D'Vl ‘lf'l"
' , ,‘.

l
w
v

.v-y- ' 1" vv‘ v'v‘r v»

.qu-wu-nu..u.._"..y

 

S ;v :3}: I»)..J 3)..
. hull. 15.4.5...

“7.19%.!
1.9

J”...

u.

.

. . 3.
@-

.,.o 2

Lung 3,..1
. 1::

Hitl‘l‘

‘ .V...(It:..#-:.L
e) .‘lfiisa-l
3.? to} Elli-D»... ‘3
121.31.! a
6.75

. V! in .5
a Lavina.” Sfip.
.V .1
0o ‘ (.9.

it it... rnnfinflfldvfi!rlfifl.

. 9.. uuflflflmhvfis.s...z.£ .
rat-:- . 5'53: .
33... Zr?

fl

in
.251!
3"!!!ng iii?
.55. ..ei.ItuJu.» .-
l l‘ u 0 i]: v
:05: .. ; r‘lixuoolsnnllH
001...)...35tr? ..
u ‘2

.

1.. .ci

.. ti.

.lxii-Ei: A.
0... Cl

. 3x31“... X...

”up-lo. (‘3‘;

 

 

 

”"leSlS

iiiililiiiiii'iiiiiiililiii

3 1293 00885 3289

 

 

This is to certify that the

dissertation entitled

Crystal Structure Studies of Kringle
Domains and A Peptidomimetic Inhibitor
Complex of d-Thrombin

presented by

Tswei—Ping Wu

has been accepted towards fulfillment
of the requirements for

Ph.D Chemistry

degree in

 

 

AU? .ij

Major professdr

 

Date 8/2/93

 

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

Cr.

31'

 

 

Crystal Structure Studies of Kringle Domains
and a. Peptidomimetic Inhibitor Complex of

a-Thrombin

By

Tswei-Ping Wu

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemistry

1993

ABSTRACT

Crystal Structure Studies of Kringle Domains
and a Peptidomimetic Inhibitor Complex of

a-Thrombin
By

Tswei-Pz'ng Wu

The crystallographic structures of the PGK4-c-aminocaproic acid (ACA) com-
plex, PGKI and tPAK2 have been determined and show that each of them possesses
a lysine binding site composed of three distinct regions according to electrostatic
properties: (1) a negatively charged region containing two negatively charged aspar-
tic acid residues, (2) a distinct positively charged region due to one (in tPAK2) or two
(in PGK4-ACA and PGKl) side chain amino groups and (3) a hydrophobic region,
which is composed of aromatic residues, separates the oppositively charged regions.
The lysine binding sites of apo-PGK4 and PGK4-ACA have been compared and
most likely the lysine binding site is preformed, lysine binding does not require con-
formational changes of the host. In the tPAK2 structure, three molecules are found
in the asymmetric unit. The crystal structure shows a strong interaction between
a lysine residue of one molecule and the lysine binding pocket of the neighboring

molecule. This interaction mimics the ligand binding interaction found in the PGK4-

“[873.be
V

A9". r, .
R‘n‘wd 5V

‘
0.1.? v‘
“A‘s!

w:

ACA complex. The overall conformation of the PGKl structure is very similar to
that of apo-PGK4. The largest deviations are due to insertion of Gln59 in PGKl and
intermolecular interactions occuring around the two—fold axis relating the segment
between two neigboring PGKl molecules.

a-Thrombin displays remarkable specificity, effecting the removal of fibrinopep—
tides A (FPA) and B (FPB) of fibrinogen through the selective cleavage of two Arg-
Gly bonds among 181 Arg/Lys-Xaa bonds in fibrinogen. A model for the thrombin-
bound structure of FPA has been proposed based on NMR data and computer-assisted
molecular modeling. In order to obtain a better understanding of the interplay be-
tween the primary sequence and the conformation required for thrombin substrates
and inhibitors, a crystallographic investigation of the FPA mimetic (F PAM) com-
plexed with human a-thrombin was undertaken. Crystals of ternary complexes
of FPAM, hirugen and thrombin were grown and three-dimensional intensity data
were measured to 2.5 A resolution. The crystallographic structure of FPA and its
chloromethylketone derivate bound to thrombin were determined. Although there
are differences between these structures, in the above modeled FPA structure, and
that of the crystal structure of FPAM bound to thrombin, the (15, 1,!) angles in the
critical region of Pl-P2-P3 in all of the structures are similar to those of bovine
pancreatic trypsin inhibitor (BPTI) in the BPTI-trypsin complex and D-Phe-Pro-
Arg—chloromethylketone (PPACK) in the PPACK-thrombin structure.

To my parents

iv

ACKNOWLEDGMENTS

I wish to express my sincere appreciation to Professor Alexander Tulinsky for his
advice, guidance and financial support throughout the course of this work. My ap-
preciation is also extended to all present and former members of Professor Tulinsky’s
research group for their assistance, friendship and good well. I especially thank Dr.
Pappan Padmanabhan for the numerous discussions and Dr. Pushpa Padmanabhan
for her expert crystallization technique. The completion of this work was made possi-
ble through the supplying of plasminogen kringle 4 protein sample, kringle 1 protein
sample, and the coordinates of tisste plasminogen kringle 2 structure from Dr. Miguel
Llinas, Dr. Francis Castellino and Dr. Bart de Vos, respectively. Thanks are also
extended to Dr. Mike Kahn for synthesizing mimetic of fibrinogen peptide A for us.

My gratitude and appreciation for life goes to my parents and parents-in-law for
their endless love and support. Last, but not in any way least, I would like to thank
my husband Tyan-Shu for his unending understanding and encouragement, especially
in my difficult time. His patience for teaching me in using IATEX to word-process this
dissertation is greatly appreciated.

LIST OF '
lBTOF
1 lhbnn

2 Experi
2
3

l C:

53 l").

'33.

3.;

33 D

'2.

‘3.

3-4 n

2.

‘2.

“2.3- 5.

3 Strut!

3.1 l

3-‘3 .\1

3.3 it

4 5mm

4.1 I;

4.2 \l

4.3 I
5

TABLE OF CONTENTS

LIST OF TABLES vii
LIST OF FIGURES ix
1 Fibrinolysis and Kringles 1
2 Experimental and Computational Methods 9
2.1 Crystallization ............................... 9
2.2 Data Collection .............................. 12
2.2.1 Diffractometcr ........................... 12

2.2.2 Area Detector ........................... 15

2.3 Data Reduction .............................. 18
2.3.1 Diffractometer Data ....................... 18

2.3.2 Area Detector Data ........................ 19

2.4 Molecular Replacement .......................... 22
2.4.1 Rotation Search .......................... 24

2.4.2 Translation Search ........................ 26

2.5 Structure Refinement ............................ 27

3 Structure of the ACA Complex of Human Plasminogen Kringle 4 31

3.1 Introduction ................................ 31
3.2 Materials and Methods .......................... 35
3.3 Results and Discussion .......................... 49
Structure of Human Plasminogen Kringle l 61
4.1 Introduction ................................ 61
4.2 Materials and Methods .......................... 65
4.3 Results and Discussion .......................... 74

Structure of Tissue—type Plasminogen Activator Kringle 2 Domain 100
5.1 Introduction ................................ 100

vi

5.? Exp

6 Compar

7 The Str
a-Thrm
7.1 Int:
752 EX;
7.3 RE
7.4 Dir

A Amino

BIBLIOG

5.2 Experimental Procedure ......................... 105
5.3 Results and Discussion .......................... 107

6 Comparison of Different Kringles 133

7 The Structure of A Designed Peptidomimetic Inhibitor Complex of

a-Thrombin 143
7.1 Introduction ................................ 143
7.2 Experimental Procedures ......................... 151
7.3 RESULTS ................................. 155
7.4 Discussion ................................. 170
A Amino Acid Shorthand Used in the Thesis. 173
BIBLIOGRAPHY 174

vii

[Q
,_,_4

352
3.3
3.4

3.6-

(f,-

D.
II}

N.

W 9‘

.C‘FSETG

(I)
A ."

F?

.',.
>v'

1.1

2.1

3.1
3.2
3.3
3.4
3.5
3.6

3.7
3.8
3.9

4.1
4.2

4.3
4.4
4.5
4.6

4.7

LIST OF TABLES

Comparison of primary structure of homologous kringles. The number-
ing of the residues is standardized to PG K5. Abbreviations used: PT,
prothrombin, PG, plasminogen, uPA, urokinase—type plasminogen ac-
tivator; HGF, hepatocyte growth factor; tPA, tissue—type plasminogen
activator; FXII, factor XII .........................

Rotation matrix in terms of Eulerian angles 01, 02, 03 ..........

Summary of crystal data of monolinic K4—ACA ............
Averaging background in 29 shells. ...................
Decay rate, 5, calculated in each 20 range. ...............
Weights of Reflections and R Values of the Final Refinement .....
Summary of Final Least-Squares Parameters and Deviations .....
New Intramolecular Hydrogen Bonds of K4—ACA. Hydrogen atoms
were assigned geometrically idealized positions. Donor atom is denoted
(D), aceptor atom (A) ........ ' ...................
Hydrogen Bonds Between K4 and ACA .................
RMS difference between the K4—ACA and Apo—K4 ..........
Differences in lysine-binding sites between K4—ACA and Apo—K4

Summary of crystal data of tetragonal K1. ...............
Distribution of reflection intensities and R—factors in various resolution
shells. ...................................
Weights of Reflections and R Values of the Final Refinement .....
Summary of Final Least—Squares Parameters and Deviations .....
Secondary structural elements of PGKI structure ............
Hydrogen Bonds found in the lysine binding site of the PGKl structure.
The prime designates a symmetry—related molecule. A donor atom is
denoted (D), an acceptor atom (A) ....................
Intermolecular interactions of PGKl < 3.7A ...............

viii

6

24

44
47
48

53
56
56
58

68

72
73
77

87
94

9"

“CAD

CJ'

0‘

9"

Lu-

. -

4.8

5.1
5.2
5.3

5.4
5.5

5.6

5.7

6.1

7.1

7.2

7.3
7.4
7.5

Deviations of the positions of residues in the lysine binding site between

the PGKl and PGK4 structures. .................... 98

A summary of the final refinement results of the X—PLOR refined struc-
ture. .................................... 106
Weights of reflections and R values of the final refinement cycle of tPAK2108
Summary of final least—squares parameters and deviations of tPAK2 . 109
Secondary structural elements of the tPAK2 structure. ........ 115
The intramolecular and intermolecular interactions present in the ly—
sine binding site of tPAK2 structure. Prime refers to neighboring
molecules. Donor atom is denoted (D), acceptor atom (A). ...... 124
RMS deviations among the three molecules in the asymmetric unit.
MA, MB, MC represent molecules A, B and C. Numbers shown in
parentheses were calculated without including Met28 .......... 126

Deviations of the positions of residues in the lysine binding site between

tPAK2 and K4—ACA ........................... 131

Hydrogen bonds of the conserved disulfide anti—parallel fl—sheets of
PGKl, PGK4, tPAK2, PTKI, and PTK2 structures. ......... 138

The primary seuence of human a—thrombin. The thrombin residue

numbers are assigned by homology with chymotrypsin. Insertions are

represented by alphabetic characters followed by numbers. ...... 146
Distribution of reflection intensities and R—factors in various resolution

sheHs. ................................... 153
Weights of reflections and R values of the final refinement ....... 157
Final least squares parameters and deviations of FPAM—thrombin. . . 158
RMS Deviations Between the Hirugen—Thrombin and FPAM-Hirugen—

Thrombin Complexes ........................... 161

ix

{Q
0»

IO
..

[.9
$.-

I‘D
g)!

5‘.)

A;

¢~"
s; «5"

1.1

1.2

1.3

1.4

2.1
2.2
2.3

2.4
2.5

2.6

3.1

3.2

3.3

LIST OF FIGURES

The model of the fibrinolytic system. Abbreviations used: PG, plas-
minogen; PM, plasmin; tPA, tissue plasminogen activator; PAI-l, plas-
minogen activator inhibitor type 1; LS, lysine sites of fibrin. .....
The three—disulfide triple—100p kringle structure. Disulfide bridges are
indicated by bold lines ...........................
Lysine ligands analogs. Abbreviations used: ACA, aminocarproic acid;
BASA, p-benzylamine-sulfonic acid; AcLys, N-acetyl-L-lysine; AM-
CHA, trans-4—(aminomethyl)-cyclohexanecarboxylic acid. ......

Comparison of the folding of PGK4 and PTFl. PTFl in bold. . . . .

Crystallization by vapor diffusion technique. (A) Hanging drop method
(B) Sitting drop method ..........................
Diagram of four—circle diffractometer ...................
Block diagram of four—circle area detector system ............
Graph of count rate vs. bias control setting. ..............
The Eulerian angles 01, 02, 03 that relate the rotated axes X’, Y’, Z’
to the original axes X, Y, Z. The rotation operation consists of 1.) a
rotation by 01 around Z—axis. 2.) a rotation by 62 around new X-axis.

3.) a rotation by 03 around new Z—axis. ................

Schematic Structure of "PROFFT” and “PROLSQ”. .........

The chain structure of plasminogen. Kringle domains are labeled Kl
- K5. Disulfide bridges are shown by bold lines. Asterisks denote
active site residues in the serine protease domain. Arrows indicate the
cleavage sites. ...............................
The primary structure of plasminogen K4. Interkringle residues are
a—c and 81—87. (see appendix for abbreviations) ............
Stereoview of modeled lysine binding site of plasminogen kringle 4.ACA

in bold. Hydrogen atoms are put at ideal positions [38] .........

10
13
16
17

32

34

36

3.4

3.6
3.7
3.5
3.9

3. i [i

3.1”

«II

“g
(J

4,3
4.4

.7)

4.6

3.4
3.5
3.6
3.7
3.8
3.9

3.10

3.11

3.12

4.1

4.2

4.3
4.4
4.5
4.6

4.7
4.8
4.9

4.10

4.11
4.12

Photograph of K4—ACA crystal. The size of the single crystal is ap-
proximately 1.2 x 0.3 x 0.1 mm ......................
Schematic diagram of K4—ACA crystal morphology with respect to
crystallographic axes ............................
Axial intensity distributions of K4—ACA crystal .............
The w—profile of reflection (5, 5, 5) taken before data collection.

Absorption correction curve of (0, 2, 0) reflection. ...........
Ramachandran plot of final Q5, t/i angles of K4~ACA. The Gly residues
are boxed. .................................
Stereoview of the electron density of the lysine—binding site of K4-
ACA. The basket contour is at 10; ACA is between Asp55, Asp57 and
Lys35, Arg71. ...............................
Stereoview of the lysine-binding site of K4-ACA. ACA is shown in
bold; hydrogen bonds are dashed. ....................
Stereoview of the comparison of the lysine—binding site of K4—ACA
and apo—K4. K4—ACA is in bold; the sulfate position in apo—K4 is
designated by S ...............................

The primary structure of human plasminogen kringle 1. Site number-
ing deletion is based on homology with PGK5 ..............
Stereoview of the modeled lysine binding site of plasminogen kringle
1; ACA in bold. Hydrogen atoms are shown at ideal positions (taken
from [38]) ..................................
Photograph of K1 crystal. Crystal length is approximately 1.2 mm. .
The Ramachandran plot of PGKl. Glycine residues are boxed. . . . .
Distribution of w angles in the PGKI structure. ............
Plot of the average thermal factors of main—chain (thin line) and side—
chain (thick line) atoms of PGKI structure. ..............
Distribution of temperature factors for water molecules .........
Distribution of occupancies for water molecules. ............
Stereoview of the internal water molecule in PGKl structure. Dashed
lines indicate hydrogen bonds .......................
Stereoview of the PGKI aromatic and proline residues forming hy-
drophobic core. ..............................
Stereoview of the PGKl lysine binding site. ..............
Electron density in vicinity of Arg34 and Arg71 of PGKl structure;

contour at 10 ................................

xi

38
39
40
41
43

50

51

55

59

62

64
67
75
76

78
80
81

82

83
85

86

«H3 Sirrt'
bluff;
ill fllil.‘
lurv~

ill) ‘1)"

.I. ,.
4.: 5m.

'1

4.1 \I‘
x l'"
v I
:j-(‘\

o...”
((2
D“
V
1

s. n
'. g].
t." i
4.-
‘ f
-

t."
O”)
,-
O

c"
In 3”
3‘ J
- I‘ "
Oi! -
5.
r.
“,9 ‘ -
a);
2
'J.\ ,
Si"?
‘5

4.13

4.14

4.15

4.16

4.17

4.18

4.19

5.1

5.2

5.3

5.4

5.5

5.6

5.7

5.8

Stereoview showing intermolecular interactions involved in the lysine
binding site. The prime denotes a symmetry—related molecule .....
RMS deviations of CA, C, N atoms between PGKl and PGK4 struc-
tures. Filled circles indicate conserved residues. ............
RMS deviations of side chains between PGKl and PGK4 structures.
Filled circles indicate conserved residues. ................
Stereoview of the optimally superimposed main chain structures of
PGKl (bold) and PGK4. ........................
Stereoview of crystal packing interactions at a two—fold symmetry axis.
Residues involved in intermolecular interactions are shown in bold. The
two—fold axis in the a—b diagonal is indicated. .............
Stereoview of intermolecular interactions found in the lysine binding
site of PGK4. The side chains from the second symmetry—related
molecule are shown in bold; the sulfate ion is designated with a filled
circle [23]. .................................
Stereoview of the superposition of the lysine binding sites of the

PGK1(bold) and PGK4 structures. ...................

Schematic presentation of the primary structure of human tissue—type
plasminogen activator. The arrow indicates the cleavage site for the
conversion of single chain tPA to two chain tPA. The active site residues
are indicated by asterisks. ........................
Primary structure of tPA kringle 2. Site numbering insertions and
deletions are based on homology with PGK5 ...............
Electron density of chloride ion in the lysine binding site of tPAK2.
Contour at 10. ..............................
Stereoview showing an internal water molecule in the tPAK2 structure.
Hydrogen bonds are represented by dashed lines .............

Stereoview showing the water molecule found in an empty pocket of

the tPAK2 structure. Hydrogen bonds are represented by dashed lines.

Stereoview of CA, C, and N structure of tPAK2. Disulfides are shown
in bold. ..................................
Stereoview of the crystal packing of the three molecules in the asym-
metric unit of the tPAK2 structure. The non-crystallographic 31 screw
ax is is perpendicular to the plane. ...................
Stereoview of the crystal packing of the three molecules in the asym-
metric unit of tPAK2 structure; The non—crystallographic 31 screw axis

is indicated. ................................

xii

88

90

91

92

95

96

97

101

103

111

112

114

116

3

In.

{J

5.9

5.10

5.11

5.12

5.13

5.14

5.15

5.16

5.17

6.1

6.2

6.3

6.4

6.5

6.6

Stereoview showing the intermolecular interactions between Lys48' and

Lys33, Asp55, and Asp57. Prime represents the neighboring molecule.

Stereoview showing the intermolecular interaction between Asn26' and
Asp55 bridged by a water molecule. Prime represents the neighboring
molecule ...................................
Stereoview of the crystal packing of the three molecules along the a—
axis. Residues involved in intermolecular interactions are shown in
bold. ....................................
Stereoview of the lysine binding site of the tPAK2 structure. .....
Intramolecular and intermolecular interactions present in the lysine
binding site of the tPAK2 structure ....................

Stereoview of the comparison of the main—chain atoms of the three
independent molecules ...........................
Stereoview of the comparison of the side—chain atoms of the three in-
dependent molecules ............................
Stereoview showing the comparison of side chains of Hi313 in three
molecules. Hisl3 in molecule B is indicated. ..............
Stereoview of superimposed lysine binding sites of tPAK2 (bold) and

K4—ACA. Lys48’ is between Asp55, Asp57 and Lys33 in tPAK2; ACA
is between Asp55, Asp57 and Lys35, Arg71 in K4—ACA. .......

Stereoview of the comparison of the CA, C, N structures of different
kringles. PGKI, PGK4, and PTKl are. shown in thin lines; tPAK2 in
bold; PTK2 in dashed. ..........................
Stereoview of the comparison of the inner disulfide bridges of PGKl,
PGK4, tPAK2, PTKl and PTK2. Disulfide bonds in bold. ......
Stereoview showing the conserved anti—parallel B—sheets located near
disulfide bridges in PGKl, PGK4, tPAK2, PTKl, and PTK2.

Stereoview comparing the lysine binding regions of PTKl (bold) and
PGK4 ....................................
Stereoview comparing the lysine binding site regions of PTK2 (bold)
and PGK4. ................................
Stereoview of the comparison of internal aromatic residues in PGKl,

PGK4, tPAK2, PTKl and PTK2 .....................

xiii

118

119

121

122

123

127

128

129

132

134

136

137

140

141

142

-'

A:

A

(-.

'1
hr'fl

-4“
.1
J'r
KM“:

7.1

7.2

7.3

7.4

7.5

7.6

7.7

7.8

7.9

7.10

7.11

7.12

The conversion of prothrombin to thrombin. CHO represents a carbo-
hydrate side chain; a, b, and c represent the cleavage sites of bovine
and human prothrombin; b’ represents an additional cleavage site in
human prothrombin. ...........................
The structure of the fibrinogen molecule. The symmetric molecule is
composed of a dimeric central domain containing the N—teminal of all
six chains (0, ,8, 7), two connecting coiled coils, two terminal domains,
and two A0 polar appendages. Four carbohydrate clusters (CHO)
occur, and are located on each 7 chain near the central domain and
on the ,3 chains of each terminal domain. Primary cross—linking sites
(XL) can be found near the C—terminal of the 7 chain and in the A0
polar appendages (taken from [76]) ....................
Scheme for synthesis of FPAM. DEAD, diethyl—azodicarboxylate;
EDC, 1—ethyl—3 (3—dimethylaminopropyl) carbodiimide; HOBT, hy-
droxybentriazole; DMAP, diinethylaminopyridine; DMF, dimethylfor-
mamide ...................................
Stereoview of the FPA M structure docked in the thrombin active site.
FPAM is in bold. .............................
Photograph of FPAM crystal. Crystal size is approximately 0.5 x0.45 x
0.1 mm ...................................
Numbering of and special restraints used for FPAM in restrained least
squares refinement. Since part of F PAM is a non—amino acid group,
special bond, angle and planar 1,-4 distance restraints were applied
during refinement to maintain geometry. ................
Ramachandran plot of a, 'in angles of FPAM—thrombin structure.
Glycines are not displayed .........................
Stereoview of the comparison of the active sites of thrombin in the
FPAM and hirugen complexes. Hirugen—thrombin, broken lines; active
site is unoccupied. ............................
Stereoview of the electron density corresponding to FPAM in the
thrombin complex. Basket contour at 10 .................
Stereoview of FPAM bound in active site of thrombin. FPAM in bold;
hydrogen bonds, broken. ......... ' ............ . . . . .
Stereoview of the comparison of FPAM (bold) and PPACK in their
thrombin complexes. ...........................
Stereoview of the comparison of FPA (bold) and FPA-chloromethyl

ketone in their thrombin complexes ....................

xiv

144

147

149

152

156

159

160

162

163

164

167

7.13 Stereoview of the comparison of FPAM (bold) and FPA in their throm-

bin complexes ................................ 169

XV

CHAPTER 1

Fibrinolysis and Kringles

In the blood coagulation cascade, fibrin is formed to block the flow of blood from a
severed vessel. After it has fulfilled its function, it is digested to soluble fragments by
the enzyme plasmin; we call this process of dissolution of the fibrin blood clot “fib-
rinolysis”. The fibrinolytic system, schematically represented in Figure 1.1, consists
of three main components: plasma zymogen plasminogen (PG), its activated prod-
uct, the proteolytic enzyme plasmin (PM) and tissue plasminogen activator (tPA).
As proposed by Wiman and Collen [1], when fibrin is formed, a small amount of
plasminogen is always specifically bound to it. Plasminogen activators present in the
blood or released from the vascular endothelium are adsorbed on the fibrin surface
and efficiently activate the plasminogen (Figure 1.1). Plasmin released from the clot
is rapidly inactivated by the fast—acting plasmin inhibitor ag—antiplasmin ; therefore,
other plasma proteins are not attacked by plasmin. Since plasminogen and inhibitor
ag—antiplasmin are abundantly present in blood at rather stable levels, the blood fibri-
nolytic activity is determined by the balance between tPA and plasminogen activator
inhibitor type 1 (PAI—l).

