@6. The first 35 residues (Alal-Leu35) of the N-terminal y-carboxy glutamic acid-domain (Alal-Cys48) of fragment 1 are disordered as are two carbohydrate chains of M r z 5000; the latter two combine to render 40% of the structure disordered. The folding of the kringle of fragment 1 is related to the close intramolecular contact between the inner loop disulfide groups. Half of the conserved sequence of the kringle forms an inner core surrounding these disulfide groups. The remainder of the sequence conservation is associated with the many turns of the main chain. The Pro95 residue of the kringle has a cis conformation and Tyr74 is ordered in fragment 1, although nuclear magnetic resonance studies indicate that the comparable residue of plasminogen kringle 4 has two positions. Surface accessibility calculations indicate that none of the disulfide groups of fragment 1 is accessible to solvent.

Keywords: prothrombin;

kringle;

1. Introduction Blood coagulation is a cascade process involving successive proenzyme activation reactions. In the penultimate step, prothrombin is converted to the serine protease thrombin catalyzed by factor Xa in the presence of membrane-bound factor Va, Ca+2 and phospholipid surface (Nesheim et al., 1980; Krishnaswamy et at., 1986; Mann, 1987). Bovine prothrombin (M, = 72,000) is a 579 residue singlechain glycoprotein composed of three independent domains, fragment 1 (Fig. 1) (residues 1 to 156, :Wr = 23,000), fragment 2 (residues 157 to 274, M, =

t Permanent address: Department of Physics. Indian Institute of Science. Bangalore012. India. $ Author to whom all correspondence should he addressed. 0 Permanent address: Department of Chemistry, University of Toledo, Toledo, OH 43606, IJ.S.A. )I Permanent address: Ahhott Laboratories, Abbott Park. IL 60064, l1.S.A.

synehrotron;

cis-proline;

disorder

14,000) and prethrombin 2 (residues 275 to 579, M, = 37,000). Further cleavage of prethrombin 2 at the Arg323-Be324 peptide bond by factor Xa results in a-thrombin, the most catalytically potent form of thrombin, consisting of two chains linked by a disulfide group. The most conspicuous feature of the sequence of fragment 1 is a three disulfide group, triple loop structure that has become known as a kringle (Fig. 1). The highly conserved kringle sequence appears 16 times in seven different proteins: once in urokinase (Steffens et al., 1982; Gunzler et al., 1980), factor XII (McMullen & Fujikawa, 1985) and vampire bat salivary plasminogen activator (Garde11 et al., 1989); twice in tissue type plasminogen activator (Pennica et al., 1983) and prothrombin (Magnusson et al., 1975); four times in hepatocyte growth factor (Nakamura et al., 1989); and five times in plasminogen (Magnusson et aE., 1976). In addition, sequencing of apoIipoprotein (a), the protIein of lipoprotein (a), a low density-like lipoprotein, shows that, remarkably, it contains 37 highly homo-

5x2

7’. I’. iSeshadri et, al.

R@*‘T;~L$$R~~Y B P 6P H s’ 130 5 8o t c”0-i ;D :C ; A

%R$ AF-n JU

L 5 A T D

20 T?

r’P

%

f N M,

- L

A

t K R 10 v

L R nK w(N -Y

JR.501 Pr -5 R c LKL~ELw

K

120

I ’ 00 N- -CHO 5

8 I

T 1

1

y

0

NE G

I QD b

“V

150

TVCV . L . I PR

60 ‘ FGKNAl Figure 1. Sequence of bovine prothrombin fragment open-circles, Gl& residues; CHO,-carbohydrate.

1. Residues conserved in 2nd kringle of prothrombin

are circled:

K-dependent carboxylase. The Glat residues vary in

similar to that of the kringle of fragment 1, except for a one residue insertion. We now report the highly refined 2.25 A resolution X-ray crystallographic structure of bovine fragment 1 based on combined measurements made with a diffractometer and photographic methods using the synchrotron beam at CHESS. With the higher resolution data, it has been possible to finalize the structure and the hydrogen bonding network of the kringle,

number

uncover

logous copies of plasminogen kringle 4 and one of the kringle 5 domain (McLean et aZ., 1987). Residues 1 to 48 of prothrombin (Fig. 1) are homologous with the N-terminal sequences of factors VII, IX and X and proteins designated as C, S and Z (Soriano-Garcia et al., 1989). All contain y-carboxy-Glu residues produced in post-translational

modification from

of Glu

residues

9 to 12, of which

by a vitamin

the first

ten are

highly conserved. The function of the Gla-domain (residues 1 to 48) is phospholipid binding in the presence of Ca2+ and it displays more conservation (65 to 7OyA) than that observed among kringles (-50%). The structure of bovine prothrombin fragment 1

has been reported at 2.8 A and its kringle (1 A = 0.1 nm) resolution (Park & Tulinsky, 1986; Tulinsky et al., 1988a) and the structure of the Gla-domain has been solved in its Ca2+ binding et conformation in Ca ‘+ fragment 1 (Soriano-Garcia al., 1989). Since then, the folding of human plasminogen kringle 4 has been determined by nuclear magnetic

resonance

(n.m.r.)

in solution,

where

it

was shown to be similar to the kringle of fragment 1 but apparently stabilized by a somewhat different set of hydrogen bonding interactions (Atkinson & Williams, 1990). However, the highly refined 1.9 w resolution crystallographic structure of plasminogen kringle 4 (Mulichak & Tulinsky, 1990) is very t Abbreviations used: Gla, y-carboxy glutamic acid; n.m.r., nuclear magnetic resonance; PEG, polyethylene glycol; r.m.s., root-mean-square; apo-fragment 1 designates the absence of Cazf

a proline

residue

in the cis conformation.

define the water structure in the near vicinity of the surface and fix the localization of Tyr74, which is flipping between two conformat’ions in t,hc n.m.r. solution structure.

