- Email: [email protected]
Conformational changes in human prothrombin as detected by antibody populations
Biochimica et Biophysica Acta, 996 (1989) 95-102 Elsevier
95
BBA 33369
Conformafional changes in human prothrornbin as detected by antibody populations H e r b e r t K . F . L a u * a n d R o b e r t D. R o s e n b e r g Charles A. Dana Research Institute and Harvard Medical School, Boston, and the Department of Biology and Whitaker College, Massachusetts Institute of Technology, Cambridge, MA (U.S.A.)
(Received 7 July 1988)
Key words: Prothrombin fragment 1; Calcium; Conformational change; (Human)
The amino-terminal pepfides of human prothrombin corresponding to residues 1-51 and 52-156 have been isolated from a thrombin digest of prothrombin fragment 1. The products of digestion were purified by means of barium citrate and ammonium sulfate precipitations, followed by gel filtration and hydroxyapatite chromatographies. They were identified by their molecular sizes as well as their amino acid compositions. Peptides 1-51 (FIA) and 52-156 (F~B) were used as affinity iigands for the isolation of antibody populations from antisera that were elicited against human prothrombin or prothrombin fragment 1. These antibody populations displayed restricted specificity for the respective ligands as shown by competitive radioinununoassays. They were used to study the confonnafional changes in prothrombin and fragment 1. The F~A-specific antibody populations d:tected a confonnafional change which is stabilized by calcium ions and which has a transition midpoint at --0.2 mM calcium ion concentration. The Ft B-specific antibody populations identified a different confonnafional change which is destabilized by calcium ions and which has a transition midpoint at ~ 0.5 mM calcium.
introduction Human prothrombin has a molecular mass of approx. 70 kDa and like other coagulation zymogens, is composed of modular structures related to different functions. These are fragment 1 which contains ten 3,-earboxyglutamic acids whose post-translational formation requires vitamin K and whose function is believed to be binding of metal ions [1-6], fragment 2 whose function is to bind factor V/Va [7], and prethrombin 2, which gives rise to thrombin after cleavage at a disulfide bond [8,9]. The amino acid sequence of prothrombin has been delineated [1] and the crystal structure of fragment 1 has been reported [10-13]. Metal ions are important, in the presence of factor V/Va and phospholipid surface, for the activation of
Abbreviations: F1, human prothrombin activation fragment 1; 1:1+2, prothrombin activation fragmt:at 1 and 2; FIA, prothrombin N-terminal peptide 1-51; FIB, prothrombin N-terminal peptide 52-156; SDS-PAGE, sodium dodecyl sulfate-polyacrylam/de gel electrophoresis; PBS, phosphate buffered saline (0.05 M sodium phosphate/0.5 M NaCI (pH 7.5); SDS, sodium dodecyl sulfate. Correspondence (and present address): H.K.F. Lau, Department of Biochemistry, University of Hong Kong, Sassoon Road, Hong Kong.
prothrombin by factor Xa [14]. It has been shown that the interactions of the membrane surface with vitamin K-dependent proteins are affected by metal ions [15-17]. On the other hand, prothrombin itself also undergoes metal-dependent conformational changes when studied by physical methods [18-20] or immunochemical means [21]. The use of antibodies as structural probes for various antigenic determinants have been successfully applied to study the activation of fragment 2 and fragment 1 + 2 of human prothrombin [22] and the human thrombin-antithrombin complex [23]. Similar approach has also been adopted to delineate alterations of the three-dimensional structure of bovine and human pro.. thrombins [24-28]. We report here the isolation of the N-terminal residues 1-51 and 52-156 of human prothrombin, the use of these peptides in affinity chromatography to obtain four different antibody subpopulations, and the use of these antibodies in studies of the interaction of calcium ions with prothrombin and prothrombin activation fragment 1. Materials and Methods All chemicals employed were of reagent grade. DEAE-cellulose and hydroxyapatite were purchased from Bio-Rad. Sephadex G-25, G-100, Sepharose 4B
0167-4838/89/$03.50 © 1989 Elsevier Science Pubfishers B.V. (Biomedical Division)
96 and QAE-50 Sephadex were obtained from Pharmacia. Echis carinatus venom were purchased from Sigma. Hirudin was obtained from Pentapharm (Basel). Rabbit IgG and complete Freund's adjuvant were bought from Miles. Rabbit brain thromboplastin was provided by Ortho Diagnostics and purified as described [29].
Purification of prothrombin and prothrombin activation fragments Human prothrombin was prepared according to Shapiro et al. [30] with minor modifications [22]. The product appeared homogeneous and protein of the highest specific activity was used. Human thrombin was purified to homogeneity with a specific activity of 2800 N.I.H. units/mg by a procedure published before [31]. Prothrombin fragment F 1+2 was prepared according to Aronson et al. [32] and the trace amounts of contaminating prothrombin and prethrombin 2 were removed by means of heparin-bound Sepharose and QAE-50 Sephadex [22]. To obtain F 1, a by-product of large-scale preparation of human thrombin was used. This material was kindly provided by Dr, J.W. Fenton of the New York State Department of Health, Albany, N.Y., and contained a mixture of prothrombin activation fragments and intermediates. A typical experiment started by precipitating 1.74 1 of this material with 14.5 mM in barium citrate. The precipitate was washed with ice-cold water and dissolved in 400 ml of 0.2 M sodium citrate. The solution was then made 40~ saturated ammonium sulfate and the resulting precipitate discarded. The supernatant was then made 80~ in saturated ammonium sulfate, and the protein precipitated was dissolved in 30 ml 0.1 M NaCI/0.02 M Tris (pH 7.4) 6 ml of this material containing ~ 6 mg protein was chromatographed on a column of Sephadex G-100 (1.6 × 100 cm) which was developed at 8 m l / h in the same buffer. Fractions of 2.5 ml were collected. The slower-moving component contained a protein of approx. 21 kDa and was rechromatographed under the same experimental conditions to remove traces of contaminating proteins. To obtain prothrombin N-terminal residues 1-51 (FIA) and prothrombin N-terminal residues 52-156 (FIB), the method of Walz et al. [33,34] was adopted with modifications. 45 ml of Fl (1 mg/ml) was digested for 24 h at 37°C with 0.9 ml thrombin (4.6 mg/ml). Afterwards 20 mM dithiothreitol was added, the pH adjusted to 8.0, and the reaction continued for 20 h at 37 ° C. The products were carboxymethylated with 22 mM iodoacetamide for 3 h at 37 °C and at pH 8.9, and then precipitated with 20 mM barium citrate. This separated the reaction mixture into precipitate and supernatant fractions. No further purification was taken for the precipitate but the supernatant (47 ml) was extensively di~ysed against 0.1 M Tris/0.1 M NaC1 (pH 7.4) and chromatographed on a hydroxyapatite
column (0.9 x 3 cm). After loading the sample, the column was washed with the dialysis buffer followed by 0.1 M potassium phosphate (pH 6.8) at a flow rate of 15 m l / h and 2 ml fractions were collected.
Immunization and processing of antisera Four New Zealand white rabbits were immunized with 100/tg of F1 in complete Freund's adjuvant. Two goats were injected with 500/~g of prothrombin without the use of adjuvant. Hyperimmune antisera from these animals were obtained by repeated injections and antisera IgG fractions were obtained as described previously [22,23]. In order to obtain specific antibody populations against F1A or F1B, two rabbit antisera IgG raised against F I (R5 and R9) were affinity-fractionated. 4 mg of FlA and 6 mg of F1B were separaltely bound to 10 ml of Sepharose 4B using cyanogen bromide [35]. The gels were washed with 1 M acetic acid/0.5 M NaCI (pH 2.4) and then equilibrated with 0.05 M sodium phosphate/ 0.5 M NaCI (pH 7.5) (PBS) before use. Then 150 mg R5 and 100 mg R9 anti-F1 were filtered through the F1Aand FIB-bound Sepharose, respectively. The columns were washed with PBS until the 280 nm absorbance of the eluates were below 0.02, and the bound species were eluted with the acetic acid/NaCl buffer. The eluted materials were neutralized and dialysed against 0.13 M NaCI/0.02 M Tris (pH 7.4) (buffer A) and were designated R5 anti-FiA and R9 anti-FiB, respectively. In order to isolate goat anti-F1A or anti-FiB, the F fractions of a goat anti-prothrombin were used as the starting material. Two aliquots of 30 mg goat IgG were separately affinity-fractionated on the same F~A- or F l B-bound Sepharose columns, followed by the washing and elution procedure as above.
Buffers Buffers used in the binding studies were prepared with double-distilled water after passing through the Ultra Pure Water System (Hydro Services and Supplies, Braintree, MA) and contained -0.075 ppb Ca 2+ or Mg 2+ according to the manufacturer. Buffers prepared with this water had been used to study the Ca 2+dependent intrinsic fluorescence of F~ according to Nelsestuen [17] and Prendergast and Mann [18]. The fluorescence of F 1 in buffer A described a sigmoidal curve which plateaued at a fluorescence change of - 41% of the original fluorescence and had a transition midpoint at 0.25 mM C a C I 2. These observations were in agreement with those published before [17] and indicated that the quality of the water was of adequate purity.