The interactions between PM, PG and tPA with lysine residues of fibrin are 10-
calized at lysine binding sites that reside in the kringle domains of these molecules

(Figure 1.1). The interactions are thought to occur with exposed lysine residues of the

 

 

inactive complex inactive complex

PAI— l

antiplasmin

- -
——> .
---

 

Fibrin Degradation Products

Figure 1.1. The model of the fibrinolytic system. Abbreviations used: PG, plasmino-
gen; PM, plasmin; tPA, tissue plasminogen activator; PAI-l, plasminogen activator
inhibitor type 1; LS, lysine sites of fibrin.

A4 .
c—r-p»

._n_ a...

't' 'v J.
n

.1...

 

.r“ 0
‘~.’.".\03 'v
\ “
‘fi
. 64" -
A. \ (4"
n ‘-
I".
~.
A i’?‘l
A; “Q ‘
- §‘
§
.,.
1' §
\‘ <.‘ "s
~<1 c..—
.
.
‘V
k,-" ~
4.:
i.
.
ql .

 

fibrin peptide chain [2]. The affinity between lysine binding sites and lysine residues
not only localizes the PG and tPA to the fibrin clot, but also keeps plasmin formed
on the fibrin surface from being inactivated by antiplasmin.

There are several lysine binding sites in plasminogen, one binding site with high
affinity and four sites with low affinity [3, 4, 5]. The strong binding site is located
within the first kringle of plasminogen, which can bind specifically to fibrin through
this site. Since the lysine binding sites of plasminogen and plasmin mediate their in-
teractions with fibrin and antiplasmin, the interactions suggest that the lysine binding
sites may play a crucial role in the regulation of fibrinolysis.

The kringle domains have three disulfide triple loop patterns (Figure 1.2) and each
consists of 80-85 residues of molecular weight approximately 10,000 daltons. Limited
proteolysis of prothrombin (PT) and PG is known to yield fragments corresponding
to kringles, indicating that kringles are independent structural units [6, 7]. Also,
these fragments retain their original lysine binding function after being isolated and

they are thought to be independent functional domains [7, 8, 9].

Figure 1.2. The three—disulfide triple—loop kringle structure. Disulfide bridges are
indicated by bold lines.

Kringles are present in several non—catalytic regions of trypsin—type proteases of
blood plasma that are involved in blood coagulation and fibrinolysis [10]. It has been

shown that kringles occur singly in urokinase—type plasminogen activator (uPA) [11],

factor XII [12], as pairs in prothrombin [6], tissue—type plasminogen activator [13],
as 4 copies in hepatocyte growth factor (HGF) [14], and as 5 copies in plasminogen
[7]. Most interestingly, there are 38 kringles found in apolipoprotein[a] [15], 37 of
which display 75—85% conservation with kringle 4 of plasminogen.

Although the different kringles have the same disulfide triple loop pattern, their
binding specificities diverge in binding different proteins or low—molecular weight
ligands. For instance, the kringle 5 (K5) fragment of PG binds to benzamidine—
Sepharose [16], fragment 2 of PT , which corresponds to the second kringle of the PT,
has an intact binding site for factor V. [8]. In addition, K1 and K4 of PG [9, 17] and
K2 of TPA [18] not only bind to fibrin and lysine but also to some w-aminocarboxylic
acids, including 5—aminocarproic acid (ACA), p-benzylamine-sulfonic acid (BASA),
N-acetyl-L-lysine (AcLys) and trans-4-(aminomethyl)-cyclohexanecarboxylic acid
(AMCHA) (Figure 1.3). All of the ligands have positive and negative charged groups
about 6.8 A apart and have been found to have an antifibrinolytic effect in vivo [19].

A comparison of kringle sequences found in some proteins is given in Table 1.1,
from which it can be seen that the kringles of the different proteases show a high degree
of sequence homology; in fact, they appear to be more closely related than the protease
parts themselves [20]. The comparison also shows that about 25% of the residues are
conserved absolutely among kringles from different proteins. These conserved residues
are found around Cys22—Cys63 and Cys51-Cys75, which are essential for the kringle—
fold. This observation supports the hypothesis that conserved residues in all kringles
are essential for the folding autonomy of the domains.

The first kringle structure solved by X—ray crystallography was the K1 of pro-
thrombin fragment 1 (PTFl) [21, 22]. Recently, human plasminogen kringle 4 has
also been determined [23] and shows a similar three dimensional kringle structure to
that of PTFl. A comparison of the folding of these two kringles is shown in F ig-

ure 1.4, from which it can be seen that the structures of the kringles are somewhat

O lClCl.ClCl.Cl.\. cu

CH2 ' CH3-CO -NH—CH

| |
IHZ - CH2
in
2
l

T”
CH2 CH2
l l CH2
CH 2 CH 2 H I
I . l . | 1 NH;
NH3 NH 3 NH3
E ACA AcLys AMCHA BASA
CH
1
CH2
|
CH2
|
CH2
l
CH2
C CH
CH ,_ 2
2 H N 5 +\\. NH l
I + 2 2 NH +
NH 3 ' 3
Hexylamine Benzamidine Benzylamine

Figure 1.3. Lysine ligands analogs. Abbreviations used: ACA, aminocarproic acid;
BASA, p-benzylamine-sulfonic acid; AcLys, N-acetyl-L-lysine; AMCHA, trans-4-
(aminomethyl)-cyclohexanecarboxylic acid.

,4‘

1
s

I
b-

A

-
.F
O.

I.”

v.-

~oi.

A c c Q Q ¢ 4‘
"I... rm VI... .03..

I.“ In“ A
.7 2. a. «J
.0 .u “N ‘UQ

Table 1.1. Comparison of primary structure of homologous kringles. The numbering
of the residues is standardized to PG K5. Abbreviations used: PT, prothrombin, PG,
plasminogen, uPA, urokinase—type plasminogen activator; HGF, hepatocyte growth
factor; tPA, tissue—type plasminogen activator; FXII, factor XII.

1 10 2O 30 40

l l l l l
tPA K2 GNSD CYFGNGSAYR GTHSLTESGA SCLPWNSMIL IGKVYTAQNP
tPA K1 DTRAT CYEDQGISYR GTWSTAESGA ECTNWNSSAL AQKPYSGRRP
uPA K1 DKSKT CYEGNGHFYR GKASTDTMGR PCLPWNSATV LQQTYHAHRS
PG K4 TPVVQD CYHGDGQSYR GTSSTTTTGK KCQSWSSMTP HRHQKT-PEN
PG K1 VYLSE CKTGDGKNYR GTMSKTKNGI TCQKWSSTSP HRPRFS-PAT

PG K2 LECEEE CMHCSGENYD GKISKTMSGL ECQAWDSQSP HAHGYI-PSK
PG K3 SGPTYQ CLKGTGENYR GNVAVTVSGH TCQHWSAQTP HTHNRT-PEN
PG K5 TPSEED CMFGNGKGYR GKRATTVTGT PCQDWAAQEP HRHSIFTPET
HGF K1 NKDYIRN CIIGKGRSYK GTVSITKSGI kCQPWSSMIP HEHSFL-PSS

HGF K2 VE CMTCNGESYR GLMDHTESGK ICQRHDHQTP HRHKFL-PER
HGF K3 DVPLETTE CIQGQGEGYR GTVNTIWNGI PCQRHDSQYP HEHDMT-PEN
HGF K4 HGQD CYRGNGKNYM GNLSQTRSGL TCSMWDKNME DLHRHI-FWE

PT K1 AACLEGN CAEGLGTNYR GHVNITRSGI ECQLWRSRYP HKPEIN-STT
PT K2 SPPLEQ CVPDRGQQYQ GRLAVTTHGL PCLAHASAQA KALSKH-QDF
F XII CYDGRGLSYR GLARTTLSGA PCQPWASEA- --TYRNVTAE

SO 60 7O 80
I I l l

tPA K2 SAQALGLGKHNY CRNPDGDA-K PWCHVLKNRR LTWEYC-DVPSC ST
tPA K1 DAIRLGLGNHNY CRNPDRDA-K PWCYVFKAGK YSSEFC-STPAC SEG
uPA K1 DALQLGLGKHNY CRNPDNRR-R PHCYVQVGLK PLVQEC-MVHDC ADG
PG K4 YP-NAGLT-MNY CRNPDADK-G PWCFTTDPSV -RWEYC-NLKKC SGT
PG K1 HP-SEGLE-ENY CRNPDNDPQG PHCYTTDPEK -RYDYC-DILEC EEE
PG K2 FP-NKNLK-KNY CRNPDREL-R PWCFTTDPNK -RWELC-DIPRC TTP
PG K3 FP-CKNLD-ENY CRNPDGKR-A PWCHTTNSQV -RWEYC-KIPSC DSS
PG K5 NP-RAGLE-KNY CRNPDGDVGG PHCYTTNPRK -LYDYC-DVPQC AAP
HGF K1 YR-GKDLQ-ENY CRNPRGEEGG PWCFTSNPEV -RYEVC-DIPQC SEV
HGF K2 YP-DKGFD-DNY CRNPDGQP-R PWCYTLDPHT -RHEYC-AIKTC ADNTMNDT
HGF K3 FK-CKDLR-ENY CRNPDGSE-S PWCFTTDPNI -RVGYCSQIPNC DMS
HGF K4 PD-ASKLN-ENY CRNPDDDAHG PWCYTGNPLI -PHDYC-PISRC EGDTTPTIVNL
PT K1 HP-GADLQ-ENF CRNPDSSITG PWCYTTDPTV -RRQEC-SIPVC GQD
PT K2 NS-AVQLV-ENF CRNPDGDEEG VWCYVAGKPG -DFGYC-DLNYC BEA
F XII QARNWGLGGHAF CRNPDNDI-R PWCFVLNRDR LSWEYC-DLAQC

 

Figure 1.4. Comparison of the folding of PGK4 and PTFl. PTFI in bold.

y . -
fir» -‘ o-
n f

‘5‘? " u- JA
4... .-.0

l
l““ G .
° r~ - . ,

.4 _.
.C'.-. u .4.
I" \ ' -
.‘JL Al‘f‘ .
-o I".. .
\ ' I ‘v-rJ
“‘ ‘“"4. I

‘7“ .
O c
l P .‘
-.. I.
l v
.O'N .-
~-.’: L :I‘\ ‘
- 3-. ‘
I I
'F A"' v u .
s... "u: ‘ ‘4.
,, .
2‘237‘? -r ‘
. .u Gui. I
D?“
L .- V.
“ ~ v.

symmetrical and very similar. The combination of the second half of the first outer
loop (A) and the second inner loop (D) is almost the mirror symmetry counterpart of
the whole second outer loop (B). The folding of the kringle also shows a complicated
duplication by operation of two 90° rotations and a translation. Since apolipopro-
tein[a] (apo[a]) carries 38 copies of the kringle, 37 of which are highly homologous but
not identical to PGK4, it is possible to model kringle structures based on the PGK4
structure and further investigate kringle-kringle interactions of apo[a].

The dissertation is organized as following: the experimental and computational
methods used for crystallizing and solving the crystal structures will be introduced
in chapter 2. In chapters 3,4 and 5, structures of K4ACA, PGKl and tPAK2 will be
given and discussed. Comparison of the above three kringle structures and PTFl and
PTF2 will be presented in chapter 6. Chapter 7 describes the structure of a designed

F PAM complex of a-thrombin.

CHAPTER 2

Experimental and Computational

Methods

2.1 Crystallization

The most important factor in solving protein structures is to have satisfactory single
crystals. Crystals can be grown from a saturated solution in any of several ways, all
of which serve to raise the solute concentration above that which can be supported
by the solution: slow evaporation, slow cooling, vapor diffusion. The best crystals are
usually produced when the solution is free from mechanical vibration and allowed to
evaporate without disturbance.

Among the crystallization methods, vapor diffusion techniques (Figure 2.1) are
probably the most widely used to grow protein crystals. As shown in the Figure 2.1,
a drop of protein containing precipitant is equilibrated against a reservoir in which the
concentration of precipitant is higher than that of the protein solution. Equilibration
proceedes by diffusion of the volatile species (water or organic solvent) until vapor
pressure in the droplet equals that of the reservoir. Since the drops used in this
method can be as small as 2p], the vapor diffusion technique is very well suited to

screening a large number of conditions when only a small quantity of material is

10

silicone grease seal

 

 

 

 

 

 

 

 

 

 

 

reservoir of precipitating agent

Figure 2.1. Crystallization by vapor diffusion technique. (A) Hanging drop method
(B) Sitting drop method.

 

 

.“

[JA‘ A

.. ~v 'V

‘ ‘ ,f“

‘ 4. r ~ ‘
.

‘.
-h—a ‘
~ :-

5....

'r. ‘ ‘ ‘
.16 l s
. n
‘9 r O
y .
CA L— -
l
1
v. r - —:
. \ H :\ d
4.. .. _,, ha
. 0
.1D or -
d; . -‘
-oxi

 

.,~' .
J n... .7 ‘
L\N .

A A. 4

1““, '

.1» '
‘Al I
‘

 

a‘?‘“-‘-'
-nl\ '
0 "an.‘ HP '

11

available.

The vapor diffusion technique employs both hanging dr0p and sitting drop meth-
ods (Figure 2.1). For the hanging drop method, a microdroplet of mother liquor is
suspended from the underside of a microscope cover slip, which is then placed over a
small well containing 1 ml of the precipitating solution. An important point is that
the cover slips must be thoroughly and carefully coated with silicone to ensure proper
drop formation and prevent spreading. The wells are most conveniently supplied by
disposable plastic Linbro plates that have 24 wells. These plates provide the advan-
tage that they can be swiftly and easily examined under a microscope and stored
compactly.

The sitting drop method shown in Figure 2.1 is usually used to get larger crystals.
The seal between the lid and the rim of a clean box is coated with some silicone
grease. A clean siliconized glass vial is then placed at the bottom of the box with the
open end up. One ml of the well solution is added by pipette. A 20111 drop of the
well solution is placed in the bottom of the vial and the same amount of the protein
solution is then added with slight mixing. The lid is then placed on the box and the
box is sealed.

When crystallization is finally induced, it is usually so rapid that only microscopic
crystals result. Using microseeding methods and by changing crystallization parame-
ters, such as protein concentration, precipitant concentration, large and single crystals
can be produced. Once a good crystal is chosen, it is mounted in a sealed glass cap-
illary in contact with its mother liquor and properly aligned for further experimental

work.

0."_ {‘9
,—
u.
.
"I r“ _

 

F.

tr '.r

x-_

it" o

“l
.g _

.> ‘
\

a“.

‘Q

.

1' '
.3-

 

12
2.2 Data Collection

Two kinds of detectors are now used for 3—D intensity data collection: the diffractome-
ter and the area detector. The former measures reflection intensities individually; the

latter has the ability to measure many reflections at once.

2.2.1 Diffractometer

Three-dimensional intensity data were collected on a N icolet P3/ F four—circle diffrac-
tometer (Figure 2.2) using a graphite monochromator with CuKo, radiation (1.54 A)
from a rotating anode X—ray tube. The diffractometer possesses four circles that can
be used to adjust the orientation of the crystal and counter, which include the d), x,
w, and 20 angles. In order to measure any reflection, the reciprocal lattice point must
be brought onto the sphere of reflection in the equatorial plane of the diffractometer
and the detector must be positioned at the proper value of 20. In order to achieve
this, the w, 43, X angles must be determined and set correctly for each reflection.
Therefore, each diffracted X—ray beam is defined by 20, w, (25, and x angles as well as
by Miller indices h, k, l. The high sensitivity of the diffractometer to angle settings
allows it to achieve an accuracy of alignment and parameter measure that exceeds
the accuracy usually obtained photographically by an order of magnitude.

The method used to measure the reflection intensities was the Wyckoff w—step
procedure [24], most widely used in protein crystallography. The crystal is slowly
moved through the reflection position, and the profile of the intensity of the reflection
and the background are recorded by the stationary detector. In this procedure, every
reflection is scanned by a limited number at small intervals in w, as opposed to the full
profile scan from background to background. Background measurements are made on
either side of the peak by offsetting by a certain w value. The scan method therefore

allows a slight mis—setting for the crystal. The breadth of a single peak at half the

13

 

Figure 2.2. Diagram of four—circle diffractometer.

vi».
.\

 

14

maximal height defines the region of the peak which should be scanned. The w profile
not only can be a basis for measurement but also reports the quality of a crystal. A
split profile may indicate that the crystal is cracked and that two closely related but
distinct crystal orientations are contributing to the profile.

Since the macroscopic dimensions of a crystal are usually not equal, absorption
due to the crystal and the medium around it is unavoidable. As a result, empirical
corrections for absorption have become common, especially in protein crystallogra-
phy, where many reflections are involved. The most quantitative approximation, as
suggested by North at al. [25], involves setting x to 90° and measuring the indepen-
dent intensity of one or more reflections as a function of the ¢ range. This allows us
to determine the optimum region of d) for carrying out the data collection. Reflec-
tions on a principal axis which coincides with the (15 axis are selected to do the (15 scan
because the Bragg planes are then always in the reflecting position for all values of
(13. The observed variation can easily be converted to an approximate correction as a
function of g6 and applied to the entire data set.

Many crystals, especially those of protein and other macromolecules, show a more
or less steady decrease in diffraction intensity during the process of data collection.
This deterioration, which is due to direct X-ray exposure, affects the consistency of
the measurements. In order to make the decay correction, a number of measurements
of crystal deterioration during data collection are monitored, representing intensities
in various resolution ranges. These include a 2—dimensional data set from 2°—15°
in 29, collected before and after the data collection, and several monitor reflections
measured throughout the data collection after every 100 reflections about every hour.
If the intensity of any monitor falls below 80% of the initial measured intensity,
the crystal is considered misaligned and this triggers the recentering of the crystal,
followed by the calculation of a new orientation matrix and cell parameters by the

least squares method.

 

A
u

but!

 

» e.
.P.
.A«

 

 
   

 

\{a'

.f.<

 

15

2.2.2 Area Detector

The rapidly increasing interest in the X—ray crystallography of macromolecules has led
steadily to the study of larger and larger unit cells. These not only have a greater total
number of reflections to be measured, but also cause more to be reflecting at a given
time. As a result, area detectors have been devised which combine the sensitivity,
accuracy, and convenience of the diffractometer but measure many reflections at once.
The Siemens Xentronics Area Detector with graphite monochromated CuKa radi-
ation and a Rigaku RU200 rotating anode X—ray generator is used for data collection
in this laboratory. A schematic of the area detector system is shown in Figure 2.3.
The area detector mounts on a dovetail track on the diffractometer, permitting ad-
justment of the detector to specimen distance. Area detector operation and control
are performed by the frame buffer, in which the FRAMBO (FRAMe Buffer Oper-
ation) program is used to collect data and display frames. The output signals of
the detector are decoded by the Position Decoding Circuit (PDC) to create a digital
position of each X—ray event measured by the detector. The resulting position is sent
to the frame buffer for processing. The primary purpose of the frame buffer is to
receive data from the PDC and convert it into frame information. The frame buffer
processor also drives a high resolution color display which shows the output from the
area detector in two forms: first, as a real-time color display showing the diffraction
pattern building as it is collected; and second, as a data frame showing the results of
several seconds (or minutes) of output integrated in the file as a single frame.
Before 3—D X-ray diffraction intensities are measured, some calibrations of the
instrument must be performed, first, the bias adjustment. Since the x—ray signal
height is varied by adjusting the detector bias, the bias should be set at a proper
value. From Figure 2.4 it can been seen that while lowering the bias slowly there will

be a point at which the rate meter reading falls. This is the lower window setting.

16

 

 

 

 

 

 

 

 

 

 

I Goniometer 4—Circle
Controller Goniometer
_l

 

PDC I Area
I Detector

 

 

 

 

 

 

Color
Display

 

 

 

 

 

 

 

 

 

 

Frame Buffer #— Ethernet

I

Keyboard I

 

 

 

Mouse

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Pn'nte — .
t Frame
. Processor
I Mouse“ r—J
Color
Keyboard _, Disvlay

 

 

 

 

 

Figure 2.3. Block diagram of four—circle area detector system.

 

COUNT RATE

ON RATEMETER

17

SET BIAS HERE

SETTING I SETTING

\

 

DETECTOR BIAS —>

Figure 2.4. Graph of count rate vs. bias control setting:

18

Increasing the bias setting should cause the rate meter to peak and then fall again
at the upper window setting. The bias should be set midway between the upper and
lower window. The bias setting can be performed with a 55Fe calibration source. The
second calibration collects a flood—field image taken from an 55Fe source to generate
a lookup table which corrects the detector response to an uniform field. A third
correction collects an image from the iron source with a precisely machined brass
plate mounted in front of the active surface of the detector. The data obtained from
this calibration frame are used to generate a calibration table for conversion from

detector address in pixels to true positions in centimeters.

2.3 Data Reduction

2.3.1 Difl'ractometer Data

The purpose of data reduction is to convert each reflection intensity 1(hkl) to its
corresponding structure factor amplitude |F(hkl)|. The intensity data are reduced
using a program called P—DATA, written by Dr. C. D. Buck in this laboratory. The
program calculates the structure factor amplitudes from diffraction intensities based

on the equation:
|F(hlcl)|2 = CONST x ABS x DEC x LORPOL x 1(hkl) (2.1)

where C ON 3 T is a constant used to scale the data, ABS is the absorption correction
factor, DEC is a correction for intensity decay as a function of X—ray exposure time,
LORPOL is the Lorentz—polarization correction factor which is a function of 20 and
monochromator, and 1(hkl) is the background—corrected intensity of a reflection.
Before applying the P-DATA program, the background of the intensity data is

averaged in shells of 20, and d, if there is a 45 dependence. Background—corrected

.nO'JV“
”41""

. , I‘
in"

(1‘

.,N

‘9

L.”

. “:u r

v. ..

Nb:- Ir!
oIL‘

..

.5-

aflp-
I‘G‘

cc...-

‘6 r.

19

intensities are calculated from the equation:

sum of background counts

 

[(intensity) = total scan count — x scan rate (2.2)

background to scan ratio

Intensities are considered to be observed according to the minimum intensity (1min)
which is taken as 2 times the magnitude of the average of negative intensity ((Ineg)).

