Quantification and Structure Elucidation of in Vivo Bevacizumab

Article pubs.acs.org/molecularpharmaceutics

Quantification and Structure Elucidation of in Vivo Bevacizumab Modification in Rabbit Vitreous Humor after Intravitreal Injection Kai On Chu,†,‡ David Ta Li Liu,† Kwok Ping Chan,† Ya Ping Yang,† Gary Hin Fai Yam,† Michael S Rogers,‡,§ and Chi Pui Pang*,† †

Department of Ophthalmology and Visual Sciences and ‡Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Hong Kong § Hong Kong Surgical Specialists, Hong Kong S Supporting Information *

ABSTRACT: Off-label and intravitreal use of bevacizumab, a recombinant immunoglobulin against VEGF, has been practiced widely for ophthalmic treatments. However, longitudinal data of its intravitreal status is unavailable due to a lack of reliable methods for bevacizumab determination. Thus its pharmacokinetics and pharmacodynamics are uncertain. We developed and validated a high performance liquid chromatographic method to determine bevacizumab in vitreous humor and utilized a novel strategy to assess in vivo temporal binding changes by affinity chromatography. Mass spectrometry and tandem mass spectrometry detection were used for structural evaluation. The coefficient of variation (CV) for intrabatch imprecision varied from 0.5 to 14.3% and for interbatch imprecision from 1.9 to 11.6%. The linearity was over 0.9982, lower limit of quantification 1.95 μg, recoveries over 95%, and accuracy between 90 and 112% over the range of 1.95−250 μg of bevacizumab in 100 μL of vitreous humor. Blank vitreous humor showed no interference peak. It was stable at room temperature for 5 h. Bevacizumab elimination in the vitreous followed first order kinetics with half-life as 5.7 days and elimination rate as 0.1221 day−1. Peptide mapping and tandem mass spectrometry revealed structural modifications of the in vivo bevacizumab mainly on the heavy chain in both variable and constant regions 7 days after intravitreal injection. Minor changes were also discovered on the light chain. Affinity chromatography showed significant affinity changes in samples 21 days after intravitreal injection. The changes were consistent with structural modifications as found in endothelial cell migration assays results. In conclusion, we have established a robust chromatographic method for determination of bevacizumab and strategies with affinity chromatography and molecular mass detection that revealed bevacizumab structural and possible functional changes in vitreous. KEYWORDS: quantification, characterization, bevacizumab, immunoglobulin, vitreous humor

■

using VEGF beads. It was subsequently filtered, bound with Rphycoerythrin immunoglobulin, and detected by fluorescence. Bakri et al.10 made an ELISA with VEGF immobilized on the plate. Bevacizumab was detected after binding with acridinium labeled anti-human IgG. However, none of these assay methods has been validated. Also, the presence of matrix can seriously interfere with antibody binding and signal response. Besides quantification, these assays did not provide information on the structural modifications of bevacizumab after intravitreal injection. It is known that when pharmaceutical monoclonal antibodies are stored at room temperature or 37 °C for some time, the structure changes according to C-terminal lysine variation,11 glycosylation heterogeneity,12 N-terminal glutamine cyclization,13 or oxidation and deamidation of asparagine.14 If these modifications occur in purified pharmaceutical solutions,

INTRODUCTION

Bevacizumab, brand name Avastin, is a recombinant monoclonal antibody against human vascular endothelial growth factor (VEGF) that has been approved by the FDA for treatment of colorectal cancer.1 It is also widely applied as an off-label treatment for various eye diseases including neovascular age-related macular degeneration (AMD),2,3 diabetic retinopathy,4,5 and central or branch retinal vein occlusion (CROVO, BRVO)6,7 as an alternative to the FDA-approved drug, ranibizumab, due to the much higher cost of the latter. The long-term clinical outcome and pharmacological effects on eye diseases of bevacizumab are still under investigation. Recently, a few reports have been published on the pharmacokinetics of bevacizumab in rabbit and human eyes.8−10 Concentrations of bevacizumab have been determined by immunoassays. Krohne et al.8 performed an ELISA assay to capture human IgG. After binding with biotinylated VEGF165, peroxidase-conjugated streptavidin was used to induce the color change. Beer et al.9 isolated bevacizumab © 2012 American Chemical Society

Received: February 22, 2012 Accepted: October 24, 2012 Published: October 24, 2012 3422

dx.doi.org/10.1021/mp3005403 | Mol. Pharmaceutics 2012, 9, 3422−3433

Molecular Pharmaceutics

Article

it is likely that similar changes might be accentuated in vivo, leading to changes in conformation and tertiary structure that affect the overall binding affinity.15,16 Structural changes of the antibody after interacting with vitreous collagen fibrils may subsequently affect its biological activities. In this study, we attempt to establish a methodology to reliably determine bevacizumab in vivo and its structural modifications. The vitreous humor does not normally contain immunoglobulins. We used a HPLC protein G column to capture human IgG and quantified it by diode array detector. To prove the identity, we used the retention time, UV spectrum, and high resolution LC-QTof system to monitor the structural modification of intact and digested bevacizumab before and after intravitreal injection. LC-QTof has been widely applied as a tool for quality control of pharmaceutical monoclonal antibody products because of its high resolution and mass accuracy for high mass molecules.17,18

recoveries were obtained by comparing the calibrated amount at each level calculated from the standard curve A to the actual spiked amount in vitreous humor (n = 4). Accuracy was evaluated by comparing the measured amount calculated by a calibration curve B to the spiked amount in vitreous humor (n = 4). The stability of bevacizumab in vitreous was assessed by storing it at room temperature or in ice for 5 h. It was quantified with HPLC and verified by LCMS for integrity. Specificity was evaluated by testing vitreous humor from 50 rabbits. Vitreous samples of 15 patients with different eye defects were also studied. To test presence of endogenous antibodies after multiple injections, the eyes of 3 rabbits were injected with 50 μL of saline. After 4 h and 1, 4, 7, 14, and 21 days, 50 μL of vitreous humor was collected for analysis. ELISA Assay on Bevacizumab in Vitreous Humor. Commercially available ELISA kit for human IgG assay from Bethyl Laboratory Inc. (Montgomery, TX) was used for comparison with results obtained from this new HPLC method. Standard curve of bevacizumab concentrations from 7.8 to 125 ng/mL as recommended by the supplier was processed 6 times. The assay was performed according to the supplier’s instructions. Vitreous samples, 100 μL each, were spiked with bevacizumab from 1.95 to 250 μg (Table 4). Each sample was processed four times according to the protocol for intrabatch precision and accuracy test. Three batches were processed for interbatch precision and accuracy assessment. Sample Collection for Temporal Structural Change Analysis. Bevacizumab was spiked into vitreous humor and incubated ex vivo at 37 °C for 0, 7, and 14 days, isolated by HPLC, and structurally characterized by liquid chromatography mass spectrometry (LCMS) and further liquid chromatography tandem mass spectrometry (LCMSMS). Bevacizumab was injected intravitreally into the eyes of 3 rabbits. After 1, 4, and 7 days, vitreous humor was withdrawn and processed similarly. Bevacizumab, 31.25 μg, was spiked in 50 μL of vitreous humor as positive control for structure verification. Similar collection strategy was used in peptide mapping except that the time points were as follows: Preinjection and 4 h, 1 day, 4 days, 7 days, 11 days, 14 days, and 21 days after intravitreal injection. They were either immediately stored at −80 °C or processed by isolation, reduction, deglycosylation, and tryptic digestion before LCMS analysis. Sample Preparation for Characterization. Bevacizumab in standard solution or vitreous humor was isolated by affinity chromatography. After neutralization by ammonium carbonate and volume reduction by vacuum evaporation, bevacizumab was extracted by C18 tips for LCMS analysis. Omix monolithic C18 tips were used for buffer exchange to replace conventional ultrafiltration membrane, size exclusion cartridges, centrifugal filtration,18,19 and solvent precipitation methods so as to improve bevacizumab cleansing and recovery that was complicated by sticky vitreal matrix. For fragmentation analysis, bevacizumab isolates were either reduced by 10 mM DTT directly or deglycosylated by 5 U PNFase F/Rapidgest SF with subsequent 10 mM DTT reduction. For peptide mapping, bevacizumab isolates were denatured by Rapidgest SF, reduced by 10 mM DTT, alkylated by 13 nM iodoacetamide, and subsequently digested by trypsin (trypsin/protein 1:25) overnight with 2 mM calcium chloride. All the digests were then extracted by C18 tips before LCMS and LCMSMS analyses.