Competitive assays These were carried out in radioimmunoassays using a second antibody system to separate free from bound
97 antigen [22,23]. F~ was iodinated to ~ 7.1.10 6 cpm/#g according to Greenwood et al. [36]. 50 #l 125I-labeled F 1 containing - 4 0 0 0 cpm (~ 1.8.10 -1° M) was preincubated with 5 0 / d of different concentrations of competing antigens for 60 min at room temperature. Then i00 ~I Gf ~_5 anti-F,A (1.10 -8 M) or R9 anti-F1B (3.8- 10 -8 M) was added, and the incubation eenfinued for 20 to 24 h at 4 ° C. The second antibody system wh:'.ch consisted of 200 #1 of sheep anti-rabbit IgG and 40/~g of nonimmune rabbit IgG, was added the next day. The precipitates were washed and counted after overnight incubation. The buffer in this assay was 0.155 M NaCl/0.005 M EDTA/0.0255 M sodium phosphate/20 mg per ml bovine serum albumin (pH 7.4). The competing antigens used were prothrombin, F 1, F 1A, F !B and F 1+ 2, in concentrations between 10-10 M to 10 -5 M. The molecular masses of prothrombin, thrombin, F l, F1 + 2 were assumed to be 70, 36, 21 and 32 kDa, respectively. Estimation of crossreacfivity and computation of the slopes of the respective logit-log dose response curves were obtained by fitting the raw data to a 'four parameter' model as described before [37,38].
Binding studies The same radioimmunoassay was utilized, except that no competing antigens were added. For antibody populations derived from goat, the second antibody system consisted of 200/~1 of rabbit anti-goat IgG and 40 #g of nonimmune goat IgG. All radiolabeled ligands, specific antibodies and the second antibody systems were either diluted into or dialysed against buffer A containing the appropriate concentrations of Ca 2+ or EDTA before use. The assays were carried out in triplicates and the results averaged. Three types of binding assays were performed:
(1) Antibody concentration-dependent binding experiments. Various concentrations of anti-F~A or anti-F~B were reacted with a fixed concentration of radiolabeled F~ or prothrombin, in the presence of either 10 mM EDTA or 10 mM CaCI 2. The fixed amount of 12SI-F1 used in this assay contained ~ 4000 cpm ( ~ 8.5.10-10 M) and was mixed with various concentrations of R5 anti-FlA (1.25.10 - 9 M to 1.25.10 -7 M), R9 anti-F1B (1.1.10 -l° M to 1.1.10 -8 M), goat anti-FiA (5.3. 10 -1° M to 5.3-10 -8 M), or goat anti-FiB (7.5.10 -1° M to 7.5.10 -8 M). The fixed amount of ~25I-labeled prothrombin contained -- 4000 cpm ( - 1.7.10-10 M), and was mixed with the same amounts of goat anti-F1A or goat anti-F~B antibodies as given above. (2) Ca 2 +-dependent binding experiments. Fixed concentrations of anti-F~A or anti-F~B antibodies were reacted with fixed concentrations of either radiolabeled F 1 or prothrombin, in the presence of 10 mM EDTA or 0.1 to 10 mM of CaC12. 1251-F1 (1.8.10 -1° M) was
diluted into buffers containing the various amounts of Ca 2+ and was allowed to react with R5 anti-F~ (1.7. 10 -l° M), R5 anti-F1A (2.2.10 ..9 M), R9 anti-F1 (1.7. 10 -9 M), R9 anti-FiB (1.3.10 -.9 M), goat anti-prothrombin (1.7.10 -8 M), goat anti-FiA (1.8.10 ..7 M), or goat anti-F~B (2.8- 10 -7 M). ~25I-prothrombin ( - 1.8 10 -1° M) was reacted similarly with goat anti-FlA (5.3• 10k 8) or goat anti-FiB (7.5.10 -9 M). •
(3) Radioiodinated ligand concentration-aependent binding experiments. Fixed concentrations of anti-F~A or anti-F~B were reacted with various amounts of ~25I-F~ or ~25I-prothrombin in buffers containing either 10 mM EDTA or 10 mM CaCI z. ~25I-F~ of 2.10 3 cpm to 5 • 10 5 cpm (1.5.10 -~° M to 3.8-10 -8 M) were diluted into buffer A containing 10 mM EDTA or 10 mM CaCI: and incubated with R5 anti-F~A (2.10 -9 M), R9 antiFIB (1.1 "10 -9 M), goat anti-F1A (1.9.10 -8 M), or goat anti-F~B (2.4.10-7 M). For binding of antibodies to radioactive prothrombin, 4.7.10 -~1 M to 9.4.10 -9 M of 125I-prothrombin were used to react with 8.10 -8 M of goat anti-F~A or 7.5.10 -9 M of goat anti-F~B. The data were plotted according to Scatchard [39] in the form of r/c versus r, where r is the concentration of the bound ligand per mole of the specific antibody and c is the concentration of the free ligand. We have tested the stabilities of the radioactively labeled F1 and prothrombin by incubating them for 48 h with the usual reagents used in the binding assay (including both second antibody systems) and examining their apparent molecular weights in SDS-polyacrylarrfide gel electrophoresis. In either case, a single radioactive peak corresponding to the original protein or peptide was observed, indicating no degradation had occurred during the binding experiments.
Gel electrophoresis The disc gel electrophoresis procedure of Davies et al. [40] as modified by Rosenberg and Waugh [29] was utilized to establish the homogeneity of polypeptides with respect to charge. SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was performed according to Laemmli [41] using a 15% crosslinked gel. Proteins were stained with Coomassie blue or by the periodic acid Schiff's base method [42,43]. Molecular weight markers included bovine serum albumin (68 kDa), ovalbumin (40 kDa), myoglobin (16 kDa). cytochrome c (12 kDa) and cyanogen bromide-cleaved peptides of myoglobin and cytochrome c.
Protein determination Protein concentrations were obtained by measurements at 280 nm, assuming 13.2, 16.2, 11.9 and 12.5 for 1 mg/ml of human prothrombin, thrombin, F1 and F1 ÷2, respectively [22], or determined by the method of Lowry et al. [44].
98
Amino acid analysis The amino acid compositions of F 1, F1A and FiB were determined by treatment wi ~'~ 6 M HC1 for 24, 48 and 72 h at l l 0 ° C , and the,, examination of the products in a Beckman 121M amino acid analyzer. Results
Preparation of FI, F~A and F~B F 1 was obtained from a n-,~xture of prothrombin fragments by means of barium citrate and ammonium sulfate precipitations, followed by gel filtration on Sephadex G-100. The matrix separated the sample into two peaks. Fractions containing the smaller molecules were pooled, concentrated and rechromatographed on the same column to yield an apparently homogeneous F 1 preparation which migrated with an apparent molecular mass of ~ 35 kDa. Upon SDS-PAGE under nonreduced conditions, a single Coomassie blue-staining band was visible which showed an apparent molecular mass of ~ 21 kDa. This observation agreed with Bensen and Hanahan [45] who found that prothrombin fragments containing "t-carboxyglutamic acids migrated with a higher apparent molecular mass on gel filtration. In the presence of reducing agent, a second faintly siained band appeared immediately below the major one. This second band had been observed before [33], and may be due to partial proteolysis of F l within one or more of its five disulfide bonds [10,11]. The amino acid composition of the Fl prepared is shown in Table I which is in excellent agreement with the published data [10,11,33, 341. F I A and FtB were obtained from thrombin digestion of Ft. The proteolytic fragments were reduced and
I ABLE 1
A~aino acid compositions of prothrombin fragments Fi, F!A and FIB An~no acid a
F!
F1A
1:1B
Asp Thr Ser Glu Pro Gly Ala Val Met lie Leu Tyr Phe Lys His Arg
15.3 19.7 10.8 24.3 9.4 10.8 13.0 8,9 0.9 4,9 8,1 5.5 4.7 4.7 2.1 12.7
3.6 6.7 3.6 12.5 0 1.0 7.0 3.3 0 0 3,8 1.6 3,7 1.8 0 2.5
13.0 13.3 7.1 12.9 10.5 9.7 5.9 4,8 0.7 3.8 5.3 3.3 1.3 2.4 2.3 9.0
a Number of amino acid residues per mole of prothrombin fragment. Tryptophan not determined.
a!kylated and then precipitated with barium citrate. The precipitate displayed a single band upon SDS-PAGE under nonreducing conditions with an apparent molecular mass of ~ 6 kDa. But like F1, there was a minor stained band migrating slightly faster than the major one when it was electrophoresed under reducing conditions. The amino acid composition of this peptide is shown in Table I. Glutamic acids accounted for -- 1 / 4 of the total amino acid residues, most of which were presumably y-carboxylated. The observation that this molecule was precipitated by barium ions, its molecular size and its amino acid composition were consistent with the assignment of this peptide being FlA. The supernatant of the thrombin-digested F1 contained two polypeptides when analyzed by SDS-PAGE and these were separated by chromatography on a hydroxyapatite column. The material not adsorbed contained a single peptide which showed an apparent molecular mass of ~ 15 kDa on SDS-PAGE under both reduced and nonreduced conditions. Upon gel filtration on Sephadex G-100, it also migrated with the same molecular mass. Its amino acid composition (Table I), apparent molecular mass and solubility in barium salt were consistent with this peptide being FiB.
