Site-Specific Pegylation of G-CSF by Reversible Denaturation

cysteine 17 of granulocyte colony stimulating factor (G-CSF), which is known to be partially buried in a hydrophobic protein pocket. The PEG linking w...
0 downloads 0 Views 289KB Size
Download PDF
1824

Bioconjugate Chem. 2007, 18, 1824–1830

Site-Specific Pegylation of G-CSF by Reversible Denaturation Francesco M. Veronese,*,† Anna Mero,† Francesca Caboi,‡ Mauro Sergi,‡ Christian Marongiu,‡ and Gianfranco Pasut† Department of Pharmaceutical Science, University of Padua, via Marzolo 5, 35100 Padua, Italy, and Bio-Ker S.r.l., Piscinamanna 09010 Pula (CA), Italy. Received April 13, 2007; Revised Manuscript Received August 3, 2007

A new strategy has been developed for extending the possibility of poly(ethylene glycol) (PEG) modification to accessible thiol groups of biologically active proteins. In particular, thiol-reactive PEGs have been coupled to the cysteine 17 of granulocyte colony stimulating factor (G-CSF), which is known to be partially buried in a hydrophobic protein pocket. The PEG linking was accomplished by partial protein denaturation with 3 M guanidine · HCl in the absence of any reducing agent in order to preserve the native protein’s disulfide bridges. PEG coupling occurred also, but at a lower degree, by using a 3 M solution of urea as the denaturing agent. Following the PEGylation, which was carried out in the unfolded state, the conjugated protein was refolded using dialysis or gel filtration chromatography to eliminate the denaturant. Different thiol-reactive PEGs and polymer molecular weights (5, 10, or 20 kDa) were investigated for G-CSF conjugation under denaturation. The secondary structure of the protein in the G-CSF-PEG conjugates, evaluated using circular dichroism and biological activity assay in cell culture, was maintained with respect to the native protein. Unexpectedly, conjugation enhanced the G-CSF tendency to aggregate, a problem that was overcome by a proper formulation.

INTRODUCTION Granulocyte colony stimulating factor (G-CSF), the major regulator of granulopoiesis in vivo, represents a pharmaceutically useful protein that stimulates proliferation, differentiation, and survival of cells of the granulocyte lineage (1, 2). E. coli expressed G-CSF differs from the natural material in its lack of glycosylation and the N-terminal methionyl residue incident to bacterial expression (3, 4). Clinical use of G-CSF (reviewed in refs (5, 6)), started in 1986 with the first clinical trials in cancer patients treated with chemotherapy (7), and it is now widely used to treat neutropenia, a condition that occurs in a wide variety of disease settings, including congenital defects, bone marrow suppression following pharmacological manipulation, infection, and also in cancer patients undergoing cytotoxic chemotherapy. In 1991, the United States Food and Drug Administration approved Neupogen (Filgrastim, rh-met-huG-CSF) for use by those patients suffering from neutropenia during or after chemotherapy. Treatment with G-CSF can enable a higher-dose-intensity schedule which may allow better antitumor effects (8). It was later approved worldwide for use in bone marrow transplantation and more recently for treatment of severe chronic congenital neutropenia. Neupogen has a potential application in mobilizing peripheral blood progenitor cells for transplantation (9), but it presents a short in vivo half-life mainly due to two factors: rapid renal clearance and receptor mediated clearance by granulocytes, the last process that provides a self-regulation control on the G-CSF concentration. To overcome these shortcomings, some methods have been proposed, and PEGylation is among these. PEGylation is intended to be a chemical linking of poly(ethylene glycol) chains to a drug (10–13). PEG conjugation masks the protein’s surface and * Corresponding author. Francesco M. Veronese, Department of Pharmaceutical Science, University of Padua, via Marzolo 5, 35100 Padua, Italy. Phone +39(0)49-8275694; Fax +39(0)49-8275366; E-mail: [email protected]. † University of Padua. ‡ Bio-Ker S.r.l.

