Protoplasma (2001) 218:95-103 PROTOPLASMA �9 Springer-Verlag 2001 Printed in Austria

Production of a recombinant human basic fibroblast growth factor with a collagen binding domain

J. A. Andrades 1, J. A. Santamaria 1, L. I?. Wu 2, F. L. Hall 2, M. E. Nimni 2, and J. Becerra 1'*

l Department of Ceil Biology and Genetics, Faculty of Sciences, University of M~ilaga, Mfilaga and 2 Surgical Research Laboratories, School of Medicine, University of Southern California, Los Angeles, California

Received July 31, 2000 Accepted February 9, 2001

Summary. Basic fibroblast growth factor (bFGF) is a potent in vitro mitogen for capillary endothelial cells, stimulates angiogenesis in vivo, and may participate in tissue repair. Basic FGF is found in abundance in tissues such as brain, kidney, and cartilage. This study reports the expression, purification, and renaturation of a biologi- cally active human basic fibroblast growth factor fusion protein (hbFGF-F1) from Escherichia coli. A prokaryotic expression vector was engineered to produce a tripartite fusion protein consisting of a purification tag, a protease-sensitive linker and collagen binding domain, and a cDNA sequence encoding the active fragment of hbFGF. The expressed hbFGF-F1 and hbFGF-F2 (it contains the collagen binding domain), located in inclusion bodies, were solubi- lized with 6 M guanidine-HC1 and renatured by a glutathione redox system and protracted dialysis under various experimental condi- tions. The purification of the recombinant proteins was achieved by binding the His-tag of the fusion protein on a nickel-nitrilotriacetic acid metal chelate column. The biological activity of the re- combinant growth factor was demonstrated by its ability to stimu- late proliferation of human vein endothelial cells, monitored by [~H]thymidine incorporation, where commercial recombinant human bFGF (rhbFGF) served as a positive control. Purified rhbFGF-F1 and rhbFGF-F2 constructs exhibited proliferative activ- ity comparable to commercial rhbFGF. The high-affinity binding was demonstrated by the binding of [3H]collagen to the rhbFGF-F2 protein immobilized on a Ni-nitrilotriacetic acid column. The rhbFGF-F2 fusion protein bound to collagen-coated surfaces with high affinity. Taken together, these results demonstrate that biolog- ically active rhbFGF fusion proteins can be recovered from trans- formed bacteria by oxidative refolding; thus, providing a means for their high-yield production, purification, and renaturation from microorganisms. Furthermore, we demonstrate that the auxiliary collagen binding domain effectively targets the recombinant growth factor to type I collagen. These studies advance the technology nec- essary to generate large quantities of targeted bFGF fusion proteins for specific biomedical applications.

* Correspondence and reprints: Departamento de Biologfa Celular y Gen6tica, Facultad de Ciencias, Campus Universitario de Teatinos, 29071 Mfilaga, Spain.

Keywords: Protein expression; Protein purification; Protein renatu- ration; Collagen; Basic fibroblast growth factor fusion protein.

Introduction

The fibroblast growth factor (FGF) family comprises at least twenty different monomeric proteins with mol- ecular weights of 16,000 to 18,000 (Moses et al. 1995). They bear 55% amino acid sequence identity (Esch et al. 1985, Gimenez-Gallego et al. 1985) and are encoded by distinct genes (Abraham et al. 1986a, b; Jaye et al. 1986). FGFs were originally detected and purified from extracts of whole tissues, most notably from the brain and pituitary gland (Bohlen et al. 1984, Maciag et al. 1984). FGFs have also been found to be associated with other tissues, such as the extracellular matrix underlying the vascular endothelium (Baird and Ling 1987, Vlodavsky et al. 1987), and within vas- cular endothelial cells and vascular smooth muscle cells (Schweigerer et al. 1987). FGFs are also synthe- sized by certain tumor cell lines (Klagsbrun et al. 1986, Shing et al. 1984), suggesting a role for these cytokines in tumor angiogenesis as well as in autocrine stimula- tion of tumor growth. Tissue repair following injury may require the actions of FGFs released from dam- aged cells and therefore exhibit properties expected of wound cytokines (Pierce et al. 1992). To date, four types of FGF receptors have been described (Gorlin 1997).

The basic fibroblast growth factor (bFGF or FGF- 2) is a multipotential factor that stimulates and sup- ports a variety of normal cell functions such as proliferation, migration, and differentiation. It stimu-

96 J.A. Andrades et al.: A recombinant human basic fibroblast growth factor

lates the proliferation of a large number of cells of mesodermal and neuroectodermal origin including fibroblasts, vascular endothelial and smooth muscle cells, chondrocytes, myoblasts, glial and rat neuronal precursor cells (Gospodarowicz 1989, Lindner 1995). bFGF is a chemotactic factor for a number of cell

types (Sato and Rifkin 1988, Zenjari et al. 1997). In culture, bFGF promotes differentiation of endothelial

cells (Gospodarowicz et al. 1976), chondrocytes (Kato and Gospodarowicz 1985), preadipocyte fibroblasts

