Site-Specific Protein PEGylation: Application to Cysteine Analogs of Recombinant Human Granulocyte Colony-Stimulating Factor

INTRODUCTION

Granulocyte colony-stimulating factor (G-CSF) is a 19 kDa glycoprotein that stimulates the proliferation, maturation and functional activation of cells of the granulocyte lineage. Recombinant G-CSF is widely used to ameliorate neutropenia resulting from meyelosuppressive chemotherapy and bone marrow transplantation, and to mobilize peripheral blood progenitor cells for transplantation and blood banking (1, 2). G-CSF has a short circulating half-life in vivo, which necessitates that the protein be administered to patients by daily subcutaneous injection. Development of modified forms of G-CSF that last longer in the body and can be administered less frequently is of significant interest to patients and healthcare providers.

Covalent modification of G-CSF with polymers such as polyethylene glycol (PEG) increases the circulating half-life of the protein in animals and humans (3–6). Covalent attachment of PEG to a protein increases the protein’s effective size and reduces its rate of clearance from the body. Previous studies used amine-reactive PEGs to modify G-CSF at exposed amine groups on lysine residues and the N-terminal amino acid. This approach is not optimal for G-CSF because the protein contains four lysine residues in addition to the N-terminal amino acid, and the lysine residues are located in regions of G-CSF that have been implicated in receptor binding (7 – 10). Not unexpectedly, modification of G-CSF with amine-reactive PEG reagents reduces in vitro biological activity of the protein by 3 - to 50-fold, depending upon the number and sizes of the attached PEG molecules (3, 4). Loss of in vitro bioactivity is greatest when G-CSF is modified with large PEGs, e.g., 20 kDa-PEGs, which are the most useful PEGs for extending the protein’s half-life. Amine-PEGylated-G-CSF also is heterogeneous, comprising a complex mixture of at least four isoforms and multiple molecular weight species, all of which have different specific activities (3–5). Less heterogeneous PEG-G-CSF conjugates with less severe loss of in vitro bioactivity can be prepared by preferential attachment of PEG to the N-terminus of G-CSF (5, 6, 11).

An alternative method for PEGylating proteins covalently attaches PEG to cysteine residues using cysteine-reactive PEGs such as maleimide PEGs. At near neutral pH, these PEG reagents selectively attach to the thiol groups of “free” cysteine residues, i.e., cysteine residues not involved in disulfide bonds. Since most proteins do not contain surface-exposed free cysteine residues, it is possible to target attachment of a PEG molecule to a specific site in a protein by introducing a free cysteine residue into a protein using site-directed mutagenesis, followed by modification of the added cysteine residue with a maleimide-PEG (site-specific PEGylation; 12). If the target site is non-essential, PEGylation will not significantly alter in vitro bioactivity of the protein. The resulting PEG-protein conjugate also has a homogeneous structure, which simplifies analytical characterization of the product. In addition, protein structure: function information can be used to aid in the rational design of PEG-protein conjugates. In this report we demonstrate the feasibility of using site-specific PEGylation technology to design long acting PEGylated G-CSF cysteine analogs with preserved in vitro bioactivity and improved therapeutic efficacy in vivo.

Materials and Methods

Construction of G-CSF Cysteine Analogs

Plasmid pBBT165, which contains a cDNA encoding human G-CSF has been described (13). DNA encoding the leader sequence of the Escherichia coli heat-stable enterotoxin (STII) gene (14) was fused to the coding sequence for mature G-CSF using polymerase chain reaction (PCR) methods (15, 16). At the same time, a TAA stop codon was added following the carboxy-terminal residue, P174, codons for prolines at positions 2, 5, and 10 were all changed to CCG, the arginine codon at position 22 was changed to CGT, the leucine codon at position 15 was changed to CTG, and the leucine codon at position 18 was changed to CTC. The G-CSF (C17S) and G-CSF (C17S) cysteine analogs were constructed using site-directed PCR-based mutagenesis (15, 16). Correct DNA sequences for G-CSF and G-CSF cysteine analogs were confirmed. The STII-G-CSF, STII-G-CSF (C17S), and STII-G-CSF (C17S) cysteine analog genes were cloned into plasmid pCYB1 (New England BioLabs, Beverly, MA) for expression in E. coli.

