeling products that often degrade separation efficiency.22,23 These effects were determined from a series of FQ-labeled samples prepared at different D/P molar ratios. CE-SDS was used to characterize detection sensitivity, resolution, and labeling artifacts. LC/ESI-MS was used to measure the degree of labeling at each ratio. Prior to MS analysis, antibodies were treated with PNGase F to reduced its mass heterogeneity linked to a complex array of N-linked glycan structures;24 deglycosylated rMAbs were labeled in 4 M guanidine. Figure 1A presents the CE-SDS analyses of nonreduced FQlabeled rMAbs prepared using D/P molar ratios ranging from 10:1 to 120:1. The expanded view illustrates detection of minor peaks consisting of antibody mass variants and high molecular weight (HMW) aggregates. Reconstructed whole mass spectra in Figure 1B confirm multiple FQ-labeled rMAbs. A summary of the labeling degree is listed in Table 1 in terms of distribution range and median (see Supporting Information Figure S-1 and Table S-1A for more details). The composition of the rMAb sample prepared at a D/P ratio of 10:1 consisted of unreacted antibody and FQ-rMAb conjugates having 1:3 labels/protein (median ) 2 labels/protein). Unreacted antibody remained present upon increasing the D/P ratio to 20: 1, whereas the distribution increased to 1:5 labels/protein (median ) 3), which enhanced fluorescence or corrected peak area (CPA, defined as peak area divided by the migration time) of the main peak by 1.7-fold. From the MS data we concluded that a minimum D/P ratio of 30:1 was required to have no unreacted antibody remaining. Fluorescence continued to increase at D/P ratios of 40:1 and 50:1, which generated FQ-rMAb conjugates having (21) Beale, S. C.; Hsieh, Y.-Z.; Wiesler, D.; Novotny, M. J. Chromatogr., A 1990, 499, 579-587. (22) Richards, D. P.; Stathakis, C.; Polakowski, R.; Ahmadzadeh, H.; Dovichi, N. J. J. Chromatogr., A 1999, 853, 21-25. (23) Stoyanov, A. V.; Ahmadzadeh, H.; Krylov, S. N. J. Chromatogr., B: Biomed. Appl. 2002, 780, 283-287. (24) Jefferis, R. Biotechnol. Prog. 2005, 21, 11-16.

5966 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Table 1. Summary of Degree of Labeling for Nonreduced and Reduced Samples nonreduceda

reducedb

degree of labeling

degree of labeling

D/Pc

distributiond

mediane

D/P

distribution

median

10:1 20:1 40:1 50:1 80:1 120:1 150:1

0-3 0-5 1-8 2-10 4-12 7-15 10-18

1 2 4 6 8 11 14

1:1 3:1 5:1 10:1 21:1 42:1

0-2 0-2 0-3 1-5 2-7 5-11

1 1 2 3 5 8

a Degree of labeling for the intact antibody calculated from LC/ ESI-MS analysis. b Degree of labeling for antibody light chain calculated from MALDI-MS analysis. c Dye-to-protein molar ratio. d Range of dyeantibody conjugates. e Middle value of the distribution range.

medians of 4 and 6 labels/protein, respectively. In fact, a linear correlation between D/P ratio and fluorescence (R ) 0.999) was observed through the range of 10:1 and 60:1 (Supporting Information Figure S-2). At D/P ratios g80:1, excessive labeling resulted in quenched fluorescence signal which was likely caused by transfer of excited-state energy from closely neighboring dye molecules.25,26 Complete labeling of all lysine residues in proteins requires tremendous effort and causes severe fluorescence quenching.16 Instead, labeling targets the most labile residues but consequently forms multiple labeled conjugates with unique electrophoretic mobilities, which degrades separation efficiency and resolution.22,27,28 For multiple labeled rMAbs, the resolution was best for the D/P ) 10:1 sample composed of 1FQ-, 2FQ-, and 3FQrMAb conjugates (asterisk in Figure 1A). This sample had the lowest degree of labeling and the sharpest peak. As the degree of labeled conjugates increased, the width of the main peak increased and caused a slight loss of resolution. In other words, the method offers flexibility: when sensitivity is desired then a D/P ratio on the higher portion of the linear curve should be used; however, this has limitations in resolution. If high resolution is desired then using a lower D/P ratio will limit the degree of labeling but generate lower sensitivity. Potential adverse side reactions induced by the process of labeling proteins with hydrophobic dyes were also examined. A labeled protein has a higher tendency to aggregate compared to its nonlabeled form.19 Labeling antibodies with 5-TAMRA increased protein aggregation 35-fold for dye-antibody conjugates having up to 20 labels/protein.20 In contrast, antibodies labeled with FQ in this study showed aggregate levels at a constant 1.4% (2% RSD) for all D/P ratios tested, including the rMAb sample incorporated with up to18 labels/protein. FQ is less hydrophobic and unlikely to cause aggregation during conjugation. Furthermore, antibody stability, measured by the consistency of intact antibody (96.6%, 0.04% RSD) and fragments (1.9%, 2% RSD), was constant throughout the range of D/P ratios, indicating labeling(25) Brinkley, M. Bioconjugate Chem. 1992, 3, 2-13. (26) Stryer, L.; Haugland, R. P. Proc. Natl. Acad. Sci. U.S.A. 1967, 58, 719726. (27) Zhao, J. Y.; Waldron, K. C.; Miller, J.; Zhang, J. Z.; Harke, H.; Dovichi, N. J. J. Chromatogr., A 1992, 608, 239-242. (28) Craig, D. B.; Dovichi, N. J. Anal. Chem. 1998, 70, 2493-2494.