The absorption correction applied in the P—DATA program is based on the tables
of [max/ I ((15) versus 45. The data in these tables are derived from the observed (1')—
dependence of reflections at X = 90°. The absorption correction may vary with 20
and up to 6 tables are allowed.

The decay correction factor, DEC, corrects the intensity deterioration of a reflec-
tion as a function of exposure time. Using a 2—D set of data measured before and
after data collection and 3 monitor reflections, , we can get a decay factor for the

crystal based on following equation:

1
DEC = —-——— 2.3
l — S - t ( )
where S = —s/ 1°, 1° is the intensity at zero time, and s is the slope of the intensity

versus exposure time. Since the check reflections and 2-D data set have different 20
values, a 20-dependent DEC is derived from the S versus 26 plot and all reflections

collected at different times and 20 values are corrected by the decay factor.

2.3.2 Area Detector Data

X—ray diffraction intensities collected from area detectors are reduced and scaled with
the XEN GEN packages of programs [26]. Programs are supplied to refine, integrate,
merge and scale data, producing a comprehensive listing of integrated intensities and

a statistical analysis of the data. There are many steps involved in data reduction.

20

(A) Defining the active pixels on the detector face

A pixel is inactive if it is outside the overall active area of the detector or is obscured
by a beamstop or some other objects in the path between the crystal and detector.
The program BORDER reads a set of frames which are collected under the same
geometric condition. In each frame the program determines the mean count value

over the entire detector face according the equation:

20013323)
(C(71)) = w

 

(512 x 512) (2'4)

where n is the frame number. For each pixel it adds a number to the appropriate
element in a 512 x 512 buffer. The number it adds will be the number of counts in
that pixel in that frame unless the count value is greater than twice the mean for the
entire frame. In this case, twice the mean is added in. Then the value at pixel (x,y)

in the buffer after frame n will be:
Z(a:..y,n) = Z(:r,y,n — 1) + min (C(x,y,n),2 X (C(n))) (2.5)

Any pixel whose Z value is above a defined fraction of the mean is assumed to be
part of the active area of the-detector. This effectively defines beamstop shadows and

other peculiarly—shaped regions.

(B) Creating the calibration file

In this step the detector pixels are converted to actual distances in centimeters by the
program CA LIBRA TE. It reads a fiducial—plate data frame made on the frame buffer
by collecting several million counts with an 55Fe source and then finds the centroids
of the spots on the fiducial image and indexes them relative to one another. From

this information it then constructs a mapping of detector pixels into centimeters from

it

an“.

N 4.

r ..

.4.

r-‘v

..‘

1v.

21

a reference position.

(C) Generating a list of bright spots

The program SPOTS reads a series of contiguous data frames and constructs a list
of bright spots for the crystal and detector refinement. A spot is regarded as bright
if it includes at least one pixel which is N sigma brighter than the local background.
Usually N can be adjusted to generate approximately 20 bright spots per data frame,
which is a reasonable number. Therefore, the stronger the diffraction the larger is the

number N used.

(D) Calculating the orientation matrix and the detector parameters

Once a set of bright spots centroids have been assembled, an initial set of unit cell
and crystal orientation parameters can be obtained with the REFINE program. It
then refines these parameters and the position of the detector and the rocking—curve
behavior of the crystal. All reflections with three indices closer to integers than the
error limit are used in a linear refinement to minimize the differences between the

integerized and unintegerized indices.

(E) Computing integrated data

The program IN TEGRA TE reads a series of contiguous data frames and computes
the integrated intensities of the Bragg reflections both by profile-fitting and by simple
summation. Both include these corrections : (1) The Lorentz effect; (2) Polarization;
(3) Variations in exposure time from frame to frame; (4) Dead—time loss; (5) The con-
tribution to the profile of pixels not included in the summation or the profil&fitting.
The program not only integrates intensities but also refines the crystal parameters

based on the observed data.

,4... ,

UGG

7‘1

‘1

1"

n —_ v
of" n

.‘ .-

A

‘4

22

(F) Reformatting the integrated reflections

In order to make integrated reflections compatible with available structure solution
software, the program REDUCE reads the integrated reflections and then writes a file
containing most of the pertinent information and none of the unnecessary information
about each observation. Also, the output file created by REDUCE is sorted in hkl

order and contains symmetry—related observations.

(G) Merging the integrated reflections

The program MRMERCE reads the reflections from REDUCE, then merges together

data from various orientations of one crystal.

(H) Scaling merged data

A linear least—squares scaling of the merged intensity data is performed by the pro-
gram SCALEI based on minimizing the differences among intensities of symmetry—
related observations. SCALEI breaks the data into 5° ranges and splits each 5° range
into two pseudo—films. One, two, or three scaling parameters per film are calculated

by the program, and they are used to calculated scaled intensities for the observations.

(I) Creating merged intensities or structure amplitudes

The final step of data reduction is to write an output file of either intensities or
structure factors which is performed by the program MAKEM U. Once this output is

obtained, structure determination can be carried out in further work.

2.4 Molecular Replacement

To solve the structure of any macromolecule, an electron density map must be gener-

ated for model—building and structure interpretation. However, calculating such elec-

 

u. .‘.

"-vrz A
A1. .
i...
r.—
.u

 

 

- I‘
[L

.

9
..~I
-.,J 1

.“I

.

 

‘v

V
D6
‘4 1L
'l—v

23

tron density maps is not straightforward, since they require the phase angles which
can not be measured during data collection. .In this investigation, molecular replace-
ment [27] provides the best way for determining initial phases of macromolecular
crystal structures. In this technique, a molecule of known structure (search model)
is used to probe the diffraction data of an unknown crystal structure; angles and
translation vectors are found which will locate the probe structure in the unit cell of
the unknown structure to generate an initial phasing model for crystallographic struc-
ture fitting and refinement. Molecular replacement is thus concerned with finding the
three rotational and three translational parameters that specify the orientation and
position of the molecule in the crystal cell with respect to the symmetry elements.
If the model is sufficiently similar to the unknown structure, molecular replacement
can be rather straightforward to apply.

Molecular replacement is performed by calculating Patterson functions of the

known model and the unknown structure based on the equation:
P(uvw) = (1/V)Z|F|2cos(27r(hu + [co + 1w)) (2.6)

where V is the volume of the unit cell, and [F]2 is proportional to the intensity of
the diffracted beam and u, v and w are the coordinates of a point within the unit.
cell. Since no phase information is required for this function, it can be calculated
from both unknown structure and search model. The molecular replacement method
consists of three stages: (1) Rotation: determination of relative orientation of the
independent molecules; (2) Translation: using information from (1) to position the
molecule in the unit cell in the correct orientation; (3) After completing (1) and (2),

the equivalence between the point X’ in the unknown structure and the point X in

24

the model can be expressed by
X' = [C]X + d (2.7)

where [C] is the rotation matrix determined in stage (1) and d is the translation vector

determined in stage (2).

2.4.1 Rotation Search

The rotation search is carried out using the SEARCH routine in the program “PRO-
TEIN” package [28]. The rotation function (Equation 2.8) measures the degree
of coincidence between Patterson functions from unknown structure P1 and search

model P2 which is rotated by a matrix C with respect to unit cell. ’
R = [191(1) x c x any) do (2.8)

Maximum values of R occur at a certain matrix C which gives close agreement be-
tween vector sets of the unknown structure and model. The matrix C (Table 2.1)
is usually expressed in terms of three Eulerian angles 01, 02 and 03 as illustrated in

Figure 2.5.

Table 2.1. Rotation matrix in terms of Eulerian angles 01, 02, 03.
I -— sin 01 cos 0; sin 03 cos 01 cos 92 sin 03 sin 02 sin 03 I
+ cos 01 cos 03 + sin 01 cos 03

— sin 01 cos 02 cos 03 cos 01 cos 0; cos 0;; sin 62 cos 0;;
— cos 01 sin 03 — sin 91 sin 03

 

 

I sin 01 sin 02 — cos 01 sin 02 cos 62 I

92

 

 

Figure 2.5. The Eulerian angles 01, 02, 03 that relate the rotated axes X’, Y’, Z’ to
the original axes X, Y, Z. The rotation operation consists of 1.) a rotation by 01
around Z-axis. 2.) a rotation by 02 around new X—axis. 3.) a rotation by 03 around
new Z—axis.

 

 

.
'v 0
ll
1
I
1...t..
, ..
I .
_ .a
on, .-
l
A4nA
‘i.
P e

Ln

'u

 

26

In macromolecular structures, the number of Patterson vectors required for rota-
tion function calculation depends on the number of atoms in the model and in the
unknown structure. The Patterson vectors of the search model are rotated and inter-
polated to the nearest grid in the Patterson of the unknown molecule and a product
function calculated. This process is repeated for each set of angles. The angles of
the rotation matrix corresponding to the highest value of the product function corre-
spond to the transformation needed to bring the model and th unknown to the same

orientation.

2.4.2 Translation Search

Once the relative orientation of the search model and the unknown structure is known,
the translational parameters can be determined. A translation function that has been
used widely is one that employs a linear correlation coefficient to determine the correct
position of an oriented molecule in the crystal cell. The correlation coefficient (C’) is

represented as:

_ 2 (mar — I < a > I’) x (Ian? — l < F. > 12))
721W - I < P. >12>2>< 2(1F.r-I< F. > n"

 

C’ (2.9)

 

Like the R factor, it is basically a measure of the agreement between observed , [Fo|,
and calculated quantities, [Fe].

The calculation of correlation coefficients has been implemented in the program
“BRUTE” [29]. It moves the search model over a grid of points in the crystal
cell. At each point the symmetry-related positions are generated and the structure
factors are calculated. The amplitudes of these calculated structure factors are then
compared with the observed values using the correlation coefficient, C’, as well as
the conventional R factor. Therefore the overall calculation time is a function of

the number of symmetry operations, the number of reflections and the number of

1....

 

 

“
‘1
.“‘
~.,

,

 

 

27

grid points. The most important feature of the program is its ability to adjust the
orientation of the model. This allows the errors in the orientation obtained from the '

rotation function to be corrected.

2.5 Structure Refinement

The crystal structure can be refined employing the restrained least-squares refinement
method either implemented in the program “PROLSQ” [30] or “PROFF T” [31] after
applying the rotational matrix and translational vector. This method is a very general
and powerful technique for optimizing the fit of a model to a set of observations. For
PROLSQ, however, application to even moderate—resolution protein structures can
result in high computational costs, as structure factors and deviations between obser-
vation and calculation are computed directly from contributions by each reflection.
Therefore, “PROLSQ” has been modified with the fast—Fourier-transform algorithms
added throughout the computation and implemented in the program “PROF F T”.
This approach reduces by up to a hundredfold the computation time for a single
cycle. I

The principle of least squares states that the “best”set of model parameters is that
which minimizes the sum of the weighted squares of the difference between observation

and calculation. It can be stated by the following mathematical terms:

re f lections

as = 2: want §“’°({x})]2 (2.10)
h

where {x} is the set of parameters, F}, and fh are observed and calculated values
and Wh, the applied weights, are the inverses of the variances of the observations.
The parameters included in the function are structure factors, bond distances, an-

gle distances, torsion angles, 1-4 planar distances, planarities, chiralities, nonbonded

 

 

~..L

w .
l P

.n ..

'7‘; \‘.

.vy— ..

 

 

 

28

contacts, and thermal parameters. Sometimes appropriate restraints are placed on
the occupancy factor values of atoms in activity effectors, water, and other solvent
molecules.

The refinement process used in “PROLSQ” or “PROF FT” are summarizedby
the schematic shown in Figure 2.6. The program reads diffraction data (h, k, 1,
F) and scattering factors (f(s)) prepared by SCATT, initial atomic coordinates(x, y,
z, B) and the restraint specifications (ideal values) prepared by PROTIN, parameter
shifts from previous refinement cycles, and control card-images. It then augments the
normal-equation elements pertinent to each of the stereochemical restraints and the
structure factor observations. Program MERGE then combines the shifts produced by
either “PROLSQ” or “PROFFT” with the atomic coordinates to produce an updated
coordinate file.

The most commonly used measure of the degree of match between observed and

calculated structure amplitudes is the residual R:

 

Z IIFzzi — lFifiz’c
R z hkl (2.11)
23 W27

hkl

 

If the value of the residual after the second Fourier synthesis is smaller than its former
value, it indicates that the second electron density map is likely to be a more reliable
representation of the actual molecular structure than the original trial model.

The intermittent model building performed on an Evans and Sutherland P3390 in-
teractive stereographics by using program F RODO [32] is carried out after each stage
of refinement. During the model building, the structure is manually adjusted to bet-
ter positions according to (2|Fol — IFCI) and (IFOI — [F'c|) maps and the Ramachandran
plot.

At the beginning of the refinement, an average isotropic thermal parameter B is

 

 

29

 

 

 

 

 

 

 

 

 

 

 

 

 

SACTT —> Reflexns ‘ ’ P P
R R
O O
Atoms
PROTIN —> and —> 0"
Geom. F L
X,Y.Z,B IDEAL T Q
MERGE
X,Y,Z,B

Figure 2.6. Schematic Structure of “PROFFT” and “PROLSQ”.

30

applied to the model and structure factors are weighted by using constant 0;: which is
approximately half of the discrepancy between F0 and Fe. During the refinement, the
variables can be changed alternatively to loose or tight restraints which depend on
the agreement of the geometric variables and structure factors. 'Water molecules are
added gradually. Peaks are selected as water for structure factor calculations only
if they are within 2.5—4.0 A of the protein or another water molecule. New water
molecules included into the refinement are initially assigned an occupancy of 0.75
and an overall B obtained from the last refinement stage. In the last stages of the
refinement, the remainder of the data are assigned in seven shells of sinB/A based on

(”Flo ‘ lFch>/2 of the range:
. 0
of = Ufa + afb * (Sln X —1/6) (2.12)

In this way, the weights of low and high resolution data can be controlled by adjusting
the coefficient, Ufa, and slope, Ufb. Lastly when the cyclic refinement is no longer
giving new information and the R value converges to a reasonable value, refinement

is terminated and the final model obtained.

h

\.
.

‘G
i
.

,
u

‘s

.

I.

_

CHAPTER 3

Structure of the ACA Complex of

Human Plasminogen Kringle 4

3. 1 Introduction

Plasminogen (Figure 3.1) is a glycoprotein produced by the liver with a molecu-
lar weight of about 92,000 daltons and carbohydrate content about 2%. A single
plasminogen chain has 790 residues with Glu and Asp-residues on the amino- and
carboxyl— terminals. Sometimes a plasminogen variant that results from autolytic
cleavage at the Lys76—Lys77 site. also exists with lysine as the amino- terminal residue
[33] This “Lys—plasminogen” is recognized as a degradation form of the native Glu-
form and is more readily activated by plasminogen activators. On the molecular level,
activation involves cleavage of an Arg560—Va1561 bond of the proenzyme to yield the
proteolytically active plasmin which has a disulfide—linked two—chain structure. The
B or light chain from Va1561 to Asn790 consists mainly of the protease domain which
is a trypsin-like serine protease. The A or heavy chain from Glul to Arg560 is formed
by a tandem array of five kringles. These kringle domains are involved mainly in the
interactions of PC with substrate and regulatory molecules. There are 48 cysteine

residues in the plasminogen molecule, forming 24 disulfide bridges, two of which link

31

Kl

1‘?

Serine Protease Domain

Figure 3.1. The chain structure of plasminogen. Kringle domains are labeled K1 —
K5. Disulfide bridges are shown by bold lines. Asterisks denote active site residues
in the serine protease domain. Arrows indicate the cleavage sites.

 

33

the A- and B- chains (Figure 3.1).

The first and fourth kringles can be obtained from the heavy chain by controlled
proteolysis and affinity chromatography on a lysine-conjugated support gel [7, 9]. The
experiments of the binding of 6—aminohexanoic acid to K1 and K4 have been carried
out [9] and showed that K1 has the higher affinity for lysine binding (Ka=60mM‘1)
than does K4 (Ka=28mM’1). According to this observation, the first kringle is prob-
ably the main contributor to fibrin binding. A weak lysine binding site is also carried
by K5 [34] which prefers ligands not having a free carboxylate function and therefore
may interaCt with lysine side chains of proteins such as benzylamine and hexylamine.
It has been proposed that K5 may play a key role in binding internal lysyl side chains
of intact fibrin to the plasmin catalytic site, which cleaves Lys-X bonds of fibrin
generating a number of C—terminal lysine residues.

Chemical modifications of K4 have been carried out in order to identify residues
essential for ligand binding. Modification of Asp57 and Arg71 (Figure 3.2) with 1,2-
cyclohexanedione and 1—ethyl-3—(3—dimethylaminopropyl)—carbodiimide [17] showed
the loss of lysine—Sepharose affinity. The importance of these two residues for ligand
binding is related to the electrostatic interaction of the charged groups with ligands.

In addition to the- polar interactions between the kringles and the lysine analogs,
hydrophobic interactions between the ligand methylene groups and aromatic amino
acids of kringles have also been studied. Reaction of Trp72 with dimethyl(2—hydroxy—
5-nitrobenzyl)sulfonium bromide [35] destroyed lysine—Sepharose affinity. From N MR
NOE experiments [36, 37], it has been found that the side chains of Trp62, Phe64 and
Trp72 are perturbed most by ligand presence and that Trp25, His31, His33, Tyr41
and Tyr74 are also affected, but to a lesser extent. From the above observations,
residues Asp57, Trp62, Phe64, Arg71, and Trp72 might be located on the lysine
binding surface, where they interact with ligand directly and strongly [17, 35].

The lysine binding subsite structure was first modeled by Tulinsky et al. [38]

— I

   

34

 

Figure 3.2. The primary structure of plasminogen K4. Interkringle residues are a—c
and 81—87. (see appendix for abbreviations)

   

'—

35

based on the structure of PTFl, which additionally implied that Asp55 might be
important to the lysine binding (Figure 3.3). This was then confirmed with the
crystallographic structure of PGK4 [23, 39]. However, a cofacial intermolecular
interaction between the binding sites of neighboring molecules excluded access of the
site to the small ligands like ACA by diffusion into crystals, thus precluding the
formation of the K4—ACA complex. This led us to crystallize PGK4 in the presence
of ACA ligand [40]. The crystallographic structure of PGK4-ACA complex not only

establishes the nature of ligand binding but provides a good model for fibrin binding.

3.2 Materials and Methods

A. Crystallization

The human plasminogen kringle 4 was generated from plasminogen by elastase diges-
tion and affinity chromatography on lysine—Sepharose [7]. The product is a heteroge-
neous mixture of approximately 70% C—terminal Ala85 and about 30% C—terminal
Ala85-Ser86—Val87. The PGK4 used for crystallization was kindly provided by Dr.
M. Llinas in the form of a lyophilized powder, which was stored at freezer tempera-
tures in a dry environment.

Soaking experiments failed to produce K4-ACA crystals [41]. This led us to
crystallize the PGK4 complex in the presence of ACA ligand. The K4 was first
dissolved in distilled water at a protein concentration of approximately 10 mg/ml,
and ACA at 25 mg/ ml in water was added. This solution was then used in growing
crystals. Crystals of K4-ACA were grown by the sitting drop method using 30%
PEG 8000, 0.12 M ammonium sulfate, with 1.4% dimethyl formamide (DMF) as
an additive at pH 6.0 in the presence of 25mM ACA. Although the conditions were
practically the same as those which produce orthorhombic apo—K4 crystals [42], K4—

ACA crystallized in a different crystal form. The unit cell parameters and other

36

 

Figure 3.3. Stereoview of modeled lysine binding site of plasminogen kringle 4.ACA
in bold. Hydrogen atoms are put at ideal positions [38].

37

crystal data are shown in Table 3.1. Crystals appeared within a week after seeding
with monoclinic apo—K4 crystals. Most of the crystals had a plate—shaped form and
tended to grow together. Single and separate crystals could be obtained occasionally

as shown in Figure 3.4.

Table 3.1. Summary of crystal data of monolinic K4—ACA

 

 

 

 

 

 

space group P21
crystal system monoclinic
molecules / asymmetric unit 1
molecular weight (dalton) 10000
a (A) 42.21
b (A) 35.46
c (A) 25.43
fl . 102.95°
vm (35%;) 1.89
solvent fraction (%) 34.5

 

B. Data Collection

The crystal used for intensity data collection had dimensions of 0.8 x 0.4 x 0.2 mm.
The crystal was mounted in a siliconized capillary of diameter 1.0 mm with the b
axis parallel to the length of the capillary (Figure 3.5). Three—dimensional intensity
data were measured to 2.25 A resolution using a Wyckoff w-step procedure [24] with
a N icolet P3/ F diffractometer at 2.5 kW (50kV, 50mA). Axial intensity distributions
of the crystal are displayed in Figure 3.6. From the w—profile of a reflection (5,
5, 5) (Figure 3.7), a scan range of 0.2° and an offset value of 035° from the peak
center for background measurement were chosen. A scan speed of 0.4° / min was used
for measurements, with a background to scan ratio of 0.1. Each measurement was

scanned for seven steps and the highest five counts were summed as the integrated

38

 

 

Figure 3.4. Photograph of K4—ACA crystal. The size of the single crystal is approx-
imately 1.2 x 0.3 x 0.1 mm.

39

 

 

 

 

 

 

 

 

Figure 3.5. Schematic diagram of K4—ACA crystal morphology with respect to crys-
tallographic axes.

43%: <O<|vx mo acogfimsmmv 38:35 _m_x< .cd Saw—m

 

 

 

 

40

 

 

 

(1)601

 

 

 

 

(”601

 

 

 

 

 

in
3

0— a.

.p

Op

 

 

 

 

(1)501

 

 

 

41

 

 

 

600
A 500
«2%
a 400
3
8
E" 300 _,
E’
d)
E
200 —~
100 —

 

 

12.63 12.83 13.03 13.23 13.43 13.63

Omega

Figure 3.7. The w—profile of reflection (5, 5, 5) taken before data collection.

   

1 P"

 

42

intensity.

Before 3—D intensity data were collected, an absorption correction was determined
for every 10° in d) by measuring intensities for reflections (0, 2, 0) and (0, 0, 10). Since
both of them showed the same trend, only reflection (0, 2, 0) was used in absorption
correction. A plot of lam/Ho) versus ((5 is shown in Figure 3.8; the 180° range between
gt 2 90° to 270° was chosen for data collection. The intensity data were collected in
three 20 shells of (32°—40°), (26°—32.2°), (2°—26.2°), in the order of highest to lowest
resolution range.

In order to make an intensity decay correction, three check reflections were mea-
sured after every 100 reflections, a 2—D hOI data set from 2° to 15° was collected
before and after data collection, and 12 reflections having a 20 range between 18° to

25° were measured periodically during data collection.

C. Data Reduction

The background readings were first averaged in 20 shells as shown in Table 3.2.
Background was also averaged with respect to 45 angle and showed no angular variation
(data are not shown). From Table 3.2 it can be seen that the average corrected
negative intensity (<Ineg>) is approximately -3. As mentioned in chapter 2, 1min was
chosen to be 6. Of a total of 3381 reflections, 3198 which were greater than twice 1min
were considered observed.

In order to get the decay correction factor, DEC, 5 values in each 26 range were
calculated, including the 3 check reflections, the 2—D data set, the 12 reflections (in 3
ranges), and the data are shown in Table 3.3. According to this table, a decay slope,

s, of 1.0 ‘4 hr‘1 was used for the decay correction.