Preparation and specificities of anti-F1A and anti-F1B Using FIA and FIB obtained above, affinity chromatographic columns were set up to isolate specific antibody populations against these fragments. - 0.15 to 0.2% of rabbit R5 and R9 anti-F l were adsorbed, but -270 of the goat anti-prothrombin antibodies were adsorbed by either FIA- or FiB-bound Sepharose columns under these experimental conditions. To test for the specificities of these antibodies, we used prothrombin, F1, F1 +2, FIA and FIB to displace 125I-Fi in the competitive radioimmunoassay for R5 anti-FlA. As expected, F 1 was the most effective in displacing the radioactive ligand. Prothrombin was quite effective, as only twice the molar excess was needed to displace 50~ 12SI-F1. It took 40-times molar excess of F IA and 400-times of FIB to compete in the same assay, while F l +2 did not react with R5 anti-F1A at all. The average slopes of their logit-log dose response curves, except for Fl + 2, were -0.7. The same prethrombin fragments were used to test for their crossreactivities for R9 anti-FiB. In this case, prothrombin and F 1+ 2 had similar slopes of -0.55 and were, respectively, 50- and 100-times less effective than F1 in competing with 125I-F1. FiB, however, competed with 125I-F1 for the antibody population only at a large molar excess (2000-fold), and with a reduced slope of -0.4. This crossreactivity was still superior to FIA which did not react with anti-FiB at all. Therefo~'e these antibodies showed restricted specificities towards the particular antigen that was used as the affinity ligand.
99
Binding studies of R.5 anti-Fi A and R9 anti-F1B antibodies y - C a r b o x y g l u t a m i c acids are believed to be responsible for binding Ca 2 + in prothrombin, and F 1A contains all ten of these residues while Fz B does not contain any. It would therefore be of interest to utilize the specific antibodies as molecular probes to study Ca2+-induced conformational changes in F~. We first investigated the b i n d i n g of the antibodies to ~251 F~ in the presence of either E D T A or Ca 2+. Fig. 1A shows that each concentration of R5 anti-F~A precipitated more F~ in the presence of Ca 2+ t h a n in the absence of Ca 2+. On the contrary, the opposite was true for R9 anti-F~B (Fig.
A
20
~
1.6
Ca 1.4 1.2
1.0
'.24./
• 0
60
A
, //--t-/P-r---
,
1.6] B
1B). N e x t we investigated the bindings of ~25I-F] to unfractionated R5 a n t i - F t, R5 anti-F]A, unfractionated R9 anti-F] and R9 anti-FiB, as a function of Ca 2+ concentrations. Fig. 2A shows Ca 2 +-dependent bi~ dings of F~ to R5 anti-F~ and R5 anti-F t A, with the ~raasition midpoints at 0.28 m M C a 2+, although / 0 anti-F~ appeared to precipitate m o r e F~ than R5 anti-F] A. Fig. 2B shows the binding of F~ to R9 anti-F~ also increased
//-.-//~ o
1.8
1.0
2.0
i'-¢/-r-¢;, = 3.0 5.0 iO.O
CALCIUM CONCENTRATION (mM)
Fig. 2. Calcium dependent binding of R5 and R9 antibodies to F~. (A) R5 anti-F] (¢) or R5 anti-F]A (o) was reacted with ]251-F] in buffers containing 0 to 10 mM CaCI 2 in a radioimmunoassay as detailed in Materials and Methods. The data are expressed as the ratio of the percentage bound F] to total F] in the presence of Ca 2+ divided by the percentage of bound F] to total F~ in the absence of Ca 2+. (B) R9 anti-F1 (O) or R9 anti-FiB (0) was reacted with 12Sl-F] in buffers containing 0 to 10 mM CaC! 2. The experimental procedure and calculation were performed as given above.
with increasing amounts of Ca 2+, reaching a half-maximal point at 0.18 m M Ca 2+. But the binding of F ! to R9 anti-F] B, in contrast, decreased with increasing a m o u n t s of Ca 2+, and its transition midpoint was higher at 0.45 m M C a 2+.
A
tll
' o 2-
Q
K
,o
7
0
!
2 Antibody
100
80 q
!
0.2 concentration
0.02 ( 108M
) 0.,
B
--"""o
i
,
t
2 r
3
6-
6O
x
:3 0
7
m or qO
O
2-
!
0
0 1.0 Antibody
0.1 concentration
0.01 ( 108M )
Fig. 1. Effect of calcium on the antibody-dependent binding to 1::1. Various concentrations of R5 anti-FiA (A) and R9 anti-FiB (B) were allowed to react with 125I-Fi in buffers containing either CaCI2 (O) or EDTA (0) in a radioimmunoassay as detailed in Materials and Methods. The percentage bound F1 per total F] was expressed as a function of the antibody concentrations.
I
Fig. 3. Interaction of R5 and R9 antibodies with FI in the presence of CaCI 2 or EDTA. These data are expressed using a Scatchard analysis where c is the concentration of free F! and r is the concentration of F! divided by the concentration of the respective antibody population. (A) R5 anti-FlA was reacted with leSI-F! in the presence of CaCl2 (e) or EDTA (o). (B) R9 anti-FiB was reacted with lesI-F1 in the presence of CaCI 2 (O) or I~DTA (o).
100 In order to quantitate these Ca2+-dependent effects, fixed concentrations of antibodies were incubated with various concentrations of 125I-F1 in either EDTA- or CaCl2-containing buffers. As shown in Fig. 3, the Scatchard plots were all nonlinear. R5 anti-FtA could be described to display two classes of binding sites in the presence of C a 2+, one with an average association constant (KA) of 1.4.10 9 M -1 and the other 1.4.108 M-1. The high-affinity binding sites constituted - 38% of the total available binding sites (Fig. 3A, Table IIA). In the absence of Ca 2 +, R5 anti-F~ A also displayed two classes of binding sites. The high-affinity sites had the same K A as when Ca 2 + were present, but they accounted for only - 2~ of the total available sites, with low-affinity binding predominated in this case (Fig. 3A, Table IIA). R9 anti-FIB, on the other hand, bound 1251-FI with the same high- and low-affinity constants in the presence or absence of Ca 2+, except that the number of high-affinity sites was bigger in the absence of Ca 2+ (Fig. 3B, Table IIA).
1.6 1.4 O rn
1.2
m 1.0 0.8
,
o
0.6
1.0
1.6
B
o
CO ,-.... m
TABLE Ii
Scott'hard analysis of the interaction of either 1251.FI or l"51-prothrombin with different antibody populations High K^ (M - ! )
Approx. ,~ total sites having
Low K A ( M - n)
high K^ (A) Using 12Sl-Fi as the ligand for binding antibodies: R5 anti-FiA Ca 2+ 1.4,10 ~ 38 EDTA 1,3,10 9 2 R9 anti-F, B Ca 2 + 3.5.10 9 33 EDTA 3.1.10 9 39 Goat anti-FtA Ca 2+ 1.3,101° 30 EDTA 1.7.10 9 30 Goat anti-FzB Ca 2+ 2.0.10 9 15 EDTA 1.8.10 9 31
1.4. l0 s 1.4.10 s 1.6. l0 s 2.1.10 s 1,5.10 ~ 1.6.10 8 1.5.107 7.3-107
(B) Using 1251-prothrombin as the ligand for binding antibodies: Goat anti-FiA Ca 2+ 3.8-10 9 30 6.0. l0 s EDTA 2.5. l0 9 35 4.1. l0 s Goat anti-FiB Ca 2+ 1.8- l0 9 33 1.8. l0 s EDTA 4.7-109 40 5.2.10 s
3.0
5.0
(mMI
m
1.2
/
1"0'
O~..x:k
• I/'"®
~
0.8
o #-o
0.6
Binding studies of goat anti-F! A and anti-Fj B When various concentrations of goat anti-F~A was diluted into buffers containing either 10 mM EDTA or CaCI 2, it was found that the antibody population bound more 125I-F~ in the presence of Ca 2+ than in its absence. The opposite was true for goat anti-FmB (data not shown). The binding of goat anti-FiA to 125I-F1 was dependent on Ca 2+ with a half-maximal point at 0.24 mM Ca 2+. On the other hand, the binding of goat anti-F! B tO 1251-F! decreased as Ca 2+ increased, while the halfmaximal point was also higher than at 0.51 mM Ca 2+ (Fig. 4).
2.0
Calcium c o n c e n t r a t i o n
! 1.0
m 2.0
~#..a 3.0
5.0
Calcium concentration (raM) Fig. 4. Calcium-dependent binding of goat anti-FiA and anti-FiB to F ! and prothrombin. These experiments were carried out and calculation expressed as given in Fig. 2. (A) Goat anti-FiA (0) or anti-FIB ( 0 ) was reacted with tasI-FI in buffers containing 0 to 5 mM CaCI 2. (B) Goat anti-FIA (e) or anti-FiB ( 0 ) was reacted wi:h t251-prothrombin in buffers containing 0 to 5 mM CaCI 2-
In order to determine whether prothrombin also exhibited these Ca 2+-dependent effects, ~25l-prothrombin was used as the radioactive ligand for binding goat antibodies. Similar results to those using ~251-F! as the radioactive ligand were obtained (Fig. 4B). Scatchard analyses of the binding between the goat antibodies with F1 and prothrombin were then performed. The results are presented in Table If. Again all Scatchard plots were nonlinear and two classes of binding sites could be used to describe each curve. F 1 bound goat anti-FiA with about 8-fold higher affinity in the presence of Ca 2+ (1.3.101° vs. 1.7.10 9 M - I ) than in its absence. On the other hand F 1 bound goat anti-F~B with twice as many high-affinity sites in the absence of C a 2+ (31 vs. 15t~).