Figure 2. (A) CE-SDS separations of reduced FQ-labeled rMAb samples prepared at different D/P ratios. The asterisk points to the post-LC clip product used for resolution assessment. (B) MALDI-MS analysis of the FQ-labeled light chain; mass peaks are assigned with the number of FQ conjugates per rMAb molecule. The dashed trace is the nonlabeled control material.

induced fragmentation was absent. These results were validated and highly comparable to unlabeled samples prepared for a UV method (97.2% intact peak, 1.7% fragments, and 1.1% aggregates). From evaluation of the data above, we concluded that the optimum D/P molar ratio is 50:1 for preparing nonreduced FQlabeled rMAbs. At this ratio, rMAbs were conjugated with 6 ( 4 FQ labels, which provides sufficient balance between sensitivity and resolution while avoiding sample preparation artifacts. Reduced rMAb Labeling Scheme. Initially, intact labeled rMAb samples were treated with excess DTT to prepare the reduced form. However, CE-SDS showed a 10-fold increase in HMW peaks for these reduced samples compared to samples detected by UV (data not shown). The HMW peaks can be attributed to combinations of antibody light and heavy chains covalently attached through nonreducible links. This artifact of labeling was controlled using a two-step reaction scheme: (i) denatured protein is reduced in an SDS/DTT buffer and (ii) free thiols are alkylated with IAM during labeling. To aid optimization of the two-step reduced method, a series of FQ-labeled rMAb samples prepared with different D/P molar ratios were characterized by CE-SDS and MALDI-MS analyses.

Figure 2A presents the CE-SDS analyses of reduced FQlabeled rMAbs prepared using D/P ratios of 1:1-42:1. The main peaks detected are free light chain (LC) and heavy chain (HC). The expanded electropherograms illustrate several proteolytic clips and post-HC species (see Supporting Information Figure S-3 for identities). MALDI-MS spectra of FQ-labeled LC are shown in Figure 2B; labeled HC was not reported due to its poor ionization. A summary of the labeling degree for the LC is presented in Table 1 (see Supporting Information Table S-1B for more details). Labeled LC prepared under reducing conditions showed similar relationships observed for labeled intact antibody. The degree of labeling increased with increasing D/P ratios. From the MS data we concluded that a minimum D/P molar ratio of 8:1 was needed to have no unreacted LC remaining. There was a strong correlation with CPA for LC (R ) 0.998) and HC (R ) 0.996) (see Supporting Information Figure S-4) with quenching being likely the cause of deviation above D/P ) 25:1. Peak widths broadened and resolution decreased with increasing D/P ratios. However, this loss of resolution was minimal and comparable to unlabeled antibody detected by UV (see Supporting Information Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5967

Figure 3. Percent corrected peak area (% CPA) of intact antibody as a function of buffer pH for rMAb samples treated with (9) and without (2) NEM during the labeling reaction.