Expression, Purification and In Vitro Analysis of G-CSF Cysteine Analogs

pCYB1 plasmids encoding the various G-CSF proteins were transformed into E. coli strain W3110. Saturated overnight cultures were diluted to an optical density at 600 nm of ~ 0.05 in Luria broth containing 100 μg/mL ampicillin. When the optical densities of the cultures reached 0.5–0.7, isopropyl-β-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0.5 mM and the cultures incubated overnight for 16 h at 28°C. In some experiments the cells were subjected to osmotic shock (17) to release the soluble contents of the periplasm. For most experiments, the cells were pelleted by centrifugation and stored at −80° C. Cell pellets from 400 mL cultures were treated with 5 mL of B-PER^TM bacterial protein extraction reagent according to the manufacturer’s protocols (Pierce Chemical Company, Rockford, IL). This mixture was treated with lysozyme (200 μg/mL) for 10 min, and MgCl₂ (10 mM final concentration) and protease-free DNAse (2 μg/mL) were added. Insoluble material was collected by centrifugation, washed with water and recentrifuged. The resulting pellet was dissolved in 20 ml of 8 M urea, 25 mM cysteine in 20 mM Tris Base. This mixture was stirred for 30 min at room temperature, then diluted into 100 mL of 40 mM sodium phosphate, 40 μM copper sulfate, 15% glycerol, pH 8.0. After 2 days at 4°C, the pH of the refold mixture was adjusted to 4.0 with HCl and the mixture centrifuged. The supernatant was loaded onto a 5 mL S-Sepharose HiTrap column (Amersham Biosciences Corporation, Piscataway, NJ) equilibrated in 40 mM sodium phosphate pH 4.0 (Buffer A) and bound proteins eluted with a linear salt gradient from 0–100% Buffer B (500 mM NaCl, 40 mM sodium phosphate, pH 4.0). Column fractions were analyzed by non-reducing SDS-PAGE. Fractions containing the G-CSF proteins and no visible impurities were pooled. In vitro bioactivities of the G-CSF proteins were measured using the murine NFS-60 cell line, essentially as described (13). Serial three-fold dilutions of the protein samples were prepared and analyzed in triplicate.

Preparation of PEGylated G-CSF (C17S) Cysteine Analogs

G-CSF (C17S) cysteine analogs were incubated at pH 8.5 for 1 h at room temperature with a 10-fold molar excess of TCEP (Tris[2-carboxyethylphosphine]hydrochloride; Pierce Chemical Company) and a 20-fold molar excess of 5 kDa maleimide-PEG or 20 kDa maleimide-PEG (Nektar, Inc., Huntsville, AL). Wild type G-CSF and G-CSF (C17S) were exposed to the same reaction conditions. At the end of the incubation period, each PEGylation mixture was diluted 10-fold with 40 mM sodium phosphate (monobasic) and the pH was adjusted to 4.0. PEGylated protein was separated from non-PEGylated protein by S-Sepharose column chromatography, using the conditions described above. The purified PEGylated proteins contained less than 0.1 endotoxin units/mg, as determined using QCL-1000 Limulus Amebocyte Lysate kits (BioWhittaker, Walkersville, MD).

Biochemical Methods

N-terminal protein sequencing was performed by Macromolecular Resources (Colorado State University, Fort Collins, CO) using an Edman degradation procedure. Protein samples were analyzed by SDS-PAGE on precast Tris-glycine polyacrylamide gels (Invitrogen Corporation, Carlsbad, CA) and stained with Coomassie Blue. 2-Mercaptoethanol was added to a final concentration of 1–5 % (V/V) in sample buffer to reduce the proteins. Protein concentrations were determined using a Bradford protein assay kit (Bio-Rad Laboratories, Richmond, CA), using bovine serum albumin as the protein standard. Western blots were performed using standard procedures. The primary antibody was a polyclonal goat anti-human G-CSF antisera (R&D Systems, Inc., Minneapolis, MN) and the secondary antibody was an alkaline phosphatase-conjugated rabbit anti-goat IgG antisera (Pierce Chemical Company). Western blots were developed using an NBT/BCIP staining kit (Promega Corporation, Madison, WI). A human met-G-CSF standard (E. coli-expressed) was obtained from R&D Systems, Inc.