increases the hydrodynamic volume of the polypeptide, thus reducing its renal clearance, preventing the approach of antibodies or antigen processing cells, and decreasing the protein degradation by proteolytic enzymes. Finally, PEG conveys to molecules its physicochemical properties and therefore also modifies the biodistribution and solubility of peptide and nonpeptide drugs. A PEGylation form of G-CSF, Neulasta, was already introduced in the market in 2002; in this case, PEG chains of 20 kDa have been selectively linked to the G-CSF amino terminus (14). The derivative presents a better pharmacokinetic profile and increased patient compliance. Interestingly, a recent study has also highlighted that this conjugate has a lower propensity to aggregate than the native G-CSF (15). In particular, the pathway of aggregation is not changed with respect to the native G-CSF, but the kinetic rate is slowed by PEGylation. This means that, after a proper time of incubation, both native G-CSF and PEG–G-CSF reach the same level of aggregation (15). The process involves first a conformational change that forms an aggregation-prone protein structure (16), which then leads to an initial noncovalent aggregation followed by the formation of covalent dimers and oligomers. These covalently linked forms are due to disulfide bridging and shuffling, which are promoted by the lone free cysteine in the G-CSF sequence (15, 16). As a confirmation, a G-CSF mutant obtained by switching the Cys17 with an Ala residue still forms aggregates, but only noncovalent ones. Starting from these data, we tried to combine the beneficial effects obtained with G-CSF PEGylation (15), i.e., the reduction in aggregation kinetic rate, and the elimination of the free thiol of Cys17 that promotes the formation of covalent aggregates (16). To reach this goal, we performed the PEGylation of G-CSF selectively at the level of Cys17 using thiol-reactive PEGs. The lone free Cys17 in rhG-CSF, the other four being involved in two disulfide bridges (Cys36-Cys42, Cys64-Cys74), is not accessible in the protein native folding. In fact, Arakawa reported that this cysteine cannot be carboxymethylated by a hydrophilic agent, iodoacetic acid, under native conformation, but it becomes fully reactive when rhG-CSF is denatured (17). In other experiments, the free thiol has been shown to react very slowly with a

10.1021/bc070123+ CCC: $37.00  2007 American Chemical Society Published on Web 10/18/2007

Thiol Site-Specific Pegylation of G-CSF

hydrophobic modifying agent, dithiobis-nitrobenzoic acid (DTNB) under nondenaturing conditions. It is postulated that the free cysteine may reside inside a hydrophobic environment of the protein, and therefore, it is inaccessible to chemical modification by hydrophilic agents, whereas it is slightly more accessible to the hydrophobic ones. Conformational studies of G-CSF have also demonstrated that the –SH at position 17 is buried in a cavity of the protein, only partially exposed to solvent. Therefore, we looked for conditions of specific conjugation between a single chain of PEG and the Cys17 of G-CSF, once this is unmasked by partial denaturation. Furthermore, we tested the different reactivity of three thiol-reactive PEGs: the vinylsulfone–PEG (PEG–VS), the orthopyridyldisulfide–PEG (PEG– OPSS), and the maleimide–PEG (MAL–PEG). Different polymer molecular weights were investigated also. As a continuation of a preliminary report (18, 19), in this paper we describe in more detail the studies involved in the development of this new PEGylation strategy and its application in the preparation of new active PEG–G-CSF conjugates, where only one chain of PEG is bound to the Cys17. Unexpectedly, the obtained conjugates showed a more marked tendency to form soluble aggregates in solution when compared to native G-CSF. Preliminary formulation studies demonstrated that the addition of albumin or β-hydroxypropylcyclodextrin can overcome this aggregation propensity.

