From Amgen Incorporated, Amgen Center, Thousand Oaks, California 91320. The folding and oxidation of recombinant human granulocyte colony-stimulating ...
Vol. 267. No. 13,Issue of May 5, pp. 8770-8777, 1992 Printed in U.S.A.
THEJOURNAL OF BIOLOGICALCHEMISTRY 01992 by The American Society for Biochemistry andMolecular Biology, Inc.
Folding and Oxidation of Recombinant Human Granulocyte Colony Stimulating Factor Produced in Escherichia coli CHARACTERIZATION OF THE DISULFIDE-REDUCED INTERMEDIATES AND CYSTEINE ANALOGS*
+
SERINE
(Received for publication, October 14, 1991)
Hsieng S. Lul, Christi L. Clogston, Linda 0. Narhi, Lee Anne Merewether, Wayne R. Pearl, and Thomas C. Boone From Amgen Incorporated, Amgen Center, Thousand Oaks, California 91320
The folding and oxidation of recombinant human species, since they are usually short-lived. As described by granulocyte colony-stimulatingfactor solubilized from Creighton and others (3-7), a model system that includes an Escherichiacoli inclusion bodies was investigated. oxidative refolding of a disulfide-reduced protein allows one During the folding process, two intermediates, I1 and to investigate the pathway of protein folding in detail. There Iz,were detectedby kinetic studies using high performance liquid chromatography. I1 exists transiently and are two possible advantages in investigating a disulfide-condisappears quickly with the concomitant formation of taining protein. First, the intrachain S-S bond is a natural 12. In contrast,IZ requires a longer time to fold into the covalent cross-link which is closely correlated with protein final oxidized form, N. CuS04 catalysis increases the conformation, and the intramolecular disulfide bond formafolding rate of I2 from 11,while CuS04 and elevated tion reflects the proximity of two relevant sulfhydryls in temperature (37“C) have a dramatic effect on the folding rate of N from Iz. These observations suggest the intermediate forms (3-5). Second, proteins in disulfide-reduced states are not shortlived allowingthe intermediates to following sequential oxidative folding pathway. be trapped by chemical modification during folding. Human granulocyte colony-stimulating factor (hG-CSF)’ is ‘1 rate-limiting ’N one of the hemopoietic growth factors which plays an imporPeptide map analysis of the iodoacetate-labeled intant role in stimulating proliferation, differentiation, and termediates revealed that I1 represents the fully refunctional activation of bloodcells (8). Human G-CSF is duced granulocyte colony-stimulating factor containcapable of supporting neutrophil proliferation in vitro and in ing 6 free cysteines; I2is the partiallyoxidized species containing a single C y ~ ~ ~ - disulfide C y s ~ ~bond; and N, vivo (9,lO). Thehuman and murine G-CSF genes have been the final folded form, has two disulfide bonds. The cloned and characterized and are about75% homologous(11, physicochemical properties and biological activities of 12). Large quantities of recombinant hG-CSF (rhG-CSF) have 11,Iz,N, and several Cys + Ser analogs made by site- been produced in genetically engineered Escherichia coli and directedmutagenesis werefurther investigated. In have been successfullyused in human clinical studies to treat guanidine hydrochloride-induced denaturation studcancer patients suffering from chemotherapy-induced neutroies, the disulfide-reduced intermediates and the ana- penia (13-15). E. coli-produced rhG-CSF is a 175-amino acid logs missing either of the disulfide bonds are confor- polypeptide chain containing an extra Met (at position -1) mationally less stablethan those of the wild type molat its NH2 terminus. The molecule also contains a free cysecule or the analog with the free Cys at position 17 teine at position 17 and two intramolecular disulfide bonds, changed to Ser. Recombinant human granulocyte colC y ~ ~~-C and y sC~~~s ‘ j ~ - C y(16). s ~ ~The twodisulfide bonds ony stimulating factor lacking either disulfide bond or both has overall secondary and tertiary structures dif- form two small loops which are separated by 21 amino acids. ferent from those of the wild type molecule and exhibits Like other bacteria-derived recombinant proteins, rhG-CSF lower biological activity. These studies show that di- produced in E. coli requires an oxidative folding procedure in sulfide bond formation is crucial for maintaining the order to recover its biological activity (17). In this paper, we molecule in a properly folded and biologically active describe the kinetic study of a folding and oxidation procedure form. for thereduced rhG-CSF solubilized from the inclusion bodies as well as the isolation and characterization of the disulfidereduced intermediates. To establish the role of disulfide bond formation in the folding of biologically active rhG-CSF, we The mechanistic study of protein folding is important in also describe the biological and physicochemical characterizaunderstanding the structure and function of proteins (1, 2). tion of intermediates and analogs made by site-directed muThe difficulty in elucidating a protein folding pathway lies in tagenesis at the Cys residues. measuring the structural properties of intermediate protein
fast‘’
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence should be addressed.