■

EXPERIMENTAL SECTION Reagents and Materials. Bevacizumab (Avastin), 25 mg/ mL, and proteomics grade trypsin were obtained from Roche (CA, USA). Dithiothretol (DTT), iodoacetamide, PNGase F, and VEGF were from Sigma (St. Louis, MO, USA). Rapidgest SF, epoxy activated packing, and Waters AP mini-glass column were from Waters (Milford, MA, USA). All reagents were of the highest analytical grade. Sample Collection for Quantification. Vitreous samples were collected with informed consent from patients undergoing surgery at the Hong Kong Eye Hospital. Blank rabbit vitreous humor samples were obtained from the enucleated eyes after sacrificing of the rabbits in a CO2 chamber. All procedures were approved by the Ethics Committee of the Chinese University of Hong Kong for human and animal research. Instrumentation for Quantification. Vitreous humor was analyzed by a Waters high performance liquid chromatograph (HPLC) equipped with a Waters 616 pump, 717 autosampler, and 996 photodiode array under 260−300 nm. Sample molecules were separated on a Poros G/20 column, 4.6 × 50 mm, at 2 mL/min (Applied Biosystems, CA) at 35 °C. Mobile phase A contained 50 mM sodium dihydrogen phosphate and 0.15 M sodium chloride (pH 7.0) whereas mobile phase B contained 2% acetic acid and 0.15 M sodium chloride. The elution program: 0−3 min A (100%−100%), B (0%−0%); 3−4 min A (100%−0%), B (0%−100%); 4−7.5 min A (0%−0%), B (100%−100%); 7.5−8.5 min A (0%−100%), B (100%−0%); 8.5−12 min (100%−100%), B (0%−0%). 100 μL of vitreous humor was diluted to 150 μL by mobile phase A. 145 μL was injected into HPLC. Peak height at 280 nm was used for quantification. The standard curve was constructed from 100 μL of vitreous humor spiked with 1.95− 250 μg of bevacizumab. Validation. Reproducibility and linearity were assessed by spiking bevacizumab from 250 to 1.95 μg into 100 μL of rabbit vitreous humor. Four batches of vitreous samples were evaluated for intrabatch precision. The process was repeated for four days for interbatch precision. Lower limit of quantification (LLOQ) was defined as the concentration that gave around 15% of coefficient of variation (CV) of the four replicates of the spiked vitreous samples. Standard curve A was constructed with bevacizumab amounts ranging from 1.95−250 μg dissolved in mobile phase A. Standard curve B was constructed with bevacizumab range of 1.95−250 μg dissolved in vitreous humor. The 3423



Article

Instrumentation for Characterization. Bevacizumab isolate was analyzed by microflow LC system (Agilent 1100, Palo Alto, CA, USA). It was separated by x-Bridge BEH 130 C18 2.5 μm × 300 μm × 50 mm (Waters, Milford, MA, USA) at flow rate of 10 μL/min and 80 °C. Mobile phase A was 5% acetonitrile/0.1% formic acid. Mobile phase B was acetonitrile/ ethanol/isopropanol/water (6:1.5:1.5:1) with 0.1% formic acid. Binary flow was performed as follows: 0−2 min A (85%−85%), B (15%−15%); 2−10 min A (85%−20%), B (15%−80%); 10− 13 min A (20%−0%), B (80%−100%); 13−18 min A (0%− 0%), B (100%−100%); 18−20 min A (0%−85%), B (100%− 15%); 20−30 min A (85%−85%), B (15%−15%). Sample molecules were scanned by Micromass QTof micro (MSMS) with positive ESI spray, capillary voltage 4900 V, sample cone voltage 45 V, extraction cone voltage 1.0 V, source temperature 120 °C, desolvation temperature 350 °C, cone gas flow 50 L/h, and desolvation gas flow 800 L/min. The scan range was 1000−4000 m/z, calibrated by sodium cesium iodide (NaCsI). The operation system was MassLynx 4.0 and multiple ions deconvoluted by MaxEnt 1. For heavy and light chain characterization, the elution condition was modified: 15% mobile phase B maintained for 2 min, and then increased to 35% at the 13th minute. It was raised to 75% at the 18th minute, decreased back to 15% within two minutes, and then kept for 10 min. Scan range on the MSMS was changed to 700−3500 m/z and calibrated by NaCsI. For peptide mapping, tryptic digested samples were separated by Zorbax SB-C18 3.5 μm × 150 mm × 100 μm. Mobile phase A was 5% acetonitrile in 0.1% formic acid. Mobile phase B was 90% acetonitrile in 0.1% formic acid. The flow rate was 1 μL/min at 50 °C. Mobile phase B was kept initially at 0% for 5 min, increased to 40% at the 90th minute, raised to 90% at the 160th minute and kept for 10 min, decreased back to 0% within 5 min, and maintained for 18 min for another injection. LCMS was used for peptide analysis and LCMSMS for peptide sequence analysis, both utilizing the positive ESI nanolock spray system. Capillary voltage was 2800 V, extraction concentration voltage was 0.5 V, source temperature was 150 °C, cone gas (nitrogen) flow was 30 L/h, and scan mass range was 500 to 1600m/z with 1 s scan time and 0.1 s interscan delay with continuum mode acquisition. The lockspray solution was 500 fmol/μL [Glu1]-Fibrinopeptide B (Sigma). It was scanned every 90 s. The instrument was calibrated daily by 2 μg/μL NaCsI. For the MSMS analysis, survey mode was used. Data was acquired by positive ESI mode, and the threshold for MSMS switching was 5 counts/s. The scan mass range was 500 to 1600 m/z, scan time 2 s, and interscan delay 0.1 s with continuum mode. MSMS scan range was from 100 to 1700 m/ z. Peptide charge selected for MSMS was 2+, 3+, and 4+. Lockspray was scanned every 30 s. Include mode was used to select peptides of interest as observed from the MS scan. Pharmacokinetic Study. Bevacizumab levels in vitreous humor samples on days 1, 7, and 21 postinjection (n = 6 for each day) were determined after intravitreal injection with 50 μL of Avastin, which was equivalent to 1.25 mg of bevacizumab. Endothelial Cell Migration Assay. The experimental procedures essentially followed our previous report.20 In brief, human umbilical vein endothelial cell (HUVEC) line (CRL2873, ATCC) was grown in endothelial cell growth medium with 10% FBS and antibiotics on gelatin-coated surfaces. HUVECs at confluence were starved in medium with 0.5% FBS for 4 h. A scraping tool (1 mm width) was used to remove a

