Characterization of Variable Regions of Monoclonal Antibodies by Top

Jun 26, 2007 - ... Mary Grubb , Wei Wu , Jonathan Josephs and William G. Humphreys ... Damian Houde , Joseph Arndt , Wayne Domeier , Steven Berkowitz ...
3 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2007, 79, 5723-5729

Characterization of Variable Regions of Monoclonal Antibodies by Top-Down Mass Spectrometry Zhongqi Zhang* and Bhavana Shah

Process and Analytical Sciences, Amgen Inc., Thousand Oaks, California 91320

A technique for rapid characterization of variable regions of monoclonal antibodies (mAb) is described. Several intact mAbs were analyzed on a Thermo-Fisher LTQOrbitrap high-resolution mass spectrometer (MS) by insource fragmentation. In-source fragmentation has the unique advantage of fragmenting all charge states of a protein at the same time and, thus, greatly improves the sensitivity of the fragment ions over a true MS/MS experiment, where a single charge state is isolated and fragmented. In addition, immediate fragmentation of the protein before tertiary structure formation may also facilitate protein fragmentation. This technique has been proved very useful for top-down analysis of large proteins. In-source fragmentation of mAbs generated a series of fragment ions. In addition to some small b and y ions from the light chain and heavy chain in the low m/z region, a series of b ions corresponding to N-terminal 106-120 residues of both heavy chain and light chain were observed. The cleavage sites for these b ions happen to be near the linker regions between the variable domains and the constant domains of these antibodies. These b ions, therefore, correspond to the entire variable region of each chain. Similar results were obtained for all mAbs analyzed, including both immunoglobulin G1 and G2 molecules. To further characterize the variable regions, these b ions were isolated and fragmented by collision-induced dissociation in the linear trap, followed by mass analysis in the orbitrap. Large number of product ions was observed from these b ions. Many of these product ions are internal fragments between the two disulfide-linked cysteine residues. To demonstrate the capability of the technique, several mAbs were force-oxidized by treating with tert-butyl hydroperoxide, followed by mass spectrometric analysis. In-source fragmentation and MS/MS of the variable region b ions clearly identified the locations of the oxidized methionine. Direct fragmentation of the parent ion of a large molecule, also known as “top-down” mass spectrometry, has been used for rapid characterization of small to medium-sized proteins,1-8 usually on * Corresponding author. Phone: (805)447-7783. Fax: (805)376-2354. E-mail: [email protected]. (1) Kelleher, N. L.; Lin, H. Y.; Valaskovic, G. A.; Aaserud, D. J.; Fridriksson, E. K.; McLafferty, F. W. J. Am. Chem. Soc. 1999, 121, 806-812. 10.1021/ac070483q CCC: $37.00 Published on Web 06/26/2007

