Analysis of Recombinant Human Platelet-Derived Growth Factor by

An Tran*, Hugh Parker, Viktorya Levi, and Michael Kunitani. Chiron Corporation, 4560 Horton Street, Emeryville, California 94608. Anal. Chem. , 1998, ...
1 downloads 0 Views 178KB Size
Download PDF
Anal. Chem. 1998, 70, 3809-3817

Analysis of Recombinant Human Platelet-Derived Growth Factor by Reversed-Charge Capillary Zone Electrophoresis An Tran,* Hugh Parker, Viktorya Levi, and Michael Kunitani

Chiron Corporation, 4560 Horton Street, Emeryville, California 94608

Reversed-charge capillary zone electrophoresis (RC-CZE) has been developed as a clipping (proteolysis) assay for homodimeric protein recombinant human platelet-derived growth factor (rhPDGF-BB), a major serum mitogenic factor involved in subcutaneous wound healing. When expressed in yeast, the protein is excreted as a fully folded homodimeric protein consisting of two antiparallel B chains held together by two interchain disulfide bonds. During fermentation, internal proteolysis (clipping between residues Arg32 and Thr33) and C-terminal truncation (Arg32 and Thr109) may occur. Internal proteolysis yields three potential forms of rhPDGF-BB: intact (both B chains are intact), single-clipped (one B chain is clipped), and double-clipped (both B chains are clipped). Clipping also creates new C-terminal sites for further C-terminal truncations and leads to a very complex mixture of isoforms. Routine baseline resolution of these three forms by various modes of HPLC proved unsuccessful. When the disulfide bonds of antiparallel chains are reduced, the complex peptide mixture can be analyzed by RP-HPLC; however, only the level of total clipping is identified. Since RC-CZE separation relies upon differences in molecular charge/size ratio, it can resolve the three rhPDGF-BB forms differing in the additional exposed residues. The choice of reversed-charge CZE columns (amine-coated column) allows proteins of high pI such as rhPDGF-BB (pI > 10) to be readily analyzed while minimizing protein loss from column adsorption. To simplify the electropherogram of clipped forms, the sample is treated first with carboxypeptidase B to reduce the charge microheterogeneity of partial Arg32 truncation. Analysis of rhPDGF-BB by RC-CZE yields a baseline separation between the three forms, intact and single- and double-clipped rhPDGF-BB. Platelet-derived growth factor (PDGF) is a mitogenic factor in serum that promotes the proliferation of mesenchymal-derived cells such as fibroblasts, glial cells, and smooth muscle cells in vitro.1 PDGF also induces several biological effects on cells such as cell activation, chemotaxis, and cell proliferation by binding to (1) Ross, R.; Raines, E. W.; Bowen-Pope, D. F. Cell 1986, 46, 155-169. S0003-2700(98)00329-1 CCC: $15.00 Published on Web 08/15/1998

© 1998 American Chemical Society

Figure 1. Amino acid sequence of rhPDGF-B monomer.

specific high-affinity receptors on the cell surface.2-5 The protein consists of two homologous polypeptide chains, A and B, linked by disulfide bonds and has an apparent molecular mass of ∼30 kDa. The existence of homologous polypeptide chains gives rise to the formation of both homodimeric (PDGF-AA, PDGF-BB) and heterodimeric (PDGF-AB) forms. All three dimeric forms of PDGF exist in nature.6 Saccharomyces cerevisae-derived recombinant human PDGFBB is a homodimer of two polypeptide B chains. Each B chain contains 109 amino acid residues with three intrachain and two interchain disulfide linkages. The amino acid sequence of monomeric rhPDGF-B is shown in Figure 1. The crystal structure of S. cerevisae-derived rhPDGF-BB, determined by X-ray analysis,7 has shown that the polypeptide B chain is folded into two highly twisted antiparallel pairs of β-strands and contains an unusual knotted arrangement of three intramolecular disulfide bonds, Cys16-Cys60, Cys49-Cys97, and Cys53-Cys99. The two remaining cysteine residues of the B chain, Cys43 and Cys52, are involved in interchain disulfide linkages. Cys43 of one chain forms a disulfide bridge with Cys52 of the second chain and vice versa, giving rise to the antiparallel dimer arrangement of the elongated monomers. The chemically determined disulfide linkages of Escherichia coli(2) Heldin, C. H.; Backstrom, G.; Ostman, A.; Hammacher, A.; Ronnstrand, L.; Rubin, K.; Nister, M.; Westermark, B. EMBO J. 1988, 7, 1387-1393. (3) Kundra, V.; Escobedo, J. A.; Kazlauskas, A, Kim, H. K.; Rhee, S. G., Williams, L. T.; Zetter, B. R. Nature 1994, 367, 474-476. (4) Bornfeldt, K. E.; Raines, E. W.; Graves, L. M.; Skinner, M. P.; Krebs, E. G.; Ross, R. Ann. N.Y. Acad. Sci. 1995, 766, 416-430. (5) Claesson-Welsh, L. Int. J. Biochem. Cell Biol. 1996, 28, 373-385. (6) Hart, C. E.; Bailey, M.; Curtis, D. A.; Osborn, S.; Raines, E.; Ross, R.; Forstrom, J. W. Biochemistry 1990, 29, 166-172. (7) Oefner, C.; D’Arcy, A.; Winkler, F. K.; Eggimann, B.; Hosang, M. EMBO J. 1992, 11, 3921-3926.

