Bioconjugate Chem. 2001, 12, 861−869
861
Physicochemical Characterization of Poly(ethylene glycol)-Modified Anti-GAD Antibodies Robert S. Larson, Virginie Menard, Harvey Jacobs, and Sung Wan Kim* Center for Controlled Chemical Delivery, Department of Pharmaceutics and Pharmaceutical Chemistry, University of Utah, 30 South 2000 East Room 201, Salt Lake City, Utah 84112. Received December 2, 2000; Revised Manuscript Received June 16, 2001
Monoclonal antibodies against glutamic acid decarboxylase (anti-GAD) were modified with poly(ethylene glycol) (PEG), and the resulting conjugates were characterized. Monoclonal anti-GAD antibodies were purified from ATCC HB184 hybridoma cells by either cell culture supernatant or ascites fluid from BALB/c mice. Polyclonal rabbit IgG antibodies were also used as a model protein. Polyclonal rabbit IgG or purified anti-GAD was modified by PEG (MW ) 5000 or 20000 Da) through either the lysine residues or through the carbohydrate moiety. Lysine modification was performed in PBS (pH 7.4) or 0.1 M borate (pH 9.2) by adding a molar excess (5-80) of a succinimidyl activated propionic acid terminated mPEG (SPA-PEG) while stirring at room temperature. Carbohydrate modifications were performed in PBS (pH 6.2) by first oxidizing the antibody with sodium periodate followed by incubation with hydrazide-terminated PEG followed by reduction with sodium cyanoborohydride. The degree of modification was assessed by 1H NMR or TNBS (trinitrobenzenesulfonic acid). Circular dichroism (CD) spectra were obtained for lysine-modified rabbit IgG at various degrees of modification ranging from 5 to 60 PEG per antibody. Binding was assessed using an ELISA method with GAD or rabbit anti-mouse-IgG (H+L) coated plates. The TNBS and 1H NMR analysis of the modified antibody showed reasonably similar results from 5 to 60 PEG per antibody. The 1H NMR method showed greater sensitivity at low modifications (below 20:1) and was fairly linear up to about 60 PEG per antibody. The CD spectra of the polyclonal rabbit IgG showed only small differences at variously modified antibody. The binding affinity of anti-GAD is lower for all PEG modifications with respect to unmodified anti-GAD. Modifications at pH 7.4 show lower binding to GAD than modifications at pH 9.2. Binding to GAD or anti-mouse-IgG is decreased as the degree of modification is increased. Lysine modifications showed lower binding to GAD or anti-mouse-IgG than carbohydrate modifications. Binding to GAD or anti-mouse-IgG is lower for PEG20000-modified anti-GAD with respect to PEG5000modified anti-GAD.
INTRODUCTION
Type I diabetes, or insulin-dependent diabetes mellitus (IDDM), is an autoimmune disorder in which the patient’s immune system selectively attacks and destroys the insulin-producing β-cells of the pancreas (1). Type I diabetes is an irreversible condition that usually occurs before adulthood in a patient’s life. The autoimmune process in Type I diabetes is characterized by macrophage infiltration into the pancreas followed by T-cell and macrophage activation which then leads to β-cell destruction and loss of insulin production (1-4). Once the β-cells are completely destroyed, no further insulin production is observed, and exogenous insulin must be administered to maintain normal glucose levels. There is a strong genetic component to Type I diabetes which seems to be linked to certain major histocompatibility complex (MHC or HLA) genes, and individuals with this tissue type show an increased susceptibility toward developing Type I diabetes (5-8). The MHC molecules present on antigen-presenting cells (APCs), such as macrophages, phagocytes, and B-lymphocytes, serve as sites of antigen presentation to T-cells of the immune system. When APCs phagocytize cellular debris, * To whom correspondence should be addressed. Phone: (801) 581-6801. Fax: (801) 581-7848. E-mail: Rburns@ deans.pharm.utah.edu.