MATERIALS AND METHODS General Methods. A lyophilized powder of rh-G-CSF was supplied by Bioker (Pula, Cagliari, Italy). PEG–OPSS, PEG–VS, and PEG–MAL of different molecular weights were purchased from Nektar (Huntsville, AL). 3-(4,5-Dimethylthiazol)-2-yl-2,5diphenyl tetrazolium bromide (MTT), 5,5′-dithiobis-(2-nitrobenzoicacid) (DTNB), solvents and denaturing agents, and other chemicals of analytical grade were from Sigma (St. Louis, MO). RPMI medium, fetal bovine serum, and penicillin were from Gibco (Pailey, UK). Reversed-phase chromatography was performed with a Pharmacia LKB (Gromma, Sweden) analytical HPLC system, using a Vydac C4 column (150 × 4.6 mm, particle size 5 µm) or a Phenomenex C18 column (250 × 4.6 mm, particle size 5 µm). Elution was carried out at a flow rate of 1 mL/min with a linear gradient of acetonitrile in water containing 0.05% (v/v) TFA from 40% to 70% over 25 min. The effluent was monitored by recording the absorbance at 226 nm. SDS-PAGE was carried out in a vertical slab gel apparatus (Mini-Protean II, Bio-Rad) using an SDS-polyacrylamide gel with a low degree of cross-linking (60% acrylamide, 0.8% bisacrylamide). The gels were stained with Coomassie brilliant blue R-250. Far-UV circular dichroism (CD) measurements were made at 25 °C on a Jasco J-710 spectropolarimeter (Tokyo, Japan) equipped with a thermostatically controlled cell holder. The instrument was calibrated with D-(+)-10-camphorsulfonic acid. The spectra were recorded in 10 mM acetic acid, 10 mg/mL mannitol, pH 4, at a protein concentration of 0.1 mg/mL, in 1 mm path length quartz cells. Far-UV CD spectra were analyzed in order to estimate the percentage of protein secondary structure from the equation % R-helix ) (θ208nmobs – 4000)/(33 000 – 4000) (20), where θ208 nm,obs is the observed mean residue ellipticity at 208 nm. Protein concentrations were determined spectrophotometrically on a Perkin-Elmer Lambda-20 spectrophotometer. The concentrations of stock solutions of G-CSF and G-CSF–PEG conjugates were evaluated from their absorbance at 280 nm. Extinction coefficients at 280 nm for G-CSF (0.88 mg-1 cm2) were considered unchanged for the G-CSF–PEG conjugates. Fluorescence emission spectra were recorded at 25 °C on a Perkin-Elmer model LS-50B spectrofluorometer, equipped with