Analytical Chemistry, Vol. 70, No. 18, September 15, 1998 3809

derived rhPDGF-BB reported by Haniu et al.8 agree with the assignments from X-ray analysis. S. cerevisae-produced rhPDGF-BB is a complex molecule with considerable structural heterogeneity. This heterogeneity is derived from posttranslational modifications such as glycosylation, methionine oxidation, and proteolytic processing (clipping, truncation). Each B chain of rhPDGF-BB contains a total of nine threonine residues and a low level of mannose is randomly O-linked to six of the nine possible threonine sites on each monomer. The rhPDGF-B sequence also contains one methionine residue at position 12 which can oxidize during fermentation to the corresponding methionine sulfoxide. The structure is further complicated since rhPDGF-BB may contain internal clipping between Arg32 and Thr33 and C-terminal truncation at Arg32 and/or Thr109. Cleavage between residues Arg32-Thr33 can lead to a mixture of three rhPDGF-BB forms: intact (IN-rhPDGFBB), single-clipped (SC-rhPDGF-BB, with one B chain clipped), and double-clipped (DC-rhPDGF-BB, with both B chains clipped) rhPDGF-BB. Each form may be further modified by the Cterminal truncation of Arg32 or Thr109, producing a complex mixture of isoforms. For IN-rhPDGF-BB, truncation may occur only at residue 109 leading to three different isoforms, as illustrated in panel A of Figure 2. For SC-rhPDGF-BB, truncation at both residues 32 and 109 can produce 8 different isoforms (panel B, Figure 2), and for DC-rhPDGF-BB, up to 10 isoforms (panel C, Figure 2). In addition, rhPDGF-BB contains a number of proline residues of which the Pro-Pro sequence at residues 41 and 42 is a probable site for cis-trans isomerization. Peptides containing proline residues are known to exist in both the cis and trans conformations due to the rotationally hindered peptidyl-proline bond, and these conformers have been shown to be relatively stable at subambient temperatures.9,10 The relatively slow cis-trans interconversion at low temperatures permits the separation of the cis and trans conformers of peptides containing one or more prolyl peptide bonds by adsorptive chromatography, e.g., reversed-phase chromatography.9,11-14 In rhPDGF-BB, the proline residues are located in a hydrophobic region of the primary structure and are adjacent to a tryptophan residue (Trp40) and a cysteine residue (Cys43) involved in the intermolecular disulfide linkage. Because of the rigidity of the neighboring interchain cystine disulfide, the hindered rotation of the Pro-Pro side chains themselves and the steric hindrance caused by the Trp40 side chain, one might expect the equilibration between the cis and trans conformations to be quite slow and the conformers to be chromatographically resolvable. Indeed, multiple peak formation was observed during RPHPLC analysis of E. coli-derived rhPDGF-BB15 which resulted in the appearance of four distinguishable peaks. Rechromatography (8) Haniu, M.; Rohde, M. F.; Kenney, W. C. Biochemistry 1993, 32, 24312437. (9) Jacobson, J.; Malender, W.; Vaisnys, G.; Horvath, C. J. Phys. Chem. 1984, 88, 4536-4542. (10) Lin, L.; Brandts, J. F. Biochemistry 1983, 22, 553-559. (11) Henderson, D. E.; Horvath, C. J. Chromatogr. 1986, 368, 203-213. (12) Kalman, A.; Thunecke, F.; Schmidt, R.; Schiller, P. W.; Horvath, C. J. Chromatogr., A 1996, 729, 155-171. (13) Thunecke, F.; Kalman, A.; Kalman, F.; Ma, S.; Rathore, A. S.; Horvath, C. J. Chromatogr., A 1996, 744, 259-272. (14) Schmidt, R.; Kalman, A.; Chung, N. N.; Lemieux, C.; Horvath, C.; Schiller, P. W. Int. J. Pept. Protein Res. 1995, 46, 47-55. (15) Watson, E.; Kenney, W. C. J. Chromatogr. 1992, 606, 165-170.

3810 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

Figure 2. Schematic representation of rhPDGF-BB where paired lines represent homodimeric rhPDGF-BB: (A) isoforms of intact rhPDGF-BB where b represents Thr109 truncation; (B) isoforms of single-clipped rhPDGF-BB where clipped B chains are represented by a break in the line and 9 represents Arg32 truncation; (C) isoforms of double-clipped rhPDGF-BB.

of each isolated peak at room temperature yielded a ratio of the four peaks comparable to that of the starting material. Increasing the temperature above 60 °C resulted in a single peak that, when isolated and rechromatographed at ambient temperature, produced the four-peak chromatogram comparable to rhPDGF-BB that had not been exposed to elevated temperature. These results indicated that the four peaks were in a reversible equilibrium, and the authors suggested that resolution of the Pro-Pro isomers was the likely explanation. Several studies have demonstrated that the chromatographic behavior of a three-dimensional protein is controlled by the amino acid residues located in a chromatographic contact area on the surface of that protein, and any changes in molecular structure outside that contact region may be undetected by adsorptive chromatography.16 The antiparallel homodimeric structure of rhPDGF-BB poses an unusual tertiary structural problem as it is possible that only one of the two monomers can interact with the chromatographic surface at any given time. Although a structural (16) Kunitani, M.; Johnson, D.; Snyder, L. J. Chromatogr. 1986, 371, 313-333.