2. Experimental Methods (a) Intensity data Prothrombin fragment 1 was crystallized as prisms up to 1.5 mm in maximum dimension from 32% (w/v) polyethylene glycol (PEG) 4000, 0.1 M-sodium phosphate (pH 67). The crystals were stored in 39% PEG 4000. @l M-Tris-maleate buffer (pH 75) and are tetragonal, a = b = 77.6 A, c = 852 A, space group P4,2,2, 8 molecules/unit cell but with a protein fraction of only about 40%. Due to the high solvent content and the fact, that additionally, about 40% of the structure is disordered (Park & Tulinsky, 1986), diffraction extended only marginally to 2.5 A resolution with a sealed X-ray tube at 2 kW power, with the dynamic range of the data also being low (20 to 25). This was not the case with a synchrotron beam that was employed to extend the intensity data to its limit and significantly imprave the observational range of the intensities. Intensity data at 2.25 A resolution were collected photographically at CHESS using oscillation methods. In

Bovine Prothrombin

Fragment

183

1

scaling and appropriat,e corrections for difference in scattering fail-off, the 3 data sets were merged for structure refinement using the following rejection ratio criteria: if IFI > IE’l,i” > 11.0 and (k7 < IFIo/(F& < 1.4. where (FI, and synchroand lf%ync are the averaged diffractomet)er tron measurements, the IFlo and lFISync werP averaged. For all ot,her reflect,ions. only the averaged diff’racatometer values were used except for reflections in the 2.5 to 2% A range. which were strictly synchrotron measurements. The average diffractometer data were used when the rejection ratio t’est failed. simply beca.usr the!. were cotisidered to be more ac~curat,e as a result of being measured from 2 cryst,al specimens and rorrrctrd for absorption z 1.5 to 1.6) and int,t>nsity deca) (maximum/minimum due t,o radiation exposure. The synchrotron dat,a. on the other hand. were obtained from 20 different crystals and were not explicitly corrected for absorption or drcaay. The symhrotron data additionally provided about 1500 new reflections to the 2.5 A resolution diffractometer set and 950 between 2.5 and 2.25 A resolution (-6026 of possible). bringing t,he total data set close to 8.‘,(H) observable reflections (77 O/b) merged out of about I H.000 that were measured. At present. this corresponds to about the muximwn amount of diffructior~ that is observable from surh 40 ?h disordered tetragonal fragment I crystals, which, nonetheless, will be seen below to produce a good model of the ordered struct,ure. The R-values between the various s&s were: 0.09 between sets 1 and 2. 0.14 between set I and synchrotron. and 0.16 between set 2 and synchrotron.

all, 72 unique photographs were recorded of prothrombin fragment I crystals. The crystals were mounted with the tetragonal c-axis along the spindle of the camera and intensities were scanned over a 47” spindle range in 3” oscillation increments. A total of about 20 crystals was used to collect, a unique set, with each film pack of 3 (a, fl. y) being exposed for 90 s: only 3 to 4 rxposurrs were taken of ea.ch individual crpstal. The films were scanned with an Optronics P-1000 drum scanner at Purdue Llniversity using a 100 pm raster step and the optical densities were stored on a magnetic tape. Thr digitized data were processed using the Purdue package of programs (Rossmann, 1979; Rossmann et al.. 1979). The position of the direct X-ray beam was detrrmined for each film and an A-set matrix was calculated (relating a reflection to the reciprocal lattice). An approximat’e Q-matrix was also computed at low resolution (5.0 .A) relating the camera and scanner co-ordinates. The A-set and Q-matrices were then refined with higher resolution data that ext,ended beyond 3.0 A spacings. The for refined matrices were subsequently employed processing the cr-films. A number of films were rejected from considerat,ion for one or more of the following reasons: (1) the beam position could not be loca,ted accurately, (2) the A-set or Q-matrix did not refine wit,h sufficient accuracy, and (3) the number of fully observed reflections was relatively small. Once the b-films were processed, the different film intensities were scaled to each other. After the plane-plane scaling, various other parameters such as unit cell. c$,,c#J,.~,, scale factor and mosaic&g of the crystal were refined. The total number of unique reflections accessible at 2.25 A resolutic?n is about I 1.000, of which IO.599 were measured with X36.5 (769;) having IFI values of >3a. In all, 3 difI+rrnt intensity data sets of fragment I wert’ measured and combined. The 1st set was at 2.8 14 resolut,ion. collected with a sealed X-ray tube operating at 2 k\l: on a Nicolet, 15/F diffractometer. The second was also a diff’ractometrr set. measured wit,h a different carystal. but from 3.5 to 2.6 A resolution with only 2441 observable reflections in common with the 1st set. Thr maximum intensity decay correction of the 2nd set was vrr>- small, at a factor of about 1.2. The 3rd data set was the 2.25 4 resolution synchrotron measurements just described (47 I7 (77 o/;,) o b.vrrvable in common with the 1st srt and 4101 (51 ‘?o) in common with the 2nd). After