© 2007 American Chemical Society

a Fourier transform ion cyclotron resonance mass spectrometer.8,9 Top-down characterization of large proteins is difficult, although some recent results suggest it is possible.10 Monoclonal antibodies (mAb), a class of proteins with mass of ∼150 kDa, are increasingly used as therapeutic agents. Rapid characterization of mAbs is becoming increasingly important. Although there are several classes of antibodies, all antibodies are composed of two identical heavy chains and two identical light chains; each has a variable region at its N-terminal end and a constant region at its C-terminal end. See Figure 1 for a schematic illustration of an immunoglobulin (Ig) G2 antibody. Since the constant regions are the same within each subclass of antibody, their structure and corresponding variants, including post-translational modifications, are often well characterized. However, the variable regions are different for each antibody, thus requiring an approach for rapid characterization. Traditionally, mAbs are characterized by “bottom-up” approaches. In a bottom-up approach, the protein is digested into peptides with a protease. The resulting peptides are then analyzed by LC/MS/MS. Pitfalls of the bottom-up approach include being labor-intensive, time-consuming, less than 100% sequence coverage, and artifacts introduced during digestion. Common artifacts include oxidation, deamidation, isoaspartate formation, succinimide formation, pyroglutamic acid formation, disulfide formation, disulfide scrambling, carbamylation, and protease-catalyzed amino acid rearrangements.11 This paper describes a simple technique for rapid characterization of the variable regions of mAbs. The technique employs insource fragmentation of the intact mAbs, combined with tandem mass spectrometry on a hybrid linear quadrupole ion trap-orbitrap mass spectrometer.6,12 Minimal sample preparation is required (2) Ge, Y.; Lawhorn, B. G.; ElNaggar, M.; Strauss, E.; Park, J. H.; Begley, T. P.; McLafferty, F. W. J. Am. Chem. Soc. 2002, 124, 672-678. (3) Sze, S. K.; Ge, Y.; Oh, H.-B.; McLafferty, F. W. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 1774-1779. (4) Ge, Y.; Lawhorn, B. G.; Elnaggar, M.; Sze, S. K.; Begley, T. P.; McLafferty, F. W. Protein Sci. 2003, 12, 2320-2326. (5) Kruppa, G. H.; Schoeniger, J.; Young, M. M. Rapid Commun. Mass Spectrom. 2003, 17, 155-162. (6) Macek, B.; Waanders, L. F.; Olsen, J. V.; Mann, M. Mol. Cell. Proteomics 2006, 5, 949-958. (7) Zabrouskov, V.; Han, X.; Welker, E.; Zhai, H.; Lin, C.; van Wijk, K. J.; Scheraga, H. A.; McLafferty, F. W. Biochemistry 2006, 45, 987-992. (8) Kelleher, N. L. Anal. Chem. 2004, 76, 196A-203A. (9) Bogdanov, B.; Smith, R. D. Mass Spectrom. Rev. 2005, 24, 168-200. (10) Han, X.; Jin, M.; Breuker, K.; McLafferty, F. W. Science 2006, 314, 109112. (11) Fodor, S.; Zhang, Z. Anal. Biochem. 2006, 356, 282-290. (12) Scigelova, M.; Makarov, A. Proteomics 2006, 16-21.

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5723

Figure 1. Schematic illustration of the structure of an IgG2 antibody. The light chains are colored in blue, heavy chains in red, and disulfide bonds in yellow. VL and VH represent variable domains for the light chain and heavy chain, respectively, and CL and CH represent constant domains for the light chain and heavy chain, respectively. There are three constant domains in each heavy chain (CH1, CH2, and CH3). The CH2 domains are glycosylated. Between the variable domains and the constant domains are short stretches of peptide referred to as the linker regions in this paper.