feature such as Arg32 clipping in rhPDGF-B monomer would be readily resolvable by anion exchange or reversed-phase HPLC, it is possible that the HPLC column surface could not simultaneously resolve clipping to diametrically opposing faces of the quaternary structure of the homodimeric rhPDGF-BB protein. As viewed by the HPLC column surface, the single-clipped rhPDGF-BB dimer may appear clipped half of the time and appear unclipped half of the time, resulting in only partial resolution of the three clipped forms. If the rhPDGF-BB homodimer could desorb from the column surface, rotate in-solution, and readsorb with the opposite rhPDGF-B monomer a large number of times, some statistical resolution of intact and clipped forms could be achieved. The use of an unusually high column temperature (85 °C) and nearly isocratic elution conditions (step elution) was our best attempt to maximize the HPLC separation of clipped forms but yielded only partial resolution. Unlike HPLC, electrophoresis is a “in-solution” process in which the entire protein is probed for differences in both charge and hydrodynamic size. For charge-based separation of proteins, electrophoretic methods such as gel isoelectric focusing (gel IEF), capillary isoelectric focusing (cIEF), and capillary zone electrophoresis (CZE) are commonly used. The extremely high pI (10.5) of the rhPDGF-BB protein precludes the use of conventional IEF analysis, since the pI range of commercially available ampholytes is 4-9.5. The adsorption of positively charged proteins onto the walls of uncoated fused-silica capillaries presents a major problem in the analysis of basic proteins by CZE. Although the use of neutral coated capillaries17-19 with masked or deactivated surface silanols reduces both electrostatic adsorption and electroosmotic flow (EOF), the residual silanols can still diminish the recovery of basic proteins. More recently, polycationic buffer additives and cationic covalent coatings have been shown to be specially effective in CZE analysis of proteins with high values of pI since at pH < pI both the capillary surface and the protein carry the same charge, minimizing protein loss from electrostatic column adsorption.20-25 Whereas IEF is an equilibrium separation, RCCZE is a relatively rapid, kinetic separation, and since proteins are not focused at their pI’s, precipitation is not a problem. The goal of this study was to develop a routine analysis for quantitation of intact, double-clipped, and single-clipped forms in yeast-derived rhPDGF-BB which would ignore other forms of heterogeneity (C-terminal truncation, methionine oxidation, Olinked mannosylation, and Pro-Pro isomerization). EXPERIMENTAL SECTION Materials. rhPDGF-BB samples were obtained from Chiron (Emeryville, CA). Dithiothreitol (DTT, reagent grade) and 8 M guanidine hydrochloride (Gu-HCl, Sequanal grade) were pur(17) Dougherty, A. M.; Cooke, N.; Shieh, P. In Handbook of Capillary Electrophoresis, 2nd ed.; Landers, J. P., Ed.; CRC Press: Boca Raton, FL, 1997; Chapter 24. (18) Hjerten, S. J. Chromatogr. 1985, 347, 191-198. (19) Jorgenson, J. W.; Lukacs, K. D. Science 1983, 222, 266-272. (20) Wiktorowicz, J. E.; Colburn, J. C. Electrophoresis 1990, 11, 769-773. (21) Yao, Y. J.; Khoo, K. S.; Chung, M. C. M.; Li, S. F. Y. J Chromatogr., A 1994, 680, 431-435. (22) Cordova, E.; Gao, J.; Whitesides, G. M. Anal. Chem. 1997, 69, 1370-1379. (23) Towns, J. K.; Regnier, F. E. J. Chromatogr. 1990, 516, 69-78. (24) Erim, F. B.; Cifuentes, A.; Poppe, H.; Kraak, J. C. J Chromatogr., A 1995, 708, 356-361. (25) Richards, M. P.; Beattie, J. H. J. Chromatogr., B 1995, 669, 27-37.

chased from Pierce Chemicals (Rockford, IL) while sodium bicarbonate (reagent grade) was purchased from Fluka (Ronkonkoma, MI). HPLC grade trifluoroacetic acid (TFA) was purchased from Pierce Chemicals (Rockford, IL) and HPLC grade acetonitrile from Burdick and Jackson (Muskegon, MI). Type I water was obtained from an Milli-Q Plus system (Millipore, Bedford, MA). Reversed-Phase HPLC Instrumentation and Methods. Chromatographic system 1, used for RP-HPLC analysis of nonreduced rhPDGF-BB at various temperatures, consisted of a Spectra Physics model 8800 solvent delivery pump (Thermo Separations, San Jose, CA) equipped with a WISP 712 autoinjector (Waters, Milford, MA) and a PE LC 135 diode array detector (Perkin-Elmer, Norwalk, CT). Separations were performed on a Delta-Pak C4 column (3.9 mm i.d. × 15 cm, 5-µm particle size, 300-Å pore size) (Waters) at a 1.2 mL/min flow rate using a 5-60% acetonitrile/0.1% TFA gradient over 40 min. Column temperature was controlled by the use of a glycol cooling bath (Nesslab, Portsmouth, NH) or a heating block (Jones Chromatography, Hengoed Wales, U.K.). The collected peak fractions were concentrated to dryness, reconstituted and then reanalyzed at 5 °C at a 1.0 mL/min flow rate using a gradient of 10-40% acetonitrile in 0.1% TFA over 40 min. Detection was performed at 214 nm. Chromatographic system 2, used for the high-temperature stepgradient RP-HPLC isolation of nonreduced intact and clipped forms of rhPDGF-BB, consisted of a Hewlett-Packard 1050 liquid chromatographic system (Palo Alto, CA). Separations were performed at a 3.0 mL/min flow rate using a semipreparative Zorbax 300SB-C18 column (9.4 mm × 25 cm) (Mac-Mod Analytical, Chadds Ford, PA) thermostated at 85 °C. The injection volume was 100 µL (10 mg/mL rhPDGF-BB), and detection was performed at 214 nm. Solvent A consisted of 0.1% TFA in water, and solvent B consisted of 0.09% TFA in acetonitrile. Isolation of the nonreduced intact and clipped forms of rhPDGF-BB was performed using a shallow acetonitrile/TFA step gradient over 30 min. The gradient consisted of the following steps: 25-29% B over 10 min, 29% B for 5 min, 29-30%B over 1 min, 30% B for 3 min, 30-33%B over 1 min, 33% B for 1 min, 33-50%B over 9 min. Chromatographic system 3, used for RP-HPLC analysis of reduced rhPDGF-BB (rhPDGF-B), consisted of a Waters 626 solvent delivery pump (Waters) equipped with a Waters model 717 autoinjector (Waters) and an ABI 757 HPLC detector (PerkinElmer). Sample reduction mixtures were prepared by mixing 20 µL of rhPDGF-BB (10 mg/mL) with 100 µL of 7.5 M Gu-HCl/100 mM Tris-HCl (pH 8.8), 70 µL of water, and 10 µL of 2 M DTT (prepared in 50 mM sodium bicarbonate, pH 9.0) in a 1.5-mL polypropylene vial, heated at 50°C for 5 min, and a 50-µL aliquot (50 µg of reduced protein) was subsequently analyzed by RPHPLC. Separations were performed on a Vydac C18 column (4.6 mm × 25 cm, Catalog No. 218TP54) fitted with a Vydac C18 guard cartridge column (4 mm × 5 cm, Catalog No. 218GCC54) (The Separations group, Hesperia, CA). Column temperature was maintained at 35 °C by use of a column heater (Jones Chromatography). Analyses were performed at a 1.2 mL/min flow rate using a linear gradient from 24 to 38% solvent B over 40 min. Solvent A consisted of 0.06% TFA in water and solvent B consisted of 0.055% TFA in acetonitrile. The injection volume was 50 µL, and detection was performed at 214 nm. Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3811