A /_ i

[max/14.

45

41)

35

110

25

L5

L0

05

43

 

 

Illllllllllll

50 90 130 170 210 250 290

Figure 3.8. Absorption correction curve of (0, 2, 0) reflection.

Table 3.2. Averaging background in 219 shells.

44

 

 

 

 

 

 

 

219 (BC) # reflections ( Ines)
5.0—18.0 1.4 391 -3.3
18.0-22.0 1.6 307 -2.0
22.0—25.0 1.8 302 -1.5
25.0—28.0 1.8 399 -1.5
28.0—30.5 1.6 381 -2.0
30.5—32.5 1.5 396 -l.7
32.5-34.5 1.4 385 —l.l
34.5-36.0 1.3 322 -l.5
36.0-37.5 1.2 345 -1.7
37.5-38.5 1.1 229 -1.5
38.5—40.0 1.1 392 -l.3

 

Table 3.3. Decay rate, S, calculated in each 20 range.

20 | 9.9 | 17.9 | 23.0 | 23.1 | 23.7 | 24.7 | 24.9

 

Sx1o-3|1.5| 4.5| 2.9] 2.6] 2.7| 1.3| 2.2

 

 

45

D. Molecular Replacement

The K4—ACA structure was solved by molecular replacement rotation—translation
methods by utilizing the refined PGK4 structure as a search model [23]. The rotation
search was performed in Patterson space with the SEARCH routine in the program
PROTEIN [28]: A triclinic cell of dimensions 80 X 80 X 80 A using data from 8.0
to 3.0 A resolution and an overall thermal parameter of 20 A2 were used in the
model structure factor calculation. A large triclinic cell was chosen to separate the
intramolecular vectors which depend solely on the rotational orientation. The low
resolution range cutoff is to avoid contributions from solvent molecules; the high
angle cutoff is to avoid detailed structural information. Both model and unknown
K4—ACA Patterson vectors of length greater than 3.0 A and less than 15.0 A were
selected, since rotation function calculationuses only intramolecular vectors. For
kringles having dimensions of approximately 15 x 30 x 30 A a resolution range from
5.0—20.0 A is usually used in rotation calculation. The lower limits value is selected
to avoid the Patterson function origin. The higher limit was chosen based on the size
of the kringle to exclude possible intermolecular vectors. A set of the 1500 highest
Patterson peaks was used in the rotation calculation.

In the rotation search calculation, an initial search model was rotated over the
range of 61, 0—360°, 02, 0—180°, and 03, 0-180° in 5° increments. The highest peak
was found at the Euler angle (800°, 350°, 125.0°) at 7.340 above the mean. The
refinement of this solution was carried out in 2°, 1° .and 03° increments, and the final
solution had peak height 7.6 a above the mean at (845°, 334°, 121.3°).

The position of the model in the unit cell after rotation was found using the
program BRUTE [29] by a translation search first in 0.5, and then 0.1 A increments
with data from 8.0—3.0 A resolution. Since the origin along the b—axis is arbitrary

in the monoclinic system, the translation search was applied in the XZ plane. To

46

get a. convincing solution, data in the 8.0-2.8 A resolution range were also used for
the translation search, which showed the same result as that in 8.0—3.0 A resolution
range. The final refined peak had a correlation coefficient of 0.43 (about 160 above
the mean).

To verify the rotation matrix and translation vectors, the packing of the molecule
was inspected on an Evans and Sutherland P5390 interactive computer graphics sys-
tem equipped with FRODO [32] software. An electron density map based on this
rotation—translation position clearly revealed the K4 structure and, in addition, new

electron density extending between the side chains of Asp57 and Arg71.

E. Structure Refinement

The starting PGK4 model of K4—ACA was refitted using interactive computer graph-
ics. The structure was then refined employing the restrained least—squares method
implemented in the program PROFF T [31] with intermittent model building per-
formed on an Evans and Sutherland P5390 interactive stereographics system. The
refinement proceeded in stages, each of which was followed by model building using
(2|Fol — [Fc|) and (IFOI — [Fc|) maps and examination of the Ramachandran plot. The
initial R value started at 0.43 with an overall thermal parameter of 16 A2; R decreased
to 0.26 after the first stage (2.8 A resolution). As phases improved, higher resolution
data were added to 2.5 A resolution. The unaccounted for density between Asp57 and
Arg71 now appeared convincingly to be due to ACA. Therefore, an ACA was fitted
into the density and included in the refinement. Water molecules were also added
at 2.5 A resolution. In the last stages of refinement, the remainder of the data (to
2.25 A) were included, and reflection weights were assigned in seven shells of sin G/A
based on (”Flo — |F|c|)/ 2 of the range. The final reflection weights and R values in
each range are given in Table 3.4, and a summary of refinement parameters is listed

in Table 3.5. The final K4—ACA structure has a crystallographic R value of 0.148 for

47

Table 3.4. Weights of Reflections and R Values of the Final Refinement

 

 

 

 

 

R value

Dmin(A) no. reflections 0(]F|)“ < ||Fo| — |FC|| > shell sphere

4.00 465 11.2 25.6 0.153 0.153

3.30 465 9.9 19.2 0.124 0.139

2.90 493 9.1 16.0 0.150 0.142

2.60 555 8.4 14.1 0.157 0.145
‘ 2.45 375 7.8 13.5 0.171 0.148

2.35 288 7.5 11.5 0.151 0.148

2.25 352 7.2 11.00 0.148 0.148
a. a(|F|) = 18.0 — 70.0 [(sin B/A) —(1/6)].

 

 

 

48

Table 3.5. Summary of Final Least—Squares Parameters and Deviations

 

]] ] target (a) ] rms (A) ]]

 

 

 

 

 

 

 

 

 

 

 

 

 

Distances (A)

Bond Distance 0.015 0.017
Angle Distance 0.025 0.042
Planar 1,—4 distance 0.035 0.050
Non-bonded distances (A)

Single torsion 0.50 0.20
Multiple torsion 0.50 0.32
Possible H-bond 0.50 0.32
Torsion angles (deg)

Planar 3 4
Staggered 15 23
Orthonormal 20 16
Plane groups (A) 0.02 0.02
Chiral centers (A3) 0.13 0.23
Thermal restraints ( 2)

Main chain bond 1.5 1.8
Main chain angle 2.0 2.6
Side chain bond 2.5 2.3
Side chain angle 2.5 3.6

 

 

 

 

 

49

2993 reflections between 702.25 A resolution range with 106 water molecules and an
average thermal parameter of 17.9 A2. The average occupancy of the water molecules
is about 0.75, and their average B value (19 A2) is slightly higher than that of the

protein (17.5 A2).

3.3 Results and Discussion

A. General Structure

The refined K4—ACA structure extends from residue Gln—b to Cys80 along with the
ACA ligand; there was no electron density for the Nj-terminal Val—a nor for the C—
terminal interkringle pentapeptide. A similarly disordered interkringle region was
found in the apo—K4 structure [23] and might be due in part to the interkringle
link heterogeneity discussed in section 2. There was also only little electron density
for the Thr18 and Arg32 side chains; therefore, they were refined as alanine. The
distribution of main—chain torsion angles (45,112) is shown in Figure 3.9, from which
it can be seen that all the residues conform well with allowed regions except for one
outlier (Met48); however, the latter fits in very well-defined electron density and has
the same conformation as that observed in apo-K4 [23]. Moreover, a cis—proline
was found at position 30, similar to the apo—K4 structure. Conversely, in K4—ACA
the sulfur atom of Cys75 is localized in one position while it was distributed between
two different equally occupied positions in the apo—structure giving rise to different
disulfide comformations [23]. The localization in K4-ACA could be a result of ACA
ligand binding or possibly due to different packing in the crystal structure. The
average thermal parameter of the kringle in K4—ACA is 17.5 A2, but although the
electron density of the ACA is quite good (Figure 3.10), it has an average B value
which is about twice that of the kringle (34 A2). Since the ACA is fixed in the lysine—

binding site by two doubly charged, hydrogen-bonding, ion pair interactions, the

50

 

 

 

 

 

 

 

 

 

 

 

—180 —90 0 90 180
180 l n l l '
9O ]/ //] __
I
l l
l I —
l I
\ \1
3% 0
n.
-90 _. E1 _
.4 _
at Met 48
_ _ fir _____ 1
“3° 1 l I. I I I I l I I
-180 -90 0 90 180

Figure 3.9. Ramachandran plot of final (3, 1b angles of K4—ACA. The Gly residues
are boxed.

180

—180

51

 

Figure 3.10. Stereoview of the ele '
. ctron densrty of the lysine—b' d' '
The basket contour 18 at 10; ACA is between Asp55, Asp57 aifildlfifrsslifie (.Afriéfil—ACA

52

higher B value of ACA must reflect a microheterogeneity of its positioning within the
binding site similar to that observed for hirudin in its complex with a—thrombin [43].

Of the hydrogen—bond interactions in the K4—ACA structure, all but eight are also
common to the apo—K4 structure [23]; the ones present in K4—ACA and not in apo—
K4 are listed in Table 3.6. The criteria used to identify possible hydrogen bonds were
(a) a donor—acceptor center distance of less than 3.05 A and (b) a hydrogen—bond
angle of greater than 120°. Of the new hydrogen bonds, that of Gln7N—Asp5OD2
is borderline, that of Tyr90H—AspcOD1 involves an interkringle residue, which was
disordered in the apo—structure, that of Lys20NZ—Glu73OE1 makes the only ion
pair of K4 a hydrogen—bonded one, and the hydrogen bond between Arg71N H1 and
Arg32O just satisfies the selection criteria. Also listed in Table 3.6 are two apparently
important hydrogen bonds involving Asp55/Asp57 of the lysine—binding site of both

structures, which will be addressed later.

B. Lysine—Binding Site

The lysine—binding site is a relatively open, elongated, shallow depression located on
the kringle surface that is formed by His31-Lys35, Pro54—Ly358, Pr061—Phe64, and
Arg71-Cys75. At neutral pH, there are two negatively charged residues located on
one end of the binding site (Asp55 and Asp57) with carboxylate oxygens 5.2 A from
one another; there are also two positively charged residues located at the other end
of the depression (Lys35 and Arg71) with quaternary amino groups 6.3 A from one
another. Thus, the binding site contains doubly charged anionic and doubly charged
cationic centers. The depression, lined by aromatic rings of Trp62, Phe64, and Trp72,
also provides a highly nonpolar environment between the charged centers so that the
lysineLbinding site approximates a dipolar surface as first suggested from modeling
[38]. Thus, zwitterionic ligands such as lysine and ACA can first interact at long

range in preparation of docking, ultimately being anchored by ion pair interactions.

53

Table 3.6. New Intramolecular Hydrogen Bonds of K4—ACA. Hydrogen atoms were
assigned geometrically idealized positions. Donor atom is denoted (D), aceptor atom

 

 

 

 

DISTANCES(A) ANGLES(deg)
DONOR ACCEPTOR D...A H...A DHA CAH
Gln7 N Asp5 OD2 3.08 2.22 145 156
Tyr9 OH Aspc OD1 2.63 1.69 156 104
Arg10 NH2 Asn43 O 2.89 2.30 118 157
Lys20 NZ Glu73 OE1 2.96 2.17 132 152
Ser27 OG Thr29 O 2.74 1.99 134 123
Thr37 N G1n29 O 2.89 2.02 142 101
Thr47 N . G1y45 O 2.71 1.86 137 103
Arg71 NHl Arg32 O 3.01 2.17 143 139
Trp62 NHl Asp55 OD2 2.99 2.43 116 120
Tyr74 OH Asp57 OD2 2.59 1.71 152 105

 

 

 

 

 

 

 

 

54

The close van der Waals contacts between the methylene carbons of the ligand and

the aromatic residues between the charge centers additionally assist the binding.

C. Kringle 4—ACA Interaction

A zwitterionic ACA molecule in an extended conformation lies between the doubly
charged anionic and cationic centers of K4 formed by Asp55/Asp57 and Lys35/Arg71
(pH of crystals is 6.0), which also makes four or five hydrogen bonds with these
residues (Table 3.7) and interacts with the lipophilic core formed by Trp62, Phe64, and
Trp72 through the methylenes between the zwitterion charges (Figures 3.10, 3.11).

The hydrogen bond to Asp55 is questionable since its donor—acceptor angle is so

8 mall; however, its ion pair interaction is undeniable and even appears to be stronger

than that of Asp57. The oxygen atoms of the carboxylate groups of Asp55 (OD2)
and Asp57 (OD2) are hydrogen bonded back to the kringle (Table 3.6), which might
aid in aligning and anchoring the side chains for ligand binding. The Asp57 and
A rg71 residues were originally implicated in ligand binding through modification of
Asp57 with 1—ethyl—3—(3—dimethylaminopropyl) carbodiimide and Arg71 with 1,2—
Cy Clohexanedione [17], while the participation of Asp55 was suggested on the basis of
Computer modeling [38]. However, from Figure 3.10 and Figure 3.11 and Table 3.7,
i t is clear that Lys35 also participates as an integral member of the principal bind-
i 113 residues as inferred from the apo—K4 structure, in agreement with modification
Studies which showed that blocking Lys35 decreases the affinity of K4 for lysine—
Sepharose [44]. Computer modeling also suggests a doubly charged cationic center
i n t he binding site of PGKI involves Arg34 and Arg71 [38‘]. In the C858 0f K4, the
guide coordinates for modeling Lys35 had to be based on Ile35 of PTFl, which only
e){t'ellded to CB, so that the lysyl side chain was simply modeled in an extended

e
nergy—minimized conformation. This is clearly not the case as Lys35 retreats to-

W
ard Arg71 in the binding site and gives rise to a doubly charged center in both

‘

55

the apo- and K4—ACA structures. The new finding thus places the K4 binding site
on a comparative level to the modeled PGKl site with respect to a doubly charged

positive center.

, Table 3.7. Hydrogen Bonds Between K4 and ACA

 

DISTANCES(A) ANGLES(deg)
DONOR ACCEPTOR D...A H...A DHA CAH

Lys35 NZ ACA 01 2.72 1.99 127 148
Arg71 NE ACA 02 2.71 1.83 145 147
Arg71 NH2 ACA 02 2.89 2.09 134 129
ACA NZ Asp57 OD1 2.84 2.28 113 107
ACA NZ Asp55 OD2 2.12 1.97 84 112

 

 

 

 

 

 

 

 

 

 

D . Comparison of K4—ACA and Apo—K4 Structures

The structures PGK4 and PTKI, which have already been compared elsewhere [39],
Show that the lysine—binding site is approximated fairly well by PTKl, but with
Some notable exceptions. These render PTKl to be a nonbonding kringle. The
S t 1‘ Ilcture of PGK4 has additionally been compared with the K4—ACA structure by
t he Optimal superpositioning of CA, C, and N atoms. The rms differences between the

t W0 Structures are listed in Table 3.8. The agreement between the two structures is

Table 3.8. RMS difference between the K4—ACA and Apo—K4

 

 

 

 

 

 

 

 

 

 

A (A) # atom
All protein atoms 0.57 602
Main chain 0.44 237
Carbonyl oxygens 0.60 79
Side chains 0.66 286
Sulfurs (Cys, Met) 0.44 8
a—Carbon 0.43 79

 

 

56

 

Fi

11 gum 3.11. Stereoview of the lysine-binding site of K4—ACA. ACA is shown in bold;

y c1I‘ogen bonds are dashed.

57

good for the main—chain atoms and superb (0.27 A) after removing about 15% of the
atoms having deviations greater than 10. The differences between the apo and the
complexed structure in the lysine-binding site are listed in Table 3.9 and a. stereoview
of the superposition of the binding site regions is shown in Figure 3.12. The average
rms difference in the main—chain position in the site is only 0.25 A with only one
difference as large as 0.50 A, while the average of the side groups is 0.69 A. The
largest deviations in the binding site are due to crystal packing interactions. In the
I<4——ACAqstructure, Gln34 forms a hydrogen bond with the carbonyl oxygen of Trp72
of a. neighboring molecule, giving rise to a rms deviation of 1.12 A. Other significant
deviations occur with Asp55 and Asp57 of the anionic charge center (0.75 and 1.23
A , respectively). These are the consequence of a cofacial kringle—kringle interaction
of the binding site regions of two neighbor molecules in the apo-K4 structure. The
Asp55/Asp57 of one molecule form two ion pairs with Arg32/ Arg71 of neighboring
molecules; thus, Asp55 and Asp57 in apo-K4 point somewhat away from the lipophilic
bi nding core, which leads to relatively large deviations from the K4—ACA structure.
I t i s also noteworthy that Arg32 appears to be disordered in K4—ACA structure where
i t Cannot make a similar intermolecular interaction. In the case of Lys58, it is involved
i n a. complex intermolecular interaction in the apo-structure with a sulfate ion of a
_neighboring molecule. Interestingly, the position of the sulfate anion in K4 is close
to the cationic center in K4-ACA (Figure 3.12). Lastly, the side chain of Ly335 in
apo‘K4 forms a. hydrogen—bonding ion pair with the sulfate thus making the side——
Ch aim conformation a little different from that of the cationic center in K4—ACA
( Table 3.9). The remainder of the residues in the lysine—binding site are practically
i dentical in the two structures. Thus, from all appearances, the lysine—binding pocket

i S . . .
preformed in the kringle structure, and ACA binding takes place Without requiring

a.
113, Conformational changes of the host.

58

Table 3.9. Differences in lysine—binding sites between K4—ACA and Apo—K4

 

 

 

rms A (A) rms A (A)
residue (main chain) (side chain)
His3l 0.21 0.28
Arg32 0.17 0.50
His33 0.16 0.19
Gln34 0.23 1.12
Lys35 0.26 0.70
Pro54 0.36 0.23
Asp55 0.19 0.75
. Ala56 0.50 0.35
Asp57 0.25 1.23
Ly558 0.25 1.27
Pro61 0.10 0.54
Trp62 0.26 0.20
Cys63 0.19 0.35
Phe64 0.15 0.12
Arg71 0.10 0.21
Trp72 0.23 0.44
Glu73 0.19 0.65
Tyr74 0.10 0.53
Cys75 0.20 0.29
av rms 0.25 0.69

 

 

 

 

 

59

58

58
7
35 35
S S

ure 3.12. Stereoview of the comparison of the lysine—binding site of K4—ACA and
0\K4. K4—ACA is in bold; the sulfate position in apo—K4 is designated by S.

Fig
ap

60

E. Ligand Interaction and N MR Results

The indole side chains of Trp62 and Trp72 form end—to—face contacts with the rings
of Phe64 and Tyr74, respectively (Figure 3.11); such aromatic clustering is common
in proteins and provides enhanced stability [45] to the hydrophobic depression. N MR
NOE experiments [36, 37] found that the side chains of Trp62, Phe64 and Trp72
are perturbed most by ligand presence and that Trp25, His31, His33, Tyr4l and
Tyr74 are also affected, but to a lesser extent, in good agreement with the crystal
structure. Most of these aromatic residues are near the surface in the binding site
region within which the ligand lies in an extended conformation (Figure 3.10, 3.11).
The ring of Phe64 is as close as 3.2 A from CA of ACA and approximately 3.9 A
from the end of the Trp62 indole ring; thus, it appears that the phenylalanyl ring
can be affected through the aromatic stacking interaction with the Trp62 side chain
or possibly by a substitution at the CA position of the ligand. This conforms with
P he64 having large ligand-induced chemical shifts with such ligands [37]. Although
Somewhat removed from the binding center, the tyrosyl ring of Tyr74 is positioned 3.7
A from the end of the indole of Trp72 while its hydroxyl group appears to be making
an important hydrogen bond orienting the carboxylate group of Asp57. The Tyr74
may sense ligand-binding effects indirectly from the aromatic interaction with Trp72
a'11C1/0r the hydrogen-bonding interaction with Asp57. The His33 residue, which is at
t he perimeter of the binding site, has its imidazole ring stacked parallel to the phenyl
ti 118 0f Phe64 and within 4.5 A of it. The binding of bulkier ligands such as AMCHA
( Figure 1.3) may have a greater influence on this imidazole ring. In contrast to His33,
Hi 831 is located away from the binding core and in agreement with N MR observations
W hiCh show that it only has a small chemical shift upon ligand binding [36, 37]. The
i 11 Cl<>1e ring of Trp25 is in the next layer below the surface, but its end is oriented about
4 I Q A from the face of the indole of Trp62. It also displays only minor ligand effects

 

 

61

which most likely transmit through the aromatic—aromatic interaction with Trp62;
however, it too may experience greater effects from bulkier ligands. Lastly, the side
chain of Tyr41 is located distantly from the center of the binding site. Therefore,

it can only sense the presence of the ligand through secondary effects transmitted

through residues in the immediate vicinity of the ligand.

 

CHAPTER 4

Structure of Human Plasminogen

Kringle 1

4 - 1 Introduction

H uman plasminogen kringle 1 (PGKl) and kringle 4 (PGK4) both carry lysine binding
Si tes with the former having a higher affinity for lysine binding. It has been shown
by N MR that PGKl and PGK4 have a similar three—dimensional structure [46, 47],

which is compact and globular and built around a core of hydrophobic, aromatic
ami no acid side chains [46].

The secondary structure of PGKl [48] has been predicted based on the statistical
methodology of Chou and Fasman. Of the 80 residues (Cysl — Cys80) (Figure 4.1)
of t he PGKl region and the PGKl—PGK2 interconnecting peptide (Glu81—Glu83),
approximately 4% has been predicted as a—helix, 30% of fl—sheet, 53% of fl—turn
and 1 3% of random coil. According to the results of Overhauser experiments, acid—
base titration and two—dimensional chemical shift correlated spectroscopy, it has been
Shown that Leu46 is surrounded by a cluster of interactive aromatic side chains, which

ltlclllde Trp25, Phe35, His41, Tyr50, Trp62, and Tyr64 [47]. The same results were

also found in the case of PGK4, in which His41 and Tyr64 are replaced by Tyr41

62

 

63

 

SiguI‘e 4.1. The primary structure of human plasminogen kringle 1. Site numbering
eletion is based on homology with PGK5.

64

and Phe64 in PGK4. The observation supported the hypothesis that the buried

hydrophobic core is fundamental for the overall kringle folding [49].

Since PGKl can bind to fibrin and some w—aminocarboxylic acids, considerable
effort has been made to understand the nature of the lysine binding site. From
the proton Overhauser experiment [50], it has been found that the ligand interacts
directly with Phe36, Trp62 and Tyr72. A close interaction which has also been
found between Tyr64 and Tyr72 indicates that residue 64 is positioned close to the
binding site. Chemical modifications made with 1,2—cyclohexanedione [51] showed
that both Arg32 and Arg34 are involved in fibrin binding; additionally, Arg34 is also
i nvolved in the binding of w—aminocarboxylic acids. In the PGK4 lysine binding site,
Asp55/Asp57 and Arg71/Lys35 are located on a dipolar surface to form the anionic

and cationic centers. The same arrangement, except for Lys35, was also observed in
the PGKl structure. Trexler et a1. [17] have pointed out that only PGKl and PGK4
have both Asp at 57 and Arg at 71; in addition, the lysine binding sites present in
P G K 1 and PGK4 have been shown to have the same specificity [19]. Therefore, Asp57
and Arg71 in the PGKI binding site were suggested to act as anionic and cationic
C eliters, respectively.