(Broad and Ha m 1983), and neurons (Neufeld et al. 1987). bFGF is a powerful angiogenic factor. It is prob- ably identical to several other cationic endothelial cell mitogens that have high affinity for glycosaminogly- cans (Lobb et al. 1986) and is identical to several other heparin-binding endothelial cell growth factors (Baird et al. 1985, Sullivan and Klagsbrun 1985). A pituitary

cDNA clone that encodes bovine bFGF has been iso- lated. Analysis of the nucleotide sequence of this clone showed that bFGF is probably synthesized initially

as a 155-amino-acid protein with an amino-terminal extension of nine amino acids not found on the sequenced 146-amino-acid form of the protein (Esch et al. 1985). As other members of the family, b F G F

does not appear to contain a classic signal peptide sequence, raising the question of how these growth

factors, which interact with membrane-bound recep- tors on their target cells (Neufeld and Gospodarowicz 1986), are released from the cells in which they are

synthesized. The biology of human bFGF is complex. Multiple

isoforms of the growth factor (18, 22, 23, and 24 kDa) can be expressed from a single mRNA transcript (Florkiewicz and Sommer 1989). These bFGF isoforms

appear to differentially localize to the nucleus (the 22, 23, and 24 kDa isoforms), the cytosol, or the cell

surface (the 18 kDa isoform). The three-dimensional structure of human bFGF suggests a globular confor- mation revealing separate binding domains for its

receptor and for the positively charged heparin

binding site (Nimni 1997). Though very potent, bFGF is rapidly degraded when

injected or ingested. When conventional matrix- polymer-based release devices were fabricated and bFGF released in a sustained fashion, 99% of its mito- genic activity was lost. Preservation and stabilization of b F GF was accomplished by binding the factor to heparin-Sepharose beads. This permit ted its prolonged storage, repeated handling, and encapsulation within

a microspherical controlled-release device (Edelman et al. 1991).

Takagi et al. (1991) have reported that the primary structure of the von Willebrand factor (vWF) contains a high-affinity collagen binding domain. They further delimited the collagen binding site to a decapeptide with the amino acid sequence WREPSFCALS (Takagi et al. 1992). The present study reports the genetic engi- neering and expression of rhbFGF fusion proteins

in E s c h e r i c h i a col i , followed by the purification and renaturat ion of the biologically active compounds. Moreover, we introduce and assess the potential utility

of a specific rhbFGF fusion protein, which, by virtue of an auxiliary collagen binding domain, exhibits an increased affinity for collagen matrices. We engineered two constructs, hbFGF-F1, which represents the hbFGF active fragment fused with a His-tag, and hbFGF-F2, which besides the hbFGF active fragment,

incorporates a modified collagen binding domain derived from the vWF. The results of this study advance the concept that strategic fusion proteins may be engineered, produced, and renatured to target this growth factor for specific applications.

Material and methods

Basic FGFs constructs

Cytoplasmic RNAs isolated from human umbilical vein cells were reversely transcribed to single-strand cDNA using antisense primers of the bFGF coding region. Polymerase chain reaction (PCR) ampli- fication was performed on the first strand cDNA and the resulting PCR products were separated electrophoretically in agarose. Dis- crete bands were purified from the agarose gel by Geneclean (Bio 101) and ligated into the recipient vector (Invitrogen) by a TA cloning strategy. Color-selected clones were isolated and analyzed by restriction mapping, followed by nucleotide sequence determi- nation. For hbFGF-F2, the fusion protein incorporating a high- affinity collagen binding domain, we utilized a 50-nucleotide primer, 5'-CATATGTGGCGCGAACCGAGCTTCATGGCTCTGAGCG GTGCTAGCATGGCAG-Y, designed to include two parts: 7 nucleotides from the hbFGF sequence linked in frame with 30 nucleotides enconding a modified collagen binding peptide (Tuan et al. 1996, Andrades et al. 2000). The resultant construct is ligated in frame into the recipient pET28b expression vector (Novagen), which has a histidine purification tag to the amino terminus of each fusion protein. Both pET-hbFGF-F1 and pET-hbFGF-F2 constructs were maintained in the XL blue strain of E. coli (Fig. 1).The orien- tation and reading frame of each insert was confirmed by manual DNA nucleotide sequence analysis, using the modified dideoxy chain termination methods (USB).

Basic FGFs expression

For recombinant protein expression, the pET-hbFGF-F1 and pET- hbFGF-F2 constructs were transformed into the BL21(DE3) strain

J. A. Andrades et al.: A recombinant human basic fibroblast growth factor 97

T7: promoter.

7

A

F1 :

F2 :

I , 2 3 I Purification T ECM-Binding Growth Factor

(Active Fragment)

Protease site

(His)o @ Trp-Axg-Glu-Pro-Ser.Phe-Met-Ala-Leu-Ser

B

Fig. 1A, B. Design and construction of genetically engineered hbFGF fusion proteins. A Tripartite hbFGF fusion proteins (1,2, and 3) were generated by reverse transcriptase-PCR, cloned into TA cloning vectors, and then ligated into a pET28b expression vector (Novagen) utilizing EcoRI and Ndel sites, which incorporated an N- terminal (His)6 tag into position 1. B The rhbFGF-F2 protein con- tained a high-affinity collagen binding domain modified from yon Willebrand factor (vWF). This extracellular matrix binding domain was flanked by linker regions containing glycine to minimize steric hindrance. A potential thrombin site was included in the vector design. The rhbFGF-F1 construct does not contain the collagen binding sequence

of E. coli. A single clone was inoculated as primary cultures and further diluted to 500 ml cultures with 2YT medium (1.6% [w/v] Bactotryptone, 1% [w/v] yeast extract, and 0.5% [w/v] NaCI, pH 7.4) containing kanamycin (100 gg/mI). When the A600 reading reached 0.7-1.0 after 2-3 h growing, the proteins were induced in the pres- ence of 1 mM isopropyl [3-D-thiogalactopyranoside (IPTG) for 5 h at 37 ~ The bacterial pellets were collected by centrifugation at 5000 g for 10 min, washed with buffer A (20 mM Tris-HC1, 250 mM NaCI, 0.05% Nonidet P-40, pH 8.0) and lysed by the addition of 0.4 mg of lysozyme and 5 ng of DNase per ml, 0.8 mM phenyl-

methylsulfonyl fluoride, and 10 mM [3-mercaptoethanol in 1/10 of the original volume of cell suspension. The lysis was facilitated by sonication and homogenization with a Brinkmann Polytron oper- ated at 25 K for two cycles of 30 s on ice and was followed by cen- trifugation at 10000 g for 20 min at 4 ~ to pellet the inclusion bodies.