portion of the cell monolayer to create a line of denuded area. The dislodged cells were removed. The remaining cells were treated under different conditions: (i) blank rabbit vitreous humor (n = 6 for each day), 1, 7, and 21 days after intravitreal injection (i.v.) with 100 μL of saline, mixed with culture medium (3:7); (ii) the blank vitreous humor culture medium mixture (n = 6 for each day) spiked with VEGF at 5 μg/mL as a negative control; (iii) rabbit vitreous humors (n = 6 for each day), 1, 7, and 21 days after 2.5 mg of bevacizumab i.v. injection mixed with culture medium and spiked with VEGF at 5 μg/mL; (iv) the blank vitreous humor culture medium mixture (n = 6 for each day) spiked with same concentration of bevacizumab according to the in vivo vitreous samples in (iii), which was determined by HPLC and spiked with VEGF at 5 μg/mL; and (v) the blank vitreous humor culture medium mixture spiked with PP2 (Src kinase inhibitor, 20 μM) and VEGF at 5 μg/mL as positive control. Cells in the denuded area were monitored by phase contrast microscopy. The number of cells in defined areas were aligned, counted, and expressed as number of cells per square millimeter. Six images were captured for each sample. The results between the in vivo bevacizumab injected vitreous humor and the corresponding spiked vitreous humor samples with the same concentrations were compared by Student's t test. Data Analysis. Structure modification was determined by mass differences between the theoretical mass and the protein fragments after reduction or deglycosylation. Structural modifications were assessed from the ABRF database (http:// www.abrf.org/index.cfm/dm.home). Peptide mapping was analyzed by comparing the retention times of the standard and samples after different treatments. Peptides from different samples within ±100 ppm in the same retention time window were regarded as same peptides. Samples with peptides of different patterns or with abundance higher than 20 counts/s were taken for MSMS analysis. The MSMS data were submitted to ProteinLynx Global server for protein sequence analysis. Peptide identification parameters: tolerance 100 ppm, fragment tolerance 0.1 Da, after trypsin digestion, fixed modification with carbamidomethylation of cysteine, miss cleavage 1, lockspray calibration, and deisotoping process. For peptide identification the protein sequence was searched through 3 standard databases, Uniprot_Sprot, 25_H_Sapiens, and ipi_HumanV3_27. De novo sequencing was also used for the databank search unidentified peptides, with peptide search within 0.2 Da, trypsin digestion and fixed modification as carbamidomethylation of cysteine. The sequences were subsequently blasted through standard databases, using PAM30MS scoring matrix. Basically, the MSMS automatic search results with 95% confidence level were taken further for manual inspection. The first three highest score peptides were verified if they were relevant to human IgG or anti-VEGF IgG. The selected peptides were further verified by the blast results with the top three scores. Possible peptides with doubly charged ions, at least 5 isotopically resolved y-, b-, or a-ion or associated peaks must match theoretical peptide fragments. The 5 highest intensity fragment ions must belong to y-, b-, or a-ions instead of internal fragments. The major fragments with intensity higher than 10−20% of maximum intensity must match the theoretical peptide fragments. If the major peaks had m/z larger than the doubly charged parent mass, the percentage of intensity could be lowered to 5−10%. Relatively more random 3424



Article

Table 1. Reproducibility (a) and Accuracy(b) of Vitreous Humor Spiked with Different Amounts of Bevacizumab from 1.95 to 250 μg in 100 μL of Vitreous Humor (n = 4) and Determined by the Present Affinity Chromatography Method (a) Reproducibility intrabatch precision (n = 4), CV (%) amount (μg) 250.00 125.00 62.50 31.25 15.63 7.81 3.91 1.95

interbatch precision (n = 4), CV (%) 2.30 2.09 3.95 1.76 7.25 4.29 4.39 12.21

0.51 0.99 0.63 0.50 5.17 5.55 5.71 14.33

5.07 2.54 2.50 0.66 4.11 4.21 2.61 11.50 (b) Accuracy

5.03 4.21 1.01 0.90 7.21 3.93 8.60 12.59

3.40 2.90 2.58 1.86 5.61 4.27 5.31 11.60

accuracy in each batch % (n = 4) amount (μg) 250.00 125.00 62.50 31.25 15.63 7.81 3.91 1.95

batch 1 102.1 98.1 94.5 94.7 102.6 97.3 105.7 108.7

± ± ± ± ± ± ± ±

2.3 2.0 3.5 1.5 5.8 3.5 3.3 6.0

batch 2 101.2 98.8 90.3 95.9 103.3 95.6 103.4 111.6

± ± ± ± ± ± ± ±

batch 3

0.5 0.9 0.5 0.4 4.1 4.4 4.0 6.6

104.1 102.4 93.3 96.3 100.3 98.6 105.0 104.4

(1)

kt = K a, t mL /VM

(2)

5.2 2.5 2.2 0.6 3.2 3.4 2.1 6.1

batch 4 105.9 104.1 92.6 96.1 95.1 96.9 107.6 111.2

± ± ± ± ± ± ± ±

5.3 4.3 0.9 0.8 6.0 2.9 6.7 6.4

overall accuracy (%) 103.3 100.9 92.7 95.7 100.3 97.1 105.4 109.0

± ± ± ± ± ± ± ±

4.0 3.5 2.5 1.0 5.5 3.4 4.2 6.4

points, respectively. Ka,std and Ka,t are the association equilibrium constant of the standard bevacizumab to VEGF and bevacizumab at different time points to VEGF respectively. mL is the moles of VEGF in the column. VM is the void volume.

fragmentation was allowed in the low m/z region. Triply charged ions could be considered as positive, if most fragments matched the theoretical fragments with high intensity peaks matching the b-, y- or a-ion series. Also, if the match scores were significantly higher than the second highest score and most fragment peaks matched the theoretical fragment with high intensity peaks matching the b-, y-, a-ion series, the peptide was identified.21 Functional Analysis. A Protein-Pak epoxy-activated affinity column (Waters) for VEGF was prepared according to the supplier’s protocol. In brief, 1 g of epoxy-activated packing (Waters, Milford, MA) was incubated with 30 μg of VEGF (Sigma). The column was washed, blocked by ethanolamine solution, and washed with 1 M sodium chloride. Four batches of standard bevacizumab, rabbit vitreous humor with bevacizumab 4 h and 1, 4, 7, and 21 days after intravitreal injection were chromatographed on a VEGF column. The compositions of mobile phase A and B were the same as in the quantification method but with the elution program changed to the following: 0−3 min A (100%−100%), B (0%−0%); 3−15 min A (100%−0%), B (0%−100%); 15−24 min A (0%−0%), B (100%−100%); 24−25 min A (0%−100%), B (100%−0%); 25−30 min A (100%−100%), B (0%−0%). The flow rate was 0.4 mL/min and column temperature 30 °C. Bevacizumab was detected by photodiode array detector at wavelength 260−300 nm. Temporal Affinity Change Analysis. The change of affinity of bevacizumab toward VEGF was assessed by the zonal elution method22,23 according to the retention capacity and the relative Ka at different time points. kstd = K a,stdmL /VM

± ± ± ± ± ± ± ±

relative K a, t = kt /kstd

(3)

where relative Ka,t is the relative association constant at different time point with respect to that of standard bevacizumab. With the 4 h samples as a reference point, the changes of relative (rel) Ka in different days were assessed: change of rel K a, t with respect to the baseline rel K a,4h = rel K a, t − rel K a,4h

(4)

where relative Ka,4h is the Ka at time point 4 h with respect to Ka of standard bevacizumab and relative Ka,t is the Ka at time point t with respect to Ka of standard bevacizumab.