Prothrombin also bound goat anti-F~A with slightly higher affinity in the presence of Ca 2+. In contrast, prothrombin bound goat anti-F~B with 2.5-times higher affinity (4.7.10 9 vs. 1.8-10 9 M - l ) in the absence of Ca 2+. Discussion
We have purified human prothrombin fragment 1 to apparent homogeneity and from it, the amin:~ terminal peptides FIA and F1B. They were identified by binding
101 properties to barium salt, SDS-PAGE, gel filtration and amino acid compositions. F1A and FiB were utilized as affiPAty ligands to isolate specific antibodies which showed restricted specificities toward the respective ligands and were useful probes to study the conformational changes induced by Ca 2+ in native 1::1 and prothrombin. Two types of conformational change could be detected and both were Ca2+-dependent. In the first type, Ca 2+ was able to stabilize a conformation in F1 when F 1 reacted with R5 anti-FlA (Fig. 1A), or when 1"1 reacted with unfractionated R5 and R9 anti-F1 (Fig. 2), or when Ft or prothrombin reacted with goat anti-prothrombin or goat anti-FlA (Fig. 4). Presumably Ca 2+ bound to some common conformations in prothrombin, F~ and F1A and induced conformational changes in these molecules. The binary complex¢s having the new conformation were more stable and could precipitate more of the specific antibodies. The interactions were the results of either higher affinity or larger number of high-affinity binding sites in the presence of Ca 2+. The half maxima of this transition were between 0.18-0.28 mM Ca 2+ and corresponded closely the dissociation constants obtained by direct binding of Ca 2+ to F1 [15], with circular dichroism [19], fluorescence [16,18] or immunochemical means [24,25]. Furthermore, the common epitopes on these molecules were necessarily immunodominant features and probably involve the y-carboxyglutamic acid residues or regions surrounding them. This conformational change is probably similar to that observed in the study of F 1 crystal structure. There was no density in the amino acid sequence 1-35 in the absence of metal ions [12,13], indicating a disorganized structure in the y-carboxyglutamic acid-containing region. However, addition of Sr 2+ ',ransformed the region to become electron dense and therefore presumably an ordered structure [46]. The second type of conformational change was detected by R9 anti-FiB and goat anti-F~B. It is also Ca2+-dependent, but Ca 2÷ in this case destabilized the new conformation. Ca 2+ probably formed binary complexes with F~ and prothrombin first, lowered the stability of the subsequent ternary complexes when anti-F~B combined with them, and thus gave rise to less precipitates. These interactions were resulted from either lower affinity binding or smaller number of high-affinity binding sites in the presence of Ca 2+. The conformer recognized in prothrombin and F t was probably localized on the FIB portion of these molecules, since the antibodies which recognized this conformational change did not recognize F~A in the competitive radioimmunoassay. However, it cannot be excluded that neoantigens could have arisen when F~A joins F~B in native F~ or prothrombin, and that it was this new format that was destabilized by Ca 2+. Tiffs conformational change took place with transition midpoints at
considerably higher Ca 2+ (0.45-0o61 raM). It is probably a weaker transition since it was completely masked by the first type of conformational change when unfractionated R5 and R9 anti-F1 interacted with F 1 in the presence of Ca a+ (Fig. 2). It is of interest to note that FIB consists of a short helical region and a compact, organized kringle structure [12,13], in which no metal-dependent changes have been described before. The significance of this conformational change is not known, but it could be involved in the factor Xa-dependent cleavage of prothrombin or in the binding of the zymogen of the phospholipid surface. References 1 Magnusson, S., Petersen, T.E., Sottrup-Jensen, L. and Claeys, H. (197~) in Proteases and Biological Control, Cold Spring Harbor Laboratory, pp. 123-149, Cold Spring Harbor, NY. 2 Magnusson, S., Sottrup-Jensen, L., Petersen, T.E., Morris, H.R. and Dell, A. (1974) FEBS Lett. 44, 189-193. 3 Neisestuen, G.L., Zytkovicz, T.H. and Howard, J.B. (1974) J. Biol. Chem. 249, 6347-6350. 4 Stenflo, J., Fernlund, P., Egan, W. and Roepstorff, P. (1974) Proc. natl. Acad. Sci. USA 71, 2730-2733. 5 Stenflo, J. and Ganrot, P.-O. (1973) Biochem. Biophys. Res. Commun. 50, 98-104. 6 Nelsestuen, G.L. and Suttie, J.W. (1974) Biochemistry 11, 4961-4964. 7 Nesheim, M.E., Hibbard, L.S., Tracy, P.B., Bloom, J.W., Myrmei, K.H. and Mann, K.G. (1980) in The Regulation of Coagulation (Mann, K.G. and Taylor, F.B., eds.), pp. 145-149, Elsevier/ North-Holland, New York. 8 Harmison, C.R., Landaburu, R.H. and Seegers, W.H. (1961) J. Biol. Chem. 236, 1693-1696. 9 Conway, E.M., Lau, H.K.F., Bauer, K.A. and Rosenberg, R.D. (1987) The development of a radioimmunoassay for quantitating prethrombin 2 in human plasma. J. Lab. Clin. Med. 110, 567-575. 10 Olsson, G., Andersen, L., Lindquist, O., Sjolin, L., Magnusson, S., Peterson, T.E. and Sottrup-Jensen, L. (1982) FEBS Lett. 145, 317-322. 11 Tulinsky, A., Park, C.H. and Rydel, T.J. (1985) J. Biol. Chem. 260, 107718b110778. 12 Park, C.H. and Tulinsky, A. (1986) Biochemistry 25, 3977-3982. 13 Harlos, K., Boys, C.W.G., Holland, S.K., Esnouf, M.P. and Balke, C.C.F. (1987) FEBS Lett. 224, 97-103. 14 Mann, K.G. (1984) Prog. Hemostasis Thromb. 7, 1-23. 14 Bajaj, S.P., Butkowski, R.J. and Mann, K.G. (1975) J. Biol. Chem. 250, 2150-2156. 16 Nelsestuen, G.L., Broderius, M. and Martin, G. (1976) J. Biol. Chem. 251, 6886-6893. 17 Nelsestuen, G.L. (1976) J. Biol. Chem. 251, 5648-5656. 18 Prendergast, F.G. and Mann, K.G. (1977) J. Biol. Chem. 252, 840-850. 19 Bloom, J.W. and Mann, K.G. (1978) Biochemistry 17, 4430-4438. 20 Tucker, M.M., Nesheim, M.E. and Mann, K.G. (1983) Biochemistry 22, 4540-4546. 21 Stenflo, J. and Ganrot, P.-O. (1972) J. Biol. Chem. 247, 8160-8166. 22 Lau, H.K.F., Rosenberg, J.S., Bceler, D.L. and Rosenberg, R.D. (1979) J. Biol. Chem. 254, 8751-8761. 23 Lau, H.K.F. and Rosenberg, R.D. (1980) J. Biol. Chem. 255, 5885-5893. 24 Furie, B., Provost, K.L., Blanchard, R.A. and Furie, B.C. (1978) J. Biol. Chem. 253, 8980-8987. 25 Furie, B. and Furie, B.C. (1979) J. Biol. Chem. 254, 9766-9771.
102 26 Tai, M.M., Furie, B.C. and Furie, B. (1984) J. Biol. Chem. 259, 4162-4168. 27 Owens, J., Lewis, R.M., Cantor, A., Furie, B.C. and Furie, B. (1984) J. Biol. Chem. 259, 13800-13805. 28 Borowski, M., Furie, B.C., Bauminger, S. and Furie, B. (1986) J. Biol. Chem. 261, 1496986114975. 29 Rosenberg, R.D. and Waugh, D.G. (1970) J. Biol. Chem. 245, 5049-5056. 30 Shapiro, S.S. and Waugh, D.F. (1966) Thromb. Diath. Haem. 16, 469-480. 31 Rosenberg, J.S., Beeler, D.L. and Rosenberg, R.D. (1975) J. Biol. Chem. 250, 1607-1617. 32 Aronson, D.L., Steven, L., Ball, A.P., Franza, B.R., Jr. and Finayson, J,S. (1977) J. Cfin. Invest. 60, 1410-1418. 33 Walz, D.A., Hewett, Emmett, D. and Seegers, W.H. (1977) Life Sci, 20, 79-84. 34 Walz, D.A., Hewett-Emmett, D. and Seegers, W.H .(1977) Proc. Natl. Acad. Sci. USA 74, 1969-1972. 35 Porath, J., Axen, R. and Ernback, S. (1967) Nature (Lond.) 215, 1491-1492.
36 Greenwood, F.C., Hunter, W.M. and Glover, J.S. (1963) Biochem. J. 89, 114-123. 37 Redbard, D. (1974) Clin. Chem. 20, 1255-1270. 38 Rodbard, D., Lenox, R.H., Wray, H.L. and Ramseth, D. (1976) Clin. Chem. 22, 350-358. 39 Scatchard, G. (1949) Ann. NY Acad. Sci. 51, 660-672. 40 Davies, B.J. (1964) Ann. NY Acad. Sci. 121, 404-427. 41 Laemmli, U.K. (1970) Nature (Lond.) 227, 680-685. 42 McGuckin, W.F. and McKenzie, B.F. (1958) Clin. Chem. 4, 476-483. 43 Zaccharius, R.M., Zell, T.E., Morrison, J.H. and Woodlock, J.J. (1969) Anal. Chem. 30, 148-152. 44 Lowry, O.H., Rosebrough, N.J., Farr, A.L. and Randall, R.J. (1951) J. Biol. Chem. 193, 265-275. 45 Benson, B.J. and Hanahan, D.J. (!975) Biochemistry 12, 3265-3277. 46 Harlos, K., Holland, S.K., Boys, C.W.G., Burgess, A.I., Esnout, M.P. and Balke, C.C.F. (1987) Nature (Lond.) 330, 82-84.