Animal Experiments

Animal experiments were performed with the approval of BolderPATH’s Institutional Animal Care and Use Committee. Groups of three male Sprague Dawley rats, weighing approximately 350 g each, received a single intravenous injection (lateral tail vein) of recombinant wild type G-CSF or 20 kDa-PEG-L3C at a dose of 100 μg protein/kg. At selected time points blood samples (0.3 to 0.4 mL) were drawn from the rats into EDTA anti-coagulant tubes. A portion of each blood sample was used for a complete blood cell count analysis (performed by IDEXX Laboratories, Irvine, CA). The remainder of the blood sample was centrifuged and the plasma frozen at −80°C. Blood samples were drawn at 0.25, 1.5, 4, 8, 12, 16, 24, 48, 72, 96, and 120 h post-injection. A pre-dose baseline sample was obtained 24 h prior to injection of the test compounds. Plasma concentrations of the proteins were measured using human G-CSF ELISA kits (R&D Systems, Inc.). Serial dilutions of plasma samples from one animal in each test group were analyzed in the in vitro bioassay to identify dilutions that fell within the linear range of the ELISA (39 to 2500 pg/mL). Duplicate samples of appropriate dilutions of plasma samples from all rats were then analyzed in the ELISA. Pharmacokinetic parameters were analyzed using the WinNonlin software package (Pharsight Corporation, Mountain View, CA). Statistical analyses between samples were compared using a student’s T-test with significance set at p ≤ 0.05.

RESULTS

Periplasmic Expression of Wild Type G-CSF and G-CSF (C17S) Proteins in E. coli

Intracellular expression of human G-CSF in E. coli yields a G-CSF protein containing an N-terminal methionine residue not present in the natural human protein. To express a human G-CSF protein without this non-natural methionine, we secreted G-CSF to the E. coli periplasm by fusing the mature coding sequence of G-CSF to the E. coli STII signal sequence. G-CSF contains five cysteine residues, which form 2 disulfide bonds (C36-C42 and C64-C74). The fifth cysteine residue, C17, is not involved in a disulfide bond (18). An identical construct was prepared for a G-CSF variant in which C17 is changed to serine (C17S). Both proteins were expressed under control of the IPTG-inducible tac promoter of plasmid pCYB1 in E. coli strain W3110. Following overnight induction at 28°C, SDS-PAGE of induced G-CSF wild type and C17S cultures showed expression of a new 19 kDa protein. Western blot analyses revealed that the 19 kDa protein reacted with an anti-human G-CSF antiserum and co-migrated with a commercial E. coli-expressed human met-G-CSF standard (data not shown). The latter result suggested that the STII leader peptide had been removed, which is consistent with G-CSF having been secreted to the periplasm. Removal of the STII signal sequence was confirmed by N-terminal protein sequencing studies presented below. Induced cells were subjected to osmotic shock (17) and soluble periplasmic proteins were separated from cytoplasmic, insoluble periplasmic, and cell-associated components by centrifugation. The bulk of the G-CSF proteins remained associated with the osmotic shock pellets, indicating that the secreted G-CSF proteins were insoluble or membrane-associated (data not shown).

Refolding, Purification and In Vitro Bioactivities of G-CSF wild type and C17S Proteins

E. coli cells expressing the G-CSF wild type and C17S proteins were treated with a mild detergent mixture (B-PER) to release the soluble components of the cells. Particulate material was collected by centrifugation and the insoluble proteins denatured, reduced and refolded as described in Materials and Methods. The refolded G-CSF proteins were purified to greater than 90% purity by S-Sepharose column chromatography. Both wild type G-CSF and G-CSF (C17S) eluted as single peaks from the S-Sepharose column at about 300 mM NaCl. Non-reducing SDS-PAGE analysis of column fractions across the major peaks showed a single major refold species migrating at 19 kDa for both proteins. Fractions containing the major refold species and no visible impurities were pooled. The final yields of G-CSF wild type and G-CSF (C17S) after the S-Sepharose column step were about 1.1 mg and 3.3 mg, respectively from 400 mL of culture (Table 1). Purified wild type G-CSF and G-CSF (C17S) co migrated under reducing and non-reducing SDS-PAGE conditions (Fig. 1 and data not shown). Reduced G-CSF migrates with a slightly higher apparent molecular mass than non-reduced G-CSF (19 kDa versus 17kDa, respectively), due to reduction of the protein’s two native disulfide bonds. Wild type G-CSF and G-CSF (C17S) had mean EC₅₀s of 10 and 7 pg/mL, respectively, in an in vitro cell proliferation assay using mouse NFS-60 cells (Fig. 2 and Table 1). G-CSF (C17S) was reproducibly 1.5-to 2-fold more potent than wild type G-CSF and 2- to 3-fold more potent than a commercial met-G-CSF standard in the bioassay (Table 1).