Recently, based on the three-dimensional structure of PTFl and N MR observa-
tions [50], the PGKl lysine binding site was modeled by Tulinsky et al. [38] and a
St"BI—‘teoview of the modeled structure is shown in Figure 4.2. In the modeling, con-
serVed side chain atoms between PTF 1 and PGKl were kept constant and like atoms
of 11Ilconserved residues served as guide coordinates for replaced side chains. As
Show!) in Figure 4.2, Asp55 and Asp57 serve as anionic center, and Arg34 partici-
DateS with Arg7l in making up the cationic center of the lysine binding site. The

g1‘()()Ve of the binding site is lined by the aromatic rings of Phe35, Trp62, and Tyr72,
with Tyr72 interacting with both Trp62 and, in a second aromatic plane, with Tyr64.
one boundary of the site is determined by the Pro54—Pro58 pentapeptide stretch,

 

65

 

:igul‘e 4.2. Stereoview of the modeled lysine binding site of plasminogen kringle 1;

A in bold. Hydrogen atoms are shown at ideal positions (taken from [38]).

 

66

which contains both Asp55 and Asp57. The other consists of two short, disconnected
peptide segments containing Arg34 and Arg71.

The X—ray crystal structure determination of PGKl was undertaken to provide a
better understanding of the kringle structure and the nature of lysine binding. The
structures of PGKl and PGK4 give suitable information for modeling kringles of

plasminogen, which has five kringles.

4.2 Materials and Methods

A. Crystallization

The PGKI was isolated intact via proteolytic fragmentation of plasminogen. The
PGKl used for crystallization was kindly provided by Dr. F. J. Castellino in the
form of a lyophilized powder, which was stored at freezer temperatures in a dry
environment.
Because of the homology between PGKl and PGK4, the conditions used for crys-
tallizing PGK4 crystals were first applied to PGKl. A drop of 10 mg/ ml of protein
Solution was equilibrated against a 1 ml well solution of 0.12 M ammonium sulfate,
wi th 30% PEG 8000 and 1.2% DMF as an additive, at pH 6.0. However, no crystals
f(>1‘Itned. In an effort to find a new crystallization condition for PGKl, a factorial
Search .[52] was carried out. The search consisted of a set of 31 factorial solutions and
<:()r18isted of a random but balanced distribution of common precipitants, salts, and
buffers. Crystallization was carried out using the vapor diffusion method at room
temperature: 2p! hanging drops of the protein to which an equal volume of precipi-
tant was added were used for each individual trial. The factorial solution with 0.1 M
ITla-gllesium chloride, 0.1 M (N-[2—Acetamido]—2—iminodiacetic acid, pH 6.5 )(ADA),
30% PEG 8000 produced clusters of long, fine needle—like crystals after a few days.

Trials to refine the crystallization condition by reducing the precipitant concentra-

67

tion and also introducing 1-2% of the organic additives in the well, either resulted in
clusters of needles or clear drops. In order to improve the crystal size, a macroseed-

i ng technique was adopted. However, using needles as seeds in macroseeding is more
complicated since needles have a tendency to bend while being transfered from the
original growth solution to the new solution. The stress created in the crystals during
this process may result in defects at each stress point, and each of them may act as

a. nucleation site for growth of new needles. In order to avoid such phenomenon, the
needles were broken into smaller segments by a sharp glass fiber and the resulting
fragments were used as seeds. The seeds were transfered to fresh protein drops every
two weeks, since they started dissolving upon standing for more than two weeks. It
was thought that there might have been a serine protease contaminant in the protein
sample which made the crystals dissolve. Crystallization in the presence of serine
protease inhibitors such as benzamidine (50mM), which can bind to PGKl also, was
tried. However, crystals still dissolved after 2—3 weeks. Crystallization at the cold
room temperature (4° C) was then carried out. The crystallization trays were trans-
feted to cold room immediately after the hanging drops were set up. The crystals
again dissolved but only after 5—6 weeks. A crystal of size 1.2 x 0.2 x 0.1 mm was
grown (Figure 4.3) by a macroseeding procedure after reducing the PEG 8000 con-
cent ration to 23% in the well at 4° C. This crystal which was the only one grown to
X‘ray 'diffractable size was mounted and data were collected. The unit cell param-
eters and some crystal data are listed in Table 4.1. From the above observations, it
a'I)I>ea.rs that the solubility of PGKl is temperature dependent; i.e., the solubility of

the PGKI protein increases with increasing temperature.

B ' Data Collection and Data Reduction

X\r&y diffraction intensity data were measured at 2.48 A resolution from the crystal

SI)ecirnen having dimensions 1.2 x 0.2 x 0.1 mm employing a Siemens Xentronics

68

 

Figure 4.3. Photograph of K1 crystal. Crystal length is approximately 1.2 mm.

69

Table 4.1. Summary of crystal data of tetragonal Kl.

 

space group P43212
crystal system tetragonal
a (A) 58.93
b (A) 58.93
c (A) 54.64
molecules / unit cell 8

molecules / asymmetric unit 1
molecular weight (dalton) 10000

 

 

 

 

 

vm (fig?) 2.37
solvent fraction (%) 48.0

 

A rea. Detector with graphite monochromated CuKa radiation generated by a Rigaku
RU200 X—ray generator operating at 7.5 KW (50KV, 150MA). The crystal—detector
distance was set to allow neighboring Bragg reflections to be distinguished and is

u s ually determined from

D = aMAx/8, (4.1)

Where aMAx is the largest unit cell length in A. The distance was 11.65 cm in the data
Collection because: (1) the dimensions of the unit cell of the kringle are smaller than
93 A (Table 4.1); and (2) the closest distance possible with this instrument is 11.65

Cm - The detector swing angle was set at 125°, the scan range was 0.2° per frame of
measurement and each frame was collected for 120 seconds.

The raw intensity data reduction was carried out using the XEN GEN pro-
grams [26]. A total of 17,701 observations were measured, of which 3,444 were
independent. The distribution of intensities observed for various resolution ranges
is given in Table 4.2. After reflections with I/ 0(1) < 2 were removed, a set containing

2 ’987 independent reflections remained (87% observed, RMERGE = 0~059)-

70

Table 4.2. Distribution of reflection intensities and R—factors in various resolution

shells.

 

Res. (A) #refs 0 <20 <50 <100 <200 <400 <600‘ >600

 

4.5 673 3 12 13 16 39 80 94 416

3.57 631 9 23 15 40 47 119 117 216

3.12 606 22 36 40 60 117 165 81 82

2.84 609 44 65 79 112 144 124 26 13

2.63 596 55 107 121 99 134 71 9 0

2.48 329 40 82 92 68 34 12 0 0

 

 

 

 

Totals: 3444 173 325 360 395 515 571 327 772

 

C- Molecular Replacement

Since PGKI has a different space group from PGK4, Patterson rotation/ translation
molecular replacement methods were used to obtain initial phases for structure deter-
mi nation and refinement. Molecular replacement was carried out with the program
X‘P LOR [53]. In the rotation search, the selected vectors were rotated by applica-
tion of a rotation matrix C (01, 62, 03) where 01, 02, 03 are the Eulerian angles. The

Orientation of the search model was sampled by using the Lattman angles [54]:
0+ = 01 + 03 0- = 91 — 63 02 = 02 (4.2)

The interval A for 02 is a constant, the interval for 0+ is given by A/cos(02 / 2) and
the interval for 0- is given by A/sin(02/2).
In a conventional rotation search, the highest peak is usually assumed to be the

C()1‘1‘ect orientation, but sometimes this is not true. In the program X—PLOR, the

M

 

71

highest peaks of the rotation function are refined prior to a translation search. This
refinement is performed by minimizing the target function of individual atomic coor-

dinates (r) :
E,0t(r) = Epc(7‘) + E,(r) (4.3)

where E.- is an empirical energy function that describes the geometry and non—bonded
interactions of the molecule and Epc is an effective energy term that is proportional
to the standard linear correlation coefficient between the squared amplitudes of the
observed and the calculated normalized structure factors. The normalized structure
factors are computed with the search model placed in a triclinic unit cell identical in
geometry to that of the crystal. After this refinement, the refined search model with
the highest correlation coefficient is used for a translation search.

The model employed in the molecular replacement was the entire peptide backbone
and conserved side chains of the refined PGK4 structure, representing 504 atoms or
82% of the PGK4 structure. A triclinic cell of dimensions 90 x 90 x 90 A using data
from 8 .0 to 4.0 A resolution was used in the model structure factor calculation. Both
the model and unknown PGKl Patterson vectors of length greater than 3.0 A and
less than 15.0 A were selected. The set of the 1500 highest Patterson peaks were used

in the rotation search. The highest peak was found with Euler angles (105°, 155°,
80°)- The refinement of this solution was carried out and the final solution had peak
height 3.5 0' above the mean at (112.2°, 153.7°, 888°).

TI‘a-nslation search of the rotated search model was also carried out with the

pl‘0gram X—PLOR by computing the standard linear correlation coefficient

TIP“:
yz, C = )
(4.4)

Where E05, are the normalized observed structure factors and Ecazc are the normalized

 

 

72

calculated structure factors. The search model was placed in the unit cell of the crystal
VVlth the position given by the coordinates x, y, z of the center of gravity and with the
orientation given by the rotation matrix C. Data from 8.0 to 4.0 A resolution was
used in the translation calculation through which both enantiomorphic space groups
P41212 and P43212 were examined. The latter was taken as the correct space group
because its solution was 5.70 above the mean, which was significantly larger than
that of the former space group (3.90). The correlation coefficient was 0.34, with the
next highest being 0.28. After the orientation and position of the molecule had been

determined, a rigid body refinement was carried out giving an R—factor of 0.44 at 8.0

—— 4.0 A resolution.

D. Structure Refinement

The K1 structure was refined using the program PROLSQ [30]. The refinement

was carried out in two stages, initially with data at 7.0—2.8 A, then extending the

resolution to 2.48 A. At the beginning of the refinement, an overall thermal parameter

was assigned as 20 A2. After 10 cycles, the refinement converged at a R—factor of

0-34 and the (2|Fol -— chl) and (IFOI — IFCI) electron density maps were calculated.
The maps clearly indicated the positions for side chains which were not included in
the Calculation. Model building was then performed on an Evans and Sutherland
P8390 interactive stereographics system. The R—factor was reduced to 0.22 at the
end of the first stage of refinement. The remainder of the data (to 2.48 A) were
induded and water molecules were also gradually added. The reflection weights were
a"S'signed in seven shells of sin O/A based on (HFIo — |F|c|)/ 2‘ of the range. The final
PGK1 structure has a crystallographic R value of 0.16 for 2695 reflections between
7 '0‘2-48 A resolution range with 73 water molecules and two chloride ions. The final
reflection weights and R values in each range are given in Table 4.3, and a summary

of refinement parameters is listed in Table 4.4. The average thermal parameter is

 

 

 

73

Table 4.3. Weights of Reflections and R Values of the Final Refinement

 

 

R value

Dmm(A) no. reflections U(|F|) < IIFOI - chH > shell sphere
4.70 357 24 57 0.189 0.189
3.90 393 21 41 0.140 0.164
3.45 391 19 37 0.142 0.157
3.15 381 17 30 0.155 0.157
2.92 379 15 26 0.161 0.157
2.72 406 14 23 0.165 0.158
2.40 388 12 21 0.173 0.159

 

O'(|F|) = 15— 120 (sin 0/A — 1/6); < HFOI- IFCII >= 34.

 

 

 

 

74

Table 4.4. Summary of Final Least—Squares Parameters and Deviations

 

 

 

 

 

 

 

 

[I 1 target (a) I rms (A) I]
Distances (A)
Bond Distance 0.020 0.015
Angle Distance 0.035 0.043
Planar 1,-4 distance 0.050 0.042
Non-bonded distances (A)
Single torsion 0.55 0.22
Multiple torsion 0.55 0.30
Possible H-bond . 0.55 0.27
Torsion angles (deg)
Planar 3 2
Staggered 15 23
Orthonormal 20 25
Plane groups (A) 0.02 0.02
Chiral centers (A3) 0.15 0.17
Thermal restraints (A2)
Main chain bond 1.5 0.8
Main chain angle 2.0 1.3
Side chain bond 2.0 1.2
Side chain angle 2.5 2.0

 

 

 

 

 

 

 

 

75

27.6 A2, which is higher than that of PGK4 (18 A2). This is consistent with the high

solvent content of PGKl (48% versus 38%).

4.3 Results and Discussion

A. General Structure

Nearly the entire structure of PGKl is well defined in the electron density; however,
no density was found for interkringle regions of a—c and 81—83. Thus, the refined
PGKI structure extends from residue Glu—d to Cy580. The distribution of the main-
chain torsion angles ((13, 1b) of the refined PGKl structure is shown in Figure 4.4, from
which it can be seen that all the non-glycine amino acids, except Glu48, conform sat-
isfactorily to the conformationally allowed regions. Although the residue Glu48 is
outside the allowed region, it fits in electron density very well. The same conforma-
tion is also observed in the PGK4 structure, where position 48 is methionine. The
distri bution of the omega angles of the refined PGKl structure is shown in Figure 4.5.
All residues, except one, have an angles within the 180° :1: 5° region. A cis—proline
With an an angle of 2° at position 30 is also found in this structure similar to that
in the apo—K4 structure. The secondary structural features of PGKl are listed in
Table 4.5. The closed interkringle C, D loops (Cys51—Cys75) form two distinct an-
tiparallel ,B—sheets due to close contact of two disulfide bridges in the folded state.
There are 10 reverse turns in the PGKl structure. It appears that about half of the
kringle residues are involved in these turns, of which one third are conserved among
different kringle sequences. Thus, the similar folding of kringles could be due to the
large number of conserved turns.
The final r.m.s. B—values of main-chain and side—chain atoms along the sequence
of PGKI are shown in Figure 4.6. The magnitudes and variations of B values are

llkely reflective of the positions of the residues with respect to the three-dimensional

 

 

-180 -90 0 90 180
180 . l l I. l\ l l l l
I it: ‘ 't :t fi.
fi ,, .. . ,. \
T 4* ‘* ‘n 3“ I _
z: . . . * I
l *" ' I T
t
l ,
90 a: L I |//| __
* // l
* /
_| * F l I
. l | *
———1\\ I ‘* I L *lg __
__ \ *. ,. K \d
m 0 r ‘L‘ 1‘ 1‘
Q. \ u \
\ f It \\
\ unit" I
l\\// * : i l _
I1——/__T ‘
_ I
L_______;__I __
—90 ._J
_ *Glu48
E] __
‘130 ———:L__~_‘—_' [E
l l l l l l l l
~180 -9o 0 90 180

 

 

 

 

 

 

 

 

 

 

Figure 4.4. The Ramachandran plot of PGKl. Glycine residues are boxed.

180

90

-90

-l80

           

\\ ~

\

.\.\\\

1
.x\\\\\\\\\\\ \\
- 73 -175

______

000000
22222

78

Table 4.5. Secondary structural elements of PGKI structure.

 

13 Structures Type Residues

[31 Antiparallel Ser14 —— Cy522
‘ Lysl5 — Thr21
Thr16 — Ile20

62 Antiparallel Gln23 - Tyr50
Lys24 -— Asn49
Trp25 — Glu48

fl 3 Antiparallel Arg52—Trp62
Asn53 — Pro61

,84 Antiparallel Cy863 — Cys75
Tyr64 — Tyr74
Thr65 — Asp73
Thr66 — Tyr72

 

 

 

Reverse Turns Type Residues

T1 Type II’ Thr3—Gly4-Asn5—Gly6

T2 Type I Gly6—Lys7—Asn8-Tyr9

T3 Type I Thr16—Lysl7-Asn18—Gly19
T4 Type III Lys24-Trp25—Ser26-Ser27
T5 Type III Ser37-Pro38-Ala39—Thr40
T6 Type I His41—Pro42-Ser43-Glu44
T7 Type II’ Glu47-G1u48—Asn49—Tyr50
T8 Type I Asn53—Pro54—Asp55—Asn56
T10 Type I Asp67—Pro68—Glu69-Lys70

 

 

 

 

79

\

.ongoasm 3:?” co mEosa Ass: 42:: EesoloEm was Ann: :23 55:01:me Mo 9833 75:2: owfigm 2: we so?“ .cd oEwE

mam—>52 QDQHmmm

 

 

 

 

 

 

 

 

 

 

om S. S. 3 oo mm on 3. ow mm on ma cm 3 3 m.
_ — _ _ _ p _ _ _ p p b p _ - an:
d
O
{on F
O
<
m
ram
m
m
[L a
. H.
Ion <
/ s
m
C a
Ten em,“
8
av

 

 

 

80

folding. The lowest values shown are associated with residues involved in the disulfide
groups, especially the central disulfide cluster, which is buried in the interior of the
kringle domain (Cys22-Cys63, Cys51-Cys75). The stabilizing effect of the disulfide
bridges on kringle folding is very visible. The high B values in Figure 4.6 correspond
to residues that are poorly defined and are somewhat disordered, especially residues
near the terminals of the peptide chain. The B values of the side chain groups are
generally higher than that of main chain showing that side chain atoms have more
flexibility than main chain atoms. From Figure 4.6, it can also be seen that low
values are associated with the C, D inner loops (Cys51—Cys75) of the kringle while
the larger B loop (Cys22—Cys51) which is six residues longer than A loop, displays
more flexibility than most other parts of the kringle.

The refined PGK1 structure contains 71 water molecules with average occupancy
of 0.68 and average thermal parameter of 24 A2. The distributions of thermal param-
eters and occupancies are shown in Figures 4.7 and 4.8. Of the 71 water molecules,
47 have occupancies between 0.5 and 1.0, and 10 have full occupancy. There is only
one water molecule buried within the PGK1 structure, where Gln23 O and Tyr64
N make hydrogen bonds with the water molecule of 2.80 A and 2.71 A, respectively
(Figure 4.9). The internal water molecule has the highest quality factor (Occz/ B)
ranking among all of the water molecules, with an occupancy, 1.0, and the lowest B
value of 12 A2.

The hydrophobic core formed by a number of stacked aromatic and proline
residues, centered around Leu46, is shown in Figure 4.10. Of these side chain groups,
all make close van der Waals contacts with one another and a number of the aromatic

'rings participate in stacking interactions in which one aromatic ring is pointed in a
perpendicular manner toward another. The edgehto—face interactions observed in the
PGK1 structure also include some proline residues. The edge of Pro61 is positioned

4.5 A from the tyrosyl ring of Tyr9 at an angle of approximately 100°. In addition,

 

 
  

          
 
 
  

\\
.\

\

\\

 

a m 1 1

82

 

 

 

OCCUPANCY

83

 

Figure 4.9. Stereoview of the internal water molecule in PGK1 structure. Dashed
lines indicate hydrogen bonds.

84

 

lFigure 4.10. Stereoview of the PGK1 aromatic and proline residues forming hydropho-
1c core.

85

the edge of Pro54 is directed at an angle of 95° toward the phenyl ring of Phe35,
at a separation of about 3.8 A. Such aromatic clustering is common in proteins and

reinforces structural stability [45].

B. Structure of the Lysine Binding Site

The lysine binding site observed in the PGK1 structure is similar to that of the
PGK4 structure. It is a relatively open, elongated, shallow depression and is located
on the surface of one of oblate faces of the kringle, bounded by the segments His31—
Phe35, Pro54—Pro58, Pro61—Tyr64, Arg71-Cys75 (Figure 4.11). Thus, this region
corresponds mostly to the inner loops of the kringle sequence and is formed by the
B, C and D segments (Figure 1.2). The two negatively charged Asp55 and Asp57
residues are located on one end of the lysine binding site having carboxylate oxygens
4.9 A from each other. The two positively charged Arg34 and Arg71 residues are
located on the other end of the depression and act as a cationic center. Although
Arg34 participates with Arg71 in making up the cationic center, it has a somewhat
disordered guanidinium group. As can be seen in Figure 4.12, a long channel of
electron density is observed between Arg34 and Arg71 but the amino group of the
former is outside the density map. In spite of the lack of density on the complete
guanidino group, from the electron density up to Arg34 NZ, it is clear that the amino
group of Arg34 must-be close to the guanidino group of Arg71 so that the two can act
as a cationic center. The aromatic cluster in the binding pocket, formed by the Phe35,
Trp62, Tyr64, Tyr72, and Tyr74 side chains, provides a highly nonpolar environment
between the charged centers. The indole ring of Trp62 and the tyrosyl ring of Tyr72
conform a V-shaped groove with an interplanar angle of approximately 80°, make
aromatic stacking interactions with Tyr64 and Tyr74, respectively, and produce a
stabilized structural framework for ligand binding. Thus, in PGK1, the binding site

is a well-defined area in which zwitterionic ligands such as lysine and ACA can be

M

 

86

 

55
74

62
34

 

Figure 4.11. Stereoview of the PGK1 lysine binding site.

87

      

 
 
     

 
 
  

   
   

  
 

 

       

 
   

I-"‘.:\‘r {fizwt 1'2““ . ‘V
'pf-ga . V . ‘. t” ‘ " '- h
3.5.1,: I.“ '2' \‘n "4' 2’ \}*‘?¢§ ’
. / ' ‘ U 9
1", - ,1 !IA\‘ .9 "q 1‘} ',’ CIA“ ’
1‘ 1", r \‘ ,” \I_’ \‘, ‘

V4"?! ~ ‘ -
.a" ’A‘l’z

Figure 4.12. Electron density in vicinity of Arg34 and Arg71 of PGK1 structure;
contour at 10.

88

docked by making ion pair interactions with anionic and cationic centers defined by
the side chain carboxylate groups of Asp55 and Asp57 on one side and conformed by
the guanidinium groups of Arg34 and Arg71 on the other side. The exposed lipophilic
groove on the kringle surface then interacts with the nonpolar portion of the ligand
by making close van der Waals contacts.

The most significant intermolecular interaction found in the crystal structure in-
volves the lysine binding site of one molecule and the Lysl5 side chain of a neighboring
molecule: the oxygen atoms of the carboxylate groups of Asp55 and Asp57 interact
with the ammonium group of the lysine (Figure 4.13). The positively charged Lysl5’
side chain of a symmetry mate makes hydrogen bonds with Asp55 OD1 and Asp57
OD1 at distances of 3.1 A and 3.0 A, respectively. (In addition, the latter two side
chains are oriented by hydrogen bonds between Asp55 OD2 and Trp62 N E1 and be-
tween Asp57 OD1 and Tyr74 OH. The hydrogen bonds found in the lysine binding

site are summarized in Table 4.6.

Table 4.6. Hydrogen Bonds found in the lysine binding site of the PGK1 structure.
The prime designates a symmetry—related molecule. A donor atom is denoted (D),
an acceptor atom (A).

 

Donor Acceptor D...A (A)
Tyr74 OH Asp57 OD1 2.3
Trp62 NEl Asp55 OD2 3.0
Lys15’ NZ Asp55 OD1 3.1
Lysl5’ NZ Asp57 OD1 3.0

 

 

 

 

 

 

A second interaction observed in the binding site is mediated via chloride ions.
The discovery of the two chloride ions was based on : (1) two large blocks of electron
density near the guanidino groups of Arg34 and Arg71 were observed in the difference

map and (2) charges are offset in the binding site by their presence. Since the Lysl5

89

 

1?. i«Sure 4.13. Stereoview showing intermolecular interactions involved in the lysine
l IlCiing site. The prime denotes a symmetry—related molecule.

