Eur. J. Biochem. 241, 453-461 (1996) 0 FEBS 1996 X-ray structure of native full-length human fibroblast-growth factor at 0.25-nm resolution Antonio ROMERO ', Antonio PINEDA-LUCENA2 and Guillermo GIMENEZ-GALLEG02 ' Instituto de Quimica Fisica Rocasolano, Madrid, Spain Centro de Investigaciones Biolbgicas, Madrid, Spain (Received 14 May17 August 1996) - EJB 96 070613 Acidic fibroblast-growth factor (aFGF) is one of the typical members of a group of nine polypeptides of relatively similar amino acid sequence known as the fibroblast-growth-factor family of proteins. Widely distributed throughout the organism, fibroblast-growth factors seem to be involved in numerous physio- logical processes ranging from control of cell proliferation and differentiation to modulation of animal behaviour and arterial blood pressure. This wide assortment of biological activities explains their involve- ment in numerous pathologies. Instability and low yields of the purified protein have precluded high- resolution structural studies of the physiological form of aFGF. Nevertheless, modifications introduced recently into the synthesis and purification procedures of this protein have allowed preparations of sam- ples that, as shown here, are reliable substrates to obtain crystals suitable for X-ray-diffraction studies. These analyses have allowed us to elucidate the three-dimensional structure of the physiological form of human aFGF by molecular-replacement methods, from the previously reported structure of a shortened form of bovine aFGF that was stabilized by point-directed mutagenesis. The structure was refined at a resolution of 0.25 nm to an R factor of 20.4% for 13 109 reflections between 0.6 nm and 0.25 nm, with rmsd of 1.1 pm and 1.9' from ideal bond lengths and bond angles, respectively. Human aFGF folds according to a ,&trefoil topology. This fold consist of six P-strand pairs. Three of them form a six-stranded P-barrel structure that is capped at one end by the other three pairs arranged in a triangular array. The N- terminus of aFGF up to residue Pro19 appears flexible in the structure and does not specifically interact with the rest of the molecule. Keywords: X-ray structure ; native full-length fibroblast-growth factor; crystallization ; protein tertiary structure. A group of nine polypeptides with similar amino acid se- quences are known as the fibroblast-growth factor (FGF) family of proteins. Acidic and basic FGF (aFGF and bFGF, respective- ly) are the prototypic members of the family. A wide variety of biological properties, including the induction of cell prolifera- tion and motility, induction or inhibition of cell differentiation, and a relatively extensive assortment of hormone-like activities, have been ascribed to these two polypeptides. The spectrum of cells sensitive to them extends to most mesoderm-derived and neuroectoderm-derived cells (Dickson et al., 1990; Tanaka et al., 1992; Miyamoto et al., 1993; reviewed by Baird and Bohlen, 1990; Burgess and Maciag, 1989; GimCnez-Gallego and Cuevas, 1994). FGF have been repeatedly implicated in tumour development and genetic diseases (Potgens et al., 1995; Wilkie et al., 1995). A detailed knowledge of the structural basis of the physiological properties of aFGF may be, thus, of considerable clinical relevance. FGF are also known as heparin-binding growth factors, because of their high affinity for this sulphated glycosaminoglycan. In vitro assays have shown that FGF mito- genic activity requires either heparin or heparan sulphate (Gi- mCnez-Gallego et al., 1986a; Rapraeger et al., 1991; Yayon et Correspondence to G. Gimenez-Gallego, Centro de Investigaciones Fax: +34 1 564 9065. Abbreviations. FGF, fibroblast-growth factor; aFGF, acidic FGF; Biol6gicas, Velizquez, 144, E-28006 Madrid, Spain bFGF, basic FGE al., 1991). Activation by heparin-like compounds can be emu- lated with myo-inositol hexasulphate, at least for aFGF (Pineda- Lucena et al., 1994). aFGF was originally purified from bovine brain as a mixture of two polypeptides (140 amino acids and 134 amino acids, re- spectively; Thomas et al., 1984, 1985). Later, aFGF was isolated from human brain as a mixture of two polypeptides of 140 amino acids and 139 amino acids (GimCnez-Gallego et al., 1986a). Cloning of human aFGF cDNA suggested that the puri- fied forms derive from a 154-amino-acid polypeptide (full- length aFGF) by limited proteolytic processing (Jaye et al., 1986; Burgess et al., 1986). This latter form has been identified in some preparations of bovine aFGF (Burgess et al., 1985). By analogy with bFGF (Ueno et al., 1986), it seems reasonable to assume that the 154-amino-acid aFGF is the physiological form of the growth factor, and that the shorter forms are purification artifacts. Short forms of aFGF (140, 139 and 134 amino acids) have been expressed in Escherichia coli cultures by recombinant DNA techniques in several laboratories (Linemeyer et al., 1987; Jaye et al., 1987). However attempts to express the full-length aFGF in common expression plasmids rendered only minute amounts of protein, which could be detected exclusively by Western blot analysis (Jaye et al., 1987). Ortega et al. (1991) have reported that recombinant aFGF is inactivated very rapidly by Cys oxidation when exposed to air. The minute amounts of full-length aFGF that are recovered from brain (Burgess et al., 1985) and cultures of transformed E.