■

RESULTS Quantification and Validation. In LC, affinity chromatography analysis of bevacizumab was tested from 1.95 to 250 μg in 100 μL of vitreous humor. The intrabatch imprecision (CV) varied from 0.5 to 14.3% and the interbatch imprecision (CV) from 1.9 to 11.6% (Table 1a). Linearity was over 0.9982 for spiked samples or standard solution curve. The lower limit of quantification was 1.95 μg. The recoveries of the spiked samples were over 95% (Supplementary Table S1 in the Supporting Information) while the accuracy of all 4 batches was from 90 to 112% over the range of 1.95−250 μg (n = 4) (Table 1b). No mass difference was found between the mass spectra of pharmaceutical sample and bevacizumab spiked vitreous humor sample (p > 0.05, n = 3). Blank vitreous humor showed no interference peak (Figure 1a). The mass differences of bevacizumab isolated from vitreous humor by affinity column and standard bevacizumab were within 30 ppm (0−3 Da) after deconvolution (Figure 1b). There was no obvious difference in

where kstd and kt are the retention capacity factors of standard bevacizumab peak and bevacizumab peaks from different time 3425



Article

35% after 28 days (Supplementary Table S2 in the Supporting Information). Pharmacokinetic Study. The pharmacokinetics of the in vivo vitreous samples followed first order kinetics with elimination constant (λ) as 0.1221 day−1 and half-life (t1/2) as 5.7 days. The concentration of bevacizumab can be sustained at 58.5 μg/mL after 21 days of intravitreal injection (Supplementary Table S3 in the Supporting Information). Structural Modification. Highly heterogeneous spectra were observed in intact bevacizumab on day 7 onward for ex vivo incubation samples and on day 4 onward for in vivo sample (Figure 2). Highly heterogeneous peaks were also found in

Figure 2. Mass spectra of intact bevacizumab after intravitreal injection. Mass spectra showed the structural modifications of in vivo rabbit vitreous samples on one day and four days after intravitreal injection. The mass spectrum of bevacizumab on day 7 could not be obtained due to highly heterogeneous structure.

Figure 1. Chromatograms and mass spectra of intact bevacizumab. (a) The upper chromatogram with 31.3 μg of standard bevacizumab, the middle chromatogram with 31.3 μg of bevacizumab in 50 μL of vitreous humor, and the lower chromatogram with 50 μL of blank vitreous humor. (b) the corresponding LCMS chromatograms with isolated standard bevacizumab (upper) and isolated bevacizumab from vitreous humor (lower) directly injected into LCMS. The inscribed diagrams are their corresponding deconvoluted mass spectra with scan range from the mass range of 2200−4000 m/z. Deconvolution was performed by the MaxEnt 1 algorithm. The masses of the upper MS were 149180, 149341, and 149500.5 Da while the masses of the lower MS were 149181, 149344, and 149503.5 Da respectively.

heavy chain of in vivo samples on day 7 after direct DTT reduction only or after deglycosylation/reduction (Figure 3). Intact bevacizumab was found having a mass of 149180 Da with its main isoform 149341 Da (Figure 1b). Deglycosylation of intact bevacizumab showed the mass was at 146291 Da (Supplementary Figure S3 in the Supporting Information). The mass of heavy chain was 51154.5 Da and light chain was 23451.5 Da after DTT reduction (Supplementary Figure S2b in the Supporting Information). After deglycosylation and reduction, the mass of heavy chain was 49710.5 Da and of light chain was 23452 Da (Figure 3a). Binding Affinity Analysis. The relative Ka (association constant) on day 21 of bevacizumab was greater than at 4 h, day 1, and day 7. The absolute difference between the relative Ka at day 21 and that at 4 h was significantly higher than the difference between day 1 and 4 h and between day 7 and 4 h

concentration and structure between storage in ice and at room temperature for 5 h. More than 95% of the initial amount remained. No antibody peak was found in the fifteen human, and fifty rabbit vitreous blank, and in vitreous humor of three rabbit eyes after intravitreal saline injection. The amount of spiked bevacizumab in ex vivo incubation samples decreased to 3426



Article

Figure 3. Mass spectra of deglycosylated/reduced bevacizumab after intravitreal injection: (a) spiked standard and (b) day 1, (c) day 4, and (b) day 7 in vivo samples respectively. In (a), mass 23452 m/z was the light chain, and mass 49710.5 m/z was the deglycosylated heavy chain. In (b), mass 23452.5 m/z was the light chain, and mass 49730.5 m/z was the deglycosylated heavy chain with sodium adduct. In (c), mass 23448 m/z and 23470 m/z were the light chain with modificaitons, and mass 49715.0 m/z was the deglycosylated heavy chain with oxidation. All the mass spectra were deconvoluted by the MaxEnt 1 algorithm. In (d), masses 23448.0 m/z and 23453.0 m/z were light chain with various modifications; the heavy chain could not be found.

Table 2. Change of Affinity of Bevacizumab after Intravitreal Injectiona kc

rel Kad

absolute diff rel Kae

days after i.v.b

batch 1

batch 2

batch 3

batch 4

batch 1

batch 2

batch 3

batch 4

batch 1

batch 2

batch 3

batch 4

std 4h 1 day 7 days 21 days

6.4354 6.8673 6.8579 6.8378 6.9125

6.7342 6.8310 6.9176 6.8977 6.6838

6.7800 6.8135 6.7814 6.7753 6.5737

6.7917 6.7527 6.7304 6.6842 7.0565

1.0671 1.0657 1.0625 1.0741

1.0144 1.0272 1.0243 0.9925

1.0049 1.0002 0.9993 0.9796

0.9942 0.9910 0.9842 1.0390

0.0015 0.0046 0.0070

0.0129 0.0099 0.0219

0.0047 0.0056 0.0174

0.003282 0.010076 0.044736

a

The table shows the capacity factor and relative association constant, Ka, changes after intravitreal injection at different periods from 3 rabbit eyes. The difference of relative Ka (diff rel Ka) is related to the changes of relative Ka with respect to the relative Ka at 4 h. bI.v.: Intravitreal injection. ck: Capacity factor ((tb − t0)/t0) where tb and t0 are the retention time of bevacizumab and void volume respectively. dRel Ka: Relative association constant (kt/kstd) where kt and kstd are the capacity factor at different time points and standard respectively. eDiff rel Ka: Difference of relative Ka (rel Ka,t − rel Ka,4h) where rel Ka,t and rel Ka,4h are the relative association constant at different days and relative association constant at 4 h. Significantly smaller diff rel Ka at day 1 (p = 0.031, n = 4) and day 7 (p = 0.049, n = 4) were found when they are compared to the diff rel Ka at day 21 as calculated by ANOVA post hoc Dunnett test using day 21 as control in the calculation.

(Table 2, Supplementary Figure S5 in the Supporting Information). Peptide Mapping. Same chromatography profiles and peptide patterns were found within three batches of eye samples within the same time point. Chromatography profiles and consistent peptide patterns were generally similar for the trypsin digested standard and samples after intravitreal injection from the same eye. Although bevacizumab concentrations between day 11 and 21 were so low that some clear chromatographic peaks were not observed, the peptides could

still be found in mass spectra within the retention time window (Figure 4). Some representative peptides are shown in Table 3. In general, the peptide patterns of the samples and standards from day 1 to day 7 were similar. However, from day 11 to day 21, certain patterns became different. Some peptides became undetectable. Peptides 871.4 m/z (Figure 4c) and 949.4 m/z (Figure 4d) faded gradually. Peptides 888.0 m/z at about 50.5 min and 863.5 m/z, 882.4 m/z, and 907.0 m/z at about 53.0 min disappeared on day 7 whereas peptides 1121.0 m/z at about 50.5 min, 879.4 m/z and 910.4 m/z at about 53.0 min, 3427



Article

Figure 4. Tryptic peptide obtained from standard and in vivo bevacizumab samples at different time points. Diagrams show representative mass spectra of trypsin digested peptides of bevacizumab standard and vitreous samples from day 1, day 4, day 7, day 11, day 14, and day 21 after intravitreal injection. Mass spectra show peptides 1070.5 m/z [4H+,4+], 1116.0 m/z [2H+,2+], and 1129.0 [2H+,2+] at 49.0 min (a); 799.9 m/z [2H+,2+] and 860.4 m/z [2H+,2+] at 49.9 min (b); 871.4 m/z [2H+,2+] at 56.8 min (c); and 949.0 m/z [2H+,2+] and 1102.5 m/z [2H+,2+] at 64.7 min (d). The number on the right-hand side indicates the maximum abundance in the spectra.