The abbreviations used are: hG-CSF, human granulocyte colonystimulating factor; rhG-CSF, recombinant human granulocyte colony-stimulating factor; GdnHC1, guanidine HCI; RP-HPLC, reversephase high performance liquid chromatography; DTT, dithiothreitol; TFA, trifluoroacetic acid DTNB, 5,5’-dithiobis(nitrobenzoicacid).
8770
of
folding Oxidative
N
hG-CSF
8771
half-maximal conversion of I p to N takes almost4.5 h. Kinetic studies also indicate that the oxidative foldingof form N from I, is biphasic. The folding rate from I2 to N during the first phase is faster. Thesecond phase of folding starts at approxt imately 8 h with a much slower rate. The folding kinetics of rhG-CSF at 25 and 37 "C without addition of copper sulfate were also investigated. As listed in Table 1, the initial ratesfor Iz formationa t both 25 and 37 "C are similar(1.0 x lo-' s-') and slightly slower than thefolding of rhG-CSF in the presenceof copper sulfate (1.9 x 10" s-'). However, a t 25 "C the generationof completely oxidized rhGs-'). In this CSF (form N) is relatively slow (rate = 6.6 X case, I, persists muchlonger andapproximately 20 h are required to reach half-maximal folding of form N from Ip versus 4.5 h in the presence of copper sulfate at 25 "C. At 1 37 "C the folding of form N from Ip without CuS04 is faster, 42 50 52 but the biphasic kinetics becomes apparent. The first phase Retention Time Wut no) of oxidation takes about 5 h, while the second phase takes FIG. 2. Folding of rhG-CSF at 25 "C in CuSO, monitored by place at approximately 8 h. After 23 h, the recovery of comRP-HPLC at different times. Chromatograms 1-6,20 min, 1 , 2 , 4 , 8,and 12 h, respectively. Approximately 50 pg of rhG-CSF in folding pletely oxidized rhG-CSF reachesonly about 80%. Isolation and Structural Characterization of Intermediatesmixture was injected. Intermediates I1 and I, and the final oxidized form N are indicated. Note that retention times shown in Fig. 2 are An alkylating agent, iodoaceticacid, was used to trap the slightly different from those in Fig. 1 due to the use of different C-4 intermediates that may be disulfide-reduced. The resulting columns. carboxymethylated derivatives contain more negatively charged carboxymethyl groups than the oxidized rhG-CSF MATERIALS AND METHODS' and are separable by ion-exchange HPLC using a sulfoethyl polyaspartamide silica-based column (data not shown). RESULTS To estimate the stoichiometry of labeling, the modification was also performed using [3H]C2-iodoacetic acid.Table 2 lists Folding Intermediates and Folding Kinetics of rhG-CSFFig. 1 shows the RP-HPLC elutionprofiles of the native and the labeling results for rhG-CSF and the purified intermediof 6 M GdnHC1, native rhGthe fully denatured andreduced rhG-CSFs (chromatograms 1 ates. In the absence and presence and 2, respectively). Their retention times differ by approxi- CSF gives 0 and 1 mol of label/mol of protein, respectively, mately 2.1-2.5 min. Fully denatured andreduced rhG-CSF in consistent with our previous observation (16). The fully regives 5 mol of label. The 6 M GdnHCl is retained slightly longer and elutes as a single duced anddenaturedrhG-CSF trapped I1 gives 3.74 mol of label and I2 1.9 mol of label/mol sharp peak. Fig. 1 also shows that rhG-CSF present in the crude E. coli lysate solubilized in either GdnHCl or Sarkosyl of protein. These results indicate that 4 and 2 free cysteinyl at 2-5 "C elutes as broader peaks (chromatograms 3, and 4, residues are available forthe labeling in I, and Iz,respectively. T o further characterize the structure of intermediates, the respectively) at slightly