Figure S-5). A strong linear correlation existed between peak width of the LC (R ) 0.999) and HC (R ) 0.979) and D/P ratio. Finally, labeling with FQ did not appear to generate labeling artifacts such as nonreducible species or induce LC/HC clips. On the basis of these results, the optimum D/P ratio for preparing reduced rMAbs was 10:1. Under this condition, LC was conjugated with 3 ( 2 FQ labels, signal was within the linear range, sufficient resolution was achieved for all minor components, and no labeling artifacts were observed. Effect of Buffer pH on Labeling. Reaction pH is a primary factor for rapid conjugation since derivatization rate is highly dependent on the acid-base properties of the target site.29 To aid development of an efficient FQ-labeling procedure, the optimum conjugation pH was investigated. CPA of the major peaks was evaluated using the following 0.1 M amine-free buffers: (i) citrate-phosphate, pH 3.1, 4.0, 5.2, and 6.5, (ii) phosphate, pH 7.3, (iii) bicarbonate, pH 8.3, (iv) borate, pH 9.3, and (v) carbonate, pH 10.5. The CPA-pH profile, which was similar to those reported for NDA, OPA, and CBQCA protein conjugates, exhibited the expected bell-shaped curve with maximum fluorescence at pH 9.3.29-31 The percent corrected peak area (% CPA, defined as the CPA divided by the total CPA and multiplied by 100) for LC and HC combined for >96% between pH 6.5 and pH 9.3 but dropped to 90% at pH 10.5 as a result of a 3-fold increase in degradation products. The HMW level was ∼0.9% for all buffers. On the basis of these results, the reduced FQ-labeling scheme was optimized using borate pH 9.3 buffer. In nonreduced conditions, pH 9.3 is not appropriate due to disulfide exchange. To stabilize the intact antibody, labeling is performed at pH 6.5 to minimize fragment artifacts caused by disulfide exchange; disulfide reshuffling is also minimized with NEM alkylation which works best at lower pH, as detailed in the paragraph below. Free Cysteine Alkylation. The efficiency of alkylating reagents to minimize reversible thiol-disulfide exchange is crucial (29) Montigny, P. d.; Stobaugh, J. F.; Givens, R. S.; Carlson, R. G.; Srinivasachar, K.; Sternson, L. A.; Higuchi, T. Anal. Chem. 1987, 59, 1096-1101. (30) Liu, J.; Hsieh, Y. Z.; Wiesler, D.; Novotny, M. Anal. Chem. 1991, 63, 408412. (31) Svedas, V. J.; Galaev, I. J.; Borisov, I. L.; Berezin, I. V. Anal. Biochem. 1980, 101, 188-195.

5968 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Figure 4. (A) Expanded view of CE-SDS separations of reduced FQ-labeled rMAb samples spiked with three protein size standard contaminants at the 0.05%, 0.1%, 0.5%, and 1.0% (w/w) level. Abbreviations: light chain (LC), heavy chain (HC), and nonglycosylated heavy chain (NGHC). (B) Percent corrected peak area (% CPA) accuracy of the assay for each spiked contaminant comparing the expected to the measured value; error bars are shown at the 95% confidence interval (n ) 3). (C) Expanded view of CE-SDS separation of an rMAb sample digested with trypsin for 5 min at 37 °C (dashed trace) compared to the control sample (solid trace). The rMAb-to-enzyme ratio was 100:1. Direction of the arrows indicate a loss (V) or gain (v) in peak area.

to the stability of the intact antibody.32,33 Three alkylating reagents (IAA, IAM, and NEM) were examined through a pH range of 3.19.3 based on the stability of intact FQ-labeled rMAb. Figure 3 compares the % CPA-pH profile for rMAb samples prepared with and without NEM. Efficiency of alkylation, measured by % CPA of the intact antibody, was highest for NEM-treated samples below pH 6.5. IAA- and IAM-treated samples showed fragment levels of 8.7% and 4.4%, respectively, compared to a 2.7% level obtained with NEM. This study demonstrated that nonreduced rMAb samples are highly stable when treated with NEM in citrate-phosphate (32) Gilbert, H. F. Methods Enzymol. 1995, 251, 8-28. (33) Weissman, J. S.; Kim, P. S. Science 1991, 253, 1386-1393.