Bioconjugate Chem., Vol. 18, No. 6, 2007 1825

a thermostatically controlled cell holder. Spectra were taken in Tris/NaCl buffer, pH 7.5, using 1 cm path length, exciting the samples (1.3 µM) at 295 nm, and recording the emission fluorescence in the wavelength range 303–500 nm, in the presence of different amount of positive (CsCl) and negative (NaI) quenchers. MALDI Mass Spectrometry. Measurements were performed on a REFLEX time-of-flight instrument (Bruker-Franzen Analytik, Bremen, Germany) equipped with a SCOUT ion source operating in positive linear mode. Ions, formed by a pulsed UV laser beam (nitrogen laser, λ ) 337 nm), were accelerated to 25 kV. Sinapinic acid (saturated solution in acetonitrile/water (50:50 v/v)) was used as a matrix. Samples were dissolved in 0.1% trifluoroacetic acid aqueous solution. Five microliters of the diluted sample solution was mixed with the same volume of matrix solution, and 1 µl of the resulting mixture was deposited on a stainless steel sample holder and allowed to dry before introduction into the mass spectrometer. Cell Culture. The NSF-60 cell line, which is developed from a mouse myeloblastic leukemia rich in G-CSF receptors, was kindly provided by Dr. William E. Evans St. Jude Children’s Research Hospital (USA). Cells were maintained in RPMI medium with 5% fetal bovine serum, 10 U.I./mL penicillin, and 30 U.I./mL Filgrastim at 37 °C, under 5% CO2 atmosphere. G-CSF PEGylation under Nondenaturing Conditions. Thiol-reactive PEGs (PEG–OPSS, PEG–VS, or PEG–MAL) were employed. The reactions were carried out in 0.5 M Tris buffer, pH 7.2. At room temperature, one of the active PEGs was added as a powder in a 10-fold molar excess to a rhG-CSF solution (1 mg/mL). The reaction was allowed to proceed for 24 h or in any case until reaching a constant amount of PEG–GCSF conjugate, which was then analyzed and purified by reverse-phase chromatography according to the conditions reported in Methods and Materials. The purity of the conjugate and the degree of conjugation were determined by SDS-PAGE and DTNB colorimetric assay (21). Control of the Reversibility of G-CSF Denaturation. Solutions of rhG-CSF, at 0.15 mg/mL, were prepared and incubated for 4 h in 2, 3, or 4 M guanidinium chloride (Gu · HCl) in phosphate buffer 0.05 M, pH 7.2. A 5 M solution of urea in the same buffer was also tested as a denaturant. The protein secondary structure recovery was analyzed after dialysis, against sodium acetate 10 mM pH 4.2, to remove the denaturant. The secondary structure was investigated by far-UV CD analysis. G-CSF PEGylation under Denaturing Conditions. To G-CSF (1.46 mg/mL), dissolved in one of the denaturant solutions reported above, PEG–OPSS (5 kDa), MAL–PEG (5 kDa), or PEG–VS (5 kDa) were added at 10-fold molar excess with respect to the protein. The reaction was allowed to proceed at room temperature, under argon atmosphere, for 4 h. Alternatively, PEG–OPSS at 10 or 20 kDa was used. In this case, the reaction was conducted, under argon atmosphere, for 20 h when using the 10 kDa PEG or at 40 °C for 1 h followed by for 4 h at room temperature when using the 20 kDa PEG. The reaction mixture was dialyzed against sodium acetate 10 mM pH 4.2, and the conjugate was purified by reverse-phase HPLC using the conditions reported above. The secondary structures of the products were analyzed by far-UV CD. Evaluation of in Vitro G-CSF Conjugate Activity. NSF60 cells were seeded in 96 well plates (104 cells/well), and serial dilutions of G-CSF and G-CSF–PEG conjugates were added. The plates were incubated for 67 h at 37 °C before the addition of 20 µL MTT (5 mg/mL in PBS filtered using 0.22 µm filters) to each well. After a further 5 h incubation, the media were removed and optical-grade dimethyl sulfoxide (100 µL) was added to dissolve the formed insoluble formazan dye. Absorbance at 540 nm was measured using an Elisa reader, and the

1826 Bioconjugate Chem., Vol. 18, No. 6, 2007

Veronese et al.

Figure 1. RP-HPLC chromatograms of G-CSF conjugation with thiolreactive PEGs under physiological conditions after 24 h. (A) reaction mixture of G-CSF and PEG–OPSS (5 kDa); (B) reaction mixture of G-CSF and PEG–VS (5 kDa).

Figure 2. Far-UV CD spectra of native G-CSF in physiological condition (×) and after the denaturation–refolding process. Denaturation was performed using 3 M (2) or 4 M (9) solutions of guanidine HCI or 5 M (() urea. The spectra were obtained at 25 °C in acidic solution at pH 4.2.

cell viability was expressed as a percentage of the viability versus untreated control cells.

RESULTS AND DISCUSSION G-CSF PEGylation under Nondenaturing Conditions. The direct site-specific conjugation of a linear PEG to the thiol of G-CSF Cys17 was investigated by carrying out the reaction under conditions that maintained the G-CSF tertiary structure. Thiol-reactive PEGs based on double-bond addition (PEG–VS and PEG–MAL) or disulfide exchange (PEG–OPSS) were employed. The reaction was allowed to proceed at room temperature for several hours, until reaching a constant amount of PEG–G-CSF conjugate, as verified by RP-HPLC. In Figure 1 is the RP-HPLC profile of the conjugation mixture of G-CSF and PEG–OPSS (Figure 1A) or PEG–VS (Figure 1B) after 24 h. The degree of conjugation of G-CSF was very limited, thus confirming the difficulty of reaching the buried Cys17. Under these conditions, PEG–OPSS appeared to be slightly more reactive than PEG–VS. PEG–MAL showed a similarly low degree of conjugation (data not shown). Reversible Denaturation of G-CSF. To overcome the low reactivity of the thiol group of G-CSF in the native state, we studied the conjugation under partially denaturing conditions. For this purpose, urea or Gu · HCl was used as a denaturing agent, while the native protein’s disulfide bridges were maintained in order to facilitate the refolding, to prevent PEGylation at unwanted sites, and to avoid S–S scrambling during the refolding.