95
BBA 33369
Conformafional changes in human prothrornbin as detected by antibody populations H e r b e r t K . F . L a u * a n d R o b e r t D. R o s e n b e r g Charles A. Dana Research Institute and Harvard Medical School, Boston, and the Department of Biology and Whitaker College, Massachusetts Institute of Technology, Cambridge, MA (U.S.A.)
(Received 7 July 1988)
Key words: Prothrombin fragment 1; Calcium; Conformational change; (Human)
The amino-terminal pepfides of human prothrombin corresponding to residues 1-51 and 52-156 have been isolated from a thrombin digest of prothrombin fragment 1. The products of digestion were purified by means of barium citrate and ammonium sulfate precipitations, followed by gel filtration and hydroxyapatite chromatographies. They were identified by their molecular sizes as well as their amino acid compositions. Peptides 1-51 (FIA) and 52-156 (F~B) were used as affinity iigands for the isolation of antibody populations from antisera that were elicited against human prothrombin or prothrombin fragment 1. These antibody populations displayed restricted specificity for the respective ligands as shown by competitive radioinununoassays. They were used to study the confonnafional changes in prothrombin and fragment 1. The F~A-specific antibody populations d:tected a confonnafional change which is stabilized by calcium ions and which has a transition midpoint at --0.2 mM calcium ion concentration. The Ft B-specific antibody populations identified a different confonnafional change which is destabilized by calcium ions and which has a transition midpoint at ~ 0.5 mM calcium.
introduction Human prothrombin has a molecular mass of approx. 70 kDa and like other coagulation zymogens, is composed of modular structures related to different functions. These are fragment 1 which contains ten 3,-earboxyglutamic acids whose post-translational formation requires vitamin K and whose function is believed to be binding of metal ions [1-6], fragment 2 whose function is to bind factor V/Va [7], and prethrombin 2, which gives rise to thrombin after cleavage at a disulfide bond [8,9]. The amino acid sequence of prothrombin has been delineated [1] and the crystal structure of fragment 1 has been reported [10-13]. Metal ions are important, in the presence of factor V/Va and phospholipid surface, for the activation of
Abbreviations: F1, human prothrombin activation fragment 1; 1:1+2, prothrombin activation fragmt:at 1 and 2; FIA, prothrombin N-terminal peptide 1-51; FIB, prothrombin N-terminal peptide 52-156; SDS-PAGE, sodium dodecyl sulfate-polyacrylam/de gel electrophoresis; PBS, phosphate buffered saline (0.05 M sodium phosphate/0.5 M NaCI (pH 7.5); SDS, sodium dodecyl sulfate. Correspondence (and present address): H.K.F. Lau, Department of Biochemistry, University of Hong Kong, Sassoon Road, Hong Kong.
prothrombin by factor Xa [14]. It has been shown that the interactions of the membrane surface with vitamin K-dependent proteins are affected by metal ions [15-17]. On the other hand, prothrombin itself also undergoes metal-dependent conformational changes when studied by physical methods [18-20] or immunochemical means [21]. The use of antibodies as structural probes for various antigenic determinants have been successfully applied to study the activation of fragment 2 and fragment 1 + 2 of human prothrombin [22] and the human thrombin-antithrombin complex [23]. Similar approach has also been adopted to delineate alterations of the three-dimensional structure of bovine and human pro.. thrombins [24-28]. We report here the isolation of the N-terminal residues 1-51 and 52-156 of human prothrombin, the use of these peptides in affinity chromatography to obtain four different antibody subpopulations, and the use of these antibodies in studies of the interaction of calcium ions with prothrombin and prothrombin activation fragment 1. Materials and Methods All chemicals employed were of reagent grade. DEAE-cellulose and hydroxyapatite were purchased from Bio-Rad. Sephadex G-25, G-100, Sepharose 4B
0167-4838/89/$03.50 © 1989 Elsevier Science Pubfishers B.V. (Biomedical Division)
96 and QAE-50 Sephadex were obtained from Pharmacia. Echis carinatus venom were purchased from Sigma. Hirudin was obtained from Pentapharm (Basel). Rabbit IgG and complete Freund's adjuvant were bought from Miles. Rabbit brain thromboplastin was provided by Ortho Diagnostics and purified as described [29].
Purification of prothrombin and prothrombin activation fragments Human prothrombin was prepared according to Shapiro et al. [30] with minor modifications [22]. The product appeared homogeneous and protein of the highest specific activity was used. Human thrombin was purified to homogeneity with a specific activity of 2800 N.I.H. units/mg by a procedure published before [31]. Prothrombin fragment F 1+2 was prepared according to Aronson et al. [32] and the trace amounts of contaminating prothrombin and prethrombin 2 were removed by means of heparin-bound Sepharose and QAE-50 Sephadex [22]. To obtain F 1, a by-product of large-scale preparation of human thrombin was used. This material was kindly provided by Dr, J.W. Fenton of the New York State Department of Health, Albany, N.Y., and contained a mixture of prothrombin activation fragments and intermediates. A typical experiment started by precipitating 1.74 1 of this material with 14.5 mM in barium citrate. The precipitate was washed with ice-cold water and dissolved in 400 ml of 0.2 M sodium citrate. The solution was then made 40~ saturated ammonium sulfate and the resulting precipitate discarded. The supernatant was then made 80~ in saturated ammonium sulfate, and the protein precipitated was dissolved in 30 ml 0.1 M NaCI/0.02 M Tris (pH 7.4) 6 ml of this material containing ~ 6 mg protein was chromatographed on a column of Sephadex G-100 (1.6 × 100 cm) which was developed at 8 m l / h in the same buffer. Fractions of 2.5 ml were collected. The slower-moving component contained a protein of approx. 21 kDa and was rechromatographed under the same experimental conditions to remove traces of contaminating proteins. To obtain prothrombin N-terminal residues 1-51 (FIA) and prothrombin N-terminal residues 52-156 (FIB), the method of Walz et al. [33,34] was adopted with modifications. 45 ml of Fl (1 mg/ml) was digested for 24 h at 37°C with 0.9 ml thrombin (4.6 mg/ml). Afterwards 20 mM dithiothreitol was added, the pH adjusted to 8.0, and the reaction continued for 20 h at 37 ° C. The products were carboxymethylated with 22 mM iodoacetamide for 3 h at 37 °C and at pH 8.9, and then precipitated with 20 mM barium citrate. This separated the reaction mixture into precipitate and supernatant fractions. No further purification was taken for the precipitate but the supernatant (47 ml) was extensively di~ysed against 0.1 M Tris/0.1 M NaC1 (pH 7.4) and chromatographed on a hydroxyapatite
column (0.9 x 3 cm). After loading the sample, the column was washed with the dialysis buffer followed by 0.1 M potassium phosphate (pH 6.8) at a flow rate of 15 m l / h and 2 ml fractions were collected.
Immunization and processing of antisera Four New Zealand white rabbits were immunized with 100/tg of F1 in complete Freund's adjuvant. Two goats were injected with 500/~g of prothrombin without the use of adjuvant. Hyperimmune antisera from these animals were obtained by repeated injections and antisera IgG fractions were obtained as described previously [22,23]. In order to obtain specific antibody populations against F1A or F1B, two rabbit antisera IgG raised against F I (R5 and R9) were affinity-fractionated. 4 mg of FlA and 6 mg of F1B were separaltely bound to 10 ml of Sepharose 4B using cyanogen bromide [35]. The gels were washed with 1 M acetic acid/0.5 M NaCI (pH 2.4) and then equilibrated with 0.05 M sodium phosphate/ 0.5 M NaCI (pH 7.5) (PBS) before use. Then 150 mg R5 and 100 mg R9 anti-F1 were filtered through the F1Aand FIB-bound Sepharose, respectively. The columns were washed with PBS until the 280 nm absorbance of the eluates were below 0.02, and the bound species were eluted with the acetic acid/NaCl buffer. The eluted materials were neutralized and dialysed against 0.13 M NaCI/0.02 M Tris (pH 7.4) (buffer A) and were designated R5 anti-FiA and R9 anti-FiB, respectively. In order to isolate goat anti-F1A or anti-FiB, the F fractions of a goat anti-prothrombin were used as the starting material. Two aliquots of 30 mg goat IgG were separately affinity-fractionated on the same F~A- or F l B-bound Sepharose columns, followed by the washing and elution procedure as above.
Buffers Buffers used in the binding studies were prepared with double-distilled water after passing through the Ultra Pure Water System (Hydro Services and Supplies, Braintree, MA) and contained -0.075 ppb Ca 2+ or Mg 2+ according to the manufacturer. Buffers prepared with this water had been used to study the Ca 2+dependent intrinsic fluorescence of F~ according to Nelsestuen [17] and Prendergast and Mann [18]. The fluorescence of F 1 in buffer A described a sigmoidal curve which plateaued at a fluorescence change of - 41% of the original fluorescence and had a transition midpoint at 0.25 mM C a C I 2. These observations were in agreement with those published before [17] and indicated that the quality of the water was of adequate purity.