Basic FGFs solubilization and purification

inclusion bodies were solubilized in 6 M guanidine-HCl, 0.1 M phos- phate buffer, pH 8.0, at room temperature in 1/10 of the original volume of cells and allowed to stand at room temperature for 1.5 h with occasional vortexing. The solubilized fusion proteins were separated from insoluble debris by centrifugation at 20000 g.

Purification of solubilized rhbFGF fusion protein was accom- plished by nickel-nitrilotriacetic acid (Ni-NTA) chromatography (Qiagen): the protein solution was initially purified as a batch and then loaded onto a Ni-NTA column, and the Ni-NTA beads were allowed to settle down for 30 rain. The column was washed with 2 volumes of 8 M urea in buffer A, pH 8.0, and then with 2 volumes of 8 M urea in buffer A, pH 6.5. rhbFGFs were eluted with 8 M urea in buffer A, pH 4.0. The protein concentration was monitored by the method of Bradford (Bio-Rad) with bovine serum albumin (BSA) as a standard.

Basic FGFs refolding

After nickel chelate affinity purification of the His-tagged protein, samples were slowly diluted to appropriate concentrations (0.20 mg/ml) in 8 M urea, pH 8.0. A redox buffer (buffer A with 2.0 mM reduced glutathione (GSH), 0.2 mM oxidized glutathione (GSSG), and 1 mM dithiothreitol) was formulated, chilled on ice for 30 min, and slowly added to the purified rhbFGF fusion proteins to dilute the urea to 1.3 M (6-fold dilution), and to initiate the oxida- tive refolding. After vigorous mixing, the rhbFGF fusion proteins were allowed to anneal into a thermodynamically stable structure at 4 ~ overnight. Samples were dialyzed overnight against buffer S (250 mM NaC1, 20 mM Tris-HC1, and 20% sucrose, pH 8.0). Samples were centrifuged at 10000 g for 20 rain at 4 ~ the optical density was measured to calculate the final concentration of protein, and then the samples were iyophilized and reconstituted into 0.1% BSA-4 mM HC1 prior to freezing.

Biological activities of soluble and collagen-bound rhbFGF

The biological activity of the solubilized rhbFGF-F1 and rhbFGF- F2 were tested with human vein endothelial cells (HVEC) in a series of cell proliferation assays. HVEC were cultured as monolayers in Dulbecco modified Eagle medium (DMEM) (GIBCO) supple- mented with 1% penicillin-streptomycin and 10% fetal bovine serum (FBS) (GIBCO) at 37 ~ 5% CO2 in humidified air. For cell proliferation assays, cells were trypsinized and seeded in 24-well plates (Costar) at a density of 5 x 103 cells per well in normal growth medium. After cells reached 50-60% confluence, the medium was replaced with 0.5% FBS-DMEM and incubated for 24 h to semistarve. Serially diluted renatured rhbFGF fusion pro- teins were added to the wells and incubated for 16 h and then incu- bated with [~H]thymidine (1 ~tCi/well; specific activits; 2 Ci/mmole; 74 GBq/mmole; ICN) for another 4 h. Commercial rhbFGF (R&D Systems) was used as standard control. After incubation, the cells were rinsed three times with phosphate-buffered saline (PBS) and lysed with 150 ,ul of lysis buffer (Solvable; ICN) per well at room temperature for 10 rain. The resulting cell extract was analyzed by liquid scintillation.

98 J.A. Andrades et al.: A recombinant human basic fibroblast growth factor

The biological activity of the rhbFGF-F2 fusion protein was also evaluated. In this study, 500 pl of a solution of type I collagen (3.0 mg/ml, pH 7.0), derived from pepsin-treated bovine tendon (Nimni 1980), was dried until the collagen molecules aggregated into fibrils onto each well of 24-well plates and then they countercoated with 0.2% BSA. Renatured fractions of the rhbFGF-F2 were added to the coated wells in PBS and incubated at 37 ~ for 2 h, and the wells were rinsed 3 times with DMEM. HVEC (1.5 x 1 0 3 per well in 0.5% FBS-DMEM) were then added to each well and incubated for 20 h at 37 ~ prior to the addition of [3H]thymidine for the DNA synthesis assay as described above. Serially diluted samples were used in each experiment, and the experiments were repeated at least two more times.

Collagen binding assay rhbFGF-F2

The affinity of the rhbFGF-F2 fusion protein for collagen was assessed by immobilizing the recombinant fusion protein onto an Ni-NTA column, then collagen (biosynthetically labeled with [3H]proline and purified from human fibroblasts cultures) was applied to the column and eluted with a linear gradient of either PBS (0.15 to 2 M) or urea (0 to 4 M).