Table 1.

In Vitro Bioactivities of G-CSF, G-CSF (C17S) and G-CSF (C17S) Cysteine Analogs

G-CSF Protein	Added Cysteine Location a	Yield (mg) b	EC50 (pg/ml) c
met-G-CSF d	-	-	18.6 ± 6.6
G-CSF e	-	1.1	10.2 ± 1.6
G-CSF (C17S)	-	3.3	7.2 ± 2.0
*-1C / C17S	N-terminus	2.5	7.0 ± 1.3
T1C / C17S	N-terminus	2.5	7.8 ± 2.9
L3C / C17S	Preceding A Helix	2.7	8.0 ± 2.2
A6C / C17S	Preceding A Helix	2.9	8.2 ± 3.3
S7C / C17S	Preceding A Helix	2.7	7.3 ± 3.2
E93C / C17S	B-C loop	0.2	7.6 ± 0.9
A129C / C17S	C-D loop	2.5	6.0 ± 0.0
T133C / C17S	C-D loop	1.9	6.6 ± 1.2
A136C / C17S	C-D loop	0.5	8.3 ± 1.3
A139C / C17S	C-D loop	3.1	5.2 ± 0.3
A141C / C17S	C-D loop	2.3	8.9 ± 1.1
Q173C / C17S	Following D Helix	3.5	6.2 ± 1.3
*175C / C17S	C-terminus	2.9	5.6 ± 0.4

^aG-CSF structure from reference 19.

^bAmount of protein recovered from 400 ml E. coli culture after refolding and S-Sepharose column chromatography

^cMeans ± SD for at least three assays for each protein.

^dmet-G-CSF standard (R & D Systems, Inc.)

^eWild type G-CSF prepared by us

Figure 1.

Non-reducing SDS-PAGE analysis of column pools of purified recombinant wild-type G-CSF, G-CSF (C17S) and G-CSF (C17S) cysteine analogs. MW indicates molecular weight markers.

Figure 2.

Dose response curves for representative G-CSF (C17S) cysteine analogs, wild type G-CSF and G-CSF (C17S) for stimulating proliferation of NFS-60 cells. Data are means ± standard deviations of triplicate wells from representative experiments.

Construction and Characterization of G-CSF (C17S) Cysteine Analogs

G-CSF has a compact globular structure, comprising four alpha helices joined by loops (19). The four alpha helices are termed A-D beginning from the N-terminus of the protein. The loop regions are referred to by the helices they join, e.g., the A-B loop joins helices A and B. Thirteen mutant G-CSF genes were constructed, each of which contains a single cysteine substitution or addition. Five cysteine analogs were constructed in the amino-terminal region preceding Helix A [*-1C (the addition of a cysteine residue preceding the natural amino terminus), T1C, L3C, A6C and S7C]; one cysteine analog was constructed in the B-C loop (E93C); five cysteine analogs were constructed in the C-D loop (A129C, T133C, A136, A139C, and A141C); and two cysteine analogs were constructed in the carboxy-terminal region following Helix D [Q173C and *175C (the addition of a cysteine residue following the natural carboxy-terminus)]. The G-CSF cysteine analogs were constructed in a C17S background to avoid potential difficulties and/or ambiguities that might be caused by the unpaired C17 residue in wild type G-CSF. The cysteine analogs were expressed, refolded and purified using the protocols described for wild type G-CSF and G-CSF (C17S). Protein recoveries for most of the G-CSF (C17S) cysteine analogs after the S-Sepharose column step were similar to the recovery of G-CSF (C17S) and greater than the recovery of wild type G-CSF (Table 1). Non-reducing SDS-PAGE analysis of the 13 purified cysteine analogs is shown in Fig. 1. The cysteine analogs were recovered predominantly as monomers that comigrated with wild type G-CSF and G-CSF (C17S) under reducing and non-reducing conditions. All but one of the cysteine analogs eluted from the ion-exchange column at a salt concentration (300–325 mM NaCl) similar to wild type G-CSF and G-CSF (C17S). The one exception, E93C, eluted at about 400 mM NaCl. In vitro biological activities (EC₅₀s) of the 13 cysteine analogs were indistinguishable from that of G-CSF (C17S) in the NFS-60 cell proliferation assay (Table 1). Examples of dose response curves for representative cysteine analogs are shown in Figure 2.