before mass spectrometric analysis. Several mAbs, including molecules from both IgG1 and IgG2 subclasses were tested. Similar results were obtained for all these molecules. An IgG2 molecule is used as an example throughout the paper. EXPERIMENTAL SECTION Sample Preparations. All intact mAbs, including one IgG1 molecule and six IgG2 molecules, used in this work were from Amgen (Thousand Oaks, CA). Before analysis, each antibody sample (10-50 µg of antibody) was desalted using a size exclusion column (Super SW 3000, 300 × 4.6 mm, Tosoh Bioscience, Tokyo, Japan) on an Agilent (Palo Alto, CA) 1100 HPLC system. The antibody was eluted from the column with a mobile phase containing 50% acetonitrile and 0.05% TFA at a flow rate of 0.2 mL/min. The eluted antibody was manually collected (∼100 µL collected) and infused into the mass spectrometer for analysis. For the results presented here, 10 µg of antibody was desalted. Forced Oxidation. For forced oxidation of the IgG2 molecule presented in this work, the antibody solution (18 mg/mL) was treated with 1% tert-butyl hydroperoxide (Sigma-Aldrich, Saint Louis, MO) in 10 mM sodium acetate buffer at pH 5.2. After a specific reaction time, an aliquot was removed and diluted 18fold to 1 mg/mL with ice-cold 10 mM sodium acetate buffer (pH 5.2). About 10 µL (10 µg of antibody) was desalted using the sizeexclusion column described above. Instrumentation. The mass spectrometer used in this work was a Thermo-Fisher (San Jose, CA) LTQ-Orbitrap high-resolution mass spectrometer. For the result presented here, the desalted antibody solution was infused into the mass spectrometer at a flow rate of 140 nL/min using a nanobore stainless steel emitter (Proxeon Biosystems) on a Thermo-Fisher nanoelectrospray source. In-source fragmentation, at a “source fragmentation” voltage (between the skimmer and the first multipole) of 65 V, was used to fragment each intact mAb. Other source conditions were optimized to produce the maximum amount of fragment ions. 5724 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Specifically, both the capillary voltage and tube lens voltage were set at their maximum levels (140 and 250 V, respectively) to increase fragmentation in the capillary-skimmer region. Heated capillary was set at a temperature of 320 °C. For MS/MS analysis, each ion of interest was isolated with an isolation window of 3.0 u and fragmented by collision-induced dissociation (CID) with normalized collision energy of 25%. Default values were used for most other data acquisition parameters (activation q ) 0.25, activation time 30 ms, FT AGC target 2 × 105 for MS mode and 1 × 105 for MSn mode). All spectra were acquired in the orbitrap in full profile mode with a resolution setting of 60 000 at m/z 400. Each spectrum was generated by signal averaging for 10 min (corresponding to ∼600 transients in MS mode and ∼250 transients in MS/MS mode) with a maximum injection time of 50 ms in MS mode and 2000 ms in MS/MS mode. The orbitrap was calibrated externally using a sodium trifluroacetate solution before analysis.13 Data Analysis. Fragment ions were assigned automatically by MassAnalyzer, a computer program developed in-house for automated bottom-up and top-down protein mass spectrometry data analysis. For bottom-up analysis of LC/MS/MS data, MassAnalyzer performs automated charge determination,14 peptide identification by correlating experimental tandem mass spectrum to theoretically predicted spectrum,15,16 and de novo peptide sequencing.17,18 For top-down analysis, MassAnalyzer automatically assigns protein fragment ions in a high-resolution spectrum. To search for a specific theoretical fragment ion (e.g., b1007+ or y1108+), MassAnalyzer first performs a search in the spectrum for the peak whose m/z value matches the theoretical m/z value of the fragment ion within a user-defined criterion (2 ppm in this work). Once the peak is found, the program estimates the theoretical isotope pattern of the fragment ion based on formulas developed previously16 and compares it to the experimental isotope pattern. The peak assignment is confirmed when the similarity score15 between the theoretical isotope pattern and experimental isotope pattern is above 0.94. The match between the theoretical isotope pattern and experimental pattern also ensures that the charge state of the peak is correctly assigned. When two or more assignments are made to a single peak, the following rules are used to determine the assignment that is most likely to be correct. (1) As the first rule, a fragment ion generated from cleavage of one peptide bond is more likely than fragment ions generated from cleavages of two peptide bonds (internal fragments). (2) All peptide bonds are not equally susceptible to CID. The following shows the likelihood of cleavage implemented in MassAnalyzer: (X-P, D-X) > E-X > (K-X, R-X, H-X, I-X, V-X, L-X) > X-X > (P-X, G-X),15 in which X represents any residues. All fragment assignments shown in this paper were validated manually by comparing their experimental isotope pattern to accurately calculated theoretical isotope pattern. Accurate calculation of the theoretical isotope pattern based on the elemental composition of a fragment ion is also conveniently (13) Moini, M.; Jones, B. L.; Rogers, R. M.; Jiang, L. F. J. Am. Soc. Mass Spectrom. 1998, 9, 977-980. (14) Zhang, Z.; Marshall, A. G. J. Am. Soc. Mass Spectrom. 1998, 9, 225-233. (15) Zhang, Z. Anal. Chem. 2004, 76, 3908-3922. (16) Zhang, Z. Anal. Chem. 2005, 77, 6364-6373. (17) Zhang, Z. Anal. Chem. 2004, 76, 6374-6383. (18) Zhang, Z.; McElvain, J. S. Anal. Chem. 2000, 72, 2337-2350.

Figure 2. In-source fragmentation spectrum of an IgG2 molecule. The full spectrum is shown on the top. The bottom shows the same spectrum from m/z 1500 to 1850. Major fragment ions are labeled as b or y ions with charge state. Charge states are omitted for singly charged ions. The prefix H represents heavy chain and L represents light chain. Ions labeled as a range indicate internal fragments. Not all identified fragment ions are labeled.