Transcript
Eur. J. Biochem. 241, 453-461 (1996) 0 FEBS 1996
X-ray structure of native full-length human fibroblast-growth factor at 0.25-nm resolution Antonio ROMERO ', Antonio PINEDA-LUCENA2 and Guillermo GIMENEZ-GALLEG02
' Instituto de Quimica Fisica Rocasolano, Madrid, Spain Centro de Investigaciones Biolbgicas, Madrid, Spain
(Received 14 May17 August 1996) - EJB 96 070613
Acidic fibroblast-growth factor (aFGF) is one of the typical members of a group of nine polypeptides of relatively similar amino acid sequence known as the fibroblast-growth-factor family of proteins. Widely distributed throughout the organism, fibroblast-growth factors seem to be involved in numerous physio- logical processes ranging from control of cell proliferation and differentiation to modulation of animal behaviour and arterial blood pressure. This wide assortment of biological activities explains their involve- ment in numerous pathologies. Instability and low yields of the purified protein have precluded high- resolution structural studies of the physiological form of aFGF. Nevertheless, modifications introduced recently into the synthesis and purification procedures of this protein have allowed preparations of sam- ples that, as shown here, are reliable substrates to obtain crystals suitable for X-ray-diffraction studies. These analyses have allowed us to elucidate the three-dimensional structure of the physiological form of human aFGF by molecular-replacement methods, from the previously reported structure of a shortened form of bovine aFGF that was stabilized by point-directed mutagenesis. The structure was refined at a resolution of 0.25 nm to an R factor of 20.4% for 13 109 reflections between 0.6 nm and 0.25 nm, with rmsd of 1.1 pm and 1.9' from ideal bond lengths and bond angles, respectively. Human aFGF folds according to a ,&trefoil topology. This fold consist of six P-strand pairs. Three of them form a six-stranded P-barrel structure that is capped at one end by the other three pairs arranged in a triangular array. The N- terminus of aFGF up to residue Pro19 appears flexible in the structure and does not specifically interact with the rest of the molecule.
A group of nine polypeptides with similar amino acid se- quences are known as the fibroblast-growth factor (FGF) family of proteins. Acidic and basic FGF (aFGF and bFGF, respective- ly) are the prototypic members of the family. A wide variety of biological properties, including the induction of cell prolifera- tion and motility, induction or inhibition of cell differentiation, and a relatively extensive assortment of hormone-like activities, have been ascribed to these two polypeptides. The spectrum of cells sensitive to them extends to most mesoderm-derived and neuroectoderm-derived cells (Dickson et al., 1990; Tanaka et al., 1992; Miyamoto et al., 1993; reviewed by Baird and Bohlen, 1990; Burgess and Maciag, 1989; GimCnez-Gallego and Cuevas, 1994). FGF have been repeatedly implicated in tumour development and genetic diseases (Potgens et al., 1995; Wilkie et al., 1995). A detailed knowledge of the structural basis of the physiological properties of aFGF may be, thus, of considerable clinical relevance. FGF are also known as heparin-binding growth factors, because of their high affinity for this sulphated glycosaminoglycan. In vitro assays have shown that FGF mito- genic activity requires either heparin or heparan sulphate (Gi- mCnez-Gallego et al., 1986a; Rapraeger et al., 1991; Yayon et
Correspondence to G. Gimenez-Gallego, Centro de Investigaciones
al., 1991). Activation by heparin-like compounds can be emu- lated with myo-inositol hexasulphate, at least for aFGF (Pineda- Lucena et al., 1994).
aFGF was originally purified from bovine brain as a mixture of two polypeptides (140 amino acids and 134 amino acids, re- spectively; Thomas et al., 1984, 1985). Later, aFGF was isolated from human brain as a mixture of two polypeptides of 140 amino acids and 139 amino acids (GimCnez-Gallego et al., 1986a). Cloning of human aFGF cDNA suggested that the puri- fied forms derive from a 154-amino-acid polypeptide (full- length aFGF) by limited proteolytic processing (Jaye et al., 1986; Burgess et al., 1986). This latter form has been identified in some preparations of bovine aFGF (Burgess et al., 1985). By analogy with bFGF (Ueno et al., 1986), it seems reasonable to assume that the 154-amino-acid aFGF is the physiological form of the growth factor, and that the shorter forms are purification artifacts. Short forms of aFGF (140, 139 and 134 amino acids) have been expressed in Escherichia coli cultures by recombinant DNA techniques in several laboratories (Linemeyer et al., 1987; Jaye et al., 1987). However attempts to express the full-length aFGF in common expression plasmids rendered only minute amounts of protein, which could be detected exclusively by Western blot analysis (Jaye et al., 1987). Ortega et al. (1991) have reported that recombinant aFGF is inactivated very rapidly by Cys oxidation when exposed to air.
The minute amounts of full-length aFGF that are recovered from brain (Burgess et al., 1985) and cultures of transformed E.
454 Romero et al. (ELK J . Biochem. 241)
coli (Jaye et al., 1987) and the instability of the aFGF prepara- tions (Ortega et a!., 1991) have precluded high-resolution struc- tural studies of the physiological form of aFGF. Thus, a short- ened protein (140 amino acids) that has been stabilized by sub- stitution of Cys61 and His107 with Ala and Gly, respectively, had to be used in a first approximation of the resolution of aFGF three-dimensional structure (Zhu et al., 1991 ; Arakawa et al., 1993; Ortega et al., 1991). This mutant protein has the additional peculiarity of being fully active in the absence of heparin (Ara- kawa et al., 1993). Thus, these studies did not provide data on the structural basis of the regulation of FGF mitogenic activity by heparin. Blaber et al. (1996) have recently carried out crystal- lographic studies on a 141-amino-acid human aFGF purified by means of very harsh procedures (reversed-phase chromatogra- phy, denaturation with guanidine hydrochloride, and refolding). A single crystal was obtained that was suitable for crystallogra- phy with this preparation. This crystal, which decayed rapidly on X-ray exposure, had to be used for space-group determination and data collection. Nevertheless, a structure of low data com- pleteness and high B factors could be established on the basis of these analysis.
Zazo et al. (1992) assembled an expression plasmid that di- rects very efficient synthesis of full-length aFGF in E. coli, and increased considerably the stability of the aFGF preparations by modification of the purification procedure. With protein pre- pared by means of this modified procedure they showed that the susceptibility to oxidation in air is not an intrinsic property of aFGF but a consequence of reversed-phase chromatography. Here we show that these full-length aFGF preparations are suit- able for X-ray -diffraction studies, and report the three-dimen- sional structure of this protein at 0.25 nm with an overall data completeness of 96 %. These results may allow future high-reso- lution analyses that require stable crystals, such as the study of the structural basis of aFGF activation by heparin.
EXPERIMENTAL PROCEDURES
Reagents. Heparin-Sepharose was obtained from Pharmacia, bacteriological agar, yeast extract and tryptone from Gibco- BRL, and myo-inositol hexasulphate from Sigma. Macrosep ultrafiltration cartridges (Omega type ; low-binding protein membrane; 3-kDa cutoff), were from Filtron Technology Corpo- ration. Distilled water, filtered through a Milli-Q (Millipore) water purifier to which an Organex column (Millipore) had been added, was used in all solutions.