(b) Structure

wfinrmrnf

The structure of fragment I was refined with I’ROLSQ using restrained least-squares procedures (Hrndricskson & Konnert. 1980). The starting reflection dat,a set was thp averaged 2.8 A resolution tliffrac,tometrrjs~nchrotron measurements. which also caontainrd about 500 reflections measured with synchrotron radiation alonr. The starting st)ructure had been refined at 2.8 J! resnlution using the diffractometer dat,a alone (Tulinsky PI nl.. 1988n). This st,ruc+ure corresponded to residues 36 to I56 of fragment 1; residues I to 35 and thr :! c,arbohydratr chains of the molecule are c~ompletelq- disordered. The initial R-factor was 0.35 with an overall t,hrrmal paramrtcar and it drc*rtlasrd to 0.25 in about 40 c*~.clrs of

Table 1 Summary

oj rejirwmvnt

progress

water t&solution

(A)

Reflections

K

(‘jTlCS

1197

0255

45

0

4197 5580

0.242 0.242 0.211

17 30 16

0 0 60

6080

0200

I0

95

5,580

Remarks

molecule

6080

0193

6529

0209

6 10

130 I30

8519 6529 6364

0203 0.185 WI75

r 1; I5

172 I72 172

Averaged ditfrnctometrr/yynchrotron data $us - 500 synchrotron reflections mapx Refitting (21Fl,-lFCl) and (IF],-I?‘/,) Resolution increased to 2.5 A 8.0 -2.5 A and 50-2.5 A Ap maps for water molec~ule~; water molecules > 2.7 o(Ap) Plus 500 synchrotron only reflections: round of map refitt,ing: also. 35 water molecules addrd 40 more water molecules 947 largest synchrotron measurements addrd; givrn different weight from averaged data 42 water

molecules

added

Ditrraction weights in 8 20 ranges Readjusting 28 dependent weights and removing somp outlier reflections

4x4

T. P. Seshadri

et al.

Table 2 Weights of rtfiections

and R-values

of @al

rc$nemw~t

No. of reflections 831 x74 929 828 744 823 849 651 t Reflection

weights used in final stages of refinement.

refinement, the last 20 or so with individual thermal parameters (Table 1). After a round of electron density. refitting of the molecule with the program FROM (Jones, 1982) followed by an increase in resolution to 2.5 A, R dropped slightly to 0242 (Table I). At this stage. (M-2.5 A and 5.0-2.5 A) difference electron density maps were examined for solvent molecules. Peaks > 2.7 o (Ap) that were within I.5 A of each other in both maps. and that made reasonable contacts with the protein. were selected as possible water molecules and they (60) were included in the refinement. After 16 cycles, including refinement of water occupancies, R decreased significantly to 0.211. Thermal parameters of water were also varied while occupancies were kept constant but only occasionally (maximum of 4 times) and only for a few cycles at a time). At the resolutions being employed, we considered occupancy to be more reliable than scattering fall-off.

Table 3 Summary

of jinul least-squares and deviations

parameters

Target sigma

r.m.s delta

0020 0040 0050

0020 0065 0072

0.60 0.60 0.60

0.27 0.44 0.41

A. Distances (A) Bond distance Angle distance Planar I, -4 distance 13. ‘Vowbonded

distances

Single torsion Multiple torsion Possible H-bond (‘. Torsion angles (deg.)

3 I5 20

Planar Staggered Orthonormal

3 27 I7

I). Miscellaneow

Plane groups (A) (‘hiral renters (A3) E. Thermal

restraints

Main-chain bond Main-chain angle Side-chain bond Side-chain angle

0.020 0150

0017 0.25

1.0 1.5 2.5 3.5

0.9 I .5 :+o 3.9

(A=)

The w-angles of 117 of 120 residues were within

+6” of 180”.

since the significant scattering contribution of water had to be associated with lower angles. Another round of refitting was then performed and an additional 500 synchrotron-only reflections were added along with 35 water molecules (R = 020), followed by another 40 water molrcules to bring the total to 130 (5 discarded with occupancies of 3.0) between 2.5 and 2.25 .& resolution by assigning to them a different weight from the averaged data in the least-squares calculations. Furthermore, another 42 water molecules were added and R converged to 0.203 in 15 cycles. At this stage. reflection weights were reassigned in 8 20 ranges based on (llFl,lE’l,l)/2 of the range. The R-factor decreased very significantly from @203 to 0185 as a result of plaring more weight on the diffraction pattern in the next 12 cycles. A final adjustment of the weights based on new discrepancies followed by 15 more cycles average weights produced a final R of 0175. The final reflection and R-values in the different 20 ranges are listed in Table 2 from which it will be seen that the agreement is consistent throughout the diffraction range and is especially good at lower resolution. The pertinent stages of the refinement are summarized in Table 1. The target) and root-mean-square values of the restraints of the final model structure are given in Table 3. These values cornspond to 1119 atoms defined in terms of 4649 variables determined by 6364 observations with only the 105 water molecules with occupancy of >@6 appearing t,o be significant. The average thermal parameter of fragment 1 is large at about 40 AZ but is consistent with vast solvent caontent and 40% disordered structure of fragment I csrystals. The final r.m.s. H-values of main and side-chain atoms along the sequence of fragment I are shown in Fig. 2. The smallest values are associated with tht, kringlr ((H) z 30 Fi’) and in particxular with the disulfide groups and the A, C. D loops (Fig. I ): the larger R-loop displays much more flexibility than the other parts of t,he kringle. The (1 and D loops are fairly buried in the structure and along with A are involved in some crystal packing interactions (Tulinsky rt al., 1988n). It is indeed rewarding that t.he structure preceding and following the kringle appeared as well as it did in electron density maps. This might well be related to the accuracy of the data, with which considerable care was exercised, and to some ext,ent to the unusually large volume of the crystal being at a noise level (solvent or otherwise) thus enhancing the contrast with respect to signal electron density.