Competitive assays These were carried out in radioimmunoassays using a second antibody system to separate free from bound
97 antigen [22,23]. F~ was iodinated to ~ 7.1.10 6 cpm/#g according to Greenwood et al. [36]. 50 #l 125I-labeled F 1 containing - 4 0 0 0 cpm (~ 1.8.10 -1° M) was preincubated with 5 0 / d of different concentrations of competing antigens for 60 min at room temperature. Then i00 ~I Gf ~_5 anti-F,A (1.10 -8 M) or R9 anti-F1B (3.8- 10 -8 M) was added, and the incubation eenfinued for 20 to 24 h at 4 ° C. The second antibody system wh:'.ch consisted of 200 #1 of sheep anti-rabbit IgG and 40/~g of nonimmune rabbit IgG, was added the next day. The precipitates were washed and counted after overnight incubation. The buffer in this assay was 0.155 M NaCl/0.005 M EDTA/0.0255 M sodium phosphate/20 mg per ml bovine serum albumin (pH 7.4). The competing antigens used were prothrombin, F 1, F 1A, F !B and F 1+ 2, in concentrations between 10-10 M to 10 -5 M. The molecular masses of prothrombin, thrombin, F l, F1 + 2 were assumed to be 70, 36, 21 and 32 kDa, respectively. Estimation of crossreacfivity and computation of the slopes of the respective logit-log dose response curves were obtained by fitting the raw data to a 'four parameter' model as described before [37,38].
Binding studies The same radioimmunoassay was utilized, except that no competing antigens were added. For antibody populations derived from goat, the second antibody system consisted of 200/~1 of rabbit anti-goat IgG and 40 #g of nonimmune goat IgG. All radiolabeled ligands, specific antibodies and the second antibody systems were either diluted into or dialysed against buffer A containing the appropriate concentrations of Ca 2+ or EDTA before use. The assays were carried out in triplicates and the results averaged. Three types of binding assays were performed:
(1) Antibody concentration-dependent binding experiments. Various concentrations of anti-F~A or anti-F~B were reacted with a fixed concentration of radiolabeled F~ or prothrombin, in the presence of either 10 mM EDTA or 10 mM CaCI 2. The fixed amount of 12SI-F1 used in this assay contained ~ 4000 cpm ( ~ 8.5.10-10 M) and was mixed with various concentrations of R5 anti-FlA (1.25.10 - 9 M to 1.25.10 -7 M), R9 anti-F1B (1.1.10 -l° M to 1.1.10 -8 M), goat anti-FiA (5.3. 10 -1° M to 5.3-10 -8 M), or goat anti-FiB (7.5.10 -1° M to 7.5.10 -8 M). The fixed amount of ~25I-labeled prothrombin contained -- 4000 cpm ( - 1.7.10-10 M), and was mixed with the same amounts of goat anti-F1A or goat anti-F~B antibodies as given above. (2) Ca 2 +-dependent binding experiments. Fixed concentrations of anti-F~A or anti-F~B antibodies were reacted with fixed concentrations of either radiolabeled F 1 or prothrombin, in the presence of 10 mM EDTA or 0.1 to 10 mM of CaC12. 1251-F1 (1.8.10 -1° M) was
diluted into buffers containing the various amounts of Ca 2+ and was allowed to react with R5 anti-F~ (1.7. 10 -l° M), R5 anti-F1A (2.2.10 ..9 M), R9 anti-F1 (1.7. 10 -9 M), R9 anti-FiB (1.3.10 -.9 M), goat anti-prothrombin (1.7.10 -8 M), goat anti-FiA (1.8.10 ..7 M), or goat anti-F~B (2.8- 10 -7 M). ~25I-prothrombin ( - 1.8 10 -1° M) was reacted similarly with goat anti-FlA (5.3• 10k 8) or goat anti-FiB (7.5.10 -9 M). •
(3) Radioiodinated ligand concentration-aependent binding experiments. Fixed concentrations of anti-F~A or anti-F~B were reacted with various amounts of ~25I-F~ or ~25I-prothrombin in buffers containing either 10 mM EDTA or 10 mM CaCI z. ~25I-F~ of 2.10 3 cpm to 5 • 10 5 cpm (1.5.10 -~° M to 3.8-10 -8 M) were diluted into buffer A containing 10 mM EDTA or 10 mM CaCI: and incubated with R5 anti-F~A (2.10 -9 M), R9 antiFIB (1.1 "10 -9 M), goat anti-F1A (1.9.10 -8 M), or goat anti-F~B (2.4.10-7 M). For binding of antibodies to radioactive prothrombin, 4.7.10 -~1 M to 9.4.10 -9 M of 125I-prothrombin were used to react with 8.10 -8 M of goat anti-F~A or 7.5.10 -9 M of goat anti-F~B. The data were plotted according to Scatchard [39] in the form of r/c versus r, where r is the concentration of the bound ligand per mole of the specific antibody and c is the concentration of the free ligand. We have tested the stabilities of the radioactively labeled F1 and prothrombin by incubating them for 48 h with the usual reagents used in the binding assay (including both second antibody systems) and examining their apparent molecular weights in SDS-polyacrylarrfide gel electrophoresis. In either case, a single radioactive peak corresponding to the original protein or peptide was observed, indicating no degradation had occurred during the binding experiments.
Gel electrophoresis The disc gel electrophoresis procedure of Davies et al. [40] as modified by Rosenberg and Waugh [29] was utilized to establish the homogeneity of polypeptides with respect to charge. SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was performed according to Laemmli [41] using a 15% crosslinked gel. Proteins were stained with Coomassie blue or by the periodic acid Schiff's base method [42,43]. Molecular weight markers included bovine serum albumin (68 kDa), ovalbumin (40 kDa), myoglobin (16 kDa). cytochrome c (12 kDa) and cyanogen bromide-cleaved peptides of myoglobin and cytochrome c.
Protein determination Protein concentrations were obtained by measurements at 280 nm, assuming 13.2, 16.2, 11.9 and 12.5 for 1 mg/ml of human prothrombin, thrombin, F1 and F1 ÷2, respectively [22], or determined by the method of Lowry et al. [44].
98
Amino acid analysis The amino acid compositions of F 1, F1A and FiB were determined by treatment wi ~'~ 6 M HC1 for 24, 48 and 72 h at l l 0 ° C , and the,, examination of the products in a Beckman 121M amino acid analyzer. Results
Preparation of FI, F~A and F~B F 1 was obtained from a n-,~xture of prothrombin fragments by means of barium citrate and ammonium sulfate precipitations, followed by gel filtration on Sephadex G-100. The matrix separated the sample into two peaks. Fractions containing the smaller molecules were pooled, concentrated and rechromatographed on the same column to yield an apparently homogeneous F 1 preparation which migrated with an apparent molecular mass of ~ 35 kDa. Upon SDS-PAGE under nonreduced conditions, a single Coomassie blue-staining band was visible which showed an apparent molecular mass of ~ 21 kDa. This observation agreed with Bensen and Hanahan [45] who found that prothrombin fragments containing "t-carboxyglutamic acids migrated with a higher apparent molecular mass on gel filtration. In the presence of reducing agent, a second faintly siained band appeared immediately below the major one. This second band had been observed before [33], and may be due to partial proteolysis of F l within one or more of its five disulfide bonds [10,11]. The amino acid composition of the Fl prepared is shown in Table I which is in excellent agreement with the published data [10,11,33, 341. F I A and FtB were obtained from thrombin digestion of Ft. The proteolytic fragments were reduced and
I ABLE 1
A~aino acid compositions of prothrombin fragments Fi, F!A and FIB An~no acid a
F!
F1A
1:1B
Asp Thr Ser Glu Pro Gly Ala Val Met lie Leu Tyr Phe Lys His Arg
15.3 19.7 10.8 24.3 9.4 10.8 13.0 8,9 0.9 4,9 8,1 5.5 4.7 4.7 2.1 12.7
3.6 6.7 3.6 12.5 0 1.0 7.0 3.3 0 0 3,8 1.6 3,7 1.8 0 2.5
13.0 13.3 7.1 12.9 10.5 9.7 5.9 4,8 0.7 3.8 5.3 3.3 1.3 2.4 2.3 9.0
a Number of amino acid residues per mole of prothrombin fragment. Tryptophan not determined.
a!kylated and then precipitated with barium citrate. The precipitate displayed a single band upon SDS-PAGE under nonreducing conditions with an apparent molecular mass of ~ 6 kDa. But like F1, there was a minor stained band migrating slightly faster than the major one when it was electrophoresed under reducing conditions. The amino acid composition of this peptide is shown in Table I. Glutamic acids accounted for -- 1 / 4 of the total amino acid residues, most of which were presumably y-carboxylated. The observation that this molecule was precipitated by barium ions, its molecular size and its amino acid composition were consistent with the assignment of this peptide being FlA. The supernatant of the thrombin-digested F1 contained two polypeptides when analyzed by SDS-PAGE and these were separated by chromatography on a hydroxyapatite column. The material not adsorbed contained a single peptide which showed an apparent molecular mass of ~ 15 kDa on SDS-PAGE under both reduced and nonreduced conditions. Upon gel filtration on Sephadex G-100, it also migrated with the same molecular mass. Its amino acid composition (Table I), apparent molecular mass and solubility in barium salt were consistent with this peptide being FiB.