Figure 3. Reaction mixture of G-CSF PEGylated under denaturing conditions. RP-HPLC chromatograms had been obtained using a C4 Vydac column and recorded at 226 nm: (A) native G-CSF; (B) reaction mixture of G-CSF and PEG–OPSS (5 kDa), after 4 h, in the presence of 3 M urea; (C–E) reaction mixture of G-CSF and PEG–OPSS (5 kDa), after 4 h, in the presence of 2, 3, or 4 M guanidine · HCl, respectively. Table 1. Degree of G-CSF PEGylation after 4 h Using Different PEG Reagents under Different Denaturing Conditions percentage of G-CSF conjugation using different denaturing conditions PEG reagent

2M Gu · HCl

3M Gu · HCl

4M Gu · HCl

3M urea

PEG–OPSS (5 kDa) PEG–VS (5 kDa) PEG–MAL (5 kDa)

48 15 44

84 30 81

95 50 91

31 11 29

A preliminary study was carried out to evaluate the refolding process of G-CSF after its exposure to different denaturant concentrations. In Figure 2, it is shown that the refolding is complete for concentrations of Gu · HCl up to 4 M. Denaturation using 5 M urea was more critical, and the refolding was incomplete. High salt concentrations interfered with the collection of CD data in the far-UV region. Therefore, spectra of protein samples containing guanidine · HCl could not be obtained below 200 nm. G-CSF PEGylation under Conditions of ReVersible Denaturation. The conjugation of G-CSF by PEG–OPSS 5 kDa, evaluated by reverse-phase HPLC, after 4 h under different denaturing conditions is shown in Figure 3. A fast and high degree of PEGylation at the lone free cysteine residue of G-CSF was successfully reached under the denaturing conditions of 3 or 4 M Gu · HCl. These conjugation data are in good agreement with the cyanylation results of G-CSF Cys17 as a function of Gu · HCl reported by other authors (16). PEG linking to denatured G-CSF using a 3 M solution of urea was less effective. Similar results were obtained using PEG–maleimide, while PEG–vinylsulfone had the lowest degree of coupling; in this case, after 4 h only 50% of conjugation was achieved in 4 M

Thiol Site-Specific Pegylation of G-CSF

Bioconjugate Chem., Vol. 18, No. 6, 2007 1827

Figure 4. RP-HPLC chromatograms of reaction mixture of G-CSF and PEG–OPSS at different molecular weights, obtained using a C18 Phenomenex column. The spectra were recorded at 280 nm. (A) G-CSF and PEG–OPSS (10 kDa) after 20 h. (B) G-CSF and PEG–OPSS (20 kDa) after 1 h at 40 °C and the following 4 h at room temperature.