Figure 3. Rechromatography of homodimeric rhPDGF-BB fractions collected from RP-HPLC at 5 °C. HPLC conditions as described in the Experimental Section, chromatographic system 1: (A) rhPDGFBB, starting material; (B) fraction 4; (C) fraction 6; (D) fraction 8.

Figure 5. RP-HPLC of reduced rhPDGF-BB (rhPDGF-B). HPLC conditions as described in the Experimental Section, chromatographic system 3: (A) rhPDGF-BB, reduced; (B) apex fraction from peak χ of Figure 4, reduced; (C) apex fraction from peak R of Figure 4, reduced; (D) apex fraction from peak β of Figure 4, reduced. Labeled peaks: (a′) 1-3 mannoses 33-108 rhPDGF-B and 1-3 mannoses 33-109 rhPDGF-B; (a) 33-109 rhPDGF-B; (b) 33-108 rhPDGF-B; (c) MetSO 1-31 rhPDGF-B and MetSO 1-32 rhPDGF-B; (d) 1-32 rhPDGF-B; (e) 1-31 rhPDGF-B; (f) MetSO 1-108 rhPDGF-B and MetSO 1-109 rhPDGF-B; (g) 0-3 mannoses 1-108 rhPDGF-B and 0-3 mannoses 1-109 rhPDGF-B.

Figure 4. RP-HPLC of homodimeric rhPDGF-BB at 85 °C using step-gradient elution. HPLC conditions as described in the Experimental Section, chromatographic system 2.

Identification of rhPDGF-B Polypeptides. The eight major peaks derived from the reduction of rhPDGF-BB were isolated on chromatographic system 3 using a semipreparative Vydac C18 (10 mm × 25 cm, Catalog No. 218TP510) column at a 5.5 mL/ min flow rate. The identity of the component(s) of each peak (labeled a′, a, b, c, d, e, f, and g in panel A of Figure 5) was determined by a combination of UV absorbance and fluorescence, methionine oxidation using chloramine T, N-terminal sequencing, amino acid composition analysis, and electrospray mass spectrometry. Tryptophan-containing polypeptides were identified by fluorescence detection (λex ) 280 nm, λem ) 343) using a Hitachi model L7480 HPLC fluorescence detector and by dual-wavelength UV detection (at 214 and 280 nm) using a Waters model 996 photodiode array detector. Since a single methionine is present 3812 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

only at residue 12, oxidation of methionine was investigated to identify the methionine-containing polypeptides. rhPDGF-BB was oxidized following the methods of Kunitani et al.26 and Shechter et al.27 by mixing an aliquot of rhPDGF-BB with chloramine T in an equimolar ratio and allowing the mixture to stand for various periods of time up to 90 min. The reaction products were subsequently reduced and analyzed by RP-HPLC. For sequencing experiments, portions of the collected peak fractions g and a were vacuum-dried and alkylated prior to automated N-terminal sequence analysis using an Applied Biosystems Procise sequencer. For amino acid composition analysis, the polypeptide(s) contained in each peak fraction was(were) acid hydrolyzed for 22-24 h at 110 °C in vacuo in the vapors of 6 N hydrochloric acid, containing 1% phenol as an antioxidant. The free amino acid mixtures were then separated by ion-exchange chromatography on a Beckman (26) Kunitani, M.; Hirtzer, P.; Johnson, D.; Halenbeck, R.; Boosman, A.; Koths, K. J. Chromatogr. 1986, 359, 391-402. (27) Shechter, Y.; Burstein, Y.; Patchornick, A. Biochemistry 1975, 14, 44974503.