Cloning, expression and purification. Human aFGF (154 amino acids and 140 amino acids) was expressed from pMG60 and pMG47, respectively, and purified from E. coli cultures as described by Zazo et al. (1992). The protein was concentrated and the buffer changed to 10 mM sodium phosphate, pH 6.0, 1 mM 2-mercaptoethanol and 2 mM myo-inositol hexasulphate (Pineda-Lucena et al., 1994).
Fluorescence measurements. Fluorescence-emission spectra were performed in a Perkin-Elmer LS50 luminescence spectrometer at 298 K. Samples were excited at 280 nm. The excitation and emission slits were 2.5 nm and 10 nm, respective- ly. The amount of fluorescent sample used was in the linear range of measurement of the spectrofluorometer. The spectrum of the buffer was subtracted from the spectrum of the sample.
Protein sequencing. N-terminal-amino-acid-sequence analysis was carried out in a pulse-liquid Applied Biosystem 477 microsequencer switched on-line to an Applied Biosystem 120A phenylthiohydantoin-amino-acid analyser. For analysis of the crystallized protein, three crystals were harvested in 2.5 M ammonium phosphate, 0.1 M sodium acetate, pH 7.0, washed
Table 1. Data collection and refinement statistics. R,,,, = Z)(IZ, - (I)])/Z(I). where I , is the observed scale intensity of each reflection, j , and ( I ) is the mean value of multiple observations. Crystallographic R factor, R = ZI(F,,I - lFoII/CIF,l.
~
Parameter Value ~~~
Space group Maximum resolution Number of measured reflections Number of unique reflections
Completeness ( I > 2 c) Resolution range Number of protein atoms Number of reflections used in the refinement Crystallographic R factor rmsd from ideal bond lengths rmsd from ideal bond angles rmsd from ideal dihedral angles rmsd from ideal improper angles
' For a resolution shell of > 0.25 nm. For a resolution shell of > 0.26-0.25 nm.
five times in the same buffer and directly loaded in the sequena- tor cartridge after dilution in 5 pl distilled water.
Data collection and processing. X-ray -intensity data were collected by a MAR Imaging Plate System (Hendrixkenfter) with CuKa radiation from a Enraf NONIUS rotating anode gen- erator operating at 3.8 kW, and a crystal-to-detector distance of 100 mm. Data sets were collected with 1.5'-scans/frame and ex- posure of 1200s. The intensity data were processed with the program MOSFLM, version 5.2 (Leslie, 1991) and scaled, merged and reduced with the CCP4 package (Collaborative Computational Package, 1994). A total of 67939 reflections were collected of which 15739 were unique, having intensities greater than 2r~(1) yielding an Rmergeof 5.4%. Data collection and refinement statistics of the structure are summarized in Table 1.
RESULTS AND DISCUSSION
Crystallization. Crystals of full-length aFGF were initially ob- tained by the hanging-drop vapour-diffusion method at room temperature from a protein solution of 30 mg/ml in 10 mM sodium phosphate, pH 7.0, 10 mM inositol hexasulphate. Each 10-pl drop contained equal parts of protein solution and 2.0 M ammonium phosphate, 0.1 M sodium acetate, pH 7.0. The crys- tals grew to a maximal size of about 0.2X0.2X0.08 mm within 3 months, and diffracted beyond 0.35-nm resolution. The crys- tallization process was subsequently optimized with different pH values to improve the quality of the crystals. Crystallization conditions at pH 6.5 produced crystals with adequate features for structural analysis, after three months. These crystals diffract to a Bragg spacing greater than 250pm and show almost no diffraction decay upon X-ray exposure. Their space group was identified as C2 and the cell constants as a = 11.61 nm, b = 4.86 nm, c = 9.83 nm and p = 126.65' by precession photogra- phy. On the assumption of three independent molecules (3 X 17 500 Da)/asymmetric unit, a V,,, of 0.0021 1 nm'/Da (Mat- thews, 1968) was estimated, which implies a solvent content of the crystals of 48% (by vol.).
Structure solution. The structure was solved by molecular-re- placement techniques, by use of the coordinates of bovine aFGF refined to 0.2-nm resolution (Zhu et al., 1991) as a search model. A solution was found with the AMORE package (Navaza,
Romero et al. (Eur. J. Biochem. 241) 455
V L Y K AEGEI~FTA’oLTEKFNLPPG*oNYKKPKLLYC3oSNGGHFLRlL40PDGn/DGTRD50RSD - -
G H R S L F NEYNT’ ‘oYISKKHAEKN’ZoWFVGLKKNGS13oCKRGPRTHYG140QKAILFLPLP154/SSD
P-9 p-10 p-11 p-12
C
Fig. 1. Structure of full-length aFGF. (A) Stereo ribbon diagram of the polypeptide fold illustrating the location of the phosphate-binding site, represented in ball-and-stick mode (MOLSCRIFT; Kraulis, 1991). Side chains of amino acids that are hydrogen-bonded to the phosphate are depicted. (B) P-strand distribution along the amino acid sequence. Differences in the sequence of the bovine protein used in the crystallographic studies of Zhu et al. (1991) are listed on the top of the corresponding residues. Residues specific to the mutant protein used in these studies are underlined twice. AZu corresponds to Cys in native bovine aFGE V, the first residue of the proteins studied by Zhu et al. (1991) and Blaber et al. (1996). (C) Superposition of the main chain (Ca) trace of full-length human aFGF (thick trace) and bovine aFGF (thin trace). Significant differences were found in loops Ser64-Glu67 and Glu105-Tyr108 (maximal values: Va165, 0.28 nm; Glu10.5, 0.19 nm).
1994). The rotation function for full-length aFGF gave three prominent peaks at Eulerian angles of (171.6, 137.0, 252.6), (15.9, 122.7, 92.0) and (24.4, 21.1, 77.0) using data from 1.5 nm to 0.35 nm. No further peaks appeared at values higher than the 40% of the correlation-coefficient value of the first solution. Clear solutions were found in the translation function at frac- tional-coordinate values of (0.4653, 0.0017, 0.2702), (0.2558, 0.5748, 0.61 1) and (0.1880, 0.7301, 0.1047) (correlation coeffi- cient, 61.7%; R factor 34.2%).