Preparation and In Vitro Bioactivities of PEGylated G-CSF Cysteine Analogs

Aliquots of the purified G-CSF cysteine analogs were reacted with a 5 kDa maleimide-PEG using the conditions described in Materials and Methods. A reducing agent (TCEP) was included in the PEGylation reaction to partially reduce the protein and expose the free cysteine residue, which forms a mixed disulfide with cysteine during the refold procedure. Control experiments indicated that the cysteine analogs needed to be partially reduced to PEGylate efficiently (data not shown). The PEGylation reactions were terminated by diluting the samples and adjusting the pH to 4.0. Fig. 3B, lane 3, shows non-reducing SDS-PAGE analysis of a representative PEGylation reaction using the L3C protein. The only detectable PEGylated species is the mono-PEGylated-L3C protein, which migrates with an apparent molecular mass of 28 kDA. Wild type G-CSF and G-CSF (C17S) were not modified under identical PEGylation conditions (Fig. 3C). PEGylated L3C protein was separated from non-PEGylated protein and excess PEG reagent by S-Sepharose column chromatography (Fig. 3A). The chromatograms from the S-Sepharose columns showed two major protein peaks eluting at approximately 275 mM NaCl and 300–325 mM NaCl. SDS-PAGE analysis indicated that the early-eluting major peak was mono-PEGylated L3C and the later-eluting major peak was unreacted L3C protein (Fig. 3B). Fractions containing predominantly mono-PEGylated protein (fractions 24 and 25; lanes 5 &6 in Fig. 3B) were pooled and used for bioactivity measurements of 5 kDa PEG-L3C. The other G-CSF cysteine analogs were PEGylated and purified by the identical protocol. All of the cysteine analogs reacted with the PEG reagent to yield monoPEGylated proteins. Relative PEGylation efficiencies for the cysteine analogs were estimated from the S-Sepharose column chromatogram peaks and varied from 23% to 60%, with most cysteine analogs being in the 30–60% range (Table 2). PEGylation efficiencies can be increased by optimizing reaction conditions for each cysteine analog (data not shown). Five cysteine analogs (L3C, T133C, A141C, Q173C and *175C) also were modified with a 20 kDa maleimide-PEG and purified by S-Sepharose column chromatography. The 20 kDa-PEGylated proteins eluted from the S-Sepharose column at approximately 250 mM NaCl. Non-reducing SDS-PAGE analysis of the purified PEGylated cysteine analogs are shown in Figure 4.

Figure 3.

Purification of PEG-L3C by S-Sepharose column chromatography. Panel A shows the S-Sepharose column chromatogram. Panel B is non-reducing SDS-PAGE of purified L3C (Lane 1), the L3C PEGylation reaction (Lane 2), S-Sepharose fractions 22–26 (Lanes 3–7) and fractions 28–29 (lanes 8–9). Fractions 24 and 25 (lanes 5 and 6) containing PEG-L3C were pooled. Fractions 28–29 (Lanes 8–9) contain unreacted L3C. Panel C shows non-reducing SDS-PAGE analysis of G-CSF and G-CSF (C17S) subjected to the identical PEGylation reaction conditions used for L3C. G-CSF (Lane 1), G-CSF PEGylation reaction (Lane 2), G-CSF (C17S) (Lane 3), G-CSF (C17S) PEGylation reaction (Lane 4). MW indicates molecular weight markers.

Table 2.

In Vitro Bioactivities of PEGylated G-CSF Cysteine Analogs

G-CSF Analog	% PEGylation Efficiency a	EC50s (pg/ml) b
		5 kDa-PEG	20 kDa PEG
*-1C / C17S	60	5.6 ± 0.3
T1C / C17S	45	7.0 ± 1.0
L3C / C17S	45	5.5 ± 0.6	8.8 ± 1.0
A6C / C17S	26	6.9 ± 1.0
A129C / C17S	23	7.1 ± 3.4
T133C / C17S	51	7.4 ± 3.1	9.0 ± 2.9
A136C / C17S	34	6.9 ± 1.1
A139C / C17S	34	6.8 ± 2.8
A141C / C17S	38	7.1 ± 0.6	9.3 ± 3.8
Q173C / C17S	50	7.0 ± 2.6	11.0 ± 1.8
*175C / C17S	52	11.0 ± 1.0	14.0 ± 2.3

^aPercent PEGylated protein estimated by S-Sepharose column chromatography analysis of 5 kDa-PEG PEGylation reactions

^bMeans ± SD for at least three assays for each protein

Figure 4.