model 6300 amino acid analyzer using a sodium acetate buffer system. Electrospray mass spectra were acquired on a Sciex API-III triple-quadrupole mass spectrometer (Perkin-Elmer Sciex Instruments, Thornhill, Canada) interfaced to a Michrom HPLC system (Michrom BioResources, Inc., Auburn, CA). Capillary Zone Electrophoresis Instrumentation and Methods. A BioFocus 3000 automated CE system (Bio-Rad Laboratories, Hercules, CA) was used for all analyses. Separations were performed on a Beckman eCap amine-coated fused silica capillary (50 µm i.d. × 375 µm o.d.) (Beckman Instruments, Fullerton, CA) maintained at 25 °C. The capillary, installed in a BioFocus userassembled cartridge, had an effective length of 42.4 cm and a total length of 47 cm. Samples were loaded onto the capillary by lowpressure injection (2 psi‚s). Electrophoresis was performed at a constant voltage of 319 kV/cm using the Beckman eCap 50 mM Tris-HCl buffer (pH 8.0) as the running buffer. The operating current was ∼20 µA. The electropherograms were monitored by UV absorbance at 214 nm at the anodic end. In the reversedcharge mode, the cathode and anode are at the inlet and outlet of the capillary, respectively. rhPDGF-BB samples were diluted to 1 mg/mL, and each diluted sample (50 µL) was incubated with 100 µL of 25 units/ mL carboxypeptidase B (Calbiochem, San Diego, CA) at room temperature for 30 min. The digestion was stopped by the addition of 25 µL of 0.1 N HCl. In some of the applications described, digested samples were spiked with the neutral marker benzyl alcohol, which migrates with the EOF, to account for runto-run variability in the EOF. Neat benzyl alcohol was diluted 200-fold with water, and 2 µL of the diluted solution was spiked into each 20 µL of digested sample mixture. Prior to analysis, all reagents and samples were degassed for 5 min at 14000g using a model 5415 C Eppendorf microcentrifuge (VWR, San Francisco, CA). Samples were analyzed on the same day they were prepared. Clipped forms of rhPDGF-BB were quantitated by expressing the corrected peak areas of the single-clipped, double-clipped, and intact forms as a percentage of the total corrected peak area (DC-, SC-, and IN-rhPDGF-BB) of the electropherogram. Corrected peak areas were calculated by dividing the peak area values (µV‚ s) by their respective peak migration times (in seconds). Two capillary conditioning runs were performed at the beginning of the run sequence to equilibrate the system. To ensure reproducible separations, the capillary was purged at high pressure (100 psi) first with 1 N NaOH (180 s), then with the eCap amine regenerator solution (180 s), and finally with the run buffer (180 s) prior to every injection. To avoid drying of the capillary, the inlet and outlet run buffer vials were left in the “up” position at the end of each run. Optimization of RC-CZE Assay. The electrophoretic conditions were optimized for separation buffer type, buffer pH, and run voltage. The influence of buffer pH was investigated in the pH interval 7-8 using two different buffers: 50 mM phosphate (pH 7.0) and 50 mM Tris-HCl (pH 8.0). At pH 7.0, the phosphate buffer generated a separation profile with split peaks, suggesting that other rhPDGF-BB variants may be resolved by this buffer system. Since the goal was to develop a clipping assay that would ignore other forms of heterogeneity, the phosphate buffer could not be used for this assay. Optimum separation of the three rhPDGF-BB isoform groups was achieved using the Beckman

eCap 50 mM Tris-HCl buffer (pH 8.0). The optimum run voltage for the assay was experimentally determined by performing the separations at increasing field strengths, ranging from 106 to 426 kV/cm. The conditions for treatment with carboxypeptidase B were also optimized for substrate-to-enzyme ratio, incubation time, and temperature.

RESULTS AND DISCUSSION Isolation of Intact, Single-Clipped, and Double-Clipped rhPDGF-BB by RP-HPLC. Similar to E. coli-derived rhPDGF,15 S. cerevisae-expressed rhPDGF-BB also exhibited multiple conformers that produced broad and heterogeneous chromatographic profiles when analyzed by RP-HPLC system 1. At 5 °C, six distinct peaks with several smaller shoulders were resolved (panel A of Figure 3). As the temperature was increased, the profile became less defined (data not shown). At 60 °C, the profile collapsed to a single peak with two backside shoulders. The early-eluting doublet peak (22 min) observed in the 5 °C profile of Figure 3 disappeared as the temperature was increased. The peaks generated at low temperature were characterized by rechromatography with RP-HPLC at 5 °C. Figure 3 shows the preparative chromatogram for the starting material (panel A) and the rechromatography for the collected fractions 4, 6, and 8 (panels B-D). The fractions, although enriched in the primary peak, each displayed significant amounts of other peaks present in the starting material. Thus, it appears that a slow equilibrium exists between some of the forms separated at 5 °C, resulting from the interconversion of rhPDGF-BB conformers. The isomerization of the Pro-Pro bond is probably responsible for the observed slow equilibrium among rhPDGF-BB conformers. To accelerate the rate of Pro-Pro isomerization and significantly reduce the number of conformers to only those resulting from clipping and truncation (Figure 2), we decided to separate rhPDGF-BB using RP-HPLC at 85 °C. Initially, the high-temperature RP-HPLC analysis of nonreduced rhPDGF-BB was performed using a shallow, linear TFA/acetonitrile gradient (data not shown). This method produced a complicated separation profile with over 15 poorly resolved peaks which could be fractionated into three major peak groups. Rechromatography of these three peak groups at 85 °C yielded the same distinct peak groups, suggesting that Pro-Pro isomerization is not resolved at this high temperature. Rechromatography of the collected peaks on chromatographic system 3 suggested that these peak groups are the double-clipped, singleclipped, and intact rhPDGF-BB forms, respectively. However, attempts to improve the RP-HPLC separation of clipped and intact forms using a linear gradient were unsuccessful. To further separate the clipped and intact forms of rhPDGF-BB, a gradient containing several isocratic steps was developed. Unlike the linear gradient system, the RP-HPLC analysis performed at an unusually high temperature using a step gradient (RP-HPLC system 2) partially resolved the rhPDGF-BB forms predominantly on the basis of clipping (not truncation) and yielded a less complicated separation profile than the linear gradient system. Using the step-gradient elution shown in Figure 4, the rhPDGF-BB was resolved into three distinct peaks: double-clipped (DC-rhPDGF-BB, peak R), single-clipped (SC-rhPDGF-BB, peak β) and intact forms (IN-rhPDGF-BB, peak χ). However, the three Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3813