Results

Recombinant constructs of hbFGF fusion proteins

In this study, we genetically engineered a chimeric human b F G F fusion protein ( rhbFGF-F2) which

includes a high-affinity collagen binding domain by

using reverse t ranscr iptase-PCR and recombinant

D N A technologies (Fig. 1). The strategically modified

collagen-binding decapept ide was derived f rom a

vWF domain involved in the recognition of exposed

vascular collagen sequences.Thus, the rhbFGF-F2 con-

struct, which incorporates a putat ive collagen-binding

decapept ide sequence W R E P S F M A L S , was designed

specially for targeting the resulting h b F G F fusion

protein to collagen. The cysteine of the original se-

quence was replaced conservatively by methionine, in

order that this auxiliary domain should not interfere

with the refolding process. Flanking linkers were

designed to include glycine residues to minimize steric

hindrances. The resulting genetically engineered

h b F G F fusion proteins contain a contiguous series of

functional domains: a purification tag, a thrombin pro- tease site, and a collagen binding domain followed by

the mature f ragment of h b F G E

Expression, solubilization, and purification of rhbFGF fusion proteins

Recombinan t h b F G F proteins can be expressed at high levels in bacteria t ransformed with plasmids bearing the T7 polymerase system, BL21 (DE3) strain.

Fig. 2. Expression and purification of the recombinant fusion pro- teins. Protein expression is visualized on reducing SDS-PAGE gel. 1 Molecular-mass markers; 2 and 5 bacterial proteins of transformed hbFGF-F1 and hbFGF-F2, respectively, before induction; 3 and 6 after induction by 1 M IPTG for 5 h; 4 and 7 after purification by Ni-NTA affinity chromatography, resulting in pure hbFGF-F1 and hbFGF-F2, respectively. There is a slight upshift of hbFGF-F2 due to the presence of the auxiliary collagen binding domain

In the presence of I m M IPTG, the levels of recombi-

nant hbFGFs expression approached 50% of cellular

protein, as determined by sodium dodecyl sulfate-

polyacrylamide gel electrophoresis (SDS-PAGE) and

protein staining (Fig. 2). The reducing SDS-PAGE

gel shows a representat ive induction of the 16 kDa

r h b F G F polypeptide (Fig. 2, lane 3) and 17 kDa for

the rhbFGF-F2 construct bearing the auxiliary vWF-

derived decapeptide (Fig. 2, lane 6). The bulk of the recombinant proteins were found in the insoluble

inclusion bodies harvested after cell lysis by centrifu- gation. To prevent cosedimentat ion of cells and cell

debris, the initial lysis was maximized by a combina- tion of lysozyme t rea tment (0.4 mg/ml), homogeniza- tion, and sonication. Since inclusion body proteins contain variable amounts of incorrect and/or inter-

chain disulfide bonds, 10 mM [3-mercaptoethanol was added to the lysis buffer to disrupt these bonds.

J. A. Andrades et aI.: A recombinant human basic fibroblast growth factor 99

Strong denaturants such as 6 M guanidine-HC1

and 8 M urea are both effective in solubilizing rhbFGF

fusion proteins from bacterial inclusion bodies. How- ever, 6 M guanidine was determined to be optimal in

terms of overall yields and recovery of the recombi- nant proteins.

After solubilization, we endeavored to purify the recombinant growth factors prior to renaturat ion to avoid deleterious interactions with other solubilized proteins. Also, recombinant proteins may undergo degradation by solubilized proteinases renatured in the same environment. Extensive purification of the recombinant proteins, hbFGF-F1 and hbFGF-F2, was

afforded by metal chelate (Ni-NTA) chromatography performed under denaturing conditions in the pres- ence of 6 M guanidine.

While 6 M guanidine was determined to be optimal

in terms of recovery of solubilized proteins, renatura- tion of biological activity was more successful in 8 M

urea. Therefore, to prepare the protein for the subse- quent refolding steps, the bound recombinant proteins were washed on the column with 8 M urea, pH 8.0, followed by a more stringent wash with 8 M urea, pH 6.5, followed finally by elution of the purified proteins in 8 M urea at pH 4.0 (Fig. 2, lanes 4 and 7).

Approximately 6rag of purified rhbFGF-F1 and 8 mg of purified rhbFGF-F2 were routinely recovered from a 100 ml bacterial culture following Ni-NTA purification.

Renaturation of rhbFGF fusion proteins

Since control of redox conditions is a key factor in suc- cessfully refolding rhbFGF fusion proteins, we tested various concentrations of GSH and GSSG as "oxido- shuffling" reagents to increase the rate and yield of correct protein formation. As shown in Fig. 3, the con-

centration and ratio of reduced oxidized glutathione exerts a major influence on the resulting biological activity. The specific biological activity of the rena- tured rhbFGF fusion proteins was determined to be optimal at a ratio of 2.0 mM GSH to 0.2 mM GSSG, when a proliferation assay with H V E C was carried

out. Since the protein concentration in this prolifera- tion assay was normalized to Bg/ml, the results indicate the relative specifc activity. At ratios of 10.0 mM GSH to 1.0 mM GSSG and 2.0 mM GSH to 0.2 mM GSSG the relative specific activities of rhFGF-F1 were 1400 + 500 and 3800 _+ 700, respectively, and those of rhFGF-F2 were 1200 + 200 and 3000_+500 (means

8000

6000

g 4000-

2000-

0- 0

?

5 10 15 �89 �89 fraction number

3;

Fig. 3. Binding of rhbFGF-F1 and rhbFGF-F2 proteins to collagen. The tripartite fusion proteins, rhbFGF-F1 (O) and rhbFGF-F2 (0), were first bound to the Ni-NTA column and then [3H]collagen was applied to the column. The bound collagen was eluted with a gradi- ent of 0 to 4 M urea in 0.05 M phosphate buffer at pH 7.0

with standard deviations, n = 4), respectively. By con- trast, the efficiencies of dithiothreitol and [3-mercap-

toethanol as reducing agents to control the rate of the

redox reactions were found to be limited (data not shown).