Non-reducing SDS-PAGE analysis of column pools of purified 5 kDa- and 20 kDa-PEGylated G-CSF (C17S) cysteine analogs. MW indicates molecular weight markers. G-CSF (C17S) is shown for comparison.

The 20 kDa PEG-L3C protein was subjected to N-terminal amino acid sequencing, yielding the sequence TPXGPAS (where X indicates a blank). This sequence matches the N-terminal sequence of mature G-CSF (20), indicating that the STII signal sequence is correctly processed. The presence of a blank at the third position is consistent with the presence of cysteine and/or PEG-cysteine at this position (the third amino acid in wild type G-CSF is leucine).

In vitro biological activities of the 11 purified 5 kDa PEG-G-CSF (C17S) cysteine analogs and 5 purified 20 kDa PEG-G-CSF (C17S) cysteine analogs were measured in the NFS-60 cell proliferation assay. EC₅₀s for the PEGylated G-CSF (C17S) cysteine analogs ranged from 5–12 pg/mL and were comparable to that of G-CSF (C17S) and wild type G-CSF (Table 2). Dose-response curves for representative PEG-G-CSF (C17S) cysteine analogs are shown in Figure 5.

Figure 5.

Dose response curves for representative PEGylated G-CSF (C17S) cysteine analogs, G-CSF (C17S) and wild type G-CSF for stimulating proliferation of NFS-60 cells. Data are means ± standard deviations of triplicate wells from representative experiments.

Pharmacokinetic and Efficacy Experiments in Rats

We determined the circulating half-lives of the 20 kDa PEG-L3C protein and wild type G-CSF following intravenous administration to rats at a dose of 100 μg protein/kg (Figure 6A). Plasma levels of the proteins were measured by ELISA from 0.25 h to 120 h post-injection. Wild type G-CSF cleared rapidly from the rats and reached undetectable levels by 24h post-injection. The terminal half-life of wild type G-CSF was calculated to be 1.8 h, similar to what is reported in the literature for met-G-CSF (1.82 h; 3). In contrast, the 20 kDa PEG-L3C protein possessed a significantly longer circulating half-life that was biphasic. Between 1.5 and 48 h, the half-life of 20 kDa-PEG-L3C was 13.7 h. Between 48 and 96 h, when blood neutrophil levels are elevated (see below), the 20 kDa-PEG-L3C protein was cleared more rapidly, with a half-life of 4.7 h.

Figure 6.

Changes in protein plasma levels (Panel A), blood neutrophil counts (Panel B), and white blood cell counts (Panel C) following a single intravenous injection of 100 μg protein/kg of 20 kDa-PEG-L3C and wild type G-CSF to rats. Data are means ± standard deviations for 3 rats per group. Asterisks (*) denote cell numbers that were significantly different (p< 0.05) between groups.

Both wild type G-CSF and 20 kDa PEG-L3C stimulated time-dependent increases in peripheral white blood cells and neutrophils over baseline values (Fig. 6B and C). White blood cell and neutrophil counts for the test groups receiving wild type G-CSF peaked 10–24 h post-injection and returned to baseline values by 48 h, similar to results reported for met-G-CSF (3, 13). In contrast, white blood cell and neutrophil counts for the rats receiving PEG-L3C did not peak until 48–72 h post-injection and did not return to baseline values until 96–120 h post-injection. White blood cell and neutrophil levels were significantly higher in the rats receiving 20 kDa PEG-L3C than in the rats receiving wild type G-CSF at 48 and 72h post-injection (p<0.05). No differences in red blood cell or platelet counts were noted between the groups (data not shown). Red blood cell counts in both groups decreased by about 15% during the first 12 h post-injection, possibly due to repeated blood sampling (data not shown).

DISCUSSION

Modification of proteins with amine-reactive PEGs has proven to be a useful technology for prolonging the circulating half-lives and improving in vivo efficacy of protein therapeutics (6, 21, 22). However, amine-PEGylated proteins often comprise heterogeneous product mixtures with significantly reduced (often 10–100 fold) specific activities (4, 12, 22) due to reaction of the PEG reagent with multiple lysine residues in the protein, some of which may lie at or near the active site or receptor binding site of the protein. Loss of bioactivity increases the amount and cost of drug required by patients and may fail to maximize the potential therapeutic benefits of the drug to patients. G-CSF is an example of a protein whose in vitro bioactivity is significantly reduced (up to 50-fold) by modification of the protein with conventional amine-reactive PEGs (4). G-CSF contains three lysines in helix A and a fourth lysine at the N-terminal end of the A-B loop (19, 20). Helix A and the N-terminal end of the A-B loop, along with helix D, have been identified as critical receptor binding sites in G-CSF (7–10, 23). It is likely that modification of one or more of the lysine residues in these critical regions contribute to the poor in vitro bioactivity of amine-PEGylated G-CSF.