earlier retention times than thefully I, and I2 were reduced with DTT in 6 M GdnHCl 3H-labeled denatured andreduced rhG-CSF. Thedifference in the chroand then carboxymethylated with non-radioactive iodoacetate matographic elution times among these rhG-CSF forms has allowed us to detect folding intermediates and to study folding to generate fully denatured and alkylated derivatives, which were then subjected to HPLC peptide mapping. Fig. 3 shows kinetics by RP-HPLC. Shown in Fig. 2 are the RP-HPLC chromatograms of the a typical peptide map derived from theStaphylococcus aureus solubilized rhG-CSF samples prepared at different refolding V-8 protease digestion of the I2 derivative. Peptide fractions times during incubation at 25 "C in the presence of CuS04. were pooled and aliquots were analyzed for determination of T h e populations of the three major rhG-CSF-related species peptide concentration, radioactivity counting, and NHz-terwhich elute at retention times around 45 to 48 min change minal sequence analysis. Table 3 summarizes the isolation dramatically asa function of time. Intermediate I, as depicted and characterization of the labeled and unlabeled peptides in Fig. 2 (chromatogram 1 ) is the starting reduced rhG-CSF. derived from I1 and 12. For form Ip, only peptides 7 and 8 to Glug3 andto Glug8, respectively) containing At the 20-min incubation time, Iz has already accumulated to (Leu47 and Cys74were radioactively labeled. Sequence analysis a level of 33% of the total rhG-CSF. At 1, 2, and 4 h, the Cys74were labeled, supporting generation of I2 has proceeded further with the concomitant confirmed thatbothand disappearance of I, and appearanceof the finaloxidized form the quantitative data that 2 labeled cysteines are present in N (Fig. 2, chromatograms 2-4). The folding into formN I2 (Table 2). For form 11, C y d 4 a n d C Yfound S ~ ~ , in peptides7 reaches its maximum 12 in hand is greater than 95% complete and 8, as well as and Cys4' inpeptide 2 ( L Y s ~ ~Gt o~ u ~ ~ ) (Fig. 2, chromatogram 6). 4 cysteines were radioactively labeled. This confirms that the The initial first orderfolding rate for conversion of I1 into a t positions 36,42,64, and 74 in I1 are notinvolved in disulfide I, was estimated to be 1.9 X lo-' s" and the initial rate for bonding. conversion of Ip into the finaloxidized rhG-CSF (form N) to As indicatedinTable 3, peptide 4 is the NH,-terminal beapproximately3.1 x s-' (Table 1). Thehalf-maximal peptide of rhG-CSF containing Cys at position 17. Analysis conversion of II to I2 takes approximately30 min, while the of peptide 4 derived fromboth intermediatesI1and 1,indicated that no radioactive label was present. The data suggest that, Portions of this paper (including "Materials and Methods," Figs. like the native rhG-CSF, both intermediates contain an in1, 3-7, and 9, and Tables 1-4) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard accessible free cysteine a t position 17. Preparation and RP-HPLC Analysis of rhG-CSF Analogsmagnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. The Cys + Ser analogs madeby site-directed mutagenesisof 44
46
48
of
8772
folding Oxidative
hG-CSF