Protein concentration

Associated with the kinetics and thermodynamics of

monomer renaturation, in vitro protein refolding, mis- folded intermediates, and multimeric aggregates com-

pete with the renaturation of rhbFGF. Aggregation reactions may originate from nonspecific interactions of predominantly unfolded as well as partially folded intermediates. As a consequence, an increase in the

rate of unproductive aggregation is generally observed when the concentration of unfolded polypeptide

chains is increased. Since aggregation tends to pre- dominate above a limiting protein concentration, we determined the practical limits of protein concentra- tion under standardized refolding conditions. In this study, the limiting concentration for optimal recovery of native hbFGF-F1 was found to be 20 gg/ml. For rhbFGF-F2, the limiting concentration turned out to be somewhat lower (18 gg/ml), possibly due to an increase in hydrophobic interactions associated with the auxiliary collagen-binding decapeptide.

100

Table 1. Biological activity of fusion proteins after dialyzation a

J. A. Andrades et al.: A recombinant human basic fibroblast growth factor

Protein Biological activity of protein at concentration (pg/ml): u

50 100 150 200

rhbFGF c 3000 + 540 4100 _+ 610 4700 • 635 4750 + 620 rhbFGF-F1 2780 + 490 3750 • 575 4400 + 635 4300 • 600 rhbFGF-F2 2780 + 465 3825 _+ 560 4600 + 610 4375 + 590

a Protein fractions were dialyzed with buffer A supplemented with 250 mM NaC1 and 20% sucrose b Biological activity is expressed as [3H]thymidine incorporation (cpm/well) by HVEC. Values are given as means with standard deviations for four assays rhbFGF represents commercial growth factor

Dialysis conditions

Although removal of solubilizing agents and refolding components can be accomplished by simple dialysis,

important parameters including buffer, pH, ionic strength, and stabilizing additives had also to be con- sidered for preventing overt precipitation. For re- covery of active rhbFGF-F1, which includes the (His)6

purification tag, and rhbFGF-F2, which also includes the auxiliary collagen-binding decapeptide, pH 8.0 was found to be optimal for recovering native conforma- tion and biological activity. At pH 6.5 precipitation

occurred, and at pH 11.0 no precipitation appeared. We determined 20% sucrose to be the best dialysis condition, while other additives like polyethylene glycol, high NaC1 concentration, glycerol, and L- arginine were not as effective (data not shown). Buffer

A, at pH 8.0, supplemented with 250 mM NaC1 and 20% sucrose, afforded optimal recoveries (62% + 15%) upon gradual protracted dialysis (unsupple-

mented buffer A afforded 7% + 2% protein recovery,

buffer A supplemented with 0.5 M NaC1 28% + 9%, and buffer A supplemented with 10% glycerol 32% +

9%). No organic solvents were added to the dialysis

buffer to prevent aggregation. After the final protein concentration was deter-

mined, the fusion proteins could be lyophilized and the lyophilized powders reconsti tuted in a 4 mM HC1 solution supplemented with 0.1% BSA and used in subsequent experiments. However, the recovery of bioactivity after lyophilization was found to be about two times less than after standard dialysis. Samples are routinely stored at -20 ~ under which conditions no appreciative loss of activity seems to occur.

Biological activity of renatured rhb FGFs fusion proteins

Biological activity of the renatured rhbFGF-F1 and rhbFGF-F2 fusion proteins was confirmed by the H V E C proliferation assay in which commercially available rhbFGF served as a positive control (Table

1). Dialyzed protein fractions (5 to 30 B1/ml) from the glutathione redox-refolding system, added to the cell culture for 16 h after a 24 h low-serum starva- tion, caused a proliferative activity comparable to the rhbFGF control treatments (50 to 200pg/ml) in HVEC. When rhbFGF-F2 was applied and bound to collagen-coated culture wells and then cells were seeded on top of the rhbFGF-F2-col lagen in 0.5% F BS -D MEM, no significant amount of stimulation of

cell proliferation was observed (data not shown). In the same manner, the level of stimulation observed for collagen-bound growth factor was smaller than that observed when the unbound rhbFGF-F2 protein was added directly to the culture medium. These results suggest that the tight binding of rhbFGF-F2 to colla-

gen lowered the availability and/or the rate of release of the biologically active growth factor.

Binding of rhbFGF-F2 to collagen

To demonstrate the affinity of the rhbFGF-F2 fu- sion protein for collagen, the rhbFGF-F2 was first immobilized on a Ni-NTA column and then exposed to biosynthetically labeled [3H]collagen, which was loaded subsequently onto the column. Under these conditions a large port ion of the radioactively labeled collagen was found to bind to the column. Washing the column with a linear gradient of NaC1 from 0.15 to 2 M did not release the [3H]collagen. However, appli- cation of a urea gradient (0 to 4 M) was able to quan-

J. A. Andrades et ai.: A recombinant human basic fibroblast growth factor 101

titatively elute all bound radioactivity (Fig. 3). By con- trast, when the rhbFGF (containing only the (His)6 tag and no collagen binding domain) was applied to the Ni-NTA column under identical conditions, very little [3H]collagen was retained on the column, suggesting that the auxiliary collagen binding domain in rhbFGF- F2 afforded this high-affinity interaction.