”m”!

90

side chain from the symmetry—related molecule makes hydrogen bonds with Asp55
and Asp57, the inclusion of chloride ions in the cationic center produces interactions
with positively charged residues Arg34 and Arg71 and all the charges in the binding
site are essentially offset. The observed Cl—N distances are 3.4 A (71NH2—Cll) and
5.6 A (34NH2—C12), respectively. Although the electron density of the two chloride
ions is quite good, they have partial occupancies and average B values (Cll = 0.5, 29
A2; C12 = 0.6, 35 A2). The unexpected low occupancies and high B values might be
due to the fact that two chloride ions, which are located on the kringle surface, are
shared by two molecules in two different asymmetric unit and only partially occupy
the lysine binding site. The slight disorder on the amino group of Arg34 might reflect
the longer distance of Arg34—C12 (5.6 A) than that of Arg71-Cll (3.4 A).

C. Comparison of Plasminogen K1 and Plasminogen K4

The structures of PGK1 and PGK4 have been compared by calculating the rotation—
translation components which minimize the difference in positions of the CA, C, and

N atoms of the kringle. The rms deviation for the main chain atoms between the
two structures is 0.88 A. After removing 59 atoms (25%) having deviations greater
than 10', the rms deviation was reduced to 0.49 A, which is close to the expected
error between the independent determinations. The average deviations of individual
residues are shown in Figures 4.14 and 4.15, and the stereoview of the superposition
of both backbone structures is shown in Figure 4.16. As can be seen in Figure 4.16
the overall conformation of the polypeptide chain of PGK1 is very similar to that of
P GK4 except in two regions, around residue 58 and the region from residue 32 to
45 — In the former region, the large deviations are due to insertion of Gln59 in PGK1,
v"lilich gives rise to an expansion of the C loop. In the latter region, the rms deviation
of the main chain is a surprisingly large 1.2 A; the large deviations here are most likely

dlle to different intermolecular interactions occuring around the two—fold axis relating

M

 

 

91

.mozEmoa 823:8 3365 33:0 wax—E 8232:? VKUL was HXOA :ooBson macaw Z .0 .<0 mo mcofiflnrow 32¢ .34.. ouswmm

mam—2:2 man—5mm

cm 3. S. no ow mm on nv ow. mm on mm cm 3 3 m.

 

— _ — p p p p _ — —i _ p — — p O
Ind
Z
la 0
_
.H
r5 w
A. >
m—
m o
m
Ted 2
.m
In
msm

 

 

 

92

.m05fimm0h U0>u0m=OU 0adomflfim m0—0um0 U0:~rm .m0u3a03uam «excl «0:0 HVHUAH $003003 mcmdgo 0me m0 mfiomadm>0fi m2.“ .Ev 0H=wmrm

mmmzaz HDDamm

cm 3. on we on an on av ov mm on ma cm 3 S m.

 

 

 

— — — b — _ b i- b - — n... — l,- l—D O
lad
: I~
Z
I: o
_
IN E
c <
{Wm _
>
In m
G
In.”
m
[V E
m
Imé
laud

 

 

 

93

 

 

Figure 4.16. Stereoview of the optimally superimposed main chain structures of PGK1

bold) and PGK4.

94

the segment between two neighboring PGK1 molecules (Table 4.7, Figure 4.17). Such
interactions not only affect the conformations of the side chain groups but also make
the position of the main chain backbone different from that of the PGK4 structure.
In addition, in the PGK4 structure, Arg32‘forms ligand—like interactions with Asp55
and the sulfate ion of a neighboring molecule (Figure 4.18) [23] that also may affect
the main chain conformation of the B loop (Gln23—Tyr50). Both of these observations
are in agreement with the second outer loop of kringles being relatively flexible with
different conformations in tissue plasminogen kringle 2 [55], in the second kringle of
prothrombin [56] and in the kringle of urokinase [57].

Since PGK1 and PGK4 have the ability to bind lysine, a comparison of the ly-
sine binding site of the two structures has been carried out and a stereoview of the
superposition of the binding site regions is shown in Figure 4.19. The rms differences
between the two structures in the lysine binding sites are listed in Table 4.8. The av-
erage rms difference in main chain positions in this site is 0.42 A except for residues
31—35 and 58. In the PGK1 structure, Arg34 projects into the binding pocket to
participate with Arg71 in making up the cationic ionic center of the binding site,
While in the PGK4 structure, Gln34 points to the bulk solvent without making any

interaction with other residues. Arg32, which is conserved in both structures, has a
huge deviation of 2.3 A in the side chain position. One reason for this is the different
free bond rotations of the side chain groups. Another reason is an intermolecular
i nteraction in the PGK4 structure (Figure 4.18). Besides Arg32, other residues found
in ligand—like binding in the PGK4 structure were Asp57 and Arg71. Similar to
the interaction between Arg32 and Asp55, the positively charged Arg71 is positioned
near the oppositely charged Asp57 residue of a neighboring kringle and participates
in an ion ‘pair interaction with it. Such interactions not only make the side chain
conformations of Asp57 and Arg71 different in both structures but also account for

t he side chain group of Arg71 in the PGK4 structure extending outward away from

 

95

Table 4.7. Intermolecular interactions of PGK1 < 3.7A.

 

MoleculeA MoleculeB Distance(A)

 

Arg32 NHl Pro37 O 3.67
Arg32 CG A1838 0 3.06
Arg32 CD Ala38 o 2.96
Arg32 NE Ala38 O 3.63
Arg32 CD Pro42 CG 3.49
Pro33 O Thr40 CA 3.58

Pro33 O Thr40 OGI 3.33
Arg34 CA Thr40 CG2 3.52
Arg34 C Thr40 CG2 3.20
Arg34 O Thr40 CG2 2.70

Pr038 O Arg32 NHI 3.67

Ala39 O Arg32 CG 3.06
Ala39 O Arg32 CD 2.96
Ala39 O Arg32 NE 3.63
Thr40 CA Pr033 O 3.58
Thr40 0G1 Pro33 O 3.33
Thr40 CG2 Arg34 CA 3.52
Thr40 CG2 Arg34. C 3.20
Thr40 CG2 Arg34 O 2.70

 

 

 

 

Pro42 CG Arg32 CD 3.49

 

“‘1

96

 

molecule 8 molecule 8

 

molecule A molecule A

Figure 4.17. Stereoview of crystal packing interactions at a two—fold symmetry axis.
Residues involved in intermolecular interactions are shown in bold. The two-fold axis
in the a—b diagonal is indicated.

97

 

 

Figure 4.18. Stereoview of intermolecular interactions found in the lysine binding site
of PGK4. The side chains from the second symmetry—related molecule are shown in
bold; the sulfate ion is designated with a filled circle [23].

98

 

 

Figure 4.19. Stereoview of the superposition of the lysine binding sites of the

PGK1(bold) and PGK4 structures.

99

Table 4.8. Deviations of the positions of residues in the lysine binding site between

the PGK1 and PGK4 structures.

 

 

 

 

 

 

rms A(A) rms A(A)
residue number PGK1 PGK4 (main chain) (side chain)
31 H H 0.39 0.20
32 R R 0.97 2.35
33 P H 1.47 —
34 R Q 1.46 4.66
35 F K 0.60 1.67
54 P P 0.28 0.31
55 D D . 0.59 0.55
56 N A 0.52 0.66
57 D D I 0.52 1.59
58 P K 2.22 —
. 61 P P 0.48 0.45
62 W W 0.27 0.32
63 C C 0.28 0.19
64 Y F 0.45 0.57
71 R R 0.35 1.17
72 Y W 0.46 0.82
73 D E 0.30 0.95
74 Y Y 0.24 0.58
75 C C 0.77 0.63

 

 

100

the binding site. Lastly, the inclusion of Lys35 within the PGK4 binding site as a
cationic center introduces a large rms difference of 1.6 A between this residue and the
Phe35 residue in the PGK1 structure. The two aromatic rings in the lysine binding
site of PGK4 which are replaced by analogous residues in PGK1, are expected to
behave similarly. For instance, PGK4 Phe64 and Trp72 are replaced by tyrosines
at both positions in PGK1. Figure 4.19 shows that the agreement of the positions
of the aromatic rings between the two structures is good for both main and side
chain atoms. The remainder of the residues in the lysine binding site are practically
identical in the two structures. A comparison of the two structures shows that even
though the binding site is confined by a rather fixed set of residues in both PGK1 and
PGK4, they show some flexibility. The absence of the aromatic ring at position 35 in
PGK4 makes the lysine binding site region more open and might make it kinetically
faster for ligand binding, especially for bulkier ligands such as AMCHA. On the other
hand, the association constant of AMCHA of PGK1 is about twice as large as that

of PGK4, probably because Phe35 is in the PGK1 lysine binding site.

 

CHAPTER 5

Structure of Tissue—type
Plasminogen Activator Kringle 2

Domain

5.1 Introduction

Tissue type plasminogen activator (tPA) is a serine protease that converts proenzyme
plasminogen into plasmin, which, in turn, degrades the fibrin network into soluble
products. The native tPA is a single polypeptide chain (sctPA) with 527 amino acids
and a molecular weight of about 70,000 daltons [13, 58]. Upon limited plasmin hy-
drolysis the molecule is cleaved at the Arg278—Ile279 peptide bond and converted to
a two chain activator (tctPA) linked by one disulfide bond (Figure 5.1) [59]. The
sctPA is less active than the tctPA form toward small substrates [60], but both forms
possess similar plasminogen activator activities [61]. It has been known that tPA
consists of five domains which are homologous to parts of other proteins. Starting
from the N —terminal, it consists of a finger domain, homologous with the finger do-
main responsible for the fibrin—affinity of fibronectin, a growth factor domain, similar

to those of factor X, factor IX, bovine protein C, and the epidermal growth factor

101

 

.o-W",

 

102

COOH

S 478

D 371

  

Serine Protease Domain

Growth Factor Domain

. Finger Domain
'

K. '0)

Figure 5.1. Schematic presentation of the primary structure of human tissue—type
plasminogen activator. The arrow indicates the cleavage site for the conversion of
single chain tPA to two chain tPA. The active site residues are indicated by asterisks.

 

103

domain, two kringle domains (K1 and K2), homologous to kringles of plasminogen,
urokinase, factor XII, and prothrombin, and lastly, a protease domain, similar to
those in trypsin or chymotrypsin [13].

The tPA can bind to fibrin like plasminogen and plasmin where it activates plas-
minogen to plasmin. Several studies have been done on the nature of the interaction
between tPA and fibrin. Deletion mutant experiments have been performed and
studied on the function of the individual domains of tPA, and it was concluded that
stimulation of the plasminogen activator activity by fibrin was mediated both by
the finger domain and the K2 domain [62]. A lysine binding site was found in K2
of tPA and its involvement in fibrin binding was strongly suggested [62]. The ly-
sine binding site in K2 of tPA was shown to have equal affinity for lysine analogs
with and without a free carboxylate group, suggestive that tPA, unlike plasminogen,
does not discriminate between C—terminal and interchain lysine residues [63]. From
chemical modification experiments, it was shown that one or more glutamic acid or
aspartic acid residues in K2 of tPA are involved in the interaction with fibrin and 1y-
sine analogs [64]. Mutation experiments showed that Asp55 and Asp57 (Figure 5.2)
were both essential for lysine binding, while Glu73 might be involved but was not
essential [64].

Modeling of tPAK2 based on the structure of PTFl was carried out by Tulinsky
et al. [38], where the binding pocket consists of residues Asp55, Asp57, Trp62, and
Trp72. The His64 residue, substituted for Tyr64 and Phe64 in PGK1 and PGK4,
respectively, is located just below the surface of the binding site near the expected
cationic site.‘ Therefore, it was thought to act as a cationic center and provide a
positive charge to the negative carboxylate group of the ligand.

Recently, the crystal structure of tPAK2 was determined by multiple isomorphous
replacement (MIR) and refined by the program X—PLOR at 2.4 A resolution [55]. Be-

fore applying the MIR method, molecular replacement was attempted to determine

 

104

”aeoooo
“a 5 6° m®®<§é
G 96999 ©

9 0 60 ®

3 66a 90 .6 CV)
to o <9

O
G 03 9 G
10 G G a a

0 7° '5 '3 o
9 o 9 e
3 9 m. m 40

9 a o a

me 9 9 ‘ <9
G e o 50 ®
a 2 0 ©
16° ‘33 0° 44%
. c (16660 4821 @G‘E 44b
21me a
6 ®

Figure 5.2. Primary structure of tPA kringle 2. Site numbering insertions and dele-
tions are based on homology with PGK5.

 

105

initial phases for tPAK2 by using PGK4 as a model. However, due to three molecules
in the asymmetric unit, the translation search proved to be unsuccessful. The struc-
ture of tPAK2 showed a similar overall folding to that of PGK1 and PGK4 except
for an a—helix in the tPAK2 structure. The crystal structure also showed a strong
interaction between the lysine residue of one molecule and the lysine binding site of
a non—crystallographically related neighboring molecule. The lysine pocket observed
in this structure is similar to that of PGK1 and PGK4. The negatively charged side
chains of Asp55 and Asp57 were proposed to be involved in the interaction with the
positively charged amino group of the ligand, and the hydrophobic surfaces of Trp62
and Trp72 were thought to form van der Waals contacts with the methylene groups
of the ligand. However, the important positively charged residue in this structure
was found to be Lys33, whose amino group is virtually positionally identical to that
of the guanidinium group of Arg71 in the PGK4 structure.

In addition to the X—PLOR refined structure, a solution structure of tPAK2 com-
plexed with 6-aminohexanoic acid was determined by N MR spectroscopy and dy-
namical simulated annealing calculations [65]. This work showed that the structure
of tPAK2 has a compact globular conformation characterized by a number of turns
as well as by one right-handed a—helix and five antiparallel fl—sheets. The a—helix
is formed by the Ser43—Gly45 segment, which contains three insertions in the tPAK2
(Figure 5.2). The binding site was also defined bytwo negatively charged residues,
Asp55 and Asp57, and an aromatic pocket lined by Tyr36, Trp62, His64 and Trp72
side chains. The positively charged side chain of Lys33 and Arg69 may favor interac-
tions with the carboxylate group of the ligand in the structure.

This chapter describes the refinement of the tPAK2 structure by the restrained
least—squares method implemented by the program “PROSLQ” with the diffraction
data (2.43 A resolution) used for the X—PLOR refinement. The final 2.43 A resolution

structure from the X—PLOR refinement was used as the initial model. A comparison

 

106

of the final “PROSLQ” and “X-PLOR” structures revealed that both showed similar
lysine binding sites, and the rms deviation on the main—chain atoms is as small as

0.18 A.

5.2 Experimental Procedure

A. X—PLOR Structure

Crystals of tPAK2 were grown by the vapor diffusion method by using 4mg/ml pro-
tein, 7% saturated NH4C1, and 50 mM NH4HC03, at pH 8.0. Crystals of tPAK2 are
monoclinic, space group P21; the unit cell with dimensions a = 54.80 A, b = 63.58 A, c
= 46.58 A, 6 = 106.73° has six molecules per unit cell with three molecules per asym-
metric unit. Three dimensional intensity data were originally measured with a Rigaku
AFC6 diffractometer and CuKa radiation at 9kW power (50kV, 180mA). Three crys-
tals were used to collect 2.8 A resolution data with Rmerge(l)=0.087. The initial struc-
ture of tPAK2 was determined by the technique of multiple isomorphous replacement.
The heavy atom derivatives used for crystal soaking were 6 mM, (N H4)2PtCl4, 10mM
of K2AUCl4, or a 200—fold dilution of a saturated KI solution. A 3.0 A intensity data
set was collected for each of the derivatives. The resulting electron density maps
clearly showed the boundaries for three molecules in the asymmetric unit and good
connectivity for most parts of the main chain. The starting model used in the refine—
ment was obtained by fitting the structure of PGK4 to the map, with corrections to
the sequence of PGK4 to correspond to that of tPAK2. Refinement of the 10.0—2.8 A
date set was performed with the program X—PLOR [53]. The R-factor was reduced
to 20% after three iterations of refinement and map fitting. The final R—factor was
17% between 10.0—2.8 A resolution with 20 water molecules and three chloride ions
(one in each molecule).

A much better set of 2.36 A intensity data was collected later with the Siemens

 

107

area detector. The raw data were reduced and scaled with the XENGEN program
[26]. A total of 30,526 observations were measured, of which 11,623 were independent
(Rsym = 0.068). The previous structure at 2.8 A resolution was used as a starting
model for the refinement of the new set of data. Refinement proceeded in stages,
initially including data at 10.0—2.8 A, then at 2.43 A resolution. A summary of the

final refinement results is given in Table 5.1.

Table 5.1. A summary of the final refinement results of the X—PLOR refined structure.

 

 

 

Resolution (A) 10.0-2.43
No. of reflections 9687
Crystallographic R 0.184
No. of protein atoms 2025

No. of water molecules 92

No. of chloride ions 3

RMS deviation in bond length (A) 0.016
RMS deviation in bond angle (deg) 3.3

RMS deviation in torsion angles (deg) 26

 

 

 

B. PROLSQ Structure

The X—PLOR refined structure devoid of chloride ions and water molecules was taken
as the initial model for the PROLSQ refinement. The R value started at 0.26 with an
overall thermal parameter of 16 A2, and R decreased to 0.20 after the first stage (8.0—
2.8 A). Chloride ions were added after 32 cycles in accordance with the IFOI - IFCI
map. As phases improved, the remainder of the data (to 2.38 A) were included, and
reflection weights were assigned in seven shells of sin 0//\ based on (IIFIo — |F|c|)/ 2
of the range. Since three molecules were in an asymmetric unit, non—crystallographic

symmetry restraints were applied between main—chain and side—chain atoms of the

 

 

108

three independent molecules. The final reflection weights and the R values in each
range are shown in Table 5.2, and a summary of the refinement parameters is listed
in Table 5.3. The final tPAK2 structure has a crystallographic R value of 0.145 for
8827 reflections between 8.0—2.38 A resolution with 203 water molecules and three

chloride ions and an average thermal parameter of 21.7 A2.

5.3 Results and Discussion

A. General Structure

The refined tPAK2 structure consists of 3 molecules in the asymmetric unit; nearly all
three molecules are well defined and extend from Ser-c to Ser81. Three chloride ions
were confirmed in the refined tPAK2 structure, one in each molecule. The presence
of chloride ions agrees with the observation that even relatively low concentrations
'of chloride can precipitate the protein, and in fact, crystals can not be obtained in
the absence of the chloride ion [55]. The three chloride ions, which have well-defined
electron densities (Figure 5.3) are at full occupancies and have low thermal parameters
(~16 A2). The average occupancy of the 203 water molecules is about 0.76, and their
average thermal B value is 25.6 A2, slightly higher than that of the protein (21.3 A2).
0f the 203 water molecules, two are internal and conserved in the three independent
molecules. One of the the two water molecules is buried within the tPAK2 structure
and bridges the main chain Leu23 O and His64 N atoms at distances of 2.7 A and 3.0
A, respectively (Figure 5.4). Interestingly, this water molecule was also observed in
the PGK1 (Chapter 4) and PGK4 [41] structures. The other internal water molecule
was found in an empty pocket of the protein structure, that is removed from, but
accessible to the bulk solvent. It bridges three main—chain atoms (Gly6 O, Tyr9 O,
and Asn53 N) and serves as a hydrogen donor and acceptor (Figure 5.5).

Although tPAK2 has four more residues than PGK4, it has a globular conforma-

109

Table 5.2. Weights of reflections and R values of the final refinement cycle of tPAK2

 

 

R value
dmin(A) no. reflections 0(IFI)“ < HFOI — [Fell > shell sphere
4.56 1142 23 47 0.172 0.172
3.75 1301 19 35 0.126 0.147
3.32 1299 17 31 0.129 0.142
3.01 1332 15 26 0.148 0.143
2.78 1309 13 22 0.149 0.144
2.60 1195 12 20 0.155 0.145
2.20 1249 _ 10 17 0.151 0.145

 

 

 

la. o(|F|)=26 — 258 [(sin o/A) — (1/6)].

 

 

 

 

110

Table 5.3. Summary of final least—squares parameters and deviations of tPAK2

 

[[ ] target (a) [ rms (A) ]]
Distances (A)

 

 

Bond Distance 0.020 0.015
Angle Distance 0.040 0.046
Planar 1,-4 distance 0.060 0.055

 

N on-bonded distances (A)

 

 

 

 

Single torsion 0.55 0.20
Multiple torsion 0.55 0.24
Possible H-bond 0.55 0.23
Torsion angles (deg)

Planar 3 2
Staggered 15 20
Orthonormal 20 23
Plane groups (A) 0.02 0.01
Chiral centers (A3) 0.15 0.15

 

 

Thermal restraints ( \2)

 

 

Main chain bond 1.5 0.9
Main chain angle 2.0 i 1.5
Side chain bond 2.0 1.4
Side chain angle 2.5 2.0
N on—Crystallographic Symmetry (A)

Main chain atoms 0.40 0.26

 

 

 

 

 

Side chain atoms ‘ 0.80 0.45

 

 

111

 

Figure 5.3. Electron density of chloride ion in the lysine binding site of tPAK2.
Contour at 10'.

 

112

 

Figure 5.4. Stereoview showing an internal water molecule in the tPAK2 structure.
Hydrogen bonds are represented by dashed lines.

 

113

 

Figure 5.5. Stereoview showing the water molecule found in an empty pocket of the
tPAK2 structure. Hydrogen bonds are represented by dashed lines.

 

114

tion similar to that of PGK4. The three—dimensional main chain folding is shown in
Figure 5.6, from which it can be seen that the overall structure of the kringle is re-
lated to the close contact between the inner loop disulfide groups of Cys22—Cys63 and
CysSl—Cys75, which orient nearly perpendicular to each other and serve as a core of
the overall folding. The secondary structural elements present in the tPAK2 molecule
are listed in Table.5.4. The four antiparallel ,B—sheets and seven fi—turns observed in
the crystal structure are consistent with the NMR results [66], with the exception of
the Gly6—Tyr9 turn. In addition, one a—helix turn was found in the Thr37—Ala38—
Gln40—Asn41—Pro42—Ser43—Ala44-Gln44a—Ala44b—Leu44c segment, which contains
a Pro residue and three insertions. The distortion of the a—helix that results from

insertion of a Pro residue in this helix region was also observed in this structure.

B. Crystal Packing

The tPAK2 molecule crystallized in the monoclinic system, space group P21. As
such, a crystallographic two—fold screw axis along b resides in the crystalline struc-
ture. However, since there are three molecules in the asymmetric unit, a non—
crystallographic 3-fold screw axis nearly parallel to the crystal a—axis is also present.
As is shown in Figures 5.7 and 5.8, the three molecules in the asymmetric unit are re-
lated to each other by an approximate 31 screw axis. The most significant interaction
among the three molecules is the ligand—like binding interaction which occurs in the
lysine binding sites of the molecules (Figure 5.9). The NZ of Lys48’ of a neighbor-
ing molecule hydrogen bonds to the negatively charged Asp55 OD2 and Asp57 OD2
atoms at a distance of 2.8 A. The carbonyl oxygen of Lys48’ is 2.6 A from the amino
group of the positively charged Lys33 side chain. The other intermolecular interac-
tion found between asymmetric units is a solvent—bridge between the Asp55 OD2 and
Asn26’ N D2 of the neighboring molecule (Figure 5.10). This water molecule makes
hydrogen bonds to the Asp55 OD2 and Asn26’ ND2 at distances of 2.6 A and 2.9 A,

 

115

 

Figure 5.6. Stereoview of CA, C, and N structure of tPAK2. Disulfides are shown in
bold.