Figure 5. Expanded view of CE-SDS separations of (A) nonreduced and (B) reduced preparations of stressed rMAb samples prior to labeling with FQ. Stressed conditions: (i) thermal, 55 °C for 3 weeks; (ii) thermal, 55 °C for 1 week; (iii) hydrolytic, pH 8.0, 37 °C for 3 weeks; (iv) hydrolytic, pH 8.0, 37 °C for 1 week; (v) hydrolytic, pH 4.0, 37 °C for 3 weeks; (vi) hydrolytic, pH 4.0, 37 °C for 1 week; (vii) control, thawed from -70 °C.

buffer pH 6.5. Furthermore, high sensitivity of the main peak was maintained under these conditions (see below). Alkylation is also critical for the reduced method (Supporting Information Figure S-6). The combination of labeling reagents (FQ/CN-) and heat facilitates the conversion of free cysteine to a reactive alkene product called dehydroalanine (DHA), which irreversibly forms thioether linkages.34 This effect was demonstrated with a cysteine-containing peptide treated with and without

alkylation prior to FQ labeling; the peptide samples were then analyzed by MALDI-MS and MALDI-MS/MS sequencing (data not shown). Mass analysis of the nonalkylated FQ-labeled peptide confirmed DHA conversion (-34 mass units on cysteine) and generation of several artifact species. However, the peptide treated with IAM prior to labeling gained +57 mass units on cysteine (34) Yamada, M.; Miyajima, T.; Horikawa, H. Tetrahedron Lett. 1998, 39, 289292.

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5969

Figure 6. (A) Analytical SEC separation of prep SEC pools of monomer, dimer, and HMW. Discrete pools of rMAb were obtained by preparative SEC and analyzed by analytical SEC using two sequential TOSOH TSKgel G3000SWxl columns (7.8 mm × 30 cm, 5 µm particles) with a 4.6 mm × 25 cm guard column. The columns were operated in 150 mM phosphate, 100 mM NaCl buffered at pH 6.8 with a flow rate of 0.5 mL/min; a total of 40 µg of protein was loaded onto the column. (B) Analytical dSEC separations of SEC monomer, dimer, and HMW pools from (A). Denaturing conditions: discrete SEC pools were reduced with 10 mM DTT in 3 M guanidine (100 mM Tris-HCl, pH 7.5) for 5 min at 75 °C; upon cooling, reduced samples were alkylated with 20 mM IAA for 5 min in the dark at room temperature. Excess IAA was quenched with addition of 40 mM DTT. dSEC was operated with 3 M guanidine in water; all other conditions were the same as in (A). (C) Expanded view of CE-SDS separations from each prep SEC pool indicating the % CPA of nonreducible aggregates.

(carbamidomethylation) without evidence of DHA formation or other side reactions. IAA and NEM were also tested, but IAM was selected for the reduced assay based on its efficiency in preventing conversion to DHA. Effect of Incubation Temperature and Time. Proteins are traditionally denatured at 90-100 °C to form SDS-protein complexes.3,19,35 In the case of manufactured rMAbs, introducing such high temperature would certainly cause artifacts in the form of fragmentation or dissociation which will degrade purity and impact lot release. Therefore, less harsh temperatures were evaluated through the range of 65-85 °C (2-15 min) targeting low fragmentation, fast derivatization rate, and low signal quenching. The results of this study showed that the most accurate representation of the sample was observed at 75 °C (5 min) for both nonreduced and reduced sample schemes. Under this condition, artifacts were absent, signal quenching was minimized, and derivatization rate was sufficient to generate high S/N. A similar optimization study was performed for unique rMAb molecules. Limit of Detection (LOD). The concentration LOD was determined using two approaches. For LOD determination of main species, the S/N ratio was calculated for an FQ-labeled rMAb serially diluted and extrapolated to S/N ratio of 3:1. For minor species, the LOD was determined from spiked contaminants of known concentration and extrapolating their S/N ratio to 3:1. An FQ-labeled rMAb sample was serially diluted with a 1:1 mixture of SDS sample buffer. Noise was calculated from the standard deviation of a 25 min electropherogram from a sample prepared without rMAb (blank). At the lowest dilution level, the S/N ratios for the LC and HC peaks were 870 and 1100, respectively. When extrapolated to S/N ratio of 3:1, the concentration LOD was ∼10 ng/mL (or 0.07 nM). The concentration LOD (35) Lee, H. G. J. Immunol. Methods 2000, 234, 71-81.