Discussion

In order to deliver the rhbFGF fusion protein to spe- cific target sites and to enhance the half-life of the delivered fusion protein, the rhbFGF-F2 construct was designed to contain a high-affinity collagen binding domain, to enable the fusion protein to be targeted to implantable collagen matrices and/or in vivo wound sites where collagen fibers are exposed. In the devel- opment of this collagen-targeting construct, we uti- lized a decapeptide (WREPSFMALS) modified from the region of vWF which binds specifically to collagen or gelatin (Takagi et al. 1992). This modified sequence was demonstrated to bind tightly to purified collagen and gelatin in our binding studies, indicating that the strategic substitution of Met for Cys at position 7 did not prevent the high-affinity interaction. The addition of the collagen-binding decapeptide sequence appar- ently did not interfere with the refolding process since the hbFGF-F2 fusion protein was biologically active in its ability to stimulate the proliferation of HVEC at comparable levels.

Although expression of recombinant proteins by bacterial systems promises an unlimited source of useful proteins and growth factors, proteins expressed in E. coli often accumulate in inclusion bodies as aggregates of insoluble forms, which do not possess biological activity, within the reducing environment of the bacterial cytoplasm (Walker and Gilbert 1994). In this study, two hbFGF constructs were expressed in E. coli and transformed with an expression plasmid which utilizes the T7 polymerase system to efficiently express recombinant proteins. The pET28-b-hbFGF constructs utilized in the study exhibit two notable advantages. First, the T7 polymerase system affords high levels of protein expression under the control of the IPTG-inducible lacUV5 promoter. Second, the purification of hbFGF fusion proteins was facilitated by the inclusion of an in-frame (His)6 purification tag in the expression vector.

The renaturation of protein aggregates in bacterial inclusion bodies into biologically active conforma-

tions involves a number of interdependent steps: solubilization, oxidative refolding, and withdrawal of denaturants, each of which represents a challenge of strategy and optimization (Handl et al. 1993, Carda- mone et al. 1995). In the present study, we determined that a strong ionic denaturant, 6 M guanidine, was more effective than 8 M urea in solubilizing the inclu- sion bodies. Guanidine also appeared to be preferable to urea since the latter may contain isocyanate, leading to carbamylation of free amino groups within the polypeptide. However, in evaluating the recoveries from subsequent refolding steps, we found that the rhbFGF activity recovered from guanidine-HC1 solu- tions was much lower than from urea. Others also have reported that guanidine-HC1 may interfere with the refolding process (Cardamone et al. 1995). For this reason, we utilized guanidine-HC1 initially as a dena- turant to solubilize the inclusion bodies and optimize the recovery of the recombinant proteins and then replaced the 6 M guanidine-HCt with 8 M urea in the course of affinity purification. Oxidative refolding was initiated by dilution to 1.3 M urea, pH 8.0, in an opti- mized urea-redox buffer.

Since removal of the denaturant by simple dialysis caused abundant precipitation, we performed an initial sixfold dilution of the denaturant (from 8 M urea to 1.3 M urea) with an optimized redox buffer system which maintained both the urea concentration and redox conditions for a protracted time interval (48h). The solubilized proteins rearranged into native conformations and the biological activity in- creased dramatically after this prolonged period without encountering overt protein precipitation.

In the final step, dialysis was used to remove the denaturant and the redox reagents and then centri- fugation to collect misfolded aggregates. Since protein precipitation at this step represents a major loss of native as well as misfolded rhbFGF, we endeavored to optimize these conditions. Small molecules such as glycerol, sucrose, polyethylene glycol (200-2000), BSA, L-Arg are often added in dialysis buffers to prevent precipitation (Babbitt et al. 1990, Cleland et al. 1992, Rudolph and Lille 1996). In this study, we found 20% sucrose to be advantageous. Also, salt con- centrations and pH are very important consideration during dialysis, as protein stability is often dependent on the ionization state of side residues (Schumann et al. 1993). We determined that dialysis into 250 mM NaC1, pH 8.0, 20% sucrose proved optimal for recov- ery of biologically active hbFGF fusion proteins. Con-

102 J, A. Andrades et al.: A recombinant human basic fibroblast growth factor

ceivably, the lack of native propeptide sequence (Gary and Mason 1990) and/or the presence of auxiliary domains may hinder the process of renaturation. Nev- ertheless, by manipulating redox conditions, protein concentrations, pH, ionic strength, small molecule additives, and dialysis conditions, we have improved the yields from about 0.5 % in our earlier attempts with rhbFGF-F2 to an average of between 25 and 30% rhbFGF-F2.

Basic FGF is a multifunctional protein that modu- lates the functions of a wide variety of cells of meso- dermal, neuroectodermal, and endodermal origin in vitro (Gospodarowicz et al. 1987, Burgess and Maciag 1989). The biological properties of bFGF suggest its possible role in wound healing, a complex process ini- tiated by an inflammatory phase followed by the depo- sition of a conditional matrix, collagen synthesis, and remodeling (Blitstein-Willinger 1991, Satoh et al. 1997). Since FGF is a pleotropic agent which can stim- ulate, inhibit, and modulate cellular events in a time- and concentration-dependent manner, it is important to control its delivery and bioavailability to advance its use as a potential therapeutic agent. In our study, we have verified that the active fragment of rhbFGF- F1 does bind nonspecifically to type I collagen. However, rhbFGF-F2, which contains a specific col- lagen binding domain, has a much greater affinity for collagen matrices when compared to rhbFGF-F1.