Structure refinement. The three correctly oriented and posi- tioned full-length aFGF molecules were initially refined as a rigid body with the 0.8-0.3-nm data and the program XPLOR (Briinger, 1992). After 30 cycles of rigid-body refinement and 200 cycles of positional refinement, the initial R factor of 36.8% dropped to 25.7%.
Difference ‘omit’ maps (F,-F,) showed that several regions of the molecule required manual manipulation to fit correctly the electron density. This adjustments were carried out with the
456 Romero et al. (Em J. Biochem. 241)
Fig. 2. Ramachandran plot of main chain dihedral angles of the three molecules of the unit cell of full-length-aFGF crystals. All main chain conformational angles, 0 and Y, of non-glycine residues (D) except SerSO and Va151, lie within or near the allowed low-energy regions.
program FRODO (Jones, 1978) on an ESV-10 Graphic Display. Two segments, 20-24 and 150-154, required extensive re- building to fit the electron density. The data resolution was pro- gressively increased to 0.25 nm to include all 15 777 reflections. Solvent molecules were introduced at stereochemically reason- able positions, where peaks were found in F,-F, maps above 3.0a, by means of the program MAIN (Turk, 1992). With 28 water molecules, individual restrained B factors were refined and the R factor converged to 20.8% (data, 0.8-0.25 nm). An additional round of simulated-annealing refinement was per- formed by heating the system to 2000 K. The final R value con- verged to 20.4% for 15777 reflections containing residues 24- 153, 28 water molecules and 3 phosphate anions. The rmsd of temperature factors along the bonds was 0.032 nm2. The final rmsd from ideal values of stereochemical parameters (Engh and Huber, 1991) were 1.1 pm for bond distances and 1.9" for bond angles. The overall average B factors for all subunits were simi- lar: 0.353 nmz for molecule A, 0.335 nm2 for molecule B, and 0.347 nmz for molecule C. The average temperature factors were 0.342 nm2 for all protein atoms, 0.386 nm2 for the 28 water mol- ecules, and 0.392 nm2 for the 3 phosphate molecules. The final refinement statistics are shown in Table 1.
Overall description of the structure. The final model consists of 3 molecules of the protein, 28 water molecules and 3 phos- phate anions, each attached to one full-length aFGF molecule. In all 3 crystallographically independent monomers, the first 22 residues are disordered and not visible in the electron-density map, although in the second aFGF molecule (named molecule B), residues 20-24 could be modelled on the basis of a weak electron density. A ribbon plot (Kraulis, 1991) of the molecule is shown in Fig. 1 A. The phosphate anion at the position it occu- pies with respect to the protein a-backbone, and the side chains of the amino acids to which its is hydrogen bonded are included in the plot. The protein folds according to a p-trefoil motif. This
fold is formed by six /?-strand pairs, five of them with hairpin structure (p2-/?3, /?4-/?5, P6-p7, /?8-/39, /?lO-/?ll), and an- other (/?l -/?I 2) that, although topologically equivalent to the other five and sometimes referred as the sixth hairpin, is not such (Murzin et al., 1992). Three of these pairs form a six- stranded barrel structure (pl -P12, B4-p.5, P8-pS) and the other three are in a triangular array that caps the barrel. The distribution of /?-strands along the amino acid sequence is shown in Fig. 1 B. A Ramachandran plot of the structure is shown in Fig. 2. Conformational angles generally lie in areas of low en- ergy. The predominance of /? character in the structure is appar- ent throughout the cluster in the region about (-120", 135"). The other cluster around (-60", -30") results from residues involved in turns. Asnl20 and His107 are found in the left- handed a-region. Extensive side chain to main chain and side chain to side chain hydrogen bonding occurs in the structure (Table 2). The interior of aFGF is mainly hydrophobic while the surface is highly polar. Three polar residues (His35, Tyrl11 and Serll3) point to the interior and establish hydrogen bonds with main chain or side chain groups. There are eight side chain to main chain and ten side chain to side chain contacts, which are long-range interactions between sequentially distant parts of the chains, most of them connecting strands and loops of the mole- cule.
The polypeptide fold of the full-length human aFGF reported here is similar to that described by Zhu et al. (1991) for a 140-amino-acid bovine [Ala6l,Glul07]aFGF mutant protein (Fig. 1 B). The rmsd for the Ca-atoms is 0.061 nm. Nevertheless, significant differences were encountered between the structures reported here and by Zhu et al. in the loops connecting P-strands 4-5 and 8-9, respectively (Fig. 1B and 1C). Differences in the last loop are perhaps the exclusive consequence of the substitu- tion of Gly107 by His in the mutant protein studied by Zhu et al. (1991). The structure reported here is similar to the preliminary
Romero et al. ( E m J. Biochern. 241) 457
Table 2. Side chain hydrogen-bond contacts for full-length human aFGF.
structure in solution of a 132-amino-acid form of the human protein reported by Pineda-Lucena et al. (1994).
The three-dimensional structure of a shortened form (141 amino acids) of human aFGF has been recently reported by Blaber et al. (1996). However, the single crystal available for the X-ray analysis only allowed solution of the structure at 0.2- nm resolution and 50% completeness of the data. Furthermore, inspection of the coordinates reveal that approximately 30 % of the residues have B-factors with the maximum values handled by X-PLOR (1 nm’). Apparently the differences reported above between the structures of the human aFGF presented here and the structure of bovine aFGF (Zhu et al., 1991) were not observ- able in the studies of Blaber et al. (1996). A topological location, different from that reported here, has been assigned by Blaber et al. (1996) to an amino acid stretch (residues 115-121; Fig. 1 B) that seems to be relevant for the biological activity of FGF, and for the differences in receptor-subtype specificity
between aFGF and bFGF (Baird et al., 1988; Springer et al., 1994). Thus, while Blaber et al. (1996) locates this stretch in the p-strand pair 8-9, in the structure presented here, they were assigned to the P-strand pair 9- 10, in agreement with the previ- ous reports of Zhu et al. (1991) and Pineda-Lucena et al. (1994). Superposition of the a-backbone of all these structures demon- strates (data not shown) that the disagreement is a consequence of a miscounting of the p-strand elements of the structure, and not of the lack of one of the characteristic &strands of the p- trefoil topology in the structure reported by Blaber et al. (1996). Comparison of this structure with that reported here and by Pineda et al. (1994) shows that His106 (His120 according to the numbering scheme used here) is actually Asnl20 and not a dif- ferent amino acid of the human aFGF used by Blaber et al. (1996).