■

and 795.9 m/z at about 56.0 min appeared on day 7 but were not found in standard, day 1, and day 4 (Table 3). MSMS analysis showed the peptide sequence of 593.8 m/z at 49.8 min was GPSVFPLAPSSK (Figure 5a), 879.9 m/z at 50.5 min was CVVVDGEHEDLLMSR (Figure 5b), 941.5 m/z at 50.5 min was EVQLVESGGGLVQPGGSLR (Figure 5c), and 871.4 m/z at 55.9 min was PEDVDWHEDPEVQF (Figure 5d). Peptides of 593.8 m/z, 879.9 m/z, and 941.5 m/z were not found on day 11, day 14, and day 21 whereas peptide 871.4 m/ z was retained till day 21 (Table 3). ELISA. The standard curve was not linear (Figure 6). The log function curve only marginally fit with the data (r2 = 0.9584). The intra- or interbatch imprecision (CV) of bevacizumab ranged from 1.95 to 250 μg in the spiked vitreous samples and was greater than 15% at some concentrations. The trend of variation did not show concentration dependence (Table 4). The accuracy was between 70 and 120%, independent of concentrations. Endothelial Cell Migration Assays. No significant difference of the number of cells between the bevacizumab treated vitreous humor and the corresponding spiked vitreous humors was found for the day 1 and 7 samples. However, a marginally significant difference was found between the day 21 bevacizumab treated vitreous sample and spiked vitreous samples (Figure 7).

DISCUSSION

Affinity Chromatography Quantification. ELISA type immunoassay for bevacizumab was usually inconsistent in the presence of aqueous humor matrix (16.6 to 42.5 μg/mL)8 or vitreous humor,8,9 or even in standard bevacizumab (CV ∼ 10%).24 The ELISA results in this study were consistent with such reported findings. While a log function curve could be drawn (R2 = 0.9584) (Figure 6), a small log variation reflected a large variation in concentrations. Accurate quantification was difficult since the standard curve was not linear. Even within the working range of the standard calibration curve, the CV of the calculated amount of bevacizumab in spiked vitreous samples was greater than 15% in some concentrations (Table 4). The variation was independent of concentrations and appeared random. The limit of detection was also difficult to define. The wide range of accuracy, from 70% to 120%, appearing in different batches and different concentrations indicated the uncertain reliability of the ELISA results. Furthermore, these immunoassay methods detected only the unbound bevacizumab. The affinity columns in our chromatographic procedure, however, involved a continuous washing step to remove interfering proteins before eluting the captured protein. Therefore the linearity, reproducibility, and accuracy were improved. Since the affinity column captured the Fc of the bevacizumab, it did not compete with the VEGF binding site. 3428



Article

Table 3. Representative Peptides from Different Retention Times and from Different Days mass of peptides (m/z) retention time (min)

std

day 1

day 4

day 7

day 11

50.5

53.0

1116.0 593.8 799.9 860.4 750.4 821.4 879.9 888.0 936.4 941.5

825.4 851.4 863.5 874.4

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

[2H+,2+] [2H+,2+] [2H+H+,2+] [2H+,2+]

882.4 [2H+,2+] 907.0 [2H+,2+] +

55.9 56.0 56.8 64.7

day 21

+

49.0 49.8 49.9

day 14

1116.0 593.8 799.9 860.4 750.4 821.4 879.9 888.0 936.4 941.5

825.4 851.4 863.5 874.4

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

882.4 [2H+,2+] 907.0 [2H+,2+] +

1116.0 593.8 799.9 860.4 750.4 821.4 879.9 888.0 936.4 941.5

825.4 851.4 863.5 874.4

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

1116.0 593.8 799.9 860.4 750.4

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

1116.0 [2H+,2+]

1070.5 [4H ,4+] 1116.0 [2H+,2+]

1116.0 [2H+,2+]

860.4 [2H+,2+]

860.4 [2H+,2+]

860.4 [2H+,2+]

879.9 [2H+,2+] 936.4 [2H+,2+] 941.5 [2H+,2+] 1048.5 [2H+,2+] 1121.0 [2H+,2+]

1121.0 [2H+,2+] 825.4 [2H+,2+] 851.4 [2H+,2+] 874.4 [2H+,2+] 879.4 [2H+,2+]

882.4 [2H+,2+] 907.0 [2H+,2+] +

918.0 [2H ,2+] 926.0 [2H+,2+] 871.4 [2H+,2+]

918.0 [2H ,2+] 926.0 [2H+,2+] 871.4 [2H+,2+]

918.0 [2H ,2+] 926.0 [2H+,2+] 871.4 [2H+,2+]

812.4 [2H+,2+] 871.4 [2H+,2+] 949.0 [2H+,2+]

812.4 [2H+,2+] 871.4 [2H+,2+] 949.0 [2H+,2+]

812.4 [2H+,2+] 871.4 [2H+,2+] 949.0 [2H+,2+]

910.4 918.0 926.0 871.4 795.9 812.4 871.4 949.0

[2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+] [2H+,2+]

871.4 [2H+,2+]

871.4 [2H+,2+] 949.0 [2H+,2+]

871.4 [2H+,2+] 949.0 [2H+,2+] 1102.5 [2H+,2+]

871.4 [2H+,2+] 1102.5 [2H+,2+]

ionization efficiency. Conventional methods have to be revamped for measuring low levels, below 0.31 μg/μL, of endogenous samples. In this study, besides affinity chromatography for purification and high capacity C18 Tips for extraction, we also used a special XBridge BEH 130 C18 2.5 μm × 300 μm × 50 mm capillary column to improve the sensitivity. This column is tolerant of high temperature. The small in particle size and narrow column diameter reduce mass transfer, minimize heterogeneous interaction, increase separation efficiency, and improve electrospray ionization efficiency due to allowing low flow rate (10 μL/min). Our other finding is that reduced products must be analyzed immediately after DDT reduction. Otherwise, the heavy chain signal could not be detected. It may be due to recombination of the disulfide bond after buffer exchange and acidification. Some studies performed reduction before deglycosylation.17,19 However we changed the order to deglycosylation before reduction and obtained satisfactory results. DTT may damage the deglycosylation enzyme, PNGas F, and reduce its activity in the medium. Unlike analysis of pharmaceutical products usually at the milligram level,17 protein analysis of in vivo samples requires the sensitivity below the microgram level. Therefore, in this study, microflow LC system and capillary columns were used to improve the ESI sensitivity for intact protein and peptide mapping analysis. NanoLockSpray was used to improve the mass accuracy. These new techniques allow detection of peptide patterns with good consistency for pattern comparison.