they breakdown the ingested proteins into small peptides which are then presented on the cell surface in association with these MHC molecules. Here, they can interact with the T-cells which will either recognize the complex as being native or foreign. In the case of Type I diabetes, T-cells recognize some of the peptide sequences from β-cell proteins as being foreign which cause the T-cells to become activated. This eventually leads to β-cell destruction and the onset of clinical diabetes. There have been many proteins studied which are suspected to be important in the initiation and progression of the disease (9-12). Among the most intensely studied autoantigens is the protein glutamic acid decarboxylase (GAD) which can be found at appreciable levels in the β-cells of the pancreas. It is thought that GAD is one of the most important autoantigens responsible for Type I diabetes (13-15). In fact, studies using the NOD (nonobese diabetic) mouse model show that diabetes can be eliminated if GAD production in the β-cells is eliminated (16). It has been shown that antibodies bound to proteins can affect the resulting peptides produced by enzymatic cleavage (17-19). In this manner, it would seem possible to alter the specific GAD peptides presented to T-cells by using antibodies against GAD (anti-GAD). By binding to the GAD, the anti-GAD antibodies will alter how the APCs process GAD which may lead to a difference in the
10.1021/bc000137e CCC: $20.00 © 2001 American Chemical Society Published on Web 11/02/2001
862 Bioconjugate Chem., Vol. 12, No. 6, 2001
Larson et al.
Figure 1. Schematics of the modification reactions of antibodies by PEG through (A) lysine and (B) carbohydrate.
specific peptides presented by these cells and thereby alter the immune response to GAD. In our laboratory, we have studied the effect of the administration of monoclonal anti-GAD antibodies in the progression of Type I diabetes in the NOD mouse model (20). We have found that weekly injections of anti-GAD antibodies can reduce the incidence of diabetes from greater than 80% in untreated NOD mice to less than 20% in treated mice. PEG modification of proteins has been shown to improve a number of properties of proteins. Typically, PEG-modified proteins will show an increased resistance to enzymatic attack (21-23), a reduced immunogenic response (24-26), an increase in plasma half-life (2528), and an increase in solution stability (29-31). Of particular interest is the alteration of the enzymatic degradation of PEG-modified proteins. Thus, it is thought that PEG modification of an antibody against a protein may have a greater effect at altering the resulting enzymatic peptide sequences than the antibody alone which may further decrease the incidence of diabetes. In this report, we describe the synthesis and characterization of PEG-modified, monoclonal anti-GAD. We studied the effect of the degree of modification (number of PEG chains per antibody), the site of modification (lysine residues versus carbohydrate moiety), reaction conditions (low pH versus high pH for lysine modification), and PEG molecular weight on the antibody structure, the binding affinity toward GAD, and the ability of other antibodies to recognize the anti-GAD-PEG conjugates. EXPERIMENTAL PROCEDURES
Antibody Production. Ascites fluid was produced by injecting hybridoma cells (HB184, ATCC) intraperitoneally into pristane-treated BALB/c mice. The monoclonal antibody, anti-GAD, was purified from ascites fluid by ammonium sulfate precipitation followed by diethylaminoethyl (DEAE) cellulose chromatography. Purity was gauged by native PAGE. Polyclonal rabbit IgG was obtained from Sigma and used without further purification. Antibody Modification. The reaction scheme for the antibody modifications is shown in Figure 1. The purified antibody was exchanged into the appropriate buffer for