5970

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

for the intact antibody of the nonreduced method was 50 ng/mL (or 0.3 nM). The slight loss in sensitivity is attributed to the use of lower conjugation pH. Nevertheless, the LOD for FQ-labeled rMAb proteins is comparable to those typically obtained with silver-stained proteins. In a subsequent attempt to measure LOD, rMAb samples were spiked with three protein size standards (14, 29, and 66 kDa, Sigma-Aldrich) at the 0.05%, 0.1%, 0.5%, and 1.0% (w/w) levels as shown in Figure 4A. The accuracy of the assay plotted in Figure 4B compares the experimental % CPA to the calculated % CPA. At the 0.05% (w/w) spike level, the S/N ratio for the 14 kDa peak was 30:1 (13 nM), that for the 29 kDa peak was 40:1 (9 nM), and that for the 66 kDa peak was 15:1 (3 nM). Extrapolation to a S/N ratio of 3:1 suggests the LIF detector could measure these proteins at 0.02% (w/w) or 5 ng/mL. To confirm the LOD determined above, an rMAb sample was digested with trypsin prior to FQ labeling. The 100:1 protein-toenzyme sample was incubated for 5 min at 37 °C. The CE-SDS separations of the trypsin-treated sample and the control sample are overlaid in Figure 4C. The change in % CPA correspond to 0.05%, 0.04%, 0.02%, and 0.03% at points a, b, c, and d, respectively. This study confirms that changes at the 0.02% level can be detected for FQ-labeled rMAbs. Stability-Indicating Analysis. The ability of this assay to detect variable changes was demonstrated with a series of rMAb samples subjected to hydrolytic and thermal stress prior to labeling. The CE-SDS analyses of (A) nonreduced and (B) reduced rMAb stressed samples are presented in Figure 5. Samples from 1 and 3 weeks were compared to untreated control material. As demonstrated, acidic environments accelerated the generation of fragments or clips, whereas basic conditions facilitated HMW and/or nonreducible species. Samples exposed to high temperature exhibited a combination of effects. These data

confirmed that the CE-SDS assays could successfully detect various changes in stressed samples. Analysis of Nonreducible Aggregates. Nonreducible linkages, such as thioethers, are known to exist in therapeutic antibodies resulting from post-translational modifications and intense purification processes.13 SDS-PAGE and dSEC techniques can be used to ascertain the existence of low-level aggregates in our rMAbs, but these methods exhibit poor resolution or nonlinear quantitative staining. To demonstrate superiority of CE-SDS over these conventional techniques, pools of SEC monomer, dimer, and HMW fractions were prepared for dSEC and CE-SDS analyses. The profiles in Figure 6A are the analytical SEC chromatograms emphasizing enrichment of dimer and HMW material. These pools were reduced, denatured, and separated by a dSEC method shown in Figure 6B. The results of the CESDS analysis are adjacent in Figure 6C and show higher resolution for the various nonreducible species in each SEC sample. It should be noted that this information would not have been observed by CE-SDS methods using UV detection because the amount of starting material was well below the LOD for this type of detector. CONCLUSIONS In this study, a fast, simple fluorescent labeling scheme of therapeutic rMAbs was developed and characterized for its use as a quantitative tool for analysis of size variants by nonreduced and reduced CE-SDS using LIF detection. The data demonstrates that labeling rMAbs with FQ in the presence of alkylating reagents minimized sample preparation artifacts in the form of fragmentation, disulfide reshuffling, covalent interactions, or HMW aggregation. Antibodies labeled under reducing conditions were optimized with a D/P molar ratio of 10:1 at pH 9.3. Nonreduced samples prepared at pH 6.5 were stable but required a D/P molar ratio of 50:1 to overcome slow labeling kinetics. Under these conditions, detection sensitivity of FQ-labeled rMAbs was equivalent to

5-TAMRA derivatization and SDS-PAGE silver-stained detection. Although the labeling scheme described in this article was developed for characterization of intact and reduced antibodies, this method is applicable to all proteins analyzed by CE with LIF detection. Overall, the ease of preparing fluorescent proteins using this procedure makes the assay itself attractive for users in quality control environments. Additionally, LIF detection offers stable and low baseline noise which is generally uncommon in CE-SDS applications using UV detection. This feature greatly improves run-to-run precision since data integration is not hindered from system peak interferences that are commonly detected in UV methods. Furthermore, the performance of this CE-SDS assay is superior for monitoring low-level impurities including aggregates composed of nonreducible linkages between antibody size variants that are often difficult to detect or quantify by conventional techniques. ACKNOWLEDGMENT The authors gratefully thank our Protein Science department for manufacturing custom peptides used in reagent characterization. The authors thank Sibylle Wilbert of ZymoGenetics for her contribution to the on-line SEC-ESI/MS method used for analysis of labeled antibodies. We also thank Randall Bass and Mei Han for helpful discussions during the preparation of the manuscript. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review March 19, 2007. Accepted May 22, 2007. AC0705521

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5971