Bovine

40

50

60

70

Prothrombin

00

Fragment

90

100

Residue

Figure 2. Restrained r.m.s. B-values of fragment are as in Fig. 1; CI,. a,, a-helices.

1. (-)

3. Results and Discussion (a) General An estimate of the mean error in co-ordinates has been made by examining the R-factor as a function of scattering angle (Luzzati, 1952), which suggests a value of 620 to 0.22 A. However, some atoms are probably positioned better than this while the errors of others with larger thermal parameters will be considerably larger. The folding of the kringle structure (Fig. 3) is related to the close intramolecular contact between the inner loop disulfide groups of Cys87 to Cys127 and Cysl15 to Cys139 (S87-S115, 3.89 A; S87-S139, 4.66 A; S115-S127, 4.36 ii; S127-S139, 4.59 A) (Fig. l), which leads to a pair of two stacked threedimensional loops (A-A’, B-B’, Fig. 4 (Priestle, 1988), which corresponds to A,C and B,D respectively, of Fig. l), which are approximately related to each other by two independent 90” rotations and a translation (Park & Tulinsky, 1986; Tulinsky et al., 1988a). Thus, the folding of the kringle displays a certain degree of intricate duplication. The secondary structural features of the folding are listed in Table 4. The close disulfide contacts (Fig. 5) give rise to two antiparallel b-strand regions approximately perpendicular to each other (Figs 3 and 4). The regions of highest conservation of sequence among kringle structures occur between flZ and b3

110

1

120

485

130

140

150

number

Main-chain;

( x ) side-chain; A, B, C, D kringle loop segments

(113-119) and in the /?,-strand (124-129 and 137-139) (Figs 1, 3 and 4; Table 4). The latter peptide segments are located adjacent and antiparallel to each other in the interior of the domain near the center of the kringle fold and bury the Cys87-Cys127, Cysl15-Cys139 disulfide pair. Most significantly, the conservation displayed here represents about half of that observed among kringles and is concentrated at a very small region in three dimensions. Possible implications of the conservation have already been discussed elsewhere (Tulinsky et al., 1988a). The remainder of the conservation dispIayed by kringles simply resides in the many turns required to accomplish the kringle fold (Figs 3 and 4), where

Table 4 Secondary A. a-Helix a1

a2

structure

of fragment

I

structure

Asp39-Ala47 Arg55-1~x162

13. p-structuret 2

Ser79-Thr81; H&5-Cys87 Gln88-Trp90; Argll I-Asnll3 Cys127-Thr129; Arg136-Glu138 Va114lLPro142; Va1149-Thr1.56

t In addition, there are main-chain Vall78-(‘ysllf, and Arg116-Trp126.

hydrogen

bonds between

Figure 3. Stereo view of CA, (‘. N. 0 structure

about half of the residues of the kringle are involved in one kind of t.urn or another. In earlier work at lower resolution, it was difficult to type-classify many of the turns (Tulinsky et al., 1988a). However. four have now been definitely identified in terms of four consecutive amino acid residues, dihedral angles, C” and hydrogen bond distances (Crawford et al., 1973) and are listed in Table 5. The remainder of the turns of the kringle (6) do not fall under any classifications and generally correspond to larger, less regular turns of the peptide chain. The lack of correspondence to idealized turn structures is most

Figure 4. Drawing of the folding of fragment. I. Antiparallel b-strands are labeled; 3-dimensional stackedloops are designated A. A’ and B, B’, respectively.

of fragment

I. Disulfide

groups

are shown

in bold

likely because the disulfide clustering is the ovt’t’riding force of the folding. As a result, t,he relatively short peptide segment between disulfides bridges appear to compensate by making less perfec:t turns to achieve a compact structure. The importance of’ the disulfide clustering has already been spectactularly demonstrated by the aerobic restoration of Cys87-Cys127 and Cysl15--Cys139 to reduce kringle 4 of plasminogen with concomitant refolding of the kringle and regeneration of its lysine binding site (Trexler & Patthy, 1983). A summary of the torsional angles for the kringlr and the interkringle link (145-156) is presenttbd in Figure 6 (Watenpaugh et al., 1979). The values of the C#Jand $ angles of each residue correspond to the head and tail of each arrow, respectively. Consecutive large arrows (4 z - 120” and II/ z 120”) correspond to segments of chain in extended b-strand vonformation (Balasubramanian, 1977; Srinivasan $ Yathidra, 1978). The antiparallel B-structure of the kringle, which is necessaril) composed of short segments because of the large number of turns, all fall within these regions; t,he remaining regions simply correspond to extended chain without any parallel or antiparallel cornponents. The interkringle link, which begins with a turn, is in an extended conformation and runs antiparallel to Glu l37--Cys144 (Figs 3 and 4) but onl) about three residues on either side of the turn art’ close to hydrogen bonding distances of each other. The dihedral angles of the four disulfide bridges of fragment 1 are listed in Table 6, from which it van be seen that the disulfide groups are generally in a tram conformation. Moreover. three are left-handed but do not adopt a preferred left-handed spiral geometry (Katz clr Kossiakoff, 1986). The larger CA-CA distances of 48 to 61, 87 to I27 and 115 to 139 are in agreement with the left-handed chirality. Most of t.he Cla-domain of fragment I is disordered (Alal-Leu35) in the apo-st.ructure (Park & Tulinsky, 1986: Tulinsky rt al.! 1988a). The torsional angles of the ordered region of the Gla-domain

Bovine

Prothrombin

Fragment

187

I

Figure 5. Stereo view of kringle inner loop disulfide groups.