the rhG-CSF gene were recovered by the procedures developed 208 nm. All ofthe molecules tested exhibit a-helical structure, S~~,~~ for rhG-CSF including folding and chromatographic separa- but the native molecule and the ~ ~ G - C S F [ C+~ Ser36,42] tion. rhG-CSF[Cys17+ Ser17]exhibited oxidation and folding molecule contain substantiallyhigher helical content than the similar to those of the wild type rhG-CSF andcould be isolated other species examined, i.e. 64% helix uersus 38% helix at with equivalent recovery. Incontrast, ~ ~ G - C S F [ C ~+ S ~ neutral ~ , ~ * pH, calculated using the Greenfield-Fasman equation Ser36-42] exhibited very slow folding in the absence of CuS04 (19). The native and ~ ~ G - C S F [ C ~ +SSer36s42] ~ ~ * ~ are ' the only (greater than 4 days) and was recovered in lowyield. The species that show an increase in helicity (up to 75%) at pH [ C Y S+ ~ ~Ser74]analog lacking the C y ~ ~ ~ - disulfide C y s ~ ~bond 3.5 (Fig. 7B). The far UV CD spectra of I1 and I p with or also folded moderately slowly but correctly in the absence of without iodoacetate modification are identical. CuS04 (more rapidly but incorrectly in the presence of As seen in Fig. 8A, the fluorescence emission spectrum of CuS04),but recovered in low yield. the rhG-CSF molecule at neutral pH is characterized by a Fig. 4A shows the elution of purified r h G - S c F [ c y ~+ ~ ~ single peak with a maximum at 344 nm, typical of a somewhat Ser74]from a C-4 reverse-phase HPLC column. It elutes 2.20 solvent-exposed Trp (20). Thereis no detectable Tyr fluoresmin (chromatograms2 and 3 at pH 3 and7, respectively) later cence (around 300 nm). The spectra of the two analogs are than thewild type rhG-CSF (chromatogram 1). I2 lacking the similar to thatof the native rhG-CSF, although the intensities C y ~ ~ ~ - disulfide C y s ~ ~bond elutes essentially at the same re- are greater, indicating that the Trp fluorescence might be tention time as the Cys74+ Ser74analog, suggesting that the somewhat less quenched by the surrounding environment. molecules have similar hydrophobicities. Recombinant hG- The spectra ofI1 and I2 differ slightly from those of the CSF[C~S+ ~ ~Ser36*42] * ~ * (Fig. 4B, chromatogram 4 ) elutes only respective labeled derivatives; the peaks are broader, again 0.33 min later than wild type rhG-CSF. reflecting a difference in the environment of the 2 Trp resiBiologicalActivity and Physicochemical Properties of the dues. rhG-CSF, Intermediates and Analogs-Table 4 lists the in Fig. 8B shows the fluorescence spectra at pH 3. The Trp uitro biological activities of rhG-CSF, folding intermediates, peak of native rhG-CSF is still evident, but is greatly deand analogs. The wild type rhG-CSF standard has an activity creased in intensity (the absolute scale of this figure is less of approximately 1.0 X 10' units/mg. Carboxymethylated than that in A), and a peak at 304 nm, attributable to Tyr, is intermediates I1and I p have only approximately 3 4 % activity also present. This suggests that the molecule has undergone of the wild type rhG-CSF. Both ~ ~ G - C S F [ C+ ~ Ser36.42] S~~,~~ a reversible change in conformation so that energy transfer also exhibit very low activity (1 from Tyr to Trp no longer occurs. However, such change and r h G - c S F [ c y ~+ ~ Ser74] ~ and 3%, respectively relative to the wild type molecule). In occurs only for the native molecule, since the other species contrast, rhG-CSF[Cys17+ Ser17]exhibits full in uitro biolog- show a decrease in the intensity of the Trp fluorescence, but ical activity. no change in the Tyr fluorescence. The spectrum of the rhGConformational stabilities of rhG-CSF intermediates and + has a shoulder at 304 nm, indicating CSF[C~S~ ~ Ser36*42] ,~' analogs were compared by denaturation with GdnHCl at pH a slight acid-induced conformational change. 