each modification by ultrafiltration (50 kDa MWCO). Lysine modifications were done in either a phosphate buffer at pH 7.4 or in a borate buffer at or 9.2. Antibody concentrations were 3 mg/mL. Succinimidyl propionate activated PEG (SPA-PEG) (MW ) 5 or 20 kDa) (Shearwater Polymers), see Figure 1, was dissolved in DMSO and added to the anti-GAD buffered solutions in a dropwise manner while stirring. The solutions were stirred at room temperature for 1 h and incubated at 4 °C overnight. As a model, polyclonal rabbit IgG was modified with molar excesses SPA-PEG (MW ) 5000 Da) of 5, 10, 20, 30, 40, 60, or 80. Carbohydrate modifications (32-33) were done in a phosphate buffer at pH 6.2. The carbohydrate was first oxidized by addition of sodium periodate (NaIO4) in a 1000 molar excess with respect to anti-GAD concentration. After stirring 1 h at 4 °C, the oxidation was quenched with ethylene glycol, and the reaction byproducts were removed by ultrafiltration (50 or 100 kDa MWCO). A 50 molar excess of hydrazide-PEG (Shearwater Polymers) was added to the oxidized antibody at pH 6.2. After stirring at room temperature for 1 h, an excess of sodium cyanoborohydride (NaBH3CN) was added and stirred at room temperature for another hour. The modified antibodies were then purified by ultrafiltration (50 or 100 kDa MWCO) against PBS, pH 7.4. 1H NMR Analysis. The modified antibody was exchanged into phosphate buffered D20 by ultrafiltration (50 kDa MWCO). The molar amount of antibody in 400 µL was calculated by UV absorbance at 280 nm based on an extinction coefficient of 1.38 mL/mg/cm and an antibody molecular weight of 150000 Da. Antibody concentrations were between 3 and 5 mg/mL. Stock solutions of acetone, DMSO, and sodium benzoate were made in D2O at known concentrations. Known amounts of the standards were added to the solutions, and the total volume is brought up to 700 µL with D2O. The amount of each standard added to each sample is based on the estimated number of PEG chains attached to the antibody such that the integrations of the PEG peak and the reference peaks will be close. For example, if the final molar concentrations of acetone and DMSO were 76 times larger than the estimated molar concentration of PEG5000 in the sample, the integrated peaks should be close to the same. Spectra were gathered on a Bruker 200 MHz NMR. The acetone (2.06 ppm), DMSO (2.56
Modification Anti-GAD with Poly(ethylene glycol)
ppm), PEG (3.54 ppm) and aromatic benzoate (7.31-7.72 ppm) peaks were integrated. The 1H NMR spectrum was taken, and the integration of the methylene (-CH2-) hydrogen peak of the PEG was compared to the acetone and DMSO peaks to calculate the PEG concentration. The number of PEGs per anti-GAD is determined by dividing the concentration of PEG by the concentration of anti-GAD. TNBS Assay. For TNBS analysis (34-35), the modified antibody solutions were diluted to about 3 mg/mL in 600 µL in 0.1 M borate buffer pH 9.2. To keep the antibody soluble throughout the reaction, DMSO was added to 33%. Absorbance blanks consisting of buffer and DMSO were also made. To each solution was added 8 µL of a 1 M TNBS solution. The solutions were mixed well and incubated at room temperature for 6 h. To stop the reaction, 4 µL of 12 N HCl was added to each solution. Dilutions of the samples were made by measuring between 30 and 120 µL of the sample and the bringing this volume to 120 µL by the addition of the blank solution. This was done to keep the background constant in all the dilutions. Then 1080 µL of 0.1 N HCl was added to each solution. The absorbance blank was made by adding 120 µL of the blank solution to 1080 µL of 0.1 N HCl. The absorbance at 345 nm was then taken for each sample. CD Spectroscopy. Lysine-conjugated (PEG5000) rabbit IgG was used at conjugation ratios ranging from 0 to 56 PEG per IgG. The antibody concentration of each was brought to 0.5 mg/mL in PBS (pH 7.4), as determined spectrophotometrically, 280 (mL/mg/cm) ) 1.38. The samples were degassed under vacuum pressure for 20 min prior to the CD measurements. The molar ellipticity was read between 200 and 250 nm using a 0.1 cm path length cell. Binding Affinity. The binding of the modified