Therefore, it evaluated the total bevacizumab but not unbound bevacizumab as in the ELISA assay. The chromatogram (Figure 1a) showed the standard pharmaceutical bevacizumab and spiked peaks were similar whereas the vitreous blank showed no interference, indicating the affinity column specifically analyzed bevacizumab in vitreous humor. LCMS also showed the retention time and mass spectra of the bevacizumab resulting from affinity chromatography and ex vivo bevacizumab to be comparable and without obvious interference (Figure 1b). Besides specificity, the structure of bevacizumab did not change in vitreous humor. Bevacizumab was absent in blank human vitreous humor, rabbit vitreous humor, and in vivo vitreous humor samples after saline injection. Moreover, bevacizumab was stable in vitreous humor over the 5 h collection period at 4 °C or at room temperature. This affinity chromatography method was thus deemed validated and reliable for quantification of bevacizumab in vitreous humor. Characterization Procedure Improvement. Many studies on ocular bevacizumab focused on its ocular level but not the in vivo structural changes even though many pharmaceutical antibodies underwent structural modifications on storage.14,18,25 These modifications, including glycan modification, lysine clipping, deamination, and pyroglutamate formation,14,17,28 usually involved small mass changes.25,28 Therefore, it requires high resolution mass spectrometry and size reducing manipulation for structural elucidation and modification detection.19 Also, these studies usually need a milligram level of purified antibodies25 to overcome the matrix effects and low 3429



Article

Figure 5. MSMS spectra and sequencing identification of bevacizumab tryptic peptides. (a) The MSMS spectrum of peptide 593.8 m/z [2H+,2+] at about 49.8 min. The sequence was GPSVFPLAPSSK, which was identified as Q9Y509_Human, which was located in Ig heavy chain variable region. (b) The spectrum of peptide 879.9 m/z [2H+,2+] at about 50.5 min. The sequence was CVVVDGEHEDLLMSR, which was identified as IGHG1_Human located in Ig gamma 1 heavy chain constant region. (c) The spectrum of peptide 941.5 m/z [2H+,2+] at about 50.5 min. The sequence was EVQLVESGGGLVQPGGSLR, which was identified as ACQ91141.1 located in anti-VEGF immunoglobulin heavy chain variable region. (d) The spectrum of peptide 871.4 m/z [2H+,2+] at about 56.8 min. The sequence was PEDVDWHEDPEVQF, which was identified as IGHG1_HUMAN located in Ig gamma 1 heavy chain variable region.

reduction and deglycosylation/reduction process. The mass difference was 1444 Da between a reduced chain, 51154.5 Da (Supplementary Figure S2 in the Supporting Information), and a deglycosylated and reduced heavy chain (Figure 3a), 49710.5 Da. After adding 1 Da for N-D correction, the mass was also 1445 Da. Moreover, another heavy chain with mass 51316 Da, which was 161.5 Da heavier, further supports the presence of G0F-G isoform. Furthermore, summation of the mass of light chains and heavy chains and minus 32 Da from 16 broken disulfide bonds17,28 produced 149180 Da. This mass is coincident with the intact mass. Therefore, bevacizumab is about 149180 Da with two light and two heavy chains, sixteen disulfide bonds, and two fucosylated biantennary glycans, G0F, and has a main G0F-G isoform.

Figure 6. Standard calibration curve of bevacizumab (n = 6) with recommended concentration working range from 7.8 to 125 ng/mL.

The high resolution and mass accuracy of QTof system also helped distinguishing any peptide modification. With the ProteinLynx Global Server system, the peptides could be sequenced and structural modification in bevacizumab could be identified. Structural Elucidation. Owing to the proprietary patent, no structure sequence information is disclosed as reference for accurate mass determination. The mass of intact bevacizumab was found to be about 149180 Da. It contained two isoforms at 149341 and 149500.5 Da. The isoforms had additional multiple 162 Da masses (Figure 1b). These may be due to the addition of galactoses at the end of the glycan chains.26,27 Since the mass of deglycosylated antibody was 146291 Da, its mass difference was 2889 Da compared to glycosylated antibody. After correction of two Da for N-D product, the mass of a glycan was 1445.5 Da. This mass was consistent with a typical fucosylated biantennary glycan, G0F, in monoclonal antibodies.26,27 This structure can be further supported by the

Structure Modification Studies. The amount of bevacizumab remained at about 35% of its initial value 28 days after ex vivo incubation and 15% of initial concentration 7 days after intravitreal injection, suggesting the degradation was 3430



Article

Table 4. Reproducibility (a) and Accuracy (b) of Vitreous Humor Spiked with Different Amounts of Bevacizumab in 100 μL of Vitreous Humor (n = 4) and Determined by ELISA with Dilutions and Working Range Recommended by the Instruction of the Kits (a) Reproducibility CV (%) bevacizumab amt (μg)

batch 1 (n = 4)

1.95 3.91 7.81 15.63 31.25 62.5 125 250

7.64 7.32 8.37 14.19 6.36 6.63 11.61 9.41

batch 2 (n = 4)

batch 3 (n = 4)

interbatch (n = 3)

8.23 16.63 11.55 4.09 23.30 1.49 5.38 5.26

10.95 17.10 10.76 0.59 5.25 9.59 9.47 8.68

16.72 6.76 9.66 6.23 23.00 6.03 6.76 7.62 (b) Accuracy mean ± SD (%)

bevacizumab amount (μg) 1.95 3.91 7.81 15.63 31.25 62.5 125 250

batch 1 (n = 4) 74.15 117.36 111.40 108.58 109.69 94.74 93.65 93.71

± ± ± ± ± ± ± ±

5.67 8.59 9.32 15.40 6.98 6.28 10.88 8.82

batch 2 (n = 4) 86.62 112.60 113.63 109.42 120.13 114.69 113.08 109.90

± ± ± ± ± ± ± ±

14.48 7.61 10.97 6.82 27.63 6.91 7.65 8.37

batch 3 (n = 4) 70.53 84.24 92.86 109.85 109.99 107.83 106.54 108.79

± ± ± ± ± ± ± ±

5.80 14.01 10.72 4.50 25.63 1.61 5.73 5.72

overall accuracy (n = 3) 77.10 104.73 105.96 109.28 113.27 105.75 104.43 104.13

± ± ± ± ± ± ± ±

8.44 17.91 11.40 0.65 5.94 10.14 9.89 9.04

structure of light chain may be more stable since clear mass at about 23452 Da appeared until day 7. Nevertheless, oxidation may occur in the light chain because a peak 23458 Da was detected. Another modification with 21 Da larger than the light chain in day 1 and day 4 samples should be sodium adduct (http://www.abrf.org/index.cfm/dm.home). Heavy chain showed lower intensity, likely due to its poorer ionization efficiency compared with light chain (Supplementary Figure S2 in the Supporting Information). The in vivo deglycosylated and reduced heavy chain, 49710.5 Da, also became more heterogeneous with time (Figure 3). Although intact and oxidized heavy chain in day 4 samples could be detected, the heavy chain could not be detected in day 7 samples. Similarly, the light chain was basically intact for seven days although phosphorylation, hydroxylation, and crosslinking may have occurred in the day 4 light chain. Sodium adduct was also found in day 1 heavy chain samples. Simple modification like lysine clipping, glutamyl clipping, deamination, and pyroglutamate formation commonly found in the N-terminal of light chain in pharmaceutical products14,17,28 did not appear in the in vivo samples. These findings imply that intensive modifications occurred in the heavy chain peptide. Structural modifications of bevacizumab were obvious 4 days after intravitreal injection, mainly in the heavy chain peptides and lesser in the light chain. However, modifications might also occur in the glycan. Recently, affinity chromatography has been extensively used to assess and compare biological ligand affinity interaction due to its high throughput, sensitive, and precise properties.29,30 Affinity chromatographic analysis for in vivo bevacizumab samples in this study (Table 2 and Supplementary Figure S5 in the Supporting Information) showed that the affinity change was significantly greater at 21 days postinjection when compared to that at day 1 and day 7.