(36-48) and its trailing tet)radecapeptide disulfide loop (Figs 3 and 4) is given in Figure 6. This region of fragment. 1 consists mostly of R-helix, which is evident in Figure 6, where the C#Jand $ values generally range near -40” and -70, respectively. Residues Asp39 to Ala47 of the Gla-domain form 2..5 turns of helix and the latter half of the disulfide loop (Arg5,5-Leu62) does likewise. The $,II/ distrihution of the cbomplete fragment 1 molecule is shown in Figure 7 from which it will be seen that they conform well w&h allowed regions. The contour lines correspond to the energy minima for polyalanine (Maigret et al., 1971). The two residues that deviat’e from minimal energy regions are Glu 112 and Asnl13: the electron density of these makes a very sharp turn so t,hat, they must be st,rained to fit t’he density. (b) lntramolwulor

ion pairs/hydrogen

bonds

The hydrogen bonding network of fragment 1 and the atoms and the residues involved in hydrogen bonds are listed in Table 7. The criteria used to identify the most significant hydrogen bonds were: (1) a donor-acceptor distance of < 3.0 A and (2) a hydrogen bond angle > 140”; where the hydrogen position of thr donor was calculated at an idealized

position. Notwithstanding the strir1genc.y of such conditions, 36 of 96 distances satisfied them. Two of the hydrogen bonds also result in ion pair interactions that have been reported (Tulinsky et al.. Glu99-Arg135 (99OE2 to 198Xa) involving 135NH2 = 3.08 A) and AsplOg-Argl16. Two other ion pairs occur between Lys57-Glu60 and AsplOg-Argl l l but the hydrogen bond angle of the former is not favorable. A similar pairing situation occurs in the lysine binding site region between Asp1 19 and Arg136 (Tulinsky et al., 1988a) except that here the distances are a bit unfavorable for hydrogen bonds. The AsplOg-Argll 1 ion pair does not appear to enter into stabilizing the folding of the triple loop structure of the kringle, whereas the other ion pairs coupled with the hydrogen bonding network strap the several loops together. In addition, residues Glu138 and Arg156 make a strong intermolecular hydrogen bonding ion pair interaction where the carboxylate and guanidinium groups are approximately 90” to each other. (c) A romatic

core

The folding of the kringle structure gives rise to a column of stacked aromatic and proline residues that runs through the center of the kringle and that

Table 5 Reverse turns of prothromhin

Residues Thr81 -Arg82-Ser83-Gly84 Leu89-TrpW-Arg91 Ser92 Ser121-Ile122-Thrl23-Gly124 S~rl31-Pro132-Thrl33-~ul34 For turn definitions,

see Crawford

Ol-u4 (4

C”l-D”4 (4

2.92 3.20 3.85 3.13

553 4.3 1 554 578

et al. (1973)

fragment

I

w2 (deg.) -59, -63, -73, -60,

-15 -44 -38 -22

- 100, 4 -90, -31 -66, -32 -94, 10

T2 T3 T9 TlO

Reverse turn Reverse turn Near reverse Reverse turn

I or near reverse turn I11 turn III or near reverse turn I

7’. I’. AW~4~dri et

al.

60

-60

-60

3333444444444455555555556666666 6789012345678901234567890~23456 Residues

Figure (bottom). deviation

6. (Jonformation

torsion angle plot of kringle and interkringle Head of each arrow corresponds to $ value of residue, opposite from 18Wsolid line segments; Xs. x1 values.

Table 6 Dihedral

angles of disuljide

bridges (deg.)

(‘A-CU-SGlLGGZ%CU-CA Xl x2 x3 x; x; Bridge

Xl

x2

x3

48-61

-81

66- I44 x7- 127 115-139

47 -74 -77

-54 77 -65 -132

C-C” (A)

x;

x;

- 109

-94

- 167

86 - 102 -68

-29 -178

-78 96

4.5.5 5.72

-162

-50

649

WI

link (top) and Gla-domain-disultidr loop end corresponds to 4; AOJ and i(‘C)N(IA

can be grouped into three clusters (Fig. 8). The largest cluster is the central one and forms the core of the tertiary structure of the kringle. It contains three aromatic (Trp90, Trp126, Tyr128) and two proline residues (Pro98, Pro1 18). Another cluster resides between the B and B’ loops (Fig. 3) consisting of Tyr94, Pro95, His96 and Pro132 (Fig. 8). The remaining cluster is the Tyr74, Pro125, Pro142 triplet. All the side-chains of these residues make close van der Waals’ contacts with one another and many of the aromatics make highly stabilizing edge-to-plane interactions. The first

Bovine Prothrombin

I

60

1,

I

Fragment 1

489

u 10 //I7 I

-60

0

60

120

180

Phi

Figure 7. ILamachandran plot of fragment 1. Gly residues boxed; @a-domain and cys48-61 kringle-pulses; interkringle link circles; iso-energy contours of Ala-Ala superimposed.

cluster surrounds and buries (Fig. 9) the two inner loop disulfide groups (Figs 1 and 4), thus also providing a highly hydrophobic environment. It is also worth noting that many of these residues are part of the Iysine binding subsite of fibrin binding (Tulinsky et al., 19886). However, only half of the second hydrophobic cluster is inaccessible from the surface (Figs 8 and 9). This apparent anomaly is most likely because only a fragment of prothrombin is being considered; this region of fragment 1 could be buried in the intact zymogen. A similar situation occurs with the Phe41, Trp42, Tyr45 aromatic cluster of the Gla-domain helix (Fig. 9). In this case, it has been shown that the cluster is buried when the Gla-domain folds (Soriano-Garcia et al., 1989). In any event, Tyr94 makes intermolecular contacts with the helix cluster in the crystal structure (Tulinsky & Park, 1988). The only aromatic or proline residues not clustered in this central-

stacked-column Phel14.

of

the

kringle

disulfide loop-circles;

are

Pro106

and

(d) cis-proline Another aspect of the kringle structure worthy of mention is that Pro95 is in a cis conformation. This first became apparent in our refinement of the structure of plasminogen kringle 4 at 1.9 A resolution (Mulichak & Tulinsky, 1990) and was then confirmed with several other related kringles (unpublished results of this laboratory). It led us to re-examine the final electron and difference density maps of fragment 1, which, not surprisingly, suggested a cis conformation for Pro95 Changing the conformation from tram produced significant improvements of the geometry of the kringle structure in the vicinity of Pro95 although the R-factor

7’. 1’. Seshudri _--.