Examination of the structure reported here does not suggest that substitution of Cys61 by Ala has important structural conse- quences, in agreement with both the spontaneous substitution of this residue by Ser in human aFGF (GimCnez-Gallego et al., 1986b) and the equivalent specific mitogenic activities of the wild-type bovine aFGF and the [Ala61, Gly107laFGF (Arakawa et al., 1993). The high similarities between the structures re- ported here and by Zhu et al. (1991) may suggest that Cys sub- stitution makes aFGF more stable by favouring the acquisition of the native structure when the protein refolds during its purifi- cation procedure and by eliminating a readily oxidizable group. Arakawa et al. (1989) have reported that denaturation and re- folding seem to increase considerably the specific activity of a mutant FGF protein in which all the Cys residues have been substituted. Further, aFGF in which Cys residues have been sub- stituted by Ser are considerably more active than the native pro- teins when they are purified by a protocol that includes a re- versed-phase-chromatography step (Seno et al., 1988 ; Ortega et al., 1991), a procedure that involves some refolding of the pro- tein (Katzenstein et al., 1986), but not when the reversed-phase chromatography is avoided (Navarro, M. L., personal communi- cation).
The hairpin between /I-strands 8 and 9. Segment Tyrl08- Lys126, which comprises part of @-strands 9 and 20 and the loop connecting them (Fig. 1 B), have been shown to interact with the
Fig.3. Stereo representation of the region around GlulOl, which shows that GlulOl packs closely against ’Qrlll and AsnlOS. These data suggest that the three residues should contribute appreciably to the stabilization of this region of the protein.
45 8 Romero et al. (Em J. Biochenz. 241)
Fig. 4. Stereo-plot of the 2F,-F, electron-density map (contour level, la) at the C-terminus (Va1151, Ser152 and Ser153).
FGF receptor at the cell membrane (Baird et al., 1988). It has been reported recently that mutation of the Glu residue of bFGF in the equivalent topological position as GlulOl of aFGF drasti- cally lowers the specific activity of the protein (Zhu et al., 1995). This residue belongs to p-strand 8 that forms a hairpin with p- strand 9. The region around GlulOl (Fig. 3) shows that it packs closely against Tyrl11 and Asnl09. These data strongly suggest that these three residues should contribute appreciably to the stabilization of this region of the protein. Although it is not pos- sible to eliminate the possibility that the superficial 062 of GlulOl, the only unburied part of GlulOl, belongs to the recep- tor-binding site of aFGF, it is apparent that substitution of GlulOl may cause a structural rearrangement of p-strands 9 and 10 and, thus, indirectly affect the affinity of aFGF for its recep- tor. The structural importance of Tyr l l l , GlulOl and Asnl09 for the biological activity of FGFs is emphasised by the strong conservation of the short stretch between the nine members of the family (Dickson et al., 1990; Tanaka et al., 1992; Miyamoto et al., 1993). Similarly, Serll3 which is conserved in the whole family, is hydrogen bonded to the backbone amide of Hisll6. This bond should also contribute to the correct shape of the loop that connects P-strands 9 and 10. Consequently, this bond has a strong influence in the proper recognition of aFGF by its recep- tor, since the loop is a key part of the receptor-recognizing do- main of aFGF (Baird et al., 1988; Springer et al., 1994). At the short loop that connects p-strands 8 and 9 is Hisl07. According to the data of Arakawa et al. (1993), substitution of this residue by Gly cancels the heparin dependence of the mitogenic activity of aFGF. This substitution increases the hydrophobicity of the protein (Arakawa et al., 1993). The effects of the substitution of His107 by Gly have been interpreted on the basis of the better accommodation of the later residue in this p-turn. Nevertheless, comparison of the structures reported here and by Zhu et al. (1991) shows that the p-turns with His or Gly are very similar. Furthermore, His 107 appears in the Ramachandran plot in an allowed and favourable conformational zone of the left-handed 3, , region (Fig. 2) . Thus, His107 does not seem to create any sort of structural stress when it occupies this position. According to these results, the heparin independence conferred on the pro- tein by the substitution of His107 by Gly should perhaps be correlated with the modification of the hydrophobicity caused
0.01
0 b N
2 w 0.00 g 0.01 3 8 (0 m a
0.00 0 10 20 0 10 20
TIME (min)
Fig. 5. Edman degradation of the N-terminus of the pooled protein from three crystals of the same batch as that used in the X-ray deter- mination of the three-dimensional structure of full-length aFGF. The first four cycles are shown. Peaks labelled 0 correspond to sequencing by-products dimethylphenylthiourea, diphenylurea and diphenylthiourea. Amino acids corresponding to the first four residues at the N-terminus of full-length aFGF are labelled with single-letter codes.
4
? - - i - T - 1 1 7
0.0 1 .o 2.0 3.0
[UREA1 (M)
Fig. 6. Fluorescence enhancement of a solution of 15 pg proteidml by increasing urea concentrations. (O), 139-residue aFGF; (a), 154- residue aFGF. Best-fit values for the data to a two-state transition (con- tinuous line; Pace, 1986; Santoro and Bolen, 1988) yields AGH,?” values of 5.5 kT/mol and 17.6 kJ/mol for the 154-amino-acid and 140-amino- acid aFGF, respectively.
by this substitution (Arakawa et al., 1993) and with the different orientation of the short loop that connects p-strand 8 and 9, as position 107 is very close to one of the receptor-binding regions of aFGE Nevertheless, the possibility of many other effects that involve direct interactions with the receptor of the histidine side chain, including those through hydrogen bonds or charge, can not be eliminated. Probably, the exact interpretation of the ef- fects on the heparin dependence of this substitution will require the resolution of aFGF bound to its receptor.