Figure 7. Comparison of the antiproliferative effects on HUVECs by bevacizumab treated in vivo vitreous humor 1, 7, and 21 days after intravitreal injection and the corresponding bevacizumab spiked vitreous humor with the same concentrations. The figure shows a marginally significant difference of antiproliferative effects appearing in the day 21 postinjected vitreous samples.

much faster in vivo, likely due to absorption. The ex vivo intact bevacizumab showed heterogeneous molecular peaks 7 days after incubation whereas the in vivo bevacizumab samples showed heterogeneous peaks appearing 4 days after intravitreal injection (Figure 2). Therefore the structure of bevacizumab was modified gradually within days following treatment. The resolution of the Q-Tof micro system used in this study is about 5000 (fwhm). The resolving power is limited to about 40 Da for 149200 Da antibody.18 Therefore modifications may not be readily detectable for intact protein. By reducing the antibody into light chain and heavy chain, the molecular size was reduced to 20000−50000 m/z. The resolution should be sufficient to resolve for 17 Da change. Modifications commonly found in pharmaceutical antibodies like cyclization of glutamate to pyroglutamic acid are within this range.17−19 Heavy chain peak became more heterogeneous with time in the in vivo samples, suggesting structural modifications. The 3431



Article

Peptide Mapping Supporting Structural Change. Consistent peptide patterns were found between standard and sample bevacizumab up to day 7 (Table 3) after injection, suggesting gradual modifications became significant 7 days after injection. The changes from day 11 to day 21 may not be attributed to the poor detectability of low level peptides because other peptides markers could be detected. For example, peptides 1070.5 m/z [4H+,4+] and 1116.5 m/z [2H+,2+] (Figure 4a); 826.5 m/z [2H+,2+] (Figure 4b); and 1102.5 m/z [2H+,2+] (Figure 4d) were detectable in day 14 and day 21 samples. Another strong evidence of structural modification was the fading peptides 871.4 m/z [2H+,2+] (Figure 4c) and 949.0 m/z [2H+,2+] (Figure 4d). It indicated certain regions were modified along with time. Furthermore, a subtle multiply charged high mass peptide, 1070.5 m/z [4H+,4+] appeared at day 14 at retention time about 49.0 min (Table 3) implying that the trypsin digestion sites, arginine and lysine, were modified to an indigestible form. Although the patterns were in general consistent up to day 7, some peptide fragment changes were found even on day 7. For example, 888.0 m/z [2H+,2+] at about 50.5 min and 863.5 m/z [2H+,2+] at about 53.0 min, peptides could not be found on day 7 but were detected in standard and samples from earlier days. On the other hand, peptides 1121.0 m/z [2H+,2+] at about 50.5 min appeared at day 7 onward but were not found in standard and samples from earlier days (Table 3). Structural modifications appeared at day 7 onward. In addition, the peptides 593.8 m/z at about 49.8 min, 941.5 m/z at about 50.5 min, and 880.0 m/z at 50.5 min were not detected in samples after day 7. The former two were located in the heavy chain variable regions whereas the latter was in the heavy chain constant regions (Figure 5), consistent with previous findings that the heavy chains underwent modifications no matter whether in the variable or in the constant region. Modification in the variable region suggested the affinity may start to change even at day 7. Significant affinity changes would occur at day 21. Peptide fragment 871.4 m/z at 56.8 min (Table 3) which was also in the heavy chain variable region was still appearing until day 21. Some domains may be more protected from endogenous attacks than the others. This pharmacokinetic data showed the elimination rate of bevacizumab was slow in the eye. High level of bevacizumab, 58.5 μg/mL, was sustained even after 21 days of injection. The present method can detect even lower level or longer postinjection day samples indicating it is useful for pharmacokinetic studies. To investigate if the structural modification could ultimately affect the antiangiogenic efficacy of buvacizumab, we compared the bevacizumab incubated in the eye over a period of time with the freshly spiked bevacizumab at the same concentrations by a quantitative endothelial cell migration assay. The possibility of functional changes of bevacizumab as a result of structural changes leading to affinity changes after days of in vivo incubation in the eye was explored. The antiproliferative effect of bevacizumab present in the day 1 and 7 in vivo vitreous humors showed no significant difference from the same concentration of freshly spiked vitreous samples which were intravitreally injected with saline. It may imply that structural modification was not sufficient to affect the affinity and function. However, a difference of antiproliferative effects between the in vivo day 21 vitreous samples and the corresponding spiked vitreous humor samples was found. It suggests the in vivo structure modification become significant

after a longer period of incubation time so that the modification would eventually affect the affinity and the subsequent antiproliferative (may be antiangiogenic) function. This finding supported the results of structural and affinity changes as determined by LCMS and affinity chromatography. These findings were consistent with a study by Stewart et al.29 that bevacizumab needed to be reinjected every 3 days for 1 month to equal the peak activity after a ranibizumab injection. Byeon et al. found that even when they applied an aqueous depressant, timolol-dorzolamide, to the eye to delay the elimination of bevacizumab, the visual accuracy and central retinal thickness did not improve.31 Structural modification found in this study could explain these phenomena. Many pharmaceutical studies have not reported the effects of structural modification on the binding properties.9,10 This study raised the concern not only on the clearance but also on the structural modification that may eventually affect the efficacy of the drug molecule even though the level can still be maintained. Although this study showed that structural modification of bevacizumab occurred after intravitreal injection, identification of the actual structural change and glycan modification required further investigation.

■

CONCLUSION We have developed a robust and validated method to quantify the therapeutic monoclonal antibody bevacizumab in vitreous humor, and to detect its structural modifications. An analytical strategy was also developed to evaluate in vivo structural and functional changes of biopharmaceutical antibody. It could help to infer the efficacy changes in in vivo environment. This study showed that not only did bevacizumab levels decrease due to clearance but also its structure changed in the variable and constant regions of the heavy chain. Obvious modifications were found 7 days after injection. In addition, affinity change was found 21 days after intravitreal injection. These phenomena were coincident with the declining antiproliferative effect of bevacizumab treated vitreous humor after 21 days of intravitreal injection comparing to the spiked vitreous humor with the same concentrations. It suggests that the efficacy may be affected.

■

ASSOCIATED CONTENT

S Supporting Information *

Tables of data for recovery of bevacizumab in vitreous humor, percentage of initial amount of spiked bevacizumab in vitreous humor (ex vivo) after incubation, and pharmacokinetics of bevacizumab in vitreous humor. Mass spectra, LCMS chromatograms, and relative Ka data. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Department of Ophthalmology and Visual Sciences, The Chinese University of Hong Kong, Hong Kong Eye Hospital, 147K, Argyle Street, Kowloon, Hong Kong. Tel: (852) 39435801. Fax: (852) 26244764. E-mail: [email protected]. Notes