490

-

et al.

remained essentially unchanged. The f,I(~*t ran cklnsity of ci,s Pro95 is shown in Wgurc IO. from which it can be seen that it, corresponds wchll to the> cis conformation. A43 Y45 ‘I‘Jti

5.50 A3 KFi:! R.52 ISi E63 I\67 MIX N73 s73 TX I TXI KU:! W” WI (23-I ( ‘H7 1JHX

RIII RI11 NI I3 Itl 16 IL1 I6 RI Iti KI 16 \VIP6 kVll6 (‘117 T129 Tl:lo (‘l-14 TI 50 I1S-t

NH NH NH OG NH NH NE NH NH NH NH ND% Xl):! NH ()(:I NE NH2 ot: NH NH HH NH NH2 NW NH SF: NH1 NH% NH NE1 UH iH ot:1 NH NH NH

1x39 F4 1 w42 E6X MU SRO El!1 P54 x.59 PI.42 x73 (‘61 C66 IX.5 El37 E 162 El52 El37 TX1 s79 YI%X I)109 I) 109 (‘87 WI26 11109 1’11.5 1) I on FL116 II119 El37 RI *75 t ES6 AtiT V111 IZXP

(10 (‘0 (‘0 OF.1 OEI (‘0 (‘0 (‘0 (10 (‘0 ODI (‘0 (‘0 (‘0 OE% (‘0 (‘0 OR2 (!O (‘0 (‘0 (‘0 OD’L (‘0 (‘0 c:o (‘0 ou1 co ODl (‘0 1‘0 ( ‘0 (‘0 1’0 (‘0

Figure 8. Stereo view of the arrangement

166 I69 ltil Iii

166 110 I69 Iti6 166 160 I40 164 160 150 I70 140 146 I73 I73 I59 IS4 148 154 14x I60 16%

n.m.r. studies have shown that, in plasnrinogen kringle 4, Tyr9. which is in the position equivalent to Tyr74 of fragrnent I, is fluctuating betwrrn two stat.es where in one it flips fast, while in the other is severely hindered (DeMarco et nl., 1985; Prtros rf CLI., 1988). Examination of the final electron tlt~nsity map of fragment 1 reveals that this t,yrosine ring is well ordered in the cryst,al st,ructure (Fig. 1 I ). Although Tyr74 is fairly accessible to solvent. it is also fairly constrained by a surrounding cage-like structure (Fig. 11). Nonetheless. it appears that the ring could occupy two different positions within the cage differing by torsional rotations that could lead t-o a separation of up to 2%) A; however, a frr(, flipping motion as implied by the n.rn.r. rrsuhs. would then be yuite impaired. In eit,her evrnt, thr, cage-like environment, is caonsistent, with thr 11.m.r. observations 1.0 indicate at least some accessibility main-chain; probe radius. 1.4 8.

Figure

10. Stereo view of electron density near cia Pro95. Basket contour at 1 x a(p).

to to

192

7’. I’. Seshadri

et, al.

Figure 11. Stereo view of electron density near Tyr74 (top) and Tyr74 embedded in molecular cage (bottom). (contours at, I x@(p)

protein (donor)-water (acceptor) hydrogen bonds is given in Table 8. The criteria used to identify hydrogen bonds in Table 8 were a distance of 125” and an occupancy of >0.5. The accessible surface area of fragment 1 has been calculated for a probe of radius 1.4 a (Lee & Richards, 1971) and the results are summarized in Figure 9. The average accessibility of the polypeptide chain is only 17o/o, while that of‘ the sidegroups is a high 82%. The latt>er is due to the overall discoid shape of the kringle

(z18Ax30Ax30A) and the fact that t.hc Gla-domain helix is exposed to solvent. Nonetheless, all the cystein residues of fragment 1 are buried and they and the residues in their immediate vicinity are completely inaccessible to solvent (Fig. 9). The average accessibility of the Gla-domain helix is about the same as that of the kringle while t,h(, interkringle link is considerably more exposed as is the Phe41, Trp42, Tyr45 aromatic cluster (Fig. 9). However, the latter are involved in an intrrmolecular crystal packing interaction about a crystallographic Z-fold rotation axis (Tulinsky et al.,

Bovine Prothrombin Table 8 Possible protein donor : solvent acceptor hydrogen bonds LVater

Residue ,litIllP

.Asn73 N wat159 0 I+97 NZ W&l61 0 Wat162 0 .lrg9l N Argl3ti NE - WatlXI 0 :&nlOl ND& [Vat190 0 wat 198 0 \‘a1151 s Awl.59 N1>1~ WatZlO 0 W&Z86 0 1hw110 N

(:lullP N

W&294

Argl56 NH2

W&%96 0

0

Distance (A) 2.3 1.8 1.7 3.0 34 2.3 2.4 2.1 26 29

1y Occupancy Angle (deg.) (AZ) factor 36.6 36.3 4C3 367 322 50.0 27.0 332 48.7 31.4