antiGAD to GAD as well as an anti-mouse-IgG antibody was assessed by ELISA. Glutamic acid decarboxylase (GAD) from E-coli (Sigma, EC 4.1.1.15) or rabbit-anti-mouseIgG (H+L) (Zymed) was used to coat 96-well ELISA plates by incubating each well with a 2 µg/mL GAD or 5 µg/mL anti-mouse-IgG solution at pH 9.5. After 2 h, the solution was removed, and the plates were incubated with a 1% BSA in PBS, pH 7.4 solution for 1 h. Dilutions of the anti-GAD (modified or unmodified) were made with PBS, 1% BSA, pH 7.4, and incubated in the plates for 2 h. The plates were then incubated with a 1:100 dilution of alkaline phophatase-coupled anti-mouse-IgG antibody (Zymed) for 2 h. The plates were then washed four times with 125 µL/well of PBS, 0.05% Tween20. A solution of p-nitrophenyl phosphate (Pierce) in a diethanolamine buffer, pH 9.8, was added. After sufficient color developed, the absorbance at 405 nm of each well was measured by an automated ELISA plate reader. RESULTS
The degree of modification was assessed by a TNBS method. The number of lysines was previously determined by amino acid analysis to be approximately 86, data not shown, which yields a total number of free amines about 90 due to the four N-terminal amines per antibody. The absorbances at 345 nm of the TNBS samples were normalized to the unmodified sample. Thus, the approximate number of free amines per antibody in each sample is 90 multiplied by the normalized absorbance at 345 nm. Thus, the number of PEG chains conjugated per antibody is the difference between 90 and the number of free amines.
Bioconjugate Chem., Vol. 12, No. 6, 2001 863
Figure 2. 1H NMR spectrum of a PEG5000-modified antibody with acetone, DMSO, and sodium benzoate standards.
For 1H NMR analysis, it is important that all unreacted PEG be removed before analysis. Ultrafiltration was used in this study to remove the unreacted PEG. The molecular weight cutoff of the ultrafiltration membrane used was 50 kDa. Permeability of the PEG (MW ) 5000) through the membrane was confirmed by analyzing the ultrafiltrate for PEG, data not shown. To quantitate the PEG peak of an 1H NMR spectrum, it is necessary to reference peaks of known concentration. Integration standards (acetone, DMSO, and sodium benzoate) were added to the PEG-antibody conjugate solution at known concentrations. In this manner, the integration of the PEG peak can be used to calculate its concentration based on the integrations of the reference peaks. Figure 2 shows a typical 1H NMR spectrum of a solution of a PEG-modified antibody with the three reference standards. This clearly shows that PEG (3.54 ppm) and the three reference standards of acetone (2.06 ppm), DMSO (2.56 ppm), and sodium benzoate (7.277.73 ppm) are easily resolved from each other. The integration of an 1H NMR peak is proportional to the number of resonant hydrogens at the particular chemical shift. Each peak can be defined as the product of the concentration and the number of resonant hydrogens. Thus, the concentration of PEG can be easily solved since the concentrations of the standards and the antibody are known. When the calculated PEG ratios are plotted versus the PEG reaction ratio, as shown in Figure 3, the plot reveals a linear relationship between the measured PEG conjugation ratio with respect to the activated PEG reaction ratio for the NMR calculations. When a linear regression is performed, it shows a slope of 0.73 with an R2 of 0.996 when the fit is forced through zero. This indicates that approximately 73% of the activated PEG added to the reaction conjugated to the antibody under the reaction conditions used. For comparison, the TNBS analysis results are plotted with the NMR analysis in Figure 3. In contrast to the NMR analysis, the TNBS analysis shows a nonlinear trend of PEG conjugation. The results of the two techniques match only at two points, 40 and 60 PEG reaction ratio. Since the TNBS method relies on an indirect method of subtraction, it seems to have difficulty detecting conjugation at lower reactions ratios (