Preparation and specificities of anti-F1A and anti-F1B Using FIA and FIB obtained above, affinity chromatographic columns were set up to isolate specific antibody populations against these fragments. - 0.15 to 0.2% of rabbit R5 and R9 anti-F l were adsorbed, but -270 of the goat anti-prothrombin antibodies were adsorbed by either FIA- or FiB-bound Sepharose columns under these experimental conditions. To test for the specificities of these antibodies, we used prothrombin, F1, F1 +2, FIA and FIB to displace 125I-Fi in the competitive radioimmunoassay for R5 anti-FlA. As expected, F 1 was the most effective in displacing the radioactive ligand. Prothrombin was quite effective, as only twice the molar excess was needed to displace 50~ 12SI-F1. It took 40-times molar excess of F IA and 400-times of FIB to compete in the same assay, while F l +2 did not react with R5 anti-F1A at all. The average slopes of their logit-log dose response curves, except for Fl + 2, were -0.7. The same prethrombin fragments were used to test for their crossreactivities for R9 anti-FiB. In this case, prothrombin and F 1+ 2 had similar slopes of -0.55 and were, respectively, 50- and 100-times less effective than F1 in competing with 125I-F1. FiB, however, competed with 125I-F1 for the antibody population only at a large molar excess (2000-fold), and with a reduced slope of -0.4. This crossreactivity was still superior to FIA which did not react with anti-FiB at all. Therefo~'e these antibodies showed restricted specificities towards the particular antigen that was used as the affinity ligand.
99
Binding studies of R.5 anti-Fi A and R9 anti-F1B antibodies y - C a r b o x y g l u t a m i c acids are believed to be responsible for binding Ca 2 + in prothrombin, and F 1A contains all ten of these residues while Fz B does not contain any. It would therefore be of interest to utilize the specific antibodies as molecular probes to study Ca2+-induced conformational changes in F~. We first investigated the b i n d i n g of the antibodies to ~251 F~ in the presence of either E D T A or Ca 2+. Fig. 1A shows that each concentration of R5 anti-F~A precipitated more F~ in the presence of Ca 2+ t h a n in the absence of Ca 2+. On the contrary, the opposite was true for R9 anti-F~B (Fig.
A
20
~
1.6
Ca 1.4 1.2
1.0
'.24./
• 0
60
A
, //--t-/P-r---
,
1.6] B
1B). N e x t we investigated the bindings of ~25I-F] to unfractionated R5 a n t i - F t, R5 anti-F]A, unfractionated R9 anti-F] and R9 anti-FiB, as a function of Ca 2+ concentrations. Fig. 2A shows Ca 2 +-dependent bi~ dings of F~ to R5 anti-F~ and R5 anti-F t A, with the ~raasition midpoints at 0.28 m M C a 2+, although / 0 anti-F~ appeared to precipitate m o r e F~ than R5 anti-F] A. Fig. 2B shows the binding of F~ to R9 anti-F~ also increased
//-.-//~ o
1.8
1.0
2.0
i'-¢/-r-¢;, = 3.0 5.0 iO.O
CALCIUM CONCENTRATION (mM)
Fig. 2. Calcium dependent binding of R5 and R9 antibodies to F~. (A) R5 anti-F] (¢) or R5 anti-F]A (o) was reacted with ]251-F] in buffers containing 0 to 10 mM CaCI 2 in a radioimmunoassay as detailed in Materials and Methods. The data are expressed as the ratio of the percentage bound F] to total F] in the presence of Ca 2+ divided by the percentage of bound F] to total F~ in the absence of Ca 2+. (B) R9 anti-F1 (O) or R9 anti-FiB (0) was reacted with 12Sl-F] in buffers containing 0 to 10 mM CaC! 2. The experimental procedure and calculation were performed as given above.
with increasing amounts of Ca 2+, reaching a half-maximal point at 0.18 m M Ca 2+. But the binding of F ! to R9 anti-F] B, in contrast, decreased with increasing a m o u n t s of Ca 2+, and its transition midpoint was higher at 0.45 m M C a 2+.
A
tll
' o 2-
Q
K
,o
7
0
!
2 Antibody
100
80 q
!
0.2 concentration
0.02 ( 108M
) 0.,
B
--"""o
i
,
t
2 r
3
6-
6O
x
:3 0
7
m or qO
O
2-
!
0
0 1.0 Antibody
0.1 concentration
0.01 ( 108M )
Fig. 1. Effect of calcium on the antibody-dependent binding to 1::1. Various concentrations of R5 anti-FiA (A) and R9 anti-FiB (B) were allowed to react with 125I-Fi in buffers containing either CaCI2 (O) or EDTA (0) in a radioimmunoassay as detailed in Materials and Methods. The percentage bound F1 per total F] was expressed as a function of the antibody concentrations.
I
Fig. 3. Interaction of R5 and R9 antibodies with FI in the presence of CaCI 2 or EDTA. These data are expressed using a Scatchard analysis where c is the concentration of free F! and r is the concentration of F! divided by the concentration of the respective antibody population. (A) R5 anti-FlA was reacted with leSI-F! in the presence of CaCl2 (e) or EDTA (o). (B) R9 anti-FiB was reacted with lesI-F1 in the presence of CaCI 2 (O) or I~DTA (o).
100 In order to quantitate these Ca2+-dependent effects, fixed concentrations of antibodies were incubated with various concentrations of 125I-F1 in either EDTA- or CaCl2-containing buffers. As shown in Fig. 3, the Scatchard plots were all nonlinear. R5 anti-FtA could be described to display two classes of binding sites in the presence of C a 2+, one with an average association constant (KA) of 1.4.10 9 M -1 and the other 1.4.108 M-1. The high-affinity binding sites constituted - 38% of the total available binding sites (Fig. 3A, Table IIA). In the absence of Ca 2 +, R5 anti-F~ A also displayed two classes of binding sites. The high-affinity sites had the same K A as when Ca 2 + were present, but they accounted for only - 2~ of the total available sites, with low-affinity binding predominated in this case (Fig. 3A, Table IIA). R9 anti-FIB, on the other hand, bound 1251-FI with the same high- and low-affinity constants in the presence or absence of Ca 2+, except that the number of high-affinity sites was bigger in the absence of Ca 2+ (Fig. 3B, Table IIA).
1.6 1.4 O rn
1.2
m 1.0 0.8
,
o
0.6
1.0
1.6
B
o
CO ,-.... m
TABLE Ii
Scott'hard analysis of the interaction of either 1251.FI or l"51-prothrombin with different antibody populations High K^ (M - ! )
Approx. ,~ total sites having
Low K A ( M - n)
high K^ (A) Using 12Sl-Fi as the ligand for binding antibodies: R5 anti-FiA Ca 2+ 1.4,10 ~ 38 EDTA 1,3,10 9 2 R9 anti-F, B Ca 2 + 3.5.10 9 33 EDTA 3.1.10 9 39 Goat anti-FtA Ca 2+ 1.3,101° 30 EDTA 1.7.10 9 30 Goat anti-FzB Ca 2+ 2.0.10 9 15 EDTA 1.8.10 9 31
1.4. l0 s 1.4.10 s 1.6. l0 s 2.1.10 s 1,5.10 ~ 1.6.10 8 1.5.107 7.3-107
(B) Using 1251-prothrombin as the ligand for binding antibodies: Goat anti-FiA Ca 2+ 3.8-10 9 30 6.0. l0 s EDTA 2.5. l0 9 35 4.1. l0 s Goat anti-FiB Ca 2+ 1.8- l0 9 33 1.8. l0 s EDTA 4.7-109 40 5.2.10 s
3.0
5.0
(mMI
m
1.2
/
1"0'
O~..x:k
• I/'"®
~
0.8
o #-o
0.6
Binding studies of goat anti-F! A and anti-Fj B When various concentrations of goat anti-F~A was diluted into buffers containing either 10 mM EDTA or CaCI 2, it was found that the antibody population bound more 125I-F~ in the presence of Ca 2+ than in its absence. The opposite was true for goat anti-FmB (data not shown). The binding of goat anti-FiA to 125I-F1 was dependent on Ca 2+ with a half-maximal point at 0.24 mM Ca 2+. On the other hand, the binding of goat anti-F! B tO 1251-F! decreased as Ca 2+ increased, while the halfmaximal point was also higher than at 0.51 mM Ca 2+ (Fig. 4).
2.0
Calcium c o n c e n t r a t i o n
! 1.0
m 2.0
~#..a 3.0
5.0
Calcium concentration (raM) Fig. 4. Calcium-dependent binding of goat anti-FiA and anti-FiB to F ! and prothrombin. These experiments were carried out and calculation expressed as given in Fig. 2. (A) Goat anti-FiA (0) or anti-FIB ( 0 ) was reacted with tasI-FI in buffers containing 0 to 5 mM CaCI 2. (B) Goat anti-FIA (e) or anti-FiB ( 0 ) was reacted wi:h t251-prothrombin in buffers containing 0 to 5 mM CaCI 2-
In order to determine whether prothrombin also exhibited these Ca 2+-dependent effects, ~25l-prothrombin was used as the radioactive ligand for binding goat antibodies. Similar results to those using ~251-F! as the radioactive ligand were obtained (Fig. 4B). Scatchard analyses of the binding between the goat antibodies with F1 and prothrombin were then performed. The results are presented in Table If. Again all Scatchard plots were nonlinear and two classes of binding sites could be used to describe each curve. F 1 bound goat anti-FiA with about 8-fold higher affinity in the presence of Ca 2+ (1.3.101° vs. 1.7.10 9 M - I ) than in its absence. On the other hand F 1 bound goat anti-F~B with twice as many high-affinity sites in the absence of C a 2+ (31 vs. 15t~).