0.95 0.67 0.97 I a0 (Hi5 o-63 0.58 0s I 0.87 @43

144 126 150 129 136 145 126

132

130 140

1988) and are not solvent exposed in the crystal structure. As has already been mentioned above, they are not exposed in the Ca2+ fragment I structure where the Gla-domain is completely ordered and assumes a phospholipid binding conformation (Soriano-Garcia et al., 1989). Lastly, a surprisingly large number of small Gly and Ala residues are highly inaccessible to solvent (Gly64, 71. 76, 124: Ala51. 6i, 108). There are five water molecules occupying the gap between the (:la-domain helix and the kringle &main (Figs 3 and 4). These are W171, W172. W216, W278 and W299. The gap appears to be related t,o flexibility between the two domains that are hinged by t.he Leu62---Asn65 tetrapeptide, since the gap is narrowed in the Ca2+-fragment 1 struct,ure hy virtue of a 30” rigid body shift in the position of t.he helix that pivots between the two domains about the hinge region (Soriano-Garcia et (W171. d.. 1989). Two of the water molecules Wl72) make hydrogen bonds to the protein, while two others are within hydrogen bonding distance to ()I)2 of Asp39 (iV278. W299); in addition. the latter molecules are within hydrogen bonding distance to each other. This work was supported by NlH grant HL25942. We thank t,he CHESS facility for beam time and Drs Keith Moffat and Wilfried Schildkamp for helpful advice. We also thank the Purdue University NSF Computing Faci1it.v for computing time, to Drs Greg Kramer and Tim Schmidt for help in processing the oscillation data and to Dr Pappan l’admanabhan for carrying out some final calculations. The co-ordinates of the fragment 1 structure in the Protein Data Bank. have been deposited Brookhaven Sational Laboratories. accession number IPFI.

References Atkinson. R. A. & Williams, R. J. P. (1990). Solution structure of the kringle 4 domain from human plasminogen by H nuclear magnetic resonance spectroseopy and distance geometry. 1. Mol. Riot. 212. ,54l-552. Balasubramanian. R. (1977). Kew type of representation

Fragment I for

493

mapping

Nature

chain-folding

(London),

in protein

molecules.

266, 856-857.

Crawford, ,J. I~.. Lipscomb, X. W. & Schellman. C. G. (1973). The reverse turn as a polypeptidr conformation in globular proteins. Proc. ‘Vat. .-lcad. Sci., Cr.S.A.

70. 538-542.

DeMarco. A.. Pluck, X. D., Banyai. L.. Trexler. M.. I ,aursen. R. A.. Patthy. I,., Llinas. M. & Williams R. J. 1’. (1985). Analysis and identification of aromatic signals in the proton magnetic resonance specst,rum of kringle 4 fragment from human plasminoprn. Biochemistry.

24. 748-753.

Gardell. S. J.. Duong, Le T., Diehl. R. E.. York. J. I).. Hare. T. R.. Register. B. R.. Jacobs, .J. W.. Dixon. R. A. F. & Friedman. P. A. (1989). Isolation. characterization and rDXA cloning of vampire bat salivary plasminogen activator. J. Biol. Phcm. 264. I7947- 17952. Gunzler, W. A., Steffens, G. J.. Otting, F.. Kim. S.-M. A.. Frankus, E. & Flohe. L. (1980). The primary strurture of high molecular mass urokinasr from human urine. The caomplete amino acid sequence of the A-chain. Hoppe-Seylerk Z. Ph@ol. (‘hem. 363, 1155-I 16.5. Hendrickson. N. A. & Konnert, ,I. H. (1980). Incorporation of stereochemical information into In (‘omputing in crystallographic refinement. Crystallography (Diamond, R.. Ramaseshan. S. & K., eds). pp. 13.0~13.23. Indian Vankatesan, Academy of Sciences, Bangalore. Jones, T. A. (1982). FRODO: a graphics fit,ting program macromolecules. for In ( ‘ovnputational I).. Pd.). pp. 303- 317. C’rystallography (Savre. (‘larendon Press. 0xf:ord. Katz. B. A. & Kossiakoff. A. (1986). The crystallographically determined structures of atypicaal strained disulfides engineered into subtilisin. J. Hiol. (‘hrm. 261. 1548Om15485. Krishnaswamy, S., Mann, K. G. & Xesheim. M. E. (1986). The prothrombinase-cat,alyzed activation of prot,hrombin proceeds through t,he intermediate meizothrombin in an ordered. sequential/reaction. .I. Rid. Chem 261. 89774978. Lee. 11. & Richards, F. M. (1971). Interpretation of protein structures: estimation of static acc.essibility. .I. Nol. Hid. 55. 379-400. Luzzati. \-. (1952). Trait,ement statistiyur drs erreurs dans la determination ties structures ~~ryst~allines. .1cta C’rystallogr. 5. 802~810. Magnusson. S.. Petersen, T. E., Sottrup-*lensen, I,. & (‘laeys. H. (1976). Complete primary structure of prothrombin: isolation, structure and reactivity of ten carboxylated glutamic acid residues and regulation of prothrombin activation by thrombin. Tn Proteases and Biological Control (Reich, E., Rifkin, I). 13. &, Shaw. E., eds), pp. 123.-149. (Iold Spring Harbor Laboratory Press. Cold Spring Harbor, ICY. Magnusson, S.. Sottrup-Jensen. I,.. Petersen. T. E., Dudek-Wojciechowska, G. & Claeys. H (1976). Homologous kringle structures common to plasminogen and prothrombin substrate specificity of enzymes activating prothrombin and plasminogen. tn l’roteolysis and Physiological Regulation (Robbins, I). W. & Rraw. K., eds). pp. 203. 208. Acaademic Press, Sew York. Maigret, B.. Pullman, B. (G.Perahia, D. (1971). .&Iolecular orbital calculations on the conformation of polypeptides and proteins. VII. Refined calculations on the alanyl residue ,I. Theowt. Riol. 31, 669S285.