of Gu · HCl (Table 1). The denaturant was removed by extensive dialysis towards acetate buffer 10 mM at pH 4.2. Protein unfolding accompanied by cysteine exposition and reactivity might also be obtained in water/organic solvent solutions. A G-CSF sample dissolved in a 80:20 water/DMSO solution showed a similar reactivity to PEG–OPSS (data not shown). This partial denaturation by DMSO could explain the PEGylation at the level of Cys17 of G-CSF reported by other researchers (22). The conjugation of PEG–OPSS at higher molecular weights, 10 or 20 kDa, was investigated in 4 M guanidinium chloride. Protein modification using these more hindered PEGs was achieved by increasing the reaction time for the 10 kDa polymer and both temperature and time for the 20 kDa PEG. After 4 h at room temperature, the degrees of conjugation were 85% and 45%, respectively (Figure 4A,B). The conjugate G-CSF–PEG (20 kDa), dialyzed and purified as reported above, was characterized by SDS gel electrophoresis (Figure 5). The conjugate appeared as a single band at ∼62 kDa, which does not reflect the sum of G-CSF and polymer molecular weights, because as is known, PEG has a higher hydrodynamic volume than a protein with the same molecular weight (23). The G-CSF–PEG conjugates were further characterized by MALDI-TOF mass spectroscopy, far-UV CD, fluorescence emission and finally by cell proliferative assay. Structural Characterization of the PEGylated G-CSF. The covalent conjugation was confirmed by MALDI-TOF mass spectroscopy. A broad peak, due to PEG polydispersity, centered at 24 089 kDa, was detected following G-CSF conjugation with PEG 5 kDa. This corresponds to a mono-PEGylated G-CSF (Figure 6A,B). The near- and far-UV CD spectra of native and PEGylated G-CSF, obtained using mPEG–OPSS 5 kDa, are shown in Figure 7A and B. The observed superimposition of the spectra demonstrates that refolding of PEGylated protein to the native structure took place after PEG linking also. Similar spectra were obtained with G-CSF–PEG–VS and G-CSF–PEG–MAL. Fluorescence spectra of native and conjugated G-CSF showed a red shift in λmax in the emission spectra, from 344 to 350 nm,

Figure 5. SDS gel electrophoresis of the following: lane 1, low molecular weight standards; lane 2, native G-CSF; lane 3, purified G-CSF–PEG–OPSS (20 kDa).

indicating that tryptophans 58 and 118 are more exposed to solvent in the conjugated protein. These data were confirmed by the tryptophan fluorescence quenching with CsCl and NaI of native G-CSF, G-CSF–PEG–OPSS, and G-CSF–PEG–VS (data not shown). The major solvent exposition of tryptophans in the PEGylated protein can be due to the influence of the highly hydrophilic PEG chain in the surrounding of this amino acid. In Vitro Cellular Proliferative Activity of G-CSF Conjugates. The activity of conjugated and native G-CSF was evaluated by analyzing the proliferation of a sensitive cell line, which has been developed from a mouse myeloblastic leukemia rich in G-CSF receptors: NFS-60 cells. All of the protein species were found to stimulate cell proliferation, and the final cell

1828 Bioconjugate Chem., Vol. 18, No. 6, 2007

Veronese et al.

Figure 6. MALDI-TOF mass spectroscopy of (A) native G-CSF; (B) G-CSF–PEG–OPSS (5 kDa); the conjugate was purified from the reaction mixture by RP-HPLC chromatography.

concentration was found to be proportional to the concentration of G-CSF or G-CSF–PEG conjugates present in the cell culture. This allows quantification of the biological potency of GCSF–PEG conjugates versus a G-CSF reference preparation, applying a dose–reaction ratio. The IC50 values of two G-CSF–PEG conjugates, both prepared using a 5 kDa PEG, are reported in Table 2. The first conjugate was obtained by thiol-exchange reaction using PEG–OPSS and the second by double bond addition using PEG–VS. Both conjugates maintained high biological activity; in particular, the G-CSF–PEG–OPSS was more active than G-CSF–PEG–VS. This may be ascribed to the fact that the more

hindered sulfone group of PEG–VS has a negative effect on maintenance of the native protein structure. Stability of PEGylated G-CSF in Solution. Accelerated stability studies of G-CSF were performed by incubating at 37 °C the PEGylated protein solutions. The aggregation was evaluated on the basis of UV absorption and gel filtration chromatography of the incubated solutions. Figure 8 illustrates that the PEGylated forms unexpectedly undergo more rapid aggregation than the native G-CSF. This result is in contrast to the known drug solubilizing and aggregation prevention properties of PEG. Furthermore, this does not reflect the observations with Neulasta in which PEGylation slowed the rate of aggregate

Thiol Site-Specific Pegylation of G-CSF

Bioconjugate Chem., Vol. 18, No. 6, 2007 1829

Figure 8. Aggregation profiles of native G-CSF, G-CSF–PEG–OPSS (5 kDa), G-CSF–PEG–OPSS (5 kDa) in 50 mM hydroxypropyl β-cyclodextrin, and G-CSF-PEG-OPSS (5 kDa) in albumin 0.1% (w/ v) solution. The samples, all containing 1 mg/mL in G-CSF protein equivalent, were incubated at 37 °C and analyzed at predetermined times by gel filtration chromatography.