Prothrombin also bound goat anti-F~A with slightly higher affinity in the presence of Ca 2+. In contrast, prothrombin bound goat anti-F~B with 2.5-times higher affinity (4.7.10 9 vs. 1.8-10 9 M - l ) in the absence of Ca 2+. Discussion
We have purified human prothrombin fragment 1 to apparent homogeneity and from it, the amin:~ terminal peptides FIA and F1B. They were identified by binding
101 properties to barium salt, SDS-PAGE, gel filtration and amino acid compositions. F1A and FiB were utilized as affiPAty ligands to isolate specific antibodies which showed restricted specificities toward the respective ligands and were useful probes to study the conformational changes induced by Ca 2+ in native 1::1 and prothrombin. Two types of conformational change could be detected and both were Ca2+-dependent. In the first type, Ca 2+ was able to stabilize a conformation in F1 when F 1 reacted with R5 anti-FlA (Fig. 1A), or when 1"1 reacted with unfractionated R5 and R9 anti-F1 (Fig. 2), or when Ft or prothrombin reacted with goat anti-prothrombin or goat anti-FlA (Fig. 4). Presumably Ca 2+ bound to some common conformations in prothrombin, F~ and F1A and induced conformational changes in these molecules. The binary complex¢s having the new conformation were more stable and could precipitate more of the specific antibodies. The interactions were the results of either higher affinity or larger number of high-affinity binding sites in the presence of Ca 2+. The half maxima of this transition were between 0.18-0.28 mM Ca 2+ and corresponded closely the dissociation constants obtained by direct binding of Ca 2+ to F1 [15], with circular dichroism [19], fluorescence [16,18] or immunochemical means [24,25]. Furthermore, the common epitopes on these molecules were necessarily immunodominant features and probably involve the y-carboxyglutamic acid residues or regions surrounding them. This conformational change is probably similar to that observed in the study of F 1 crystal structure. There was no density in the amino acid sequence 1-35 in the absence of metal ions [12,13], indicating a disorganized structure in the y-carboxyglutamic acid-containing region. However, addition of Sr 2+ ',ransformed the region to become electron dense and therefore presumably an ordered structure [46]. The second type of conformational change was detected by R9 anti-FiB and goat anti-F~B. It is also Ca2+-dependent, but Ca 2÷ in this case destabilized the new conformation. Ca 2+ probably formed binary complexes with F~ and prothrombin first, lowered the stability of the subsequent ternary complexes when anti-F~B combined with them, and thus gave rise to less precipitates. These interactions were resulted from either lower affinity binding or smaller number of high-affinity binding sites in the presence of Ca 2+. The conformer recognized in prothrombin and F t was probably localized on the FIB portion of these molecules, since the antibodies which recognized this conformational change did not recognize F~A in the competitive radioimmunoassay. However, it cannot be excluded that neoantigens could have arisen when F~A joins F~B in native F~ or prothrombin, and that it was this new format that was destabilized by Ca 2+. Tiffs conformational change took place with transition midpoints at
considerably higher Ca 2+ (0.45-0o61 raM). It is probably a weaker transition since it was completely masked by the first type of conformational change when unfractionated R5 and R9 anti-F1 interacted with F 1 in the presence of Ca a+ (Fig. 2). It is of interest to note that FIB consists of a short helical region and a compact, organized kringle structure [12,13], in which no metal-dependent changes have been described before. The significance of this conformational change is not known, but it could be involved in the factor Xa-dependent cleavage of prothrombin or in the binding of the zymogen of the phospholipid surface. References 1 Magnusson, S., Petersen, T.E., Sottrup-Jensen, L. and Claeys, H. (197~) in Proteases and Biological Control, Cold Spring Harbor Laboratory, pp. 123-149, Cold Spring Harbor, NY. 2 Magnusson, S., Sottrup-Jensen, L., Petersen, T.E., Morris, H.R. and Dell, A. (1974) FEBS Lett. 44, 189-193. 3 Neisestuen, G.L., Zytkovicz, T.H. and Howard, J.B. (1974) J. Biol. Chem. 249, 6347-6350. 4 Stenflo, J., Fernlund, P., Egan, W. and Roepstorff, P. (1974) Proc. natl. Acad. Sci. USA 71, 2730-2733. 5 Stenflo, J. and Ganrot, P.-O. (1973) Biochem. Biophys. Res. Commun. 50, 98-104. 6 Nelsestuen, G.L. and Suttie, J.W. (1974) Biochemistry 11, 4961-4964. 7 Nesheim, M.E., Hibbard, L.S., Tracy, P.B., Bloom, J.W., Myrmei, K.H. and Mann, K.G. (1980) in The Regulation of Coagulation (Mann, K.G. and Taylor, F.B., eds.), pp. 145-149, Elsevier/ North-Holland, New York. 8 Harmison, C.R., Landaburu, R.H. and Seegers, W.H. (1961) J. Biol. Chem. 236, 1693-1696. 9 Conway, E.M., Lau, H.K.F., Bauer, K.A. and Rosenberg, R.D. (1987) The development of a radioimmunoassay for quantitating prethrombin 2 in human plasma. J. Lab. Clin. Med. 110, 567-575. 10 Olsson, G., Andersen, L., Lindquist, O., Sjolin, L., Magnusson, S., Peterson, T.E. and Sottrup-Jensen, L. (1982) FEBS Lett. 145, 317-322. 11 Tulinsky, A., Park, C.H. and Rydel, T.J. (1985) J. Biol. Chem. 260, 107718b110778. 12 Park, C.H. and Tulinsky, A. (1986) Biochemistry 25, 3977-3982. 13 Harlos, K., Boys, C.W.G., Holland, S.K., Esnouf, M.P. and Balke, C.C.F. (1987) FEBS Lett. 224, 97-103. 14 Mann, K.G. (1984) Prog. Hemostasis Thromb. 7, 1-23. 14 Bajaj, S.P., Butkowski, R.J. and Mann, K.G. (1975) J. Biol. Chem. 250, 2150-2156. 16 Nelsestuen, G.L., Broderius, M. and Martin, G. (1976) J. Biol. Chem. 251, 6886-6893. 17 Nelsestuen, G.L. (1976) J. Biol. Chem. 251, 5648-5656. 18 Prendergast, F.G. and Mann, K.G. (1977) J. Biol. Chem. 252, 840-850. 19 Bloom, J.W. and Mann, K.G. (1978) Biochemistry 17, 4430-4438. 20 Tucker, M.M., Nesheim, M.E. and Mann, K.G. (1983) Biochemistry 22, 4540-4546. 21 Stenflo, J. and Ganrot, P.-O. (1972) J. Biol. Chem. 247, 8160-8166. 22 Lau, H.K.F., Rosenberg, J.S., Bceler, D.L. and Rosenberg, R.D. (1979) J. Biol. Chem. 254, 8751-8761. 23 Lau, H.K.F. and Rosenberg, R.D. (1980) J. Biol. Chem. 255, 5885-5893. 24 Furie, B., Provost, K.L., Blanchard, R.A. and Furie, B.C. (1978) J. Biol. Chem. 253, 8980-8987. 25 Furie, B. and Furie, B.C. (1979) J. Biol. Chem. 254, 9766-9771.
102 26 Tai, M.M., Furie, B.C. and Furie, B. (1984) J. Biol. Chem. 259, 4162-4168. 27 Owens, J., Lewis, R.M., Cantor, A., Furie, B.C. and Furie, B. (1984) J. Biol. Chem. 259, 13800-13805. 28 Borowski, M., Furie, B.C., Bauminger, S. and Furie, B. (1986) J. Biol. Chem. 261, 1496986114975. 29 Rosenberg, R.D. and Waugh, D.G. (1970) J. Biol. Chem. 245, 5049-5056. 30 Shapiro, S.S. and Waugh, D.F. (1966) Thromb. Diath. Haem. 16, 469-480. 31 Rosenberg, J.S., Beeler, D.L. and Rosenberg, R.D. (1975) J. Biol. Chem. 250, 1607-1617. 32 Aronson, D.L., Steven, L., Ball, A.P., Franza, B.R., Jr. and Finayson, J,S. (1977) J. Cfin. Invest. 60, 1410-1418. 33 Walz, D.A., Hewett, Emmett, D. and Seegers, W.H. (1977) Life Sci, 20, 79-84. 34 Walz, D.A., Hewett-Emmett, D. and Seegers, W.H .(1977) Proc. Natl. Acad. Sci. USA 74, 1969-1972. 35 Porath, J., Axen, R. and Ernback, S. (1967) Nature (Lond.) 215, 1491-1492.
36 Greenwood, F.C., Hunter, W.M. and Glover, J.S. (1963) Biochem. J. 89, 114-123. 37 Redbard, D. (1974) Clin. Chem. 20, 1255-1270. 38 Rodbard, D., Lenox, R.H., Wray, H.L. and Ramseth, D. (1976) Clin. Chem. 22, 350-358. 39 Scatchard, G. (1949) Ann. NY Acad. Sci. 51, 660-672. 40 Davies, B.J. (1964) Ann. NY Acad. Sci. 121, 404-427. 41 Laemmli, U.K. (1970) Nature (Lond.) 227, 680-685. 42 McGuckin, W.F. and McKenzie, B.F. (1958) Clin. Chem. 4, 476-483. 43 Zaccharius, R.M., Zell, T.E., Morrison, J.H. and Woodlock, J.J. (1969) Anal. Chem. 30, 148-152. 44 Lowry, O.H., Rosebrough, N.J., Farr, A.L. and Randall, R.J. (1951) J. Biol. Chem. 193, 265-275. 45 Benson, B.J. and Hanahan, D.J. (!975) Biochemistry 12, 3265-3277. 46 Harlos, K., Holland, S.K., Boys, C.W.G., Burgess, A.I., Esnout, M.P. and Balke, C.C.F. (1987) Nature (Lond.) 330, 82-84.