Further studies will be required to evaluate the in vivo and/or cell culture conditions by which the rhbFGF-F2 fusion protein is released from the col- lagen matrix and how such a timed and/or conditional release mechanism will affect cells. Together with pre- vious studies from our laboratory (Tuan et al. 1996, Han et al. 1997), this one extends the methodologies to generate large quantities of biologically active FGF from prokaryotic microorganisms. It also provides support to the concept that growth factors may be tar- geted to specific tissues, wounds, grafts, and prosthetic implants by the inclusion of auxiliary binding domains into their primary structures.

Acknowledgments

We are grateful for the financial support of National Institutes of Health (AGO2577), and of the Ministerio de Educacidn y Cultura (PB95-1134, SAF99-0133 Spain), which supported the visit of Dr. Jos6 A. Andrades to the University of Southern California. We also thank Dr. Ignacio Nufiez de Castro and Dr. Pedro Fernfindez- Llebrez for their valuable comments on the manuscript, and Ling Liu for her excellent technical assistance.

References

Abraham JA, Mergia A, Whang JL, Tumulo A, Friedman J, Hjerrild KA, Gospodarowicz D, Piddes JC (1986a) Nucleotide sequence of a bovine clone encoding the angiogenic protein, basic fibro- blast growth factor. Science 233:545-548

- Whang JL, Tumulo A, Mergia A, Friedman J, Gospodarowicz D, Fiddes JC (1986b) Human basic fibroblast growth factor: nu- cleotide sequence and genomic organization. EMBO J 5: 2523- 2528

Andrades JA, Becerra J, Hall FL, Nimni ME (2000) Factor de cre- cimiento fibroblastico basico con un dominio de union especifico a colageno. Spanish patent P200002811, November 2000

Babbitt PC, West BL, Buechter DD, Chen L, Kuntz ID, Kenyon G (1990) Removal of a proteolytic activity associated with aggre- gates formed from expression of creative kinase in Escherichia coli leads to improved recovery of active enzyme. Biotechnology 8:945-949

Baird A, Ling N (1987) Fibroblast growth factors are present in the extracellular matrix produced by endothelial cells in vitro: impli- cations for a role of heparinase-like enzymes in the neovascular response. Biochem Biophys Res Commun 142:428-435

- Morm~de P, B6hlen P (1985) Immunoreactive fibroblast growth factor in cells of peritoneal exudate suggests its identity with macrophage-derived growth factor. Biochem Biophys Res Commun 126:358-364

Blitstein-Willinger E (1991) The role of growth factors in wound healing. Skin Pharmacol 4:175-182

Bohlen R Baird A, Esch F, Ling N, Gospodarowicz D (1984) Isola- tion of brain fibroblast growth factor by heparin Sepharose affin- ity chromatography: identity with pituitary fibroblast growth factor. Proc Natl Acad Sci USA 81:6963-6967

Broad TE, Ham RG (1983) Growth and adipose differentiation of sheep preadipocyte fibroblasts is serum free medium. Eur J Biochem 135:33-39

Burgess WH, Maciag T (1989) The heparin-binding (fibroblast) growth factor family of proteins Annu Rev Biochem 58:575~506

Cardamone M, Puri NK, Brandon MR (1995) Comparing the refold- ing and reoxidation of recombinant porcine growth hormone from a urea denatured state and from Escherichia coli inclusion bodies. Biochemistry 34:5773-5794

Cleland JL, Hedgepeth C, Wang DI (1992) Polyethylene glycol enhanced refolding of bovine carbonic anhydrase B: reaction stoichiometry and refolding model. J Biol Chem 267: 13327- 13334

Edelman ER, Mathiowitt E, Langer R, Klagsbrun M (1991) Con- trolled and modulated release of basic fibroblast growth factor. Biomaterials 12:619-626

Esch FA, Baird A, Ling N, Ueno N, Hill F, Denoroy L, Kleper R, Gospodarowicz D, Bohlen R Guillemin R (1985) Primary struc- ture of bovine pituitary basic fibroblast growth factor (FGF) and comparison with the amino-terminal sequence of bovine brain acidic FGE Proc Natl Acad Sci USA 82:6507-6511

Florkiewicz RZ, Sommer A (1989) Human basic fibroblast growth factor gene encodes four polypeptides: three initiate translation from non-AUG codons. Proc Natl Acad Sci USA 86:3978-3981

Gary AM, Mason AJ (1990) Requirement for activin A and trans- forming growth factor [31 pro-regions in homodimer assembly. Science 247:1328-1330

Gimenez-Gallego G, Rodkey J, Bennet C, Rios-Candelore M, DiSalvo J, Thomas KA (1985) Brain-derived acidic fibroblast growth factor: complete amino acid sequence and homologies. Science 230:1385-1388

Gorlin RJ (1997) Fibroblast growth factors, their receptors and receptor disorders. J Craniomaxillofac Surg 25:69-79

J. A. Andrades et al.: A recombinant human basic fibroblast growth factor 103

Gospodarowicz D (1989) Fibroblast growth factor. In: Perucho M (ed) Critical reviews in oncogenesis. CRC Press, Boca Raton, Fla, pp 1-26

- Moran J, Braun D, Birdwell CR (1976) Clonal growth of bovine vascular endothelial cells: fibroblast growth factor as a survival agent. Proc Natl Acad Sci USA 73:4120-4124

- Ferrara N, Schweigerer L, Neufeld G (1987) Structural character- ization and biological functions of fibroblast growth factor. Endocr Rev 8:95-114

Han B, Hall FL, Nimni ME (1997) Refolding of a recombinant col- lagen-targeted TGF-beta2 fusion protein expressed in Escherichia coli. Protein Expr Purif 11:169-178