The /3-strand pair p1-D. The electron-density map of the re- gion of the protein corresponding to the C-terminus is shown in Fig. 4. In contrast to bovine aFGF, for which the structure could not be traced beyond Pro150 (Zhu et al., 1991), the good quality
Romero et al. ( E m J. Biochern. 241) 459
Fig. 7. Stereoview of the 2F,-F, electron-density map (contour level, la) around the phosphate-binding site. The phosphate is hydrogen bonded to the backbone amide nitrogen of Lys127, the Nb' of Am32 and the N E of Lysl32.
of the map obtained for the full-length human aFGF allowed structure prediction of the whole C-terminal segment, except for the last residue (Asp154). However, at the N-terminus, it was not possible to trace the chain before Tyr22, although in molecule B, some weak electron density was observed, which probably corresponded to Gly20 and Asn21. To check whether the scarce electron density was a consequence of proteolysis, the N-termi- nus of the batch of crystals used for X-ray-data collection was determined by Edman degradation. The sequence corresponding to the N-terminus of aFGF (Ala-Glu-Gly-Glu) is distinguishable above the background (Fig. 5), although the analysis was carried out at a sensitivity close to the limit of the sequencer. Conse- quently, it seems that the undetectability of residues 1-20 in the electron-density maps has to be attributed to a high flexibility of this region of the protein. The fragmentary electron densities observed at Gly20 and Am21 indicates the point where the spe- cific interaction of N-terminus with the remainder of the mole- cule begins. For bovine aFGF, restriction of flexibility started after Lys24 (Zhu et al., 1991), according to the numbering scheme used here. The structure of human aFGF presented here shows a hydrogen bond between Val 151 and Pro 25 that could not be observed in the structure of bovine aFGF (Zhu et al., 1991). Additional hydrogen bonding and restricted mobility of the N-terminal and C-terminal regions of the protein show that this part of the structure is somewhat stiffer in full-length aFGF than in other shortened forms. This additional rigidity is proba- bly the reason for the higher stability of the full-length aFGF than of the 140-amino-acid aFGF (Fig. 6). This difference con- trasts with the equivalent stabilities of the 140-amino-acid and the 132-amino-acid aFGF (Pineda-Lucena et al., 1994), a simi- larity that agrees with the disorder of the nine N-terminal resi- dues of the 140-amino-acid aFGF observed by Zhu et al. (1991). Our results confirm that the disorder of the N-terminus of aFGF reported by Zhu et al. (1991) is not a peculiarity of the shortened form of the protein that they solved the structure of. Since most of the N-terminus does not seem to contribute much to the main- tenance of the three-dimensional structure of the protein, the conservation of its sequence between the different animal spe- cies perhaps is related to other physiological functions as the cellular sorting of the protein (Rothman and Wieland, 1996) a n d or to specific activities of this stretch of amino acids, since dif- ferent hormone-like activities have been shown for N-terminal
fragments of FGF (Gonzilez et al., 1994; Sasaki et al., 1995, 1966; Li et al., 1996).
myo-inositol-hexasulphate-binding site. NOESY spectra of aFGF bound to myo-inositol hexasulphate show that this com- pound interacts with residues Lys126, Lys127, Arg133 and Lys142 of the protein (Pineda-Lucena et al., 1994). Our electron- density map does not show any evidence of myo-inositol hexa- sulphate bound to aFGF, although the crystallization was carried out at concentrations of myo-inositol hexasulphate at least three orders of magnitude above its dissociation constant (Pineda-Lu- cena et al., 1994). Interaction of myo-inositol hexasulphate with aFGF should be mainly electrostatic, as is the interaction of hep- arin (Thompson et al., 1994). Thus, the high ionic strength (2 M ammonium phosphate) of the crystallization conditions probably displaced myo-inositol hexasulphate from its binding site. 1.5 M NaCl fully displaces heparin, whose dissociation constant for aFGF is one order of magnitude lower than that of myo-inositol hexasulphate (Pineda-Lucena et al., 1994). Fig. 7 shows a well delimited electron density at the myo-inositol-hexasulphate- binding site that can not be attributed to this compound. Instead, a phosphate anion could be fitted well to such electron density and refined with a low B temperature factor. This phosphate anion appears hydrogen bonded to the backbone amide nitrogen of Lys127, the N6 of Am32 and the N, of Lys132. Displacement of this phosphate by myo-inositol hexasulphate without substitu- tion of its hydrogen bonds with Lysl27 and Asn32 could ac- count for the considerable difference between the chemical shifts of these residues in the 'H-NMR spectra (1 ppm) of the free and the myo-inositol-hexasulphate-bound proteins (Pineda-Lucena et al., 1996). A sulphate group has been identified in the three- dimensional structure of bFGF reported by Eriksson et al. (1991), which appears to be hydrogen-bonded to residues posi- tionally equivalent to those to which the phosphate binds in aFGF in the structure reported here.
We thank J. M. Ramirez, for critical reading of the manuscript, J. Varela, for protein sequencing, and M. Zazo, for excellent technical as- sistance. This work was partially funded hy the Direccidn Generul de Invesrigucidn CientrFcu y Te'cnica, and Funducicin Gregorio Marattin- Boehringer lngelheim S. A. and Fundacidn Futuro-Boehringer Ingelheim S. A. agreements.
460 Romero et al. (Eus J. Biockem. 241)
REFERENCES Arakawa, T., Hsu, Y.-R., Shiffer, S. G., Tsai, L. B., Curless, C. & Fos, G.
M. (1989) Characterization of a cysteine-free analog of recombinant human basic fibroblast-growth factor, Biockem. Biopkys. Res. Com- mun. 161, 335-341.
Arakawa, T., Horan, T. P., Narhi, L. O., Rees, D. C., Schiffer, S. G., Holst, P. L., Prestrelski, S. J., Tsai, L. B. & Fox, G. M. (1993) Pro- duction and characterization of an analog of acidic fibroblast-growth factor with enhanced stability and biological activity, Protein Eng. 6, 541 -546.
Baird, A., Schubert, D., Ling, N. & Guillemin, R. (1988) Receptor- and heparin-binding domains of basic fibroblast-growth factor, Proc. Nut/ Acad. Sci. USA 85, 2324-2328.