494

_-

7’. 1). Seshadri

Mann. K. G. (1987). The assembly of blood clotting csomplrxes on membranes. Trends Biochem,. S-i. 12. 229 233. M(:Lean, J. W., Tomlinson, ,J. E., Kuang, W. ,J., Eat,on, D. L., Chen, E. Y., Fless, 6. M., Scanu, A. M. & Lawn, R. M. (1987). cDNA sequence of human apolipoprotein (a) is homologous to plasminogen. Nature (London), 330, 132-137. McMullen. B. A. & Fujikawa, K. J. (1985). Amino acid sequence of the heavy chain of human a-fact,or XIIa (activated Hageman factor). .J. Biol. Chem,. 260, 5328-5341. Mulichak, A. M. & Tulinsky, A. (1990). Structure of the lysine-fibrin binding subsite of human plasminogen kringle 4. Blood Coag. Fibrinolys. 1, 673-679. ?u’akamura, T. Nshizawa. T., Hagiya. M., Seki, T., Shimonishi, M., Sigimura, A., Tashior. K. & Shimizu, S. (1989). Molecular cloning and expression of human hepatocyte growth factor. Nature (London), 342. 440.-443. Nesheim. M. E., Hubbard, L. S., Tracy. P. K., Bloom, J. W.. Myrmel, K. H. & Mann, K. (:. (1980). Participation of factor V, in prothrombinase. In 7% tlrgulation of (:oaguZation (Mann. K. G. & Taylor. F. K.. eds), pp. 145-149, Elsevier, sorth Holland. Park. (1. H. & Tulinsky, A. (1986). Three-dimensional structure of the kringle sequence: structure of prothrombin fragment 1. Biochemistry, 25. 3977-3982. Pennicsa, I)., Holmes. W. E., Kohr, W. J.. Harkins, R. N.. Yelverton, E.. Reeburg. P. H.. Heynecker. H. L., (:oeddel. 1). V. & Collen, D. (1983). Cloning and expression of human tissue-type plasminogen activator cDNA in E. coli. Nature (London), 301, 214~m221. Petros. A. M.. Gyenes, M., Patthy, L. & Llinas, M. (1988). Analysis of the aromatic ‘H NMR spectrum of chicsken plasminogen kringle 4. Arch. Biochem. Biophys. 264, 192-202. Priestlr, ,J. (1988). RIBBON: a stereo cartoon drawing program for proteins J. Appt. Crystallogr. 21. 57%576. Edited

et al

------____--.-----_

-... -~~--.

Rossmann. Xl, (2 ( 1979). Processing oscillation tlitfl~ac%iorl data for very large unit c~ells with an autontatil, clonvolution techniquta and profik titt,ing. .1. .1rjp/. ~,Yry.stallogr. 12, 225- 238. Rossmann. M. (i.. Leslie, A. (:. \V.. Abdel-Meguid. S. S. lu: Tsakihara. T. (1979). l’rocessing and post-retintment of oscillation (‘amera data. .I. ,4ppl. (‘rysfn(loyr. 12. 570 5x1. Soriano-Garc*ia. M.. Park. (‘. H., ‘fulinsky. A.. K. G. $ Skrzypc:zak-,JankurI, E. Ravichandran, (1989). Structure of (‘a’+ prothrombin fragment 1 inc.luding the conformation of t,hr (:la domain. Biochrmistry,

28, 680.568

IO.

Srinivasan, .4. It. & Yathidra, PI;. (1978). The c*ontorma~ tional wheel-A circular plot for correlating thr primary and secondary structure of a linear p$Trner with the conformational angles of its indlvldual residues. Hiopolynhers, 17, 159,% 1600. Steffens, (:. .J.. (iunzler. W A., Otting, F.. Frankus. E. Cyr Flohr, I,. (1982). The c*omplete amino acid srquenct~ of low molecular mass urokinase from human urintb. Hoppe-Seyler’s Z. Physiol. CIhe,m. 363, 1043 1058. Trexler, 141.& Patthy. I,. (1983). Folding autonomy of the kringle 4 fragment of human plasminogrri. I’roc. :~uI. dca,d. Ski., T’.S.A.

80, 2457-2461

Tulinsky. A. & Park. C. H. (1988). Structure of prothronbin fragment 1 and its relat,ion to c*alcium binding. In C~urrrnt ildvances in C.itanrin, K tlwarch (Suttie. .J. W.. ed.). pp. 295 304, Elsrvier Scien(>e Publishing 00.. Inc... Amst,erdam. Tulinsky, A.. Park, (‘. H. bz Skrzypczak-,Jartkrrn, E. (1988a). Structure of prothrombin fragment I refined at 2.8 A resolution. ,I. Mol. Hiol. 202, 88.5901. Tulinskv. A.. Park, (‘. H., Mao, R. & Llinas, M. (19W). Lysine/fibrin binding sites of kringles modeled aft,rr the structure of kringlr 1 of prothrombin. Proteins, 3. 85-96. Watenpaugh, K. D., Sieker, L. C. & .Jensen. I,. H. (1979). The structure of rubredoxin at I.2 .q resolution. J. Mol.

by R. Hubar

Biol.

131. 509.-522.