Figure 3. MS/MS of heavy-chain Hb115(7+) with identified b, y, and internal fragment ions (shown as a range of residues) labeled. The y ions are relative to the precursor Hb115 ion. Charge states are omitted for singly charged ions. The top shows the lower mass range and the bottom shows the higher mass range of the spectrum, each normalized to the most intense ion within that mass range. The location of the precursor ion (not observed) is marked by an arrow.

implemented in MassAnalyzer. MassAnalyzer outputs annotated spectra similar to Figure 7 (bottom) and fragment coverage maps similar to Figures 5 and 6.

RESULTS AND DISCUSSION Mass Determination of Intact mAbs. When intact mAbs were analyzed on the LTQ-Orbitrap, the intact mAb ions were Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5725

Figure 4. MS/MS of light-chain Lb115(7+) with identified b, y, and internal fragment ions labeled. The y ions are relative to the precursor Lb115 ion. Charge states are omitted for singly charged ions. The top shows the lower mass range and the bottom shows the higher mass range of the spectrum, each normalized to the most intense ion within that mass range. The location of the precursor ion (barely observed) is marked by an arrow.

observable but with poor signal-to-noise ratio (data not shown). The poor sensitivity of intact mAbs on the LTQ-Orbitrap is partly due to the fact that large proteins tend to fragment during ion transfer from the C-trap to the orbitrap. The intact mass of mAbs, however, can be conveniently and accurately determined on an electrospray time-of-flight instrument with an online LC/MS setup (data not shown). The chromatographic peak of interest can then be collected and infused into the LTQ-Orbitrap for top-down structural analysis with in-source fragmentation. In-Source Fragmentation of Intact mAbs. Seven intact antibodies, including both IgG1 and IgG2 subclasses, were analyzed on the LTQ-Orbitrap through in-source fragmentation. Figure 2 shows the in-source fragmentation spectrum of an IgG2 molecule as an example. It is seen that in addition to b2 to b7 ions of the light chain and b3, b4, y2 (y2 has a m/z of 173.0921 and is not shown in Figure 2), and y3 ions of the heavy chain, as well as a few internal fragment ions, a series of large ions at m/z 1200-2200 were observed. Computerized data analysis reveals that these fragment ions correspond to different charge states (6+ to 10+) of b106-b118 ions of the heavy chain and b114-b118 ions of the light chain, both with intact intrachain disulfide bonds. Similar spectra are obtained for all other IgG molecules studied in this work. For all IgG molecules studied, the cleavage sites are usually within residues 106-118 for the heavy chains and 112120 for the light chains. These cleavage sites happen to be near the linker regions between the variable domains and the constant domains for both chains of these antibodies (Figure 1), meaning that these b ions practically correspond to the entire variable regions for both the heavy chain and the light chain. The heavy chain b ions are usually less intense than the light chain b ions. The fact that the primary cleavage sites are near the linker regions between the variable domains and the constant domains 5726

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

is presumably because this region is much more flexible than other regions in the gas phase and, therefore, more susceptible to CID.19,20 Because the counterpart y ions are extremely large (∼135 kDa), they are not readily observable in the electrospray mass spectrum, partly due to the poor performance of the orbitrap for large proteins. The fact that the entire variable regions are directly observed without interference from the constant regions provides us a convenient method to characterize the variable regions, whose structures and possible post-translational modifications are usually poorly characterized because of their sequence differences among different antibodies. For top-down characterization, a true MS/MS should ideally be performed on the intact mAb by first isolating the parent ions, followed by collision-induced dissociation, in order to eliminate interferences from chemical contaminations. However, the poor sensitivity of mAb on an ion trap instrument makes this ideal approach impractical. In-source fragmentation has the advantage of fragmenting all charge states of a protein at the same time and thus greatly improves the sensitivity of the fragment ions over a true MS/MS experiment, where a single charge state is isolated and fragmented. In addition, immediate fragmentation of the protein before tertiary structure formation may also facilitate protein fragmentation (“prefolding dissociation”).21 This technique has been proved very useful for top-down analysis of large proteins.10,22 However, a relatively pure mAb sample is required for the in-source fragmentation analysis. Fortunately, because of (19) Wu, Q. Y.; Vanorden, S.; Cheng, X. H.; Bakhtiar, R.; Smith, R. D. Anal. Chem. 1995, 67, 2498-2509. (20) Zhang, Z.; Bordas-Nagy, J. J. Am. Soc. Mass Spectrom. 2006, 17, 786-794. (21) Zhai, H.; Han, X.; Breuker, K.; McLafferty, F. W. Anal. Chem. 2005, 77, 5777-5784. (22) Loo, J. A.; Edmonds, C. G.; Smith, R. D. Anal. Chem. 1991, 63, 24882499.