peaks are not baseline resolved, and the first two peaks, which contain the DC and SC forms are asymmetrical and broad, suggesting the presence of peak heterogeneity. Lack of baseline resolution is indicative of an equilibrium between peaks. Peak impurity was later confirmed by analysis of fractions collected across peaks R, β, and χ of Figure 4 by RP-HPLC system 3, as described below. The RP-HPLC analysis of the reduced nonapex peak fractions of Figure 4 demonstrated a mixture of forms. Thus, this approach could not be used to accurately quantitate the three forms of clipped rhPDGF-BB homodimer. This high-temperature step-gradient RP-HPLC method also had other limitations: it did not yield reproducible results on different HPLC pump systems. Small increases in system dwell volume resulted in a dramatic degradation of the separation, similar to the poorly resolved 15-peak profile of the linear gradient. In addition, the 85 °C chromatography required extensive helium eluant degassing, which would otherwise result in complete oxidation of both methionines to the corresponding methionine sulfoxides. Although this RP-HPLC method was clearly inappropriate for routine quantitation of clipped forms, it was useful for the analytical isolation of intact, single-clipped, and doubleclipped rhPDGF-BB forms used for peak identification of the RCCZE assay. Multiple fractionation cycles of the peak apexes using this high-temperature (85 °C) RP-HPLC method were required to obtain purified samples of each of the three rhPDGFBB forms. Identity and Purity Determination of Peaks r, β, and χ. The identity and purity of the apex fractions collected from the high-temperature step-gradient RP-HPLC analysis shown in Figure 4 were established by RP-HPLC chromatography and RC-CZE analysis. The RP-HPLC method involved the reduction of rhPDGF-BB protein with DTT and analysis of the resulting polypeptide mixture by RP-HPLC system 3. As mentioned above, rhPDGF-BB undergoes clipping between residues Arg32 and Thr33. The clipped polypeptides contain a large clip (33-109 rhPDGFB) and a small clip (1-32 rhPDGF-B) which remain attached through an intrachain disulfide bond. Upon DTT reduction, the clipped polypeptides as well as the unclipped rhPDGF-B monomers are released from the dimeric rhPDGF-BB. RP-HPLC analysis of the reduced protein mixture results in the separation of eight major polypeptide peaks shown in panel A of Figure 5. The identity of the component(s) of each peak in this figure was determined by N-terminal sequencing, amino acid composition analysis, and electrospray mass spectrometry, as described in the Experimental Section. Panel B of Figure 5 shows the RP-HPLC profile of the reduced apex fraction collected from peak χ of Figure 4. Since the chromatogram showed only peaks f and g, which corresponded to full-length (1-109) rhPDGF-B and the full-length methionine sulfoxide containing polypeptides (MetSO 1-109) rhPDGF-B and no clipped polypeptides, this fraction was designated IN-rhPDGFBB. Panel C of Figure 5 shows the RP-HPLC profile of the reduced apex fraction collected from peak R of Figure 4. Since only clipped polypeptides a, a′, b, c, d, and e were observed and no full-length rhPDGF-B was detected, the results clearly indicate that this fraction contained DC-rhPDGF-BB. Panel D shows the RP-HPLC profile of the reduced apex fraction from peak β of Figure 4. This fraction contained both two full-length polypep3814 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

Figure 6. Reversed-charge CZE separation of purified intact, singleclipped, and double-clipped rhPDGF-BB without treatment with carboxypeptidase B. Electrophoretic conditions as described in the Experimental Section except that the applied voltage was 106 V/cm: (A) apex fraction from peak R of Figure 4; (B) apex fraction from peak β of Figure 4; (C) apex fraction from peak χ of Figure 4. Labeled peaks: group I, IN-rhPDGF-BB; group II, SC-rhPDGF-BB with no Arg truncation; group III, SC-rhPDGF-BB with single Arg truncation; group IV, DC-rhPDGF-BB with no Arg truncation; group V, DC-rhPDGFBB with single Arg truncation; group VI, DC-rhPDGF-BB with double Arg truncation.

tides, peaks g and f, and clipped polypeptides a′, a, b, c, d, and e. The peak area counts of the clipped polypeptides (peaks a′, a, b, c, d, and e) were equivalent to the area counts of the monomer peaks (peaks f and g), suggesting that this fraction consisted of SC-rhPDGF-BB. However, it is not possible to estimate the purity of the single-clipped fraction using this analysis since one can never be sure whether the polypeptides of the RP-HPLC profile originated from the single-clipped polypeptides alone or they contain contaminants from neighboring fractions. Thus, the identity and purity of the three apex fractions were further confirmed by RC-CZE analysis. The RC-CZE analysis was performed as described in the Experimental Section except that fractions collected from peaks R, β, and χ were not treated with carboxypeptidase B prior to analysis. RC-CZE analysis profiles for the untreated apex fractions R, β, and χ are shown in panels A, B, and C of Figure 6, respectively. This analysis generated three peaks for fraction R, two peaks for fraction β, and one peak for fraction χ. These results are consistent with the expected number of isoform groups for DC-rhPDGF-BB, SC-rhPDGF-BB, and IN-rhPDGF-BB, as illustrated in Figure 2. Peak identities were established as described in the next section. Using this RC-CZE analysis, the purity of the SC-, DC-, and IN-rhPDGF-BB forms in the three fractions was determined to be >95%. Taken together, these results indicated that purification of the three forms of rhPDGF-BB had been achieved and the apex fractions collected from peaks R, β, and χ were DC-rhPDGF-BB, SC-rhPDGF-BB, and IN-rhPDGF-BB, re-