Figure 7. (A) Near-UV CD spectra of native G-CSF, G-CSF–PEG– OPSS (5 kDa). (B) Far-UV CD spectra of native G-CSF, GCSF–PEG–OPSS (5 kDa). UV-CD spectra were taken at 25 °C in 10 mM acetic acid, 10 mg/mL mannitol, pH 4.2. Table 2. In Vitro Activity in NSF-60 Cells of Native and Conjugated G-CSF sample

IC50 (pmol/mL)

relative activity

rhG-CSF G-CSF–PEG–OPSS (5 kDa) G-CSF–PEG–VS (5 kDa)

0.0096 0.0145 0.039

1 0.66 0.246

formation (15). On the other hand, native G-CSF and Neulasta finally lead to covalent aggregates, which are formed as a consequence of disulfide scrambling promoted by the free thiol of Cys17 (15). As expected, in our study we did not observe covalent aggregates, since by SDS treatment it was possible to form the monomeric conjugate again. Anyway, preliminary formulation studies demonstrated that it is possible to stabilize the conjugates by adding hydroxypropyl β-cyclodextrin or albumin to G-CSF–PEG solution (Figure 8).

CONCLUSION We have demonstrated the possibility to get site-specific PEGylation at the level of buried and less-accessible thiol groups of G-CSF cysteine 17. This was achieved by taking advantage of the quantitative reversible G-CSF unfolding by chemical denaturants and maintaining unaltered disulfide bridges to avoid disulfide scrambling. The conjugation was straightforward in the case of 5000 Da PEGs but needed longer time or increased temperature for the higher-molecular-weight PEG (10 and 20 kDa). The structure of the thiol-reactive moiety in the PEG chain plays a role in the biological activity of the final product. In fact, the conjugate G-CSF–PEG–OPSS retains higher activity than the G-CSF–PEG–VS. This may be because PEG, and in particular the steric hindrance of the thiol-reactive moiety, may induce some subtle local changes in the G-CSF threedimensional structure. This phenomenon was slightly highlighted by fluorescence studies, although it was not evident by CD

analysis. Furthermore, this can be correlated with the increased tendency of PEGylated G-CSF in solution to aggregate with respect to the native protein and the N-terminus PEGylated G-CSF (15). The different effect on G-CSF aggregation of this PEGylation strategy can be ascribed to the site of PEG linking. The polymer here is attached to a buried amino acid, and this can cause conformational change that enhances the rate of protein aggregation. We speculate therefore that the faster aggregation may be a consequence of protein hydrophobic residue exposition. In fact, the addition of hydroxypropyl β-cyclodextrin, which is known to mask hydrophobic residues, reduces the aggregation rate (Figure 8). On the other hand, the PEGylation at the level of Cys17 has been performed with the aim of avoiding the formation of covalent aggregates by thiol bridging, a problem faced by both the native and N-terminus PEGylated G-CSF (15). As expected, the aggregates, formed by G-CSF PEGylated at the cysteine, are in fact not covalent, and their formation can be totally prevented by a proper formulation by, for instance, using albumin (Figure 8). Therefore, this type of G-CSF–PEG has the advantage of overcoming the problem of covalent aggregation, a drawback presented by both native G-CSF and Neulasta. This research highlights the possibility of exploiting a protein reversible denaturation for allowing a more selective PEGylation or conjugation of buried residues. Other conditions, which perturb the protein structure, can also enhance the thiol accessibility of G-CSF. In fact, preliminary experiments conducted in our laboratory showed that a complete G-CSF PEGylation at the level of Cys17 could also be achieved by PEG–OPSS during a very short (