Baird, A. & Bohlen, P. (1990) Fibroblast-growth factors, in Handbook i$ experimental pharmacology, peptide growth factors and their re- cepton (Sporn, M. B. & Roberts, A. B., eds) vol. 95, pp. 369-418, Springer, Berlin.
Blaber, M., DiSalvo, J. & Thomas, K. A. (1996) X-ray crystal structure of human acidic fibroblast-growth factor, Biochemistry 35, 2086- 2094.
Briinger, A. T. (1992) X-PLOR version 3.1. A system for X-ray crystal- lograpky uncl NMR. Yale University Press, New Haven CT.
Burgess, W. H., Mehlman, T., Friesel, R., Johnson, W. V. & Maciag, T. (1985) Multiple forms of endothelial cell growth factor, J . Bid. Ckern. 260, 11 389-11 392.
Burgess, W. H., Mehlman, T., Marshak, D. R., Fraser, B. A. & Maciag, T. (1986) Structural evidence that endothelial cell growth factor /3 is the precursor of both endothelial cell growth factor a and acidic fibroblast-growth factor, Proc. Nut1 Acad. Sci. USA 83. 721 6-7220.
Burgess, W. H. & Maciag, T. (1989) The heparin-binding (fibroblast) growth factor family of proteins, Annu. Rev. Biockem. 58, 575-606.
Collaborative Computational Project, Number 4. (1994) The CCP4 suite: programs for protein crystallography, Acta Cry.stallog,: Sect. D Bid. Crystallogs 50, 760-763.
Dickson, C., Acland, P., Smith, R., Dixon, M., Deed, R., MacAllan, D., Walter, W., Fuller-Pace, F., Kiefer, P. & Peters, G. (1990) Character- ization of 111-2: a member of the fibroblast-growth factor family, J. Cell Sci. Suppl. 13, 87-96.
Engh, R. A. & Huber, R. (1991) Accurate bond and angle parameters for X-ray protein refinement, Acta Crystallog,: A47, 392-400.
Eriksson, A. E., Lawrence, S. C., Weaver, L. H. & Matthews, B. W. (1991) Three-dimensional structure of human basic fibroblast-growth factor, Proc. Nut1 Acad. Sci. USA 88, 3441-3445.
GimCnez-Gallego, G., Con, G., Hatcher, V. B. & Thomas, K. A. (1986a) Human brain-derived acidic and basic fibroblast-growth factors : amino terminal sequences and specific mitogenic activities, Biockem. Biopkys. Res. Cummun. 135, 541 -548.
GimCnez-Gallego, G., Conn, G., Hatcher, V. & Thomas K. A. (1986b) The complete amino acid sequence of human brain-derived acidic fibroblast-growth factor, Biockem. Biopkys. Res. Commun. 138,
GimCnez-Gallego, G. & Cuevas, P. (1994) Fibroblast growth factor, a protein with a broad spectrum of biological activities, Neurol. Res.
Gonzilez, A. M., Carman, L. S., Ong, M., Ray, J., Gage, F. H., Shults, C. W. & Baird, A. (1994) Storage, metabolism, and processing of I-125-fibroblast-growth factor-2 after intracerebral injection, Brain Res. 665, 285-292.
Jaye, M., Howk. R., Burgess, W., Ricca, G. A,, Chiu, LM. , Ravera, M. W., O'Brien, S. J., Modi, W. S., Maciag, T. & Drohan, W. N. (1986) Human endothelial cell growth factor: cloning, nucleotide sequence, and chromosome localization, Science 233, 541 -545.
Jaye, M., Burgess, W. H., Shaw, A. B. & Drohan, W. N. (1987) Biologi- cal equivalence of natural bovine and recombinant human a-endothe- lial cell growth factors, J . Biol. Chem. 262, 16612-16617.
Jones, T. A. (1978) A graphics model building and refinement system for macromolecules, J. Appl. Crystallog,: 15, 24-31.
Katzenstein, G. E., Vrona, S. A,, Wechlsler, R. J., Steadman, B. L., Lewis, R. V. & Middaugh, C. R. (1986) Role of conformational changes in the elution of proteins from reversed phase HPLC col- umns, Proc. Not1 Acad. Sci. USA 83, 4268-4272.
611-617.
16, 313-316.
Kraulis, P. J. (1991) Molscript: a program to produce both detailed and schematic plots of protein structure, J . Appl. Crystallogr. 24, 946.- 950.
Leslie, A. G. W. (1991) Macromolecular data processing, in Cry);srallo- graphic computing (Moras, D., Podjarny, A. D. R. & Thierry, J. C., eds) vol. V, pp. 27-38, Oxford University Press, Oxford.
Li, A. J., Oomura, Y., Hori, T., Aou, S., Sasaki, K., Kimura, H. & Too- yama, I. (1996) Fibroblast-growth factor receptor-l in the lateral hy- pothalamic area regulates food intake, Exp. Neurol. 137, 318-323.
Linemeyer, D. L., Kelly, L. J., Menke, J. G., GimCnez-Gallego, G., DiSalvo, J. & Thomas, K. A. (1987) Expression in Esckerickin coli of a chemically synthesized gene for biologically active bovine acidic fibroblast-growth factor, Bin-technology ( N Y ) 5, 960-965.
Matthews, B. W. (1968) Solvent content of protein crystals, J. Mol. Biol.
Miyamoto, M., Naruo, K., Seko, C., Matsumoto, S., Kondo, T. & Kuro- kawa, T. (1993) Molecular cloning of a novel cytokine cDNA encod- ing the ninth member of the fibroblast-growth factor family, which has unique secretion property, Mol. Cell B i d . 13, 4251 -4259.
Navaza, J. (1994) AMORE: an automated package for molecular re- placement, Acta Crystallogs A50, 157-163.
Ortega, S., Schaeffer, M.-T., Soderman, D., DiSalvo, J., Linemeyer, D. L., GimCnez-Gallego, G. & Thomas, K. A. (1991) Conversion of cysteine to serine residues alters the activity, stability and heparin dependence of acidic fibroblast-growth factor, J . B i d . Chem. 266,
Pace, C. N. (1986) Determination and analysis of urea and guanidine bydrochloride denaturation curves, Methuds Enzymol. 131, 266-- 280.