Figure 7. Reversed-charge CZE separation of rhPDGF-BB with and without treatment with carboxypeptidase B (cp-B). Electrophoretic conditions as described in the Experimental Section: (A) rhPDGFBB, cp-B untreated; (B) rhPDGF-BB, cp-B treated. Peak identifications are the same as in legend of Figure 6.

spectively. These fractions were subsequently used in RC-CZE assay development, as described below. Development of the Reversed-Charge CZE Assay. This analysis was developed for the quantitative determination of the percentages of intact, single-clipped, and double-clipped forms in the rhPDGF-BB protein. The separation mechanism of RC-CZE is a function of the charge-to-mass (or hydrodynamic size) ratio. The analysis uses a positively charged amine-coated capillary to avoid adsorption of the basic rhPDGF-BB protein to the negatively charged bare fused-silica capillary wall. Reversal of the capillary surface charge also reverses the direction of electroosmotic flow, establishing a stable anodic EOF toward the outlet of the positively charged capillary. In RC-CZE, negatively charged molecules migrate in the same direction as the EOF (negative to positive) and are detected first. Positively charged molecules migrate against the EOF. Size also contributes to the migration; molecules of the same charge but different size migrate at different rates. The smaller the size of a positively charged molecule, the greater the migration velocity since it moves more efficiently against the EOF. A typical RC-CZE electropherogram of rhPDGF-BB protein is shown in panel A of Figure 7. The first peak in the electropherogram is the neutral marker (benzyl alcohol), which migrates with the EOF. RC-CZE analysis of the rhPDGF-BB protein without carboxypeptidase treatment prior to analysis yielded a separation profile with five protein peaks. Since all rhPDGF-BB forms carry

a positive charge at pH 8.0, they migrate against the EOF and pass through the detection window after the neutral marker. The identity of each peak in panel A of Figure 7 was established using the SC-rhPDGF-BB, DC-rhPDGF-BB, and IN-rhPDGF-BB apex fractions purified from the high-temperature RP-HPLC run shown in Figure 4. Figure 6 shows the RC-CZE analysis of the purified INrhPDGF-BB, SC-rhPDGF-BB, and DC-rhPDGF-BB fractions which had not been pretreated with carboxypeptidase B. The separation profile for the intact forms is shown in panel C of this figure. Since separation by RC-CZE is based on the charge/mass (or charge/ size) ratio, the truncation of the basic Arg32 has a greater effect on the migration velocity than truncation at the neutral amino acid residue Thr109. Truncation at Thr109 does not alter the charge or appreciably the size of the rhPDGF-BB protein. Thus, we do not expect a separation between the three group I isoforms. Indeed, the RC-CZE analysis profile for the purified IN-rhPDGF-BB fraction showed only one major peak, consistent with this prediction. The RC-CZE separation profile for the purified SC-rhPDGFBB fraction is shown in panel B of Figure 6. Truncation at Arg32 (group III) produces a more acidic (or less basic) isoform while truncation only at Thr109 (group II) does not alter the molecular charge. Thus, the eight isoforms of SC-rhPDGF-BB can be resolved into two peaks by RC-CZE, and due to their charge states, one would expect the group III isoforms to migrate faster than the group II isoforms. As expected, the purified SC-rhPDGF-BB fraction generated only two major peaks (panel B of Figure 6). The RC-CZE separation profile for the purified DC-rhPDGFBB fraction is shown in panel A of Figure 6. Truncation of Arg32 at one of the clip sites (group V) makes the isoform more acidic (or less basic), while truncation of Arg32 at both of the clip sites (group VI) removes two basic amino acid residues, making these isoforms even more acidic (or the least basic). Thus, the expected migration order of the isoforms would be group VI, followed by group V, and finally group IV. As shown in panel A of Figure 6, the purified DC-rhPDGF-BB fraction generated three peaks. The slower migrating minor peaks on the leading edge of the first peak in panels A-C of Figure 6 correspond to methionine sulfoxide oxidized forms of DC-, SC-, and IN-rhPDGF-BB, respectively, as demonstrated by chloramine T oxidation of the purified fractions (data not shown). The methionine sulfoxide oxidized forms migrate slower than their unmodified counterparts, perhaps due to a slight denaturation of their tertiary structures. The separation observed between IN-rhPDGF-BB (group I) and SC-rhPDGF-BB (group II) isoforms is likely due to the increase in the hydrodynamic size (shape) of the single-clipped isoforms, presumably through random conformation of the clipped polypeptide chains (residues 26-32 and 33-40). Clipping between Arg32 and Thr33 on one polypeptide chain of the dimeric rhPDGF-BB adds one COOH and one NH2 group. This results in a similar charge for the intact and single-clipped rhPDGF-BB isoforms, but increases the hydrodynamic size of the single-clipped molecule, leading to its slower mobility. A similar argument can be made for the separation seen between the SC-rhPDGF-BB (group II) and the more unfolded (two clipped chains, largest hydrodynamic size) DC-rhPDGF-BB isoforms (group IV). Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3815