Handl CE, Harel J, Flock JI, Dubreuil JD (1993) High yield of active STb enterotoxin from a fusion protein (MBP-STb) expressed in Escherichia coli. Protein Expr Purif 4:275-281

Jaye M, Howk R, Burgess W, Ricca GA, Chiu IM, Ravea MW, O'Brien S J, Modi WS, Maciag T, Drohan WN (1986) Human endothelial cell growth factor: cloning, nucleotide sequence, and chromosomal localization. Science 233:541-545

Kato Y, Gospodarowicz D (1985) Sulfated proteoglycan synthesis by confluent cultures of rabbit costal chondrocytes grown in the presence and absence of fibroblast growth factor. J Cell Biol 100: 477-485

Klagsbrun M, Sasse J, Sullivan R, Smith JA (1986) Human tumor cells synthesize an endothelial cell growth factor that is struc- turally related to basic fibroblast growth factor. Proc Natl Acad Sci USA 83:2448-2452

Lindner V (1995) Role of basic fibroblast growth factor and platelet- derived growth factor (B-chain) in neointima formation after arterial injury. Z Kardiol 4:137-144

Lobb R, Sasse J, Sullivan R, Shing Y, D'Amore R Jacobs J, Klags- brun M (1986) Purification and characterization of heparin- binding endothelial cell growth factors. J Biol Chem 261: 1924-1928

Maciag T, Mehlman T, Freisel R, Schreider A (1984) Heparin binds endothelial cell growth factor, the principal endothelial cell mitogen in bovine brain. Science 225:932-935

Moses MA, Klagsbrun M, Shing Y (1995) The role of growth factors in vascular cell development and differentiation. Int Rev Cytol 161:1-48

Neufeld G, Gospodarowicz D (1986) Basic and acidic fibroblast growth factors interact with the same cell surface receptors. J Biol Chem 261:5631-5637

- - Dodge L, Fujii DK (1987) Heparin modulation of the neuro- tropic effects of acidic and basic fibroblast growth factors and nerve growth factors on PC-12. J Cell Physiol 131:131-140

Nimni ME (1980) The molecular organization of collagen and its role in determining the biophysical properties of the connective tissues. Biorheology 17:51-82

- (1997) Polypeptide growth factors: targeted delivery systems. Biomaterials 18:1201-1225

Pierce GF, Tarpley JE, Yanagihara D, Mustoe TA, Fox GM, Thoma- son A (1992) Platelet-derived growth factor (BB homodimer), transforming growth factor-beta 1, and basic fibroblast growth factor in dermal wound healing: neovessel and matrix formation and cessation of repair. Am J Pathol 140:1375-1388

Rudolph R, Lilie H (1996) In vitro folding of inclusion body pro- teins. FASEB J 10:49-56

Sato Y, Rifkin D (1988) Autocrine activities of basic fibroblast growth factor: regulation of endothelial cell movement, plas- minogen activator synthesis, and DNA synthesis. J Cell Biol 107: 1199-1205

Satoh H, Shino A, Sato F, Asano S, Murakami I, Inatomi N, Nagaya H, Kato K, Szabo S, Folkman J (1997) Role of endogenous basic fibroblast growth factor in the healing of gastric ulcers in rats. Jpn J Pharmacol 73:59-71

Schumann J, Bohm G, Schumacher G, Rudolph R, Jaenicke R (1993) Stabilization of creatinase from Pseudornonas putida by random mutagenesis. Protein Sci 2:1612-1620

Schweigerer L, Neufeld G, Friedman J, Abraham JA, Fiddes JC, Gospodarowicz D (1987) Capillary endothelial cells express basic fibroblast growth factor, a mitogen that promotes their own growth. Nature 325:257-259

Shing Y, Folkman J, Sullivan R, Butterfield C, Murray J, Klagsbrun M (1984) Heparin affinity purification of a tumor- derived capillary endothelial cell growth factor Science 223: 1296-1299

Sullivan R, Klagsbrun M (1985) Purification of cartilage-derived growth factor by heparin affinity chromatography. J Biol Chem 260:2399-2403

Takagi J, Jujisawa T, Sekiya F, Saito Y (1991) Collagen-binding domain within bovine propolypeptide of yon Willebrand factor. J Biol Chem 266:5575-5579

- Asai H, Saito Y (1992) A collagen/gelatin-binding decapeptide derived from bovine propolypeptide of von Willebrand factor. Biochemistry 31:8530-8534

Tuan T-L, Cheung DT, Wu L-T, Yee A, Gabriel S, Han B, Morton L, Nimni ME, Hall FL (1996) Engineering, expression and renatu- ration of targeted TGF-beta fusion proteins. Connect Tissue Res 34:1-9

Vlodavsky I, Folkman J, Sullivan R, Friedman R, Ishal-Michaeli R, Sasse J, Klagsbrun M (1987) Endothelial cell-derived basic fibro- blast growth factor: synthesis and deposition into snbendothelial extracellular matrix. Proc Natl Acad Sci USA 84:2292-2296

Walker KW, Gilbert HF (1994) Effect of redox environment on the in vitro and in vivo folding of RTEM-1 beta-lactamase and Escherichia coli alkaline phosphatase. J Biol Chem 269: 28487- 28493

Zenjari C, Boilly B, Hondermarck H, Boilly-Marer Y (1997) Nerve- blastema interactions induce fibroblast growth factor-1 release during limb regeneration in Pleurodeles waltl. Dev Growth Differ 39:15-22