Pineda-Lucena, A,, Jimenez, M. A., Nieto, J. L., Santoro, J., Rico, M. & GimCnez-Gallego, G. (1994) 'H-NMR assignment and solution strut.- ture of human acidic fibroblast-growth factor activated by inositol hexasulfate, J . Mul. Bid. 242, 81 -98.
Pineda-Lucena, A., JimCnez, M. A,, Lozano, R. M., Nieto, J. L., Santoro, J., Rico, M. & GimCnez-Gallego, G. (1996) Three-dimensional struc- ture of acidic fibroblast-growth factor in solution: effects of binding to a heparin functional analogue, J . Mol. Biol., in the press.
Potgens, A. J. G., Westphal, H. R., de Waal, R. M. W. & Ruiter, D. J. (1995) The role of vascular permeability factor and basic fibroblast- growth factor in tumor angiogenesis, Biol. Ckem. Hoppe-Seyler 376,
Rapraeger, A. C., Krufka, A. & Olwin, B. B. (1991) Requirement of heparan sulfate for bFGF-Mediated fibroblast-growth and myoblasf. differentiation, Science 252, 1705 - 1708.
Rothman, J . E. & Wieland, F. T. (1996) Protein sorting by transport vesicles, Science 272, 227-234.
Sasaki, K., Oomura, Y., Figurov, A,, Morita, N. & Yanaihara, N. (1995) An amino-terminal fragment peptide of acidic fibroblast-growth factor modulates synaptic transmission in rat hippocampal slices, Brain Rex Bull. 38, 185-191.
Sasaki, K., Oomura, Y., Li, A. J. , Hanai, K., Tooyama, I., Kimura, H., Yanaihara, N. & Hori, T. (1996) Actions of acidic fibroblast-growth factor fragments on food intake in rats, Obesity Rex 3, S697-S706.
Santoro, M. M. & Bolen, D. W. (1988) Unfolding free energy changes determined by the linear extrapolation method. 1 . Unfolding of phenylmethanesulfonyl a-chymotrypsin using different denaturants, Biochemistry 27, 8063-8068.
Seno, M., Sasada, R., Iwane, M., Sudo, K., Kurokawa, T., Ito, K. & Igarashi, K. (1988) Stabilizing basic fibroblast-growth factor using protein engineering, Biockem. Biopkys. Rex Commun. 151, 701 - 708.
Springer, B. A., Pantoliano, M. W., Barbera, F. A., Gunyuzlu, P. L., Thompson, L. D., Herblin, W. F., Rosenfeld, S. A. & Book, G. W. (1994) Identification and concerted function of two receptor binding surfaces on basic fibroblast-growth factor required for mitogenesis, J. Biol. Ckem. 269, 26879-26884.
Tanaka, A,, Miyamoto, K., Minamino, N., Takeda, M., Sato, B., Matsuo, H. & Matsumoto, K. (1992) Cloning and characterization of an an- drogen-induced growth factor essential for the androgen-dependent growth of mouse mammary carcinoma cells, Proc. Nut/ Acad. Sci. USA 89, 8928-8932.
Thomas, K. A,, Rios-Candelore, M. & Fitzpatrick, S. (1984) Purification and characterization of acidic fibroblast-growth factor from bovine brain, Proc. Nut/ Acad. Sci. USA 81, 357-361.
33,491 -497.
5842- 5846.
57-70.
Romero et al. ( E m J. Biockem. 241) 461
Thomas K. A., Rios-Candelore, M., Gimtnez-Gallego, C., DiSalvo, J., Bennett, C., Rodkey, J. & Fitzpatrick, S. (1985) Pure brain-derived acidic fibroblast-growth factor is a potent angiogenic vascular endo- thelial cell mitogen with sequence homology to interleukin 1, Proc. Nut1 Acad. Sci. USA 82, 6409-6413.
Thompson, L. D., Pantoliano, M. W. & Springer, B. A. (1994) Energetic characterization of the basic fibroblast-growth factor heparin interac- tion - identification of the heparin binding domain, Biochemistry
Turk, D. (1992) Weiterentwicklung eines Programms fur Molekulgrapkik und Electrondichte Manipulation und seine Anwendung auf versckie- dene Protein-Strukturaufklarungen, PhD thesis, Technische Uni- versitat Munchen.
Ueno, N., Baird, A,, Esch, F., Ling, N. & Guillemin, R. (1986) Isolation of an amino terminal extended form of basic fibroblast-growth factor, Biockem. Biopkys. Rex Commun. 138, 580-588.
Wilkie, A. 0. M., Morriss-Kay, G. M., Jones, E. Y. & Heath, J. K. (1995) Functions of fibroblast-growth factors and their receptors, Cur< Biol.
33, 3831-3840.
5, 500-507.
Yayon, A,, Klagsbrun, M., Esko, J. D., Leder, P. & Ornitz, D. M. (1991) Cell surface, heparin-like molecules are required for binding of basic fibroblast-growth factor to its high affinity receptor, Cell 64, 841 - 848.
Zazo, M., Lozano, R. M., Ortega, S., Varela, J., Diaz-Orejas, R., Rami- rez, J. M. & Gimtnez-Gallego, G. (1992) High-level synthesis in Esckerickia coli of shortened and full-length human acidic fibro- blast-growth factor and purification in a form stable in aqueous solu- tions, Gene (Amst.) 113, 231-238.
Zhu, X., Komiya, H., Chirino, A,, Faham, S., Fox, G. M., Arakawa, T., Hsu, B. T. & Rees, D. C. (1991) Three-dimensional structures of acidic and basic fibroblast-growth factors, Science 251, 90- 93.
Zhu, H. Y., Ramnarayan, K., Anchin, J., Miao, W. Y., Sereno, A., Mill- man, L., Zheng, J. H., Balaji, V. N. & Wolff, M. E. (1995) Glu96 of basic fibroblast-growth factor is essential for high affinity receptor binding - identification by structure-based site-directed mutagene- sis, J. Biol. Chem. 270, 21 869-21 874.