By contrast, our data indicate that it is possible to create PEGylated G-CSF (C17S) cysteine analogs with minimal, if any, loss of in vitro biological activity by targeting attachment of the PEG molecule to specific predetermined sites in the protein using site-specific PEGylation technology. We identified three regions, the N-terminal region preceding helix A, the C-D loop and the C-terminal region following helix D, where cysteine residues can be introduced without appreciable effects on in vitro bioactivity of G-CSF (C17S). All of the purified G-CSF (C17S) cysteine analogs reacted readily with maleimide PEGs and yielded mono-PEGylated proteins, whereas wild type G-CSF and G-CSF (C17S) did not react with the PEG reagents under identical conditions. The latter result strongly suggests that the PEG molecule is attached to the added cysteine residue in the cysteine analogs. The failure of C17 in wild type G-CSF to react with the PEG reagent under the conditions used in our study probably due to the fact that C17 is partially buried (19, 24). A previous report indicated that C17 in G-CSF is unreactive to alkylating agents (24). All of the PEGylated G-CSF (C17S) cysteine analogs studied retained complete or near complete in vitro biological activity, even when modified with large 20 kDa-PEGs. The large number of fully active G-CSF (C17S) cysteine analogs and PEGylated G-CSF (C17S) cysteine analogs identified probably is due to our focusing mutagenesis efforts on the non-helical regions of the protein, which lie outside of the major receptor binding sites. The finding that large polymers can be attached to many amino acids in these regions without appreciable affects on in vitro bioactivities of the proteins provides further evidence that the regions examined (the region preceding helix A, the C-D loop and the region following helix D) are largely non-essential for in vitro biological activity of G-CSF. Our data regarding the region preceding helix A are consistent with studies showing that PEG-G-CSF proteins with only moderate loss of in vitro bioactivity (2- to 3-fold) can be prepared by preferential attachment of PEG to the N-terminus of G-CSF or met-G-CSF (5,11).

Attachment of a single 20 kDa maleimide-PEG to G-CSF (C17S/L3C) dramatically improves the protein’s circulating half-life and therapeutic efficacy relative to wild type G-CSF in rats, similar to what has been reported for met-G-CSF modified with one or multiple amine-reactive PEGs (3, 4, 6). 20 kDa PEG-L3C stimulated increases in blood neutrophils and white blood cells that were significantly greater and longer lasting than the increases seen with wild type G-CSF. Kidney filtration and receptor-mediated endocytosis has been implicated as dual mechanisms regulating G-CSF metabolism and neutrophil homeostasis in vivo (25). G-CSF receptors are abundant on neutrophils and bone marrow cells. The more rapid clearance of PEG-L3C observed beginning 48h post-injection correlates with peak neutrophil levels in the rats. It is likely that metabolism of PEG-L3C by bone marrow cells and/or neutrophils is responsible for the more rapid clearance of PEG-L3C observed at these times. In studies to be presented elsewhere (Bell et al., manuscript in preparation), we find that PEG-L3C also is significantly more effective than met-G-CSF at accelerating recovery from chemotherapy-induced neutropenia following single or multiple subcutaneous injections in rats.

In summary, we have shown that it is possible to create mono-PEGylated G-CSF (C17S) proteins with preserved in vitro biological activity and improved in vivo therapeutic efficacy by targeting attachment of the PEG molecule to nonessential sites in the protein using site-specific PEGylation. Targeted attachment of the PEG molecule results in homogeneously modified PEG-G-CSF (C17S) proteins with significantly greater in vitro biological activities than PEG-G-CSF proteins prepared using conventional non-specific amine-PEGylation technologies. G-CSF has a four-helix bundle structure that is shared by over 20 different cytokines and growth factors, collectively referred to as the growth hormone supergene family (19, 26, 27). Information gained from our studies with G-CSF should prove useful for creating long-acting, site-specific PEGylated forms of other members of this important gene family.