Figure 5. Fragment coverage map of heavy-chain Hb115(7+). The m/z values of most ions are indicated within parentheses. Also indicated on the map are two disulfide-linked cysteine residues at residues 22 and 97 and the methionine residue at 114.

Figure 6. Fragment coverage map of light-chain Lb115(7+). The m/z values of most ions are indicated within parentheses. Also indicated on the map are two disulfide-linked cysteine residues at residues 23 and 88 and the methionine residue at 4.

the high-energy source conditions required for in-source fragmentation of intact antibodies (see Experimental Section), most contaminant molecules are fragmented into small ions. As a result, the b ions at m/z 1300-1850 region (7+ to 9+) are relatively free from interferences from other smaller proteins or peptides.

In addition to characterizing the variable regions of mAbs, insource fragmentation of intact mAbs can also be conveniently used to characterize the N-terminal residues of both the heavy chain and the light chain, as well as the C-terminal residues of the heavy chain, from the small b and y ions. Small y ions of the light chain Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5727

Figure 7. Part of in-source fragmentation spectra of the IgG2 samples after 4 (middle) and 96 h (top) of oxidation as compared to the unoxidized sample (bottom). The heavy-chain variable region, as indicated by Hb114-Hb118 ions, is oxidized (+16 u) after 4 h, and the light-chain variable region is not oxidized after 96 h of oxidation. The oxidation site is determined to be Met-114 on the heavy chain because Hb112 and Hb113 ions are not oxidized.

are not observable because the C-terminal residue of the light chain is a disulfide-linked cysteine residue. Characterization of terminal residues can be very important sometimes due to variations of the terminal residues such as heavy chain C-terminal lysine processing23 and N-terminal pyroglutamate formation.24,25 For example, the heavy chain b3 and b4 ions shown in Figure 2 confirm that the N-terminal glutamine residue on the heavy chain of this IgG2 molecule is converted to pyroglutamate residue. MS/MS of the Variable Regions. To further characterize the variable regions, the b ions corresponding to the variable regions of the heavy chain and light chain were isolated and fragmented (23) Lazar, A. C.; Kloczewiak, M. A.; Mazsaroff, I. Rapid Commun. Mass Spectrom. 2004, 18, 239-244. (24) Chelius, D.; Jing, K.; Lueras, A.; Rehder, D. S.; Dillon, T. M.; Vizel, A.; Rajan, R. S.; Li, T.; Treuheit, M. J.; Bondarenko, P. V. Anal. Chem. 2006, 78, 23702376. (25) Yu, L.; Remmele, R. L.; He, B. Rapid Commun. Mass Spectrom. 2006, 20, 3674-3680.

5728 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

by CID in the linear ion trap, followed by MS/MS analysis (pseudo-MS3)22 in the orbitrap. Figure 3 shows the tandem mass spectrum of the heavy-chain Hb115(7+) ion of the IgG2 molecule, and Figure 4 shows the tandem mass spectrum of the light-chain Lb115(7+) ion. Fragment ions are automatically assigned by MassAnalyzer and validated manually. Determined masses of all fragment ions match their theoretical mass within 2 ppm. For fragment ions containing disulfide bonds, the mass changes due to the presence of these disulfide bonds are all considered when calculating the theoretical mass of these ions. Because of the disulfide bridges within the variable regions, a significant number of internal fragments between the two cysteine residues are observed. For easier visualization, a fragment coverage map for the heavy chain and light chain is shown in Figure 5 and Figure 6, respectively. Out of the 114 peptide bonds in the heavy-chain variable region, 31 bond cleavages were observed, representing