The authors declare no competing financial interest. 3432



■

Article

(15) Kosky, A. A.; Razzaq, U. O.; Treuheit, M. J.; Brems, D. N. The effects of alpha-helix on the stability of Asn residues: deamidation rates in peptides of varying helicity. Protein Sci. 1999, 8 (11), 2519−2523. (16) Jefferis, R. Glycosylation of recombinant IgG antibodies and its relevance for therapeutic applications. In Cell Engineering; Al-Rubeai, M., Ed.; Kluwer Academic Publishers: New York. 2002; pp 93−107. (17) Rehder, D. S.; Dillon, T. M.; Pipes, G. D.; Bondarenko, P. V. Reversed-phase liquid chromatography/mass spectrometry analysis of reduced monoclonal antibodies in pharmaceutics. J. Chromatogr., A 2006, 1102 (1−2), 164−175. (18) Gadgil, H. S.; Pipes, G. D.; Dillon, T. M.; Treuheit, M. J.; Bondarenko, P. V. Improving mass accuracy of high performance liquid chromatography/electrospray ionization time-of-flight mass spectrometry of intact antibodies. J. Am. Soc. Mass Spectrom. 2006, 17 (6), 867−872. (19) Bailey, M. J.; Hooker, A. D.; Adams, C. S.; Zhang, S.; James, D. C. A platform for high-throughput molecular characterization of recombinant monoclonal antibodies. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2005, 826 (1−2), 177−187. (20) Cao, L. M.; Liu, H. M.; Lam, S. C.; Yam, H. F.; Pang, C. P. In Vitro Screening for Angiostatic Potential of Herbal Chemicals. Invest. Ophthalmol. Visual Sci. 2010, 51 (12), 6658−6664. (21) Chen, Y.; Kwon, S. W.; Kim, S. C.; Zhao, Y. Integrated approach for manual evaluation of peptides identified by searching protein sequence databases with tandem mass spectra. J. Proteome Res. 2005, 4 (3), 998−1005. (22) Winzor, D. J. Determination of binding constants by affinity chromatography. J. Chromatogr., A 2004, 1037 (1−2), 351−367. (23) Hage, D. S. Binding constants: determination by affinity chromatography. In Encycyclopedia of chromatography; Taylor & Francis Publish.: New York, 2005; pp 184−190. (24) Bakri, S. J.; Snyder, M. R.; Pulido, J. S.; McCannel, C. A.; Weiss, W. T.; Singh, R. J. Six-month stability of bevacizumab (Avastin) binding to vascular endothelial growth factor after withdrawal into a syringe and refrigeration or freezing. Retina 2006, 26 (5), 519−522. (25) Olivova, P.; Chen, W.; Chakraborty, A. B.; Gebler, J. C. Determination of N-glycosylation sites and site heterogeneity in a monoclonal antibody by electrospray quadrupole ion-mobility time-offlight mass spectrometry. Rapid Commun. Mass Spectrom. 2008, 22 (1), 29−40. (26) Wan, H. Z.; Kaneshiro, S.; Frenz, J.; Cacia, J. Rapid method for monitoring galactosylation levels during recombinant antibody production by electrospray mass spectrometry with selective-ion monitoring. J. Chromatogr., A 2001, 913 (1−2), 437−446. (27) Siemiatkoski, J.; Lyubarskaya, Y.; Houde, D.; Tep, S.; Mhatre, R. A comparison of three techniques for quantitative carbohydrate analysis used in characterization of therapeutic antibodies. Carbohydr. Res. 2006, 341 (3), 410−419. (28) Yu, L.; Remmele, R. L., Jr.; He, B. Identification of N-terminal modification for recombinant monoclonal antibody light chain using partial reduction and quadrupole time-of-flight mass spectrometry. Rapid Commun. Mass Spectrom. 2006, 20 (24), 3674−3680. (29) Calleri, E.; Temporini, C; Massolini, G. Frontal affinity chromatography in characterizing immobilized receptors. J. Pharm. Biomed. Anal. 2011, 54 (5), 911−925. (30) Slon-Usakiewicz, J. J.; Ng, W.; Dai, J. R.; Pasternak, A.; Redden, P. R. Frontal affinity chromatography with MS detection (FAC-MS) in drug discovery. Drug Discovery Today 2005, 10 (6), 409−416. (31) Byeon, S. H.; Kwon, O. W.; Song, J. H.; Kim, S. E.; Park, Y. S. Prolongation of activity of single intravitreal bevacizumab by adjuvant topical aqueous depressant (Timolol-Dorzolamide). Graefes Arch. Clin. Exp. Ophthalmol. 2009, 247 (1), 35−42.

ACKNOWLEDGMENTS The study is partially supported by a Special Equipment Grant 2003/04 from the Chinese University of Hong Kong to Chi Pui Pang and Chi Chiu Wang for the Liquid Chromatography tandem mass spectrometry system.

■

REFERENCES

(1) Hurwitz, H.; Fehrenbacher, L.; Novotny, W.; Cartwright, T.; Hainsworth, J.; Heim, W.; Berlin, J.; Baron, A.; Griffing, S.; Holmgren, E.; Ferrara, N.; Fyfe, G.; Rogers, B.; Ross, R.; Kabbinavar, F. Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N. Engl. J Med. 2004, 350 (23), 2335− 2342. (2) Michels, S.; Rosenfeld, P. J.; Puliafito, C. A.; Marcus, E. N.; Venkatraman, A. S. Systemic bevacizumab (Avastin) therapy for neovascular age-related macular degeneration twelve-week results of an uncontrolled open-label clinical study. Ophthalmology 2005, 112 (6), 1035−1047. (3) Ladewig, M. S.; Ziemssen, F.; Jaissle, G; Helb, H. M.; Scholl, H. P.; Eter, N.; Bartz-Schmidt, K. U.; Holz, F. G. Intravitreal bevacizumab for neovascular age-related macular degeneration. Ophthalmologe 2006, 103 (6), 463−470. (4) Haritoglou, C.; Kook, D.; Neubauer, A.; Wolf, A.; Priglinger, S.; Strauss, R.; Gandorfer, A.; Ulbig, M.; Kampik, A. Intravitreal bevacizumab (Avastin) therapy for persistent diffuse diabetic macular edema. Retina 2006, 26 (9), 999−1005. (5) Scott, I. U.; Edwards, A. R.; Beck, R. W.; Bressler, N. M.; Chan, C. K.; Elman, M. J.; Friedman, S. M.; Greven, C. M.; Maturi, R. K.; Pieramici, D. J.; Shami, M.; Singerman, L. J.; Stockdale, C. R. A phase II randomized clinical trial of intravitreal bevacizumab for diabetic macular edema. Ophthalmology 2007, 114 (10), 1860−1867. (6) Jaissle, G. B.; Ziemssen, F.; Petermeier, K.; Szurman, P.; Ladewig, M.; Gelisken, F.; Völker, M.; Holz, F. G.; Bartz-Schmidt, K. U. Bevacizumab for treatment of macular edema secondary to retinal vein occlusion. Ophthalmologe 2006, 103 (6), 471−475. (7) Priglinger, S. G.; Wolf, A. H.; Kreutzer, T. C.; Kook, D.; Hofer, A.; Strauss, R. W.; Alge, C. S.; Kunze, C.; Haritoglou, C.; Kampik, A. Intravitreal bevacizumab injections for treatment of central retinal vein occlusion: six-month results of a prospective trial. Retina 2007, 27 (8), 1004−1012. (8) Krohne, T. U.; Eter, N.; Holz, F. G.; Meyer, C. H. Intraocular pharmacokinetics of bevacizumab after a single intravitreal injection in humans. Am. J. Ophthalmol. 2008, 146 (4), 508−512. (9) Beer, P. M.; Wong, S. J.; Hammad, A. M.; Falk, N. S.; O’Malley, M. R.; Khan, S. Vitreous levels of unbound bevacizumab and unbound vascular endothelial growth factor in two patients. Retina 2006, 26 (8), 871−876. (10) Bakri, S. J.; Snyder, M. R.; Reid, J. M.; Pulido, J. S.; Singh, R. J. Pharmacokinetics of intravitreal bevacizumab (Avastin). Ophthalmology 2007, 114 (5), 855−859. (11) Harris, R. J. Processing of C terminal lysine and arginine of proteins isolated from mammalian cell culture. J. Chromatogr., A 1995, 705 (1), 129−134. (12) Saba, J. A.; Kunkel, J. P.; Jan, D. C.; Ens, W. E.; Standing, K. G.; Butler, M.; Jamieson, J. C.; Perreault, H. A study of immunoglobulin G glycosylation in monoclonal and polyclonal species by electrospray and matrix-assisted laser desorption/ionization mass spectrometry. Anal. Biochem. 2002, 305 (1), 16−31. (13) Bongers, J.; Cummings, J. J.; Ebert, M. B.; Federici, M,M; Gledhill, L.; Gulati, D.; Hilliard, G. M.; Jones, B. H.; Lee, K. R.; Mozdzanowski, J.; Naimoli, M.; Burman, S. Validation of a peptide mapping method for a therapeutic monoclonal antibody: what could we possibly learn about a method we have run 100 times? J. Pharm. Biomed. Anal. 2000, 21 (6), 1099−1128. (14) Huang, L.; Lu, J.; Wroblewski, V. J.; Beals, J. M.; Riggin, R. M. In vivo deamidation characterization of monoclonal antibody by LC/ MS/MS. Anal. Chem. 2005, 77 (5), 1432−1439. 3433


Quantification and Structure Elucidation of in Vivo Bevacizumab

Recommend Documents