7.2.Fig. 5A shows the absorbance spectra of nativeand Fig. 9 shows the hydrodynamic behavior of the native denatured rhG-CSFs. It appears that GdnHCl denaturation molecule, the Cys74+ Ser74analog,and Cys36*42 + Ser36.42 results in a blue shift of the UV spectrum, reflecting increased analog as determined by gel filtration. Recombinant hG-CSF exposure of the aromatic amino acids to the polar aqueous is a very compact molecule ( M , = 15,000 uersusthe expected solvent. The absorbance difference at 290 nm can thus be 18,800). While both analogs still elute with an apparent on the molecular weight smaller than expected, they both elute earused to estimate the effect of adenaturingagent conformational stability of rhG-CSF. From the results shown lier than the native molecule, indicating that without either in Fig. 5B, rhG-CSF appears to stay in the native state below disulfide bond, the molecule behaves somewhat larger, prob2 M GdnHCl concentration, denatures quickly above 2.5 M ably due to an increase in flexibility. denaturant, and is completely unfolded at 3.5 M GdnHCl. DISCUSSION This denaturation profile approximates a simple two-state transition. The midpoint of denaturation is at approximately The present study demonstratesthat thefolding of reduced 3 M GdnHCl (Table 4). Also indicated in Fig. 5B is the rhG-CSF proceeds through identifiable intermediates I1 and accessibility of the free Cyd7 atdifferent GdnHCl concentra- I2 (Fig. 2). The originally fully reduced (I1), partially oxidized tions as determined by DTNB reaction. The increased acces- (I2), and the final oxidized (N) forms can be separated and sibility of CysI7 is coincident with the denaturation of rhG- quantitated by HPLC for detailed kinetic studies. The disulCSF detected by the change in absorbance at 290 nm. fide-reduced intermediates as well as the disulfide-unpaired As shown in Fig. 6, similar GdnHCl denaturation studies analogs appear to behave more hydrophobically than the ~ ~Ser74],and native rhG-CSF. were also performed on I], 12, r h G - c S F [ c y ~+ rhG-CSF[Cys"l4* Ser36*42]. The denaturation transition ocCopper ion and other trace metals have been reported to curs at lower GdnHCl concentration for the intermediates catalyze air oxidation of proteins due to their ability to acceland analogs. A further decrease in absorbance was observed erate thiol oxidation at concentrations ranging from 0.1 and to proceed at higher GdnHCl concentration. The concentra- 10 PM (21-24). The optimal Cu2+concentration for rhG-CSF tions of GdnHCl that are required to achieve a midpoint oxidation is in the range of 20-40 p M (17). The kinetic data denaturation for the intermediates and analogs range from show that copper ion increases the folding rate of intermediate 1.4 to 2.1 M (Table 4). The thermodynamic constant for rhG- IQfrom I1 approximately 2-fold and increases the folding rate CSF is 5.4 Kcal/mol, a value typical of a folded globular of N from I2approximately &fold at 25 "C. Cu2+also promotes protein while all of the disulfide-reduced intermediates and oxidation to greater than 95% completion and shortens the analogs have values below 3 kcal/mol. second phase folding time (see "Results"). Increasing the Fig. 7A shows the far UV CD spectra of the native molecule, temperature to 37 "C can also accelerate folding and oxidation t h e r h G - c S F [ c y-+~ ~Ser74] ~ analog, the ~ ~ G - C S F [ C ~+S ~ of ~ .the ~ * final oxidized form N but does not appear to increase Ser"fi,42] analog, and thefolding intermediates at pH 7.5. rhG- the rate of the second phase of folding. The kinetic studies CSF is rich in a-helix, asevidenced by the minima at 222 and andthe detection of different intermediate folding forms
-
of hG-CSF
folding Oxidative
8773
A
FIG. 8. Fluorescence spectra of
I
I
rhG-CSF species at pH 7.6 (panel