27% of the total peptide bonds. Similarly, out of the 114 peptide bonds in the light-chain variable region, 53 bond cleavages were observed, representing 46% of the total peptide bonds. Although the sequence coverage is not extensive, especially for the heavy chain, the fragments cover most parts of the variable regions. Sites of any modifications inside the variable regions can usually be narrowed down to a few residues (4 residues for the heavy chain and 2 residues for the light chain on average) using these fragment ions, when tandem mass spectra of modified and unmodified variable regions are compared. Determination of Modification Sites. The capability of the described technique was demonstrated by determining the oxidation site of force-oxidized samples of the IgG2 molecule. After forced oxidation for a specific time, the samples were infused into the mass spectrometer and analyzed by in-source fragmentation followed by tandem mass spectrometry. Figure 7 shows part of the in-source fragmentation spectra of the IgG2 samples after two different lengths of oxidation time as compared to the unoxidized sample. It is seen that the heavy-chain b114-b118 ions are completely oxidized (+16u) after 4 h of oxidation, while the lightchain b ions show no oxidation after 96 h. Further examination of the spectrum indicates that although Hb114-Hb118 ions of the heavy chain are completely oxidized, Hb112 and Hb113 ions are not oxidized at all. Therefore, it is concluded that the methionine residue at heavy-chain location 114 is responsible for the oxidation. To further confirm that methione-114 is responsible for the oxidation, the heavy-chain Hb115(7+) ion was isolated and fragmented using CID, followed by mass analysis in the orbitrap. Comparing the tandem mass spectra of oxidized and unoxidized Hb115(7+) (spectra not shown), it was observed that all y ions are oxidized. Importantly, b114 is oxidized, while b107 is not, demonstrating the modification site is within residues 108-114, supporting the conclusion that methionine-114 is responsible for the oxidation. MS/MS is not required to determine the modification site in this case because methionine-114 is coincidentally positioned within the cleavage sites of in-source fragmentation. However, MS/MS is required for many other cases when the modification site is in the other parts of the variable domains. It is therefore concluded that, for the two methionine residues in the variable regions of the IgG2 molecule, methionine-114 on the heavy chain is susceptible to oxidation under forced oxidation

conditions, while methionine-4 on the light chain is resistant to oxidation. None of the four tryptophan residues on the heavy chain and one tryptophan residue on the light chain is susceptible to oxidation. The readiness of methionine-114 on the heavy chain to be oxidized is apparently because it is located near the linker region between the variable domain and constant domain and thus solvent accessible. In-source fragmentation of intact mAb, followed by tandem mass spectrometry, is shown to be a rapid and sensitive method for characterization of the variable regions of antibodies. One clear disadvantage of this technique is that it does not provide any information in the constant regions of the mAbs. When the entire molecule needs to be characterized, a bottom-up strategy or a “middle-down” strategy, involving reducing the disulfide bonds and fragmenting the heavy chain and light chain separately, is usually employed. However, when the variable regions are the primary concern, the technique described here serves as a rapid method to provide structural information in these regions. This technique is especially useful for characterizing post-translational or in-process modifications in the variable regions. Although the sequence information provided in the MS/MS data is not extensive, taking advantage of the accurate mass measurement in a high-resolution instrument, exact type of modification, and modification site can often be determined when combined with chemical knowledge. Direct analysis of the intact antibodies also minimizes artifacts created during sample manipulation. This technique can potentially be used as a high-throughput method for characterizing modifications in the variable regions of monoclonal antibodies, when combined with a quick online desalting method. ACKNOWLEDGMENT The authors thank Jason Richardson for his help in setting up the nanoelectrospray system, Hai Pan for his helpful discussions during this work, and Janice Davis and Joseph Phillips for their support, discussion, and assistance during the work and preparation of the manuscript.

Received for review March 8, 2007. Accepted May 23, 2007. AC070483Q

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

5729