Taken together, one would expect to get a total of six peaks (1 + 2 + 3 ) 6) by this analysis. As shown in panel A of Figure 7, RC-CZE analysis of the protein produced only five major peaks with a small shoulder on the leading edge of the first peak. To check for comigrating isoform groups, a mixture of the isolated SC-rhPDGF-BB and DC-rhPDGF-BB fractions was analyzed by RC-CZE and the separation profile compared to those of the individually analyzed forms. This experiment revealed that the double-clipped group IV isoforms comigrated with the singleclipped group III isoforms (data not shown). To reduce microheterogeneity due to charge differences at Arg32, the test samples were treated with carboxypeptidase B prior to analysis to completely remove Arg32 from the clipped polypeptides. Carboxypeptidase B, a C-terminal protease specific for the basic amino acids Arg and Lys, converts the single-clipped isoforms of group II (no Arg32 truncation) to group III (single Arg32 truncation by carboxypeptidase) and the double-clipped isoforms of groups IV (no Arg32 truncation) and V (single Arg32 truncation) to group VI (double Arg32 truncation by carboxypeptidase). To confirm peak identity, the purified SC-rhPDGF-BB and DCrhPDGF-BB fractions were reanalyzed by RC-CZE both before and after treatment with carboxypeptidase B, as shown in panels A and B of Figure 8. After treatment, SC-rhPDGF-BB and DCrhPDGF-BB forms produced electropherograms with only one slow-migrating major peak, indicating the formation of the more acidic single Arg32 truncated SC-rhPDGF-BB (group III) and double Arg32 truncated DC-rhPDGF-BB (group VI) isoforms, respectively. Treatment of the rhPDGF-BB protein with carboxypeptidase B prior to analysis also generated a simpler separation profile with only three peaks (panel B of Figure 7). The three baselineresolved peaks belong to isoform groups I (IN-rhPDGF-BB), III (SC-rhPDGF-BB), and VI (DC-rhPDGF-BB). Thus, the carboxypeptidase B treatment eliminated the comigration of the isoform groups III and IV, allowing complete separation of the intact, single-clipped, and double-clipped forms of rhPDGF-BB. The size of the three peaks (shoulder + main peak) represents the relative percentage of the three forms in the starting material. Assay Performance. To establish the utility of the RC-CZE analysis of rhPDGF-BB, the assay performance was evaluated for precision, linearity, and interlaboratory and intercapillary manufacturer ruggedness using rhPDGF-BB samples containing a relatively high degree of clipping. For simplicity, only the results for the intact forms (the smallest and the least precise peak) are reported as a worst case scenario. Precision. Intraday and interday variability of the corrected peak area percentages of double-clipped, single-clipped, and intact rhPDGF-BB were determined under optimum assay conditions, as described in the Experimental Section. Intraday precision was determined for repeated analyses of the same rhPDGF-BB sample digest in a single day. Five singleinjection analyses were performed over a 12-h period using a fresh set of CE reagents for each analysis. The calculated average corrected peak area percentage of the IN-rhPDGF-BB was 20.6%, with an intraday RSD of 3.1% (n ) 5). Interday precision was also evaluated through replicate analyses of a rhPDGF-BB sample on different days. Five analyses consisting of a single injection each were performed on five 3816 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

Figure 8. Reversed-charge CZE separation of purified singleclipped and double-clipped rhPDGF-BB with and without treatment with carboxypeptidase B (cp-B). Electrophoretic conditions as described in the Experimental Section except that the applied voltage was 106 V/cm: (A) apex fraction from peak β of Figure 4 (SCrhPDGF-BB), cp-B untreated; (B) apex fraction from peak β of Figure 4 (SC-rhPDGF-BB), cp-B treated; (C) apex fraction from peak R of Figure 4 (DC-rhPDGF-BB), cp-B untreated; (D) apex fraction from peak R of Figure 4 (DC-rhPDGF-BB), cp-B treated. Peak identifications are the same as in legend of Figure 6.

different days. The calculated average corrected peak area percentage of the IN-rhPDGF-BB was 22.3%, with an interday RSD of 4.4% (n ) 5). Taken together, these results show good precision for the RC-CZE assay. Linearity. Assay linearity was determined by analyzing 11 samples of rhPDGF-BB with concentrations ranging from 0.1 to 10 mg/mL prior to treatment with carboxypeptidase B. The measured total corrected peak areas of rhPDGF-BB were plotted against the 11 sample concentrations, which resulted in a leastsquares linear regression having an r2 of 0.999 with the response factors for each of the 11 concentrations having an RSD of 9.4%. Ruggedness. The interlaboratory ruggedness of the assay was evaluated by comparing the corrected peak area percentages of IN-rhPDGF-BB obtained from single analyses of 15 rhPDGF-BB samples performed in two different laboratories. Analyses in each laboratory were performed using a different BioFocus 3000 instrument. The mean ratio (Lab1/Lab2) of the corrected peak area percentage values ( SD was 1.00 ( 0.05 for the 15 samples

tested. These results establish good correspondence among the data from the two laboratories, demonstrating laboratory-tolaboratory ruggedness. The intercapillary manufacturer variability was also evaluated by analyzing a rhPDGF-BB sample treated with carboxypeptidase B on the BioFocus using capillaries from two different manufacturers: eCap amine-coated (Beckman Instruments) and Microcap positive wall capillaries (Saracep Inc., Santa Clara, CA). The ratio

of the corrected peak area percentage for the IN-rhPDGF-BB (Microcap/eCap) was 1.02, demonstrating capillary-to-capillary ruggedness of the RC-CZE assay.

Received for review March 20, 1998. Accepted July 6, 1998. AC980329R

Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3817