 

116

Table 5.4. Secondary structural elements of the tPAK2 structure.

 

6 Structures Type Residues

Bl Antiparallel Ser14 — Cys22
Leu15 — Ser21
Thr16 - Ala20

[3 2 A ntiparallel Pro24—Asn49
Trp25-His48a

fl 3 Antiparallel Arg52—Trp62
Asn53 — Pro6l

64 Antiparallel Cys63 — Cys75
His64 — Tyr74
Val65 — Glu73
Leu66 — Thr72

 

Reverse Turns Type Residues '

T1 Type II’ Phe3—Gly4—Asn5—Gly6
T2 Type I Gly6—Ser7—A1a8—Tyr9
T3 Type I Thrl6—Glul7-Ser18—Gly19
T4 Type I Pro24—Trp25—Ser26-Ser27
T5 Type III Ser27—Met28—Ile29—Leu30
T6 Type 11 Leu30-Ile3l—Gly32-Lys33
T7 Type I’ Lys66a—Asn67—Arg68—Arg69

a-Helix Thr37—Leu44c

 

 

 

 

 

 

 

Figure 5.7. Stereoview of the crystal packing of the three molecules in the asymmetric
unit of the tPAK2 structure. The non-crystallographic 31 screw ax is is perpendicular
to the plane. '

 

118

 

 

Figure 5.8. Stereoview of the crystal packing of the three molecules in the asymmetric
unit of tPAK2 structure; The non—crystallographic 31 screw axis is indicated.

 

119

902 55 oz 57 _ 902 55 02 57

\

“ N243 ~ N’248'
(3:7 6:

N52 33 NM 33

Figure 5.9. Stereoview showing the intermolecular interactions between Lys48’ and
Lys33, Asp55, and Asp57. Prime represents the neighboring molecule.

120

002 55 002 55

4f 4"

I I
I I

W'NDZ 26' WNW 26'

Figure 5.10. Stereoview showing the intermolecular interaction between Asn26’ and
Asp55 bridged by a water molecule. Prime represents the neighboring molecule.

121

respectively. Not surprisingly, the strong interaction along the non—crystallographic

3—fold screw axis (nearly parallel to the crystal a—axis) enhances crystal growth along

the a—axis [55] (Figure 5.11).

C. Structure of the Lysine Binding Site

The lysine binding site in the tPAK2 structure is located on the kringle surface and
is supported mainly by the inner kringle loop, which consists of central section of the
loop B and the central parts of loops C and D (Figures 1.2 and 5.12). The binding site
residues can be divided into three groups. The negatively charged residues Asp55 and
Asp57, with carboxylate oxygens 4.3 A from one another, serve as an anionic center
at one end of the binding site. An elongated depression, lined by the indole rings of
Trp62 and Trp72, provides a highly nonpolar environment for the methylene groups
of the ligand. The two indole rings are oriented in an antiparallel manner with an
interplanar angle of approximately 80° and form aromatic stacking interactions with
the rings of His64 and Tyr74, respectively. Finally, a cationic center is formed by
the side chain of the positively charged residue Lys33. The lysine binding site not
only is stabilized by the symmetric structural framework but also by a number of
inter— and intra—molecular interactions (Figure 5.13, Table 5.5). The OD2 atoms of
Asp55 and Asp57 are 2.8 A from the Lys48 NZ atom of the neighboring molecule. In
addition, OD1 atoms of Asp55 and Asp57 hydrogen bond back to the protein with
Trp62 N E1 and Tyr74 OH at distances of 3.0 A and 2.4 A, respectively. The carbonyl
oxygen of Lys48 of the adjacent molecule forms a hydrogen bond with the positively
charged Lys33 side chain at a distance of 2.6 A. The chloride ion, which is found in the
interface between two molecules, interacts with the main chain amide atoms of Val34,
Tyr35 and His64 with C1 to N distances of 3.2 A, 3.2 A, and 4.0 A, respectively. The
chloride ion not only makes these three intramolecular interactions, but also forms

an intermolecular interaction with N D2 of Asn26' of the next kringle. Two solvent—

122

\ 7 ‘. v 7‘ '65
‘91:: ‘&“‘¢‘ ’1 “"3! “V‘s;
as: .~,~;¢!.2.ov ‘3» savage .
v 'x‘ m" “ 5 ”I f “.2 ».;€).‘.‘..9V
(12°; '1???" :a Qf'zersé. s»
a 7 92‘s,.” 1‘». .5» ”9}'i...:’ . .
/“ s‘.‘ egg. 1‘5,
0’ V" 0 V! a

I

- l . .‘
. (or C
‘M ..:_f)‘ ”('é’ '
0‘ 3"
.I'i

 

 

~ 3’3"" ,‘wé’?’ ‘ £94 3" ”$7
t r «A. 9 Q i o t.
‘I’ ..' sofa -,_ “Ii"e‘r‘ ‘
\ 543""?-.g" . 737.3133"
0 . .. 0
. ‘1 $‘A‘:'s‘ ‘1 0 fi‘ ‘2‘)...5” ‘
~‘ ’lv \ ' ‘ ' '.
{saw-93,1-» ’ ~- '33}. 1!
~ . mam ‘ A» (6‘ “a ’M":1
5“? '1‘ ".3. -I «‘WDQ‘ L ‘9
ecl’n , ‘ ' "$5.0,
I'r’ :‘fll. "I.”d.
,\ J

Figure 5.11. Stereoview of the crystal packing of the three molecules along the a—axis.
Residues involved in intermolecular interactions are shown in bold.

123

 

Figure 5.12. Stereoview of the lysine binding site of the tPAK2 structure.

124

 

Figure 5.13. Intramolecular and intermolecular interactions present in the lysine
binding site of the tPAK2 structure.

125

bridged interactions were found between His64 N D1 and Thr71 O, Asp55 OD2 and
Asn26’ N D2 of the neighboring molecule. Interestingly, the two water molecules are
conserved in all three molecules of the asymmetric unit with unit occupancies and

average thermal parameters of only 16.0 A2.

Table 5.5. The intramolecular and intermolecular interactions present in the lysine
binding site of tPAK2 structure. Prime refers to neighboring molecules. Donor atom
is denoted (D), acceptor atom (A).

 

Donor Acceptor D...A (A)
Tyr74 OH Asp57 OD1 2.40
Trp62 N E1 Asp55 OD1 3.04
Lys48’ NZ Asp55 OD2 2.82
Lys48’ NZ Asp57 OD2 2.80

W386 O Asp55 OD2 2.61

 

,Asn26’ ND2 W386 O 2.92
Val34 N CL 3.18
Tyr35 N CL 3.20

His64 N E2 CL 4.00
His64 NDl W339 O 3.01

W339 O Thr71 O 2.32
His64 N E2 Tyr35 O 2.55
Lys33 NZ Lys48' O 2.55

 

 

 

 

 

N MR NOE experiments [65, 67] found that the side chains of Tyr36, Trp62, His64,
Trp62 and Trp72 were perturbed most by ligand presence. The Trp25 and Tyr74
aromatic rings were also shifted, but not to undergo direct contacts with the ligand.
In the crystal structure, the side chain of Hi364 is approximately 3.9 A from the
end of the Trp62 indole ring; thus, it appears that the imidazole ring is affected
through the aromatic stacking interaction with the side chain of Trp62. The tyrosyl
ring of Tyr74, which is 3.7 A from the edge of the indole ring of Trp72 and has a

hydrogen bond between its hydroxyl atom and the carboxylate group of Asp57, may

 

126

sense ligand—binding effects indirectly from the aromatic interaction with Trp72 or
the hydrogen—bonding interaction with Asp57. The indole ring of Trp25 is in the
layer below the surface with its end about 3.8 A from the face of the indole ring of
Trp62; thus, it may sense the presence of the ligand through an aromatic—aromatic
interaction with Trp62. Although the side chain of Tyr36 appears to be involved in
ligand-binding in the N MR N OE experiments, there is no obvious direct interaction
between Tyr36 and Lys48’ in the crystal structure (Figure 5.13). Therefore, the side
chain of Tyr36 may experience the presence of the ligand from secondary effects or
in the case of the bulkier ligands. Although the lysine binding site of the tPAK2 is
blocked by Lys48’ from the neighboring molecule, the lysine binding site structure
agrees well with the N MR observation results . Thus, the intermolecular binding
interaction mimics a ligand—like binding site interaction and might be useful as a

model for the tPAK2—fibrin interaction.

D. Comparison of the Three Molecules in the Asymmetric
Unit

The three molecules in the asymmetric unit have been compared, and the r.m.s.
deviations are listed in Table 5.6. From a superposition of the three molecules (Fig-
ures 5.14, 5.15), it can be seen that the agreement of the main—chain atoms among
the three molecules is quite good and almost all the side—chain groups have a similar
conformation, except Hisl3, Met28, and Arg68. In molecules A and C, the Hisl3 side
chain has the same conformation with X1 = —63° and X2 = -81°, while in molecule
B, X1 = —148° and X2 = —86°. Figure 5.16 shows that the Hisl3 ND2 makes a
hydrogen bond with the Tyr9 OH at a distance of 2.8 A in molecules A and C, but
not in B. This observation is consistent with the N OE results and acid/base titration

experiments [67], which found two side chain positions for Hisl3, one located on the

127

kringle surface (as in molecule B) and the other making intramolecular interactions
with other protein groups (as in molecules A and C). The residue of Met28, which is
located on the kringle surface and exposed to the bulk solvent, shows different confor-
mations in all three molecules (Figure 5.15). This difference results in the large r.m.s.
deviations of sulfur atoms (Table 5.6). Residue Arg68 shows different side—chain con-
formations in molecules A and B (Figure 5.15). Inspection of this region in the crystal
reveals that the side chains are involved in intermolecular interactions in molecules A
and B, but not in C. In molecule C, no electron density was observed for side chains

of the Asn67-Arg68—Arg69 segments which were refined as glycine residues.

Table 5.6. RMS deviations among the three molecules in the asymmetric unit. MA,
MB, MC represent molecules A, B and C. Numbers shown in parentheses were cal-
culated without including Met28.

 

 

 

 

 

 

 

MA & MB (A) MA & MC (A) MB 85 MC (A)

All protein atoms 0.83 0.49 0.68
Main chain 0.34 0.30 0.40
Carbonyl oxygens 0.45 0.42 0.53
Side chain‘s 1.16 0.63 0.88
Sulfurs (Cys, Met) 0.82 0.70 1.43

(0.54) (0.32) (0.70)
Carbon alphas 0.35 0.31 0.41

 

 

 

 

 

 

 

 

E. Comparison of the Structures of Lysine Binding Sites of

tPAK2 and K4—ACA

In order to compare the lysine binding sites of tPAK2 and K4—ACA, an optimal
superposition of the two structures based on the main chain atoms CA, C, N was

obtained. The r.m.s. deviations of residues in the lysine binding sites of the two

128

 

Figure 5.14. Stereoview of the comparison of the main—chain atoms of the three
independent molecules. ‘

 

129

 

Figure 5.15. Stereoview of the comparison of the side—chain atoms of the three inde-
pendent molecules.

 

130

 

Figure 5.16. Stereoview showing the comparison of side chains of His13 in three
molecules. Hisl3 in molecule B is indicated.

 

131

structures are listed in Table 5.7. A superposition of lysine binding sites of the tPAK2
and K4-ACA structures is shown in Figure 5.17. An examination of Figure 5.17
reveals that although the Ile31—Tyr35 segment of tPAK2 is different from that of
K4—ACA, the lysine binding site structures of the two kringles are in general similar,
with an anionic center formed by Asp55 and Asp57, an aromatic groove conformed
by the indole rings of Trp62 and Trp72, and a cationic center provided by the amino
group of Lys33 in tPAK2 or the guanidino group of Arg71 in K4-ACA. Comparison
of the main—chain structures in the lysine binding site gives a r.m.s. deviation of 1.5
A; and the difference reduces to 0.6 A when atoms with deviations greater than 10 are
removed. Not surprisingly, the r.m.s. deviation in the B loop is extremely large. This
significant structural difference results from the different number of residues of the B
loop in the two structures; in which tPAK2 has three more than PGK4. Except for
the 31—35 peptide segment, good agreement was observed between the two structures
for the rest of the lysine binding sites, especially the residues which are conserved. A
comparison of the position of Lys48’ and that of ACA shows that both are ducked in
the aromatic groove with their methylene groups making van der Waals contacts with
indole rings of Trp62 and Trp72. One point regarding the cationic centers of tPAK2
and PGK4 is noteworthy: on PGK4, Arg71 is the residue that interacts with the
carboxylate group of the ligand, whereas, in tPAK2, the arginine residue at position
71 is lacking but has the Lys33 side chain at the cationic center in the lysine binding
site. From Figure 5.17, it can be seen that the position of the amino group of Lys33
in the tPAK2 structure is virtually identical to that of the guanidinium group of the
Arg71 in the ‘PGK4 structure. Thus, it provides an interesting example of structural

but not sequential homology.

 

132

Table 5.7. Deviations of the positions of residues in the lysine binding site between

tPAK2 and K4—ACA

 

 

rms A(A) rms A(A)
residue number tPAK2 K4—ACA (main chain) (side chain)
. 31 I H 5.60 9.60
32. G R 4.30 —
33 K H 1.14 1.06
34 V Q 0.80 1.82
35 Y K 1.35 0.91
54 P P 0.57 0.59
55 D D 0.63 1.19
56 G A 0.64 —
57 D D 0.50 1.40
58 A K 0.55 1.71
61 P P 0.19 0.76
62 W W 0.16 0.44
63 ' C C 0.22 0.46
64 H F 0.49 0.69
71 T R 0.70 3.70
72 W W 0.46 0.96
73 E E 0.33 0.98
74 Y Y 0.45 0.54
75 C C 0.64 0.34

 

 

 

 

 

 

133
F_¢J7 f‘ _¢J7
C. '0"
' \ 5 ’ 4 .—...\ 5 ' ‘§4
4” "> * 4” a
35 / /,'/’/ 35 ‘ 1
G " , U
< L /( ‘ < k K ‘
' / x ' / \
\\ \ fi \\ \"
‘ 71 V: ‘ 71 ‘2

Figure 5.17. Stereoview of superimposed lysine binding sites of tPAK2 (bold) and
K4-ACA. Lys48’ is between Asp55, Asp57 and Lys33 in tPAK2; ACA is between
Asp55, Asp57 and Lys35, Arg7l in K4—ACA.

 

CHAPTER 6

Comparison of Different Kringles

In order to ascertain the structural changes of kringles, all the kringle structures that.
have been solved are compared and shown in Figure 6.1, which includes prothrombin
kringle 1 (PTKl, 2.25 A resolution) [22], prothrombin kringle 2 (PTK2, in PPACK—
thrombin complex at 3.3 A resolution), [56], plasminogen kringle 1 (PGK1, 2.38 A
resolution), plasminogen kringle 4 (PGK4, 1.9 A resolution) [23], tissue plasminogen
activator kringle 2 (tPAK2, 2.43 A resolution). From Figure 6.1, three different kinds
of kringle folding (thin, bold, and dashed lines) are observed and the largest deviation
among them is in the B and D loops. Except for these two regions the five kringles
show nearly identical conformations in their A and C loops having a r.m.s. deviation
of 0.5 A. This is consistent with the high degree of homology in the A and C loops
(40%) of kringles from different proteins (Table 1.1). As mentioned in Chapter 4, the
different folding between tPAK2 and PGK4 is due to four insertions in the former
structure, in which three are in the B loop and one is in the D loop. These add a
helical turn to the conformation of the B loop of the tPAK2 structure and change the
conformation of the loop dramatically. It is somewhat surprising that the PTK2 fold
differs markedly from those of the four kringles. It has a distorted two—turn helix at
Ala28—Ly335 and the hairpin fl—turn of the D loop pivots as a unit about 60° at Val65

and Asp71. The two—turn helix in the B loop might be induced by a close approach

134

 

135

 

Figure 6.1. Stereoview of the comparison of the CA, C, N structures of different
kringles. PGK1, PGK4, and PTKl are shown in thin lines; tPAK2 in bold; PTK2 in
dashed.

 

136

to the C—terminal helix of the B-chain of thrombin. The residues in the D loop of
PTK2 make 18 van der Waals contacts less than 4.0 A with thrombin. Since the
PTK2 structure was solved as the PTK2—PPACK—thrombin complex, the features
of the B and D loops of PTK2 could be either due to the complexation or inherent
to the kringle fold. If the former, the conformational change may be necessary to
maintain the structure of the complex. The structure of urokinase type plasminogen
activator kringle 1 (uPAKl), which was determined by N MR [57], reveals two helix
turns in the kringle. It is noteworthy that one of the two helicies corresponds to the
one observed in tPAK2 (Ser40—Gly45); the other one is in the region of Asn26-Gln33,
which corresponds to a helical turn in PTK2. Although the three kinds of kringles
have different conformations on their B and D loops, they all share a similar overall
size with approximate dimensions of 15 X 30 X 30 A.

Although the overall kringle folds are not absolutely the same for the five kringles,
the conformations of the central disulfides, which are perpendicular to each other, are
very similar, with r.m.s. deviation of 0.25 A (Figure 6.2). The disulfide cluster has
close intramolecular contacts between Cy322—Cys63 and Cys51-Cys75 and gives rise
to two approximately perpendicular anti—parallel stretches of 6—sheet (61 and 62)
which are conserved in these five kringles (Figure 6.3, Table 6.1). The conservation
of the two anti—parallel 6—sheets most likely results in a highly stabilized zone and
serves to support the overall folding.

As is described in Chapters 3, 4 and 5, each of the kringle domains with affinities
for w—aminocarboxylic acids (includes PGK1, PGK4, and tPAK2) possesses a binding
site composed of three distinct regions according to electrostatic properties. The first
is a negatively charged region containing two negatively charged aspartic acid residues.
The second is a distinct positively charged region due to one (in tPAK2) or two (in
PGK1 and PGK4) side chain amino groups. The third region separates the other two

and appears as a cleft between the oppositively charged regions. This hydrophobic

137

o ‘ \
M f 63 If 11‘ i if
.175 // 1375 ,. //'1
/ Q< // \\‘ 4; fi/\ //;\\‘
C u C ‘m\
/,,21 / , ; A
4/ J/

Figure 6.2. Stereoview of the comparison of the inner disulfide bridges of PGK1,
PGK4, tPAK2, PTKl and PTK2. Disulfide bonds in bold.

138

 

Figure 6.3. Stereoview showing the conserved anti-parallel 6—sheets located near

disulfide bridges in PGK1, PGK4, tPAK2,_PTK1, and PTK2.

139

Table 6.1. Hydrogen bonds of the conserved disulfide anti—parallel 6-sheets of PGK1,
PGK4, tPAK2, PTKl, and PTK2 structures.

 

Donor Acceptor N...O (A) N...O (A) N...O (A) N...O (A) N...O (A)

PGK1 PGK4 tPAK2 PTKl PTK2

61 16 N 20 O 2.65 2.93 2.90 2.57 2.76
61 22 N 14 O 2.83 2.98 2.93 3.01 2.88
62 63 N 73 O 3.17 3.18 3.02 2.89 2.98
62 73 N 63 O 3.44 3.17 2.94 2.93 3.05

 

 

 

 

 

region is composed of aromatic residues, one of which is a Trp that is conserved
even in non—binding kringle domains, and is the residue immediately preceding the
fourth Cys of the kringle. Although many residues of the PGK4 lysine binding site
are conserved in PTKl, it has no affinity to fibrin, lysine and analogous ligands. As
seen in Figure 6.4, Arg71, which serves as a cationic center in PGK4, is conserved
in PTKl, but is involved in an intramolecular ion pair interaction with Glu34 that
blocks one end of the lysine binding site. Though Asp55, an essential residue of the
anionic center of the PGK4 binding site, is also conserved in the PTKl structure, it
makes an ion pair interaction with Arg72 and precludes access to the site by lysine
or other w-aminocarboxylic acids. Furthermore, the substitution of the aromatic side
chains of Trp72 and Tyr74 with charged Arg and Glu residues in PTKl eliminates
the large hydrophobic surface that binds methylene groups of ligands. In the PTK2
(Figure 6.5) structure, the two negatively charged Asp55 and Asp57 residues and
one of the positive residues, Lys35 of PGK4, important for lysine binding, are also
conserved. However, Lys31-Lys35 of PTK2 is in a helical conformation that causes

Lys35 to project out into solvent region and disrupts this cationic center feature. Since

140

the structure of PTK2 was determined as the PPACK—thrombin complex, it is not
clear whether the different conformation of this loop is inherent to the kringle, as in
tPAK2 and the uPAKl, or whether it might be due to a conformational change upon
complexation. Since PTK2 is generally thought not to bind lysine, it is consistent
that it has a different native folding conformation. Further support for this inference
is the observation of the same two-turn helix in the uPAKl [57].

Nearly all the internal aromatic residues are conserved or are highly homologous in
the five different kringles (Table 1.1). For instance, Tyr50 of tPAK2, PGK1 and PGK4
is replaced with Phe in PTK1 and PTK2. Similarly, His64 of tPAK2 is replaced by Phe
in PGK4 and Tyr in PGK1, PTK1 and PTK2. Not surprisingly then, these aromatic
residues have very similar conformations in the five different kringles (Figure 6.6), with
an r.m.s. deviation of 0.45 A. As mentioned in Chapter 1, the high degree of internal
conservation around Cys22-Cys63 and Cys51—Cys75 maintains the hydrophobic core,
which appears to be important for the three—dimensional kringle folding.

Thus, four types of kringle folding are known: (1) PGK1, PGK4, PTK1; (2)
tPAK2; (3) PTK2 ; (4) uPAKl. Interestingly, the lengths of the loops of PTK1 and
tPAK2 are identical to those of PTK2 and uPAKl, respectively, however, the kringle
foldings of the former are different from those of the latter. It appears that kringles

in general may have inherent flexibility in their three—dimensional folding.

141

,,
:\ 55 \ :\ 55 ’ \
f K721,”< K") A

Figure 6.4. Stereoview comparing the lysine binding regions of PTK1 (bold) and
PGK4.

142

 

Figure 6.5. Stereoview comparing the lysine binding site regions of PTK2 (bold) and

PG K4.

143

S i [[9 S [3]]9
fa fso
62 62
54 54

@s @s

5 5
64 64

Figure 6.6. Stereoview of the comparison of internal aromatic residues in PGK1,

PGK4, tPAK2, PTK1 and PTK2.

CHAPTER 7

The Structure of A Designed
Peptidomimetic Inhibitor

Complex of a-Thrombin

7 .1 Introduction

The blood coagulation mechanism consists of a series of linked proteolytic reactions
in which zymogens are converted into trypsin—like enzymes. The activation events
take place on the surfaces of cells such as platelets, white blood cells, and endothelial
cells. These transformations are accelerated by non—enzymatic protein cofactors that
act either by altering the conformation of the zymogens or by binding converting
enzymes and zymogens in close proximity on the surface.
a—Thrombin is a trypsin—like serine proteinase that plays a key role both in blood
coagulation and in other physiological processes that involve catalytic functions and
non—enzymatic intermolecular associations [68, 69, 70]. In the penultimate step of the
blood coagulation cascade, thrombin is generated from prothrombin by limited pro-
teolysis [71] (Figure 7.1); thrombin then transforms fibrinogen to fibrin. The human

Cl'rthrombin used in this study is a two chain molecule which possesses a molecular

144

PROTHROMBIN
b b’ c

a
CHO CHO l CHO
Ala l I I Ser ‘——[ Thr Ile—'l—— Ser

 

 

 

 

 

Prothrombin Prethrombin 1
fragment 1 .[
Prothrombin Prethrombin 2
fragment 2 .l
A - chain B - chain
I: S-S —l
THROMBIN

Figure 7.1. The conversion of prothrombin to thrombin. CHO represents a carbo-
hydrate side chain; a, b, and c represent the cleavage sites of bovine and human
prothrombin; b’ represents an additional cleavage site in human prothrombin.

146

weight of approximately 36,600 [72, 73]. The A—chain consists of 36 residues linked
via a disulfide bond to a B—chain of 259 residues at residues 1 and 122, respectively.
The B—chain is highly homologous to other coagulation/fibrinolytic serine proteases
(Factor IXa, Factor Xa, protein C, urokinase, tissue plasminogen activator, and plas-
min) and contains the active site residues His57, Asp102, and Serl95. The sequence
of human a—thrombin is listed in Table 7.1 [74].

The primary function of thrombin is to convert fibrinogen to fibrin, which is the
major protein component of a blood clot and is reponsible for mechanically binding
together platelets and several plasma proteins into a network of fibrin polymers that
block the flow of blood from a severed vessel. F ibrinogen is a disulfidelinked dimer
of three peptide chains with stoichiometry (Aa,B6,7) and that contain 625, 461,
411 amino acids, respectively (Figure 7.2) [75, 76]. Fibrin assembly begins with the
cleavage of two Aa chains of fibrinogen at Argl6-Glyl7 with the subsequent release of
two molecules of fibrinopeptide A (F PA) [77]. The FPA deficient fibrinogen molecule
is known as the fibrin monomer. Fibrin monomers proceed to polymerize to fibrin
protofibrils, in which the monomers are laid end—to—end with an overlap equal to one—
half of the monomer’s length of 450 A [78], i.e. each protofibril contains one monomer
225 A long. The second stage of fibrin assembly involves the lateral association of
protofibrils to form fibers. This stage is usually accompanied by thrombin cleavage
of B6 chains of fibrinogen at Arg14—G1y15 and release of fibrinopeptide B (FPB).
The protofibrils are not able to form a blood clot until at least three additional
types of bonds are formed. First, the protofibrils must be linked to one another
through a lateral attachment site in the a chain [79, 80, 81]. At this point, the clot
consists of a loosely linked, unstable network. Second, the entire clot is stabilized by
the formation of covalent, end-to-end, antiparallel bonds between the C-terminal of
adjacent 7 chains in the protofibrils [82]. Finally, the fibrin structure is covalently

stabilized as the result of the action of Factor XIIIa (fibrin stabilizing factor), an

147

Table 7.1. The primary seuence of human a—thrombin. The thrombin residue num-
bers are assigned by homology with chymotrypsin. Insertions are represented by
alphabetic characters followed by numbers.

A Chain :

Thr 1H Phe ' 1G Gly 1F Ser 1E Gly 1D Glu 1c Ala 18
Asp 1A C ys 1 Gly 2 Len 3 Arg 4 Pro 5 Leu 6
Phe 7 Glu 8 Lys 9 Lys 10 Ser 1 1 Len 1 2 Glu 13
Asp 14 Lys 14A Thr 14 B Gln 14C Arg 14D Gln 14E Len 14F
Len 14G Glu 14H Ser 141 Tyr 14.1 [1e 14K Asp 14L Gly 14M
Arg 15

B Chain :

He 16 Val 17 Glu 18 Gly 19 Ser 20 Asp 21 Ala 22
Glu 23 [le 24 Gly 25 Met 26 Set 27 Pro 28 Trp 29
Glu 30 Val 31 Met 32 Len 33 Phe 34 Arg 35 Lys 36
Ser 36A Pro 37 Gln 38 Glu 39 Len 40 Len 41 Cys 42
Gly 43 Ala 44 Ser 45 Len 46 lie 47 Ser 48 Asp 49
Arg 50 Tip 51 Val 52 Len 53 Thr 54 Ala 55 Ala 56
His 57 C ys 58 Len 59 Len 60 ‘ Tyr 60A Pro 608 Pro 60C
Trp 60D Asp 60E Lys 60F Asn 60G Phe 60H Thr 60I Glu 61
Asn 62 Asp 63 Len 64 Len 65 Val 66 Arg 67 He 68
Gly 69 Lys 70 His 71 Set 72 . Arg 73 Thr 74 Arg 75
Tyr 76 Glu 77 Arg 77A Asn 78 He 79 Glu 80 Lys 81
He 82 Ser 83 Met. 84 Len 85 Glu 86 Lys 87 He 88
Tyr 89 He 90 His 91 Pro 92 Arg 93 Tyr 94 Asn 95
Trp 96 Arg 97 Gln 97A Asn 98 Len 99 Asp 100 Arg 101
Asp 102 Ile 103 Ala 104 Leu 105 Met 106 Lys 107 Leu 108
Lys 109 Lys 110 Pro 111 Val 112 Ala 113 Phe 114 Ser 115
Asp 116 Tyr 117 11¢ 118 His 119 Pro 120 v.1 121 Cys 122
Leu 123 Pro 124 Asp 1 25 Arg 1 26 Glu 127 Thr 128 Ala 129

Ala 129A Ser 1298 Len 129C Len 130 Gln 131 Ala 132 Gly 133
Tyr 134 Lys 135 Gly 136 Arg 137 Val 138 Thr 139 Gly 140

Tip 1 4 1 Gly 1 42 Asn 1 43 Len 1 44 Lys 145 Gln 1 46 Thr 147
Trp 1 48 Thr 149 Ala 1 49A Asn 1 1 493 Val 1 49C Gly 149D Lys 1 4913
Gly 150 Gln 151 Pro 152 Set 153 Val 154 Leu 155 Gln 156
Val 157 Val 1 58 Asn l 59 Len 1 60 Pro 1 61 Be 162 Val 163
G In 1 64 Arg 1 65 Pro 166 Val 167 C ys 168 Lys 169 Asp 1 70
Ser 171 Thr 172 Arg 173 He 174 Arg 175 Be 176 The l 77
Asp ‘ 1 78 Asn 1 79 Met 180 Phe 181 C ya 1 82 Ala 183 Gly 184

Tyr 184A Lys 185 Pro 186 Asp 186A Gln 186B Gly 1860 Lys 186D
Arg 187 Gly 188 Asp 189 Ala 190 C ys 191 Gln 192 Gly 193
Asp 194 Ser 1 95 Gly 196 Gly 197 Pro 198 Phe 199 Val 200

Met 201 Lys 202 Ser 203 Pro 204 Phe 204A Asn 204B Asn 205
Arg 206 'Ik'p 207 Tyr 208 Gln 209 Met 210 Gly 211 He 21 2
Val 213 Set 214 "hp 215 Gly 216 Glu 217 — 218 Gly 219
Cys 220 Asp 221 Arg 221A Asp 222 Gly 223 Lys 224 Tyr 225
Gly 226 Phe 227 Tyr 228 The 229 His 230 Val 231 Phe 232
Arg 233 Len 234 Lys 235 Lys 236 Trp 237 Ile 238 Gln 239
Lys 240 Val 241 Be 242 Asp 243 Gln 244 Phe 245 Gly 246

Gln 247

148

 

 

    
 

._ ' 1501 CENTRAL 1501
TERMIfiSL , ml DOMAIN ; _1 TERMINAL
00““ (MW 32.6001 Dow‘m
«saw 57 :00) (MW 57 200:
5011 FPA 50‘ sol
FPS
CHO in) SS RING / \
.ww 2500) SS RING .- ‘O O- . .
. ‘ / 7H. 7.,
x. %‘ / 'oo 7 ' “ "K
' , ’ morass (71 250A 0 ‘7‘ ..
m ' (SENSITIVE Masooi ‘ -
SITES COILED COIL
mw 39400)
an.
(US)
' ‘\
W In POLAR
. mace
XL ‘. (mason:
«mm

Figure 7.2. The structure of the fibrinogen molecule. The symmetric molecule is
composed of a dimeric central domain containing the N —teminal of all six chains (0:, 6,
‘7), two connecting coiled coils, two terminal domains, and two Aa polar appendages.
Four carbohydrate clusters (CHO) occur, and are located on each 7 chain near the
central domain and on the 6 chains of each terminal domain. Primary cross-linking
sites (XL) can be found near the C-terminal of the ‘7 chain and in the A01 polar
appendages (taken from [76]).

149

enzyme which introduces bonds between 6 amino group of lysine and the 7 carboxy
group of glutamine.

Various alternate methods [83, 84, 85, 86, 87] have been employed in the inves-
tigation of the mechanism of the interaction of thrombin with fibrinogen, especially
about the cleavage of the Arg—Gly peptide bond in the A01 chain of human fibrino-
gen. These studies have shown that the first six residues of FPA do not interact
with thrombin, whereas Asp7 and Phe8, which are located 10 and 9 residues away,
respectively, from the thrombin cleavage site, influence the effectivness of the binding
of synthesis peptide substrates to thrombin. Since Asp7 and Phe8 are relatively far
in sequence from the scissile bond at Arg16, several investigations have proposed that
the N —terminal of F PA binds to thrombin in a bent configuration with Phe8 close to
Arg16 [88, 89]. Through N MR results [90], it has been shown that residues Asp7,
Phe8, Leu9, Vall5, and Arg16 are involved in the interaction of the peptide with
thrombin. Furthermore, transfer N OE measurements [90] indicated that a 6—bend
might exist within the segment from Gly12 to Va115 in the FPA—thrombin complex.

In order to examine the role that this reverse turn plays in the thrombin active
site, a chloromethyl ketone inhibitor mimetic of FPA (FPAM) has been designed and
synthesized as shown in Figure 7.3 [91]. A model for the thrombin—bound structure
of F PAM has been proposed (Figure 7.4) [92] based on the crystallographic throm-
bin structure [93, 43], N MR data [94], computer assisted molecular modeling and
peptidomimetic substrates and inhibitors [91]. Subsequently, the X-ray crystal struc-
tures of human F PA bound to bovine thrombin [95] and a chloromethyl ketone FPA
derived thrombin complex [96] were determined; some variance between these and the
predicted FPAM structures were observed. In order to obtain a better understand-
ing of the interplay between the primary sequence and the conformation required for
thrombin substrates and inhibitors, a crystallographic investigation of the previously

mOdeled mimetic FPAM complexed with human oz—thrombin was undertaken [97].

OHO

Ph NAPh P9113, DEAD / ,Ph
0112012 T
mvph
\
1)Rucla.Na|04 M ,9"
2) CH .510
2N2 2 .. VP"

 

 

Cl

NH O

H NH

O=$
HN <
1) EDC. HOBT, DMAP

51311. 4:1 DMF:H20 8:0

2) H2N-Va1-Argcrosycnzc1 NH

 

v NHBO 00=<

3) Anhydrous HF P"...

 

Figure 7.3. Scheme for synthesis of FPAM. DEAD, diethyl-azodicarboxylate;
EDC, 1—ethyl—3 (3—dimethylaminopropyl) carbodiimide; HOBT, hydroxybentriazole;
DMAP, dimethylaminopyridine; DMF, dimethylformamide.

151

. 195

Figure 7.4. Stereoview of the FPAM structure docked in the thrombin active site.
FPAM is in bold.

152
7 .2 Experimental Procedures

A. Crystallization

The FPAM—thrombin—hirugen (N—acetylhirudin 53’—64' with sulfato—Tyr63’) complex
was crystallized in a similar way to that of the hirugen—thrombin complex [98]. An
approximately 10 fold molar excess of hirugen was added to a frozen 1 ml sample of
thrombin solution (2.6mg/ml) at 4°C. The solution was then diluted to 2 ml with 0.1
M phosphate buffer at pH 7.3 and 225113 of FPAM chloromethyl ketone solution at
a concentation of 2.5 mg/ ml in methanol (10 molar excess) was added. The solution
of the ternary complex was concentrated to about 5 mg/ml using a Centricon 10
miniconcentrator (MW cutoff 10K) in a refrigerated centrifuge. Crystallization was
carried out in 10,118 hanging drops against 1 ml of well solution containing 0.1 M
sodium phosphate buffer (pH 7.3), 28% PEG 8000. Autolysis was prevented by the
hirugen in crystals grown in this manner (Figure 7.5). The crystals were found to be
isomorphous to those of the hirugen thrombin complex : monoclinic, space group C2,
four molecules per unit cell, a=71.13 A, b=72.43 A, c=73.00 A, 6 = 101.09° with an

estimated protein content of 50% and VHF—2.5 A3/ dalton.

B. Data Collection

X-ray diffraction intensities were measured at 2.5 A resolution from one crystal having
dimensions 0.40 x 0.27 x 0.10 mm employing a Siemens multiwire area detector with
graphite monochromated CuKa radiation from a Rigaku RU200 rotating anode tube
Operating at 50KV and 150mA. The crystal—detector distance was 11.65 cm, the
detector swing angle was set at 12°, the scan angle was 0.2° per frame of measurement
and each frame was measured for 90 seconds. The raw data which were reduced and

scaled with the XENGEN programs [26], yielded 37,533 reflections of which 11,675

153

 

 

Figure 7.5. Photograph of FPAM crystal. Crystal size is approximately 0.5 x 0.45 x
0.1 mm .

154

were unique. The distribution of intensities observed for various resolution ranges
is given in Table 7.2. After removing reflections with I/0(I) < 2, a set containing

10,918 independent reflections remained (93% observed. Rmerge = 0.039).

Table 7.2. Distribution of reflection intensities and R-factors in various resolution
shells.

 

Res. (A) #refs 0 <20 <50 <100 <20a <400 <600 >600

 

4.54 2155 0 5 32 38 84 259 329 1408
3.60 2122 9 25 55 69 172 425 381 986
3.15 2109 24 85 133 203 374 624 378 288
2.86 2107 48 193 297 390 517 510 130 22
2.66 2112 38 245 496 487 517 293 35 1
2.50 1070 28 160 374 299 179 28 2 0

Totals: 11675 147 713 1387 1486 1843 2139 1255 2705

 

 

 

 

 

C. Refinement

The F PAM—thrombin crystal structure was solved using isomorphous thrombin co-
ordinates of the hirugen-thrombin complex [98]. Since no electron density was found
for the autolysis insertion loop from Thr147-Lysl49E in other isomorphous hirugen
complexes, the initial model only included the A chain from ThrlH to Arg15, and
Ile16-Glu146 and Gly150—Glu247 of the B chain of thrombin. The structure was
refined employing the restrained least-squares method implemented in the program

PROLSQ [30] with intermittent model building performed on an Evans and Suther-

155

land P5390 interactive stereographics system with the program FRODO [32]. The
refinement proceeded in two stages, data from 7.0—2.8 A resolution were included
initially, then data from 7.0—2.5 A resolution. Each major round of refinement was
followed by model building; (2|Fol — [Fc[) and (IFOI - [Fc|) electron density maps
were used in conjunction with the Ramachandran plot. The R value started at 28%
with hirugen not considered in the calculation and with an overall thermal param-
eter of 25.0 A2. The first (2]Fol — [Fc|) Electron density map at 2.8 A resolution
showed good density for most of the thrombin—hirugen residues and Val5—Arg16 of
the FPAM (Figure 7.6). The hirugen and FPAM were fitted into the density and
gradually included throughout further calculations. The special aromatic groups in
F PAM, not being regular amino acids, did not have normal peptide bonds connecting
to the bicyclic ring. In order for the PROLSQ program to recognize and refine the
aromatic rings, the dictionary had to be modified and some additional restraints were
applied in the control file (Figure 7.6). The R value decreased to 19.2% after the first
refinement stage (2.8 A resolution) and water molecules were located at 2.5 A reso-
lution. Peaks considered to be possible water molecules were identified by comparing
(7.0—2.5) A and (8.0-2.5) A resolution difference maps. In addition, the hirugen and
the FPAM positions were updated according to the electron density maps. The final
F PAM—thrombin—hirugen structure has a crystallographic R value of 13.8% for 10,139
reflections between 7.0 and 2.5 A resolution with 234 water molecules and an average
thermal parameter of 29 A2. The average occupancy of the water molecules is about
0.67 and their average thermal parameter is 29 A2. The final reflection weights and R
values in each range are given in Table 7.3, and a summary of refinement parameters
is listed in Table 7.4. The distribution of main-chain torsion angles (113,16) is shown
in Figure 7.7; nearly all the non-glycine amino acids fall within or close to confor-
mationally allowed regions (except GlulC and SerlE of the A-chain). A cis peptide

bond occurs at Pro37 with an an angle of 077°. The to angles of the remaining peptide

156

bonds are in a narrow range close to planarity.

7.3 RESULTS

A. Structure of Thrombin

Nearly the entire structure of thrombin is well defined; however, as in other hirugen—
thrombin complexes, little or no electron density was found for N-terminal residues
ThrlH—GlulC of the A chain and C—terminal residues Phe245—Glu247 of the B chain.
In the present structure, no electron density was observed for the sidechain atoms of
Asn62, LysllO, Gln151 and the sometimes important Glu192 residue. The structure
of thrombin in the ternary complex was compared with that of the hirugen—thrombin
complex by the optimal superposition of CA, C, and N atoms; agreement between
the two is excellent(Table 7.5). The residues in the active sites of the two complexes
have been compared in detail and a stereoview of the superposition of the active
site regions is shown in Figure 7.8. The only significant change in the active site
induced by the binding of F PAM to thrombin is associated with Trp60D. Another
large deviation in the region that occurs at the sidechain of Ile174 results in the
sidechain pointing toward the face of the phenyl ring of Phal of FPAM overall, the
residues in the catalytic site have practically the same conformations as those in
the hirugen—thrombin complex [98], where the active site is unoccupied. Thus, the
conformation of the active site of thrombin is conserved and binding of FPAM or
D—Phe—Pro—Arg—chloromethylketone (PPACK) does not induce much change in the

region.

 

157

C1
NH
A 0

H2 N N Arg 16

H NH

0
Val 15
PIN)

Ben 2 I \ E/O Gly 14

- NH

CM\ "{..EQIK‘C/c.2\ ,C..
N a {[1
i' [ Rng3
' N

—> bond distance

........ angle distance
0 planar 1,—4 distance

Figure 7.6. Numbering of and special restraints used for FPAM in restrained least
squares refinement. Since part of F PAM is a non—amino acid group, special bond,
angle and planar 1,-4 distance restraints were applied during refinement to maintain

geometry.

158

Table 7.3. Weights of reflections and R values of the final refinement

 

 

R value
dmin(A) no. reflections 0(IFI) < ||Fo| — |Fc|| > shell sphere
4.60 1388 39 80 0.146 0.146
3.85 1406 . 34 64 0.111 0.128
3.42 1423 31 56 0.118 0.125
3.13 1387 28 51 0.141 0.128
2.90 1482 26 44 0.157 0.132
2.72 1521 23 41 0.165 0.135
2.40 1532 21 37 0.163 0.138

 

 

 

00110 = 26 —1801(sin om — (1/611; (llFol -- chH) = 53

 

 

 

 

159

Table 7.4. Final least squares parameters and deviations of FPAM—thrombin.

 

 

 

 

 

 

 

 

 

[] ] target a [rms 6]]
Distances (A)
Bond distance 0.020 0.016
Angle distance 0.030 0.044
Planar 1,-4 distance 0.050 0.049
N on-bonded distances (A)
Single torsion 0.50 0.23
Multiple torsion 0.50 0.32
Possible H-bond 0.50 0.31
Torsion angles (deg)
Planar 3 2
Staggered 15 24
Orthonormal 20 31
Plane groups (A) 0.02 0.01
Chiral centers (A3) 0.15 0.18
Thermal restraints ( 12)
Main chain bond 1.5 1.1
Main chain angle 2.0 1.9
Side chain bond 2.5 1.7
Side chain angle 2.5 2.6

 

 

 

 

 

 

 

160

 

 

 

 

 

 

 

 

 

 

 

 

 

-180 —90 0 90 180
1 1 1
180 ,1 1 1- . 1\ 1 1 . 1 1 1
\
\ _
31'
1
1 __
1 , *
/
90 *1 / I *SerlE *—
‘ // I It I
r l | _
1 I I.
l | ' 1“.
‘
1 L 1 ~—
\
"' \ a ‘ T *K \1
m 0 \ 7* a 4 i\
Q‘ \ at **&h \\
\ '11 1., ‘2} at 1
_ a *a ‘7'”: 1 _
[\\/ .. an I 1
7 Pf. an 2 a
—1 H ' 1““ Ila” ** 1 _
____at_*_._.____l
a u
-90 _ __
-~ *Glu 1C ~—
.11
—180 ryi___——*——1
1 1 1 1 1 7 1 1 1 1
-180 -90 0 90 180

Figure 7.7. Ramachandran plot of <13, 111 angles of F PAM—thrombin structure. Glycines
are not displayed.

180

-180

 

161

  

C 1191
‘ 1 87%215 , ‘ 174 215 ,

2%

 

Figure 7.8. Stereoview of the comparison of the active sites of thrombin in the FPAM
and hirugen complexes. Hirugen-thrombin, broken lines; active site is unoccupied.

162

Table 7.5. RMS Deviations Between the Hirugen—Thrombin and FPAM—Hirugen—

Thrombin Complexes

 

 

 

 

 

 

 

 

 

 

A (A) atom no.
All protein atoms 0.54 1972
Main chain 0.28 738
Carbonyl oxygens 0.35 246
Side chains 0.70 988
Sulfurs (Cys, Met) 0.36 14
Carbon alphas 0.29 246

 

 

B. Structure of FPAM

The F PAM structure is nearly completely defined by electron density in the active
site of the thrombin (Figure 7.9). However, the two phenyl rings lack continuity to
the aromatic groups. The two phenyl rings are about 7.8 A from each other in the
apolar binding site region of thrombin (F igure.7.10), and each of them interacts with
thrombin through a number of hydrophobic contacts. The conformation of Arg16
to Rng3 of FPAM (Figure 7.6) is very similar to that of PPACK in the PPACK—
thrombin complex (Figure 7.11) [99], with the former making an approximate helical
turn between Gly14—Arg16, which is followed by the mimetic 6—bend of the bicyclic
ring. There are some minor differences in the main chain positions between the two
structures in the helical-like turn. The main chain nitrogen atoms of Gly14 and Arg16
make a. two strand antiparallel 6-sheet with the thrombin Ser214—Gly216 segment
(Figure 7.10) so this region, like that of PPACK, also possesses very favorable inter-
actions. Although the electron density of F PAM is generally quite good (Figure 7.9),
the mimetic has an average thermal parameter of 49 A2, which is almost twice that of
thrombin (28 A2). A similar value was observed for hirudin in the hirudin-thrombin
complex [43], where the difference was attributed to imprecision in positioning of the

inhibitor.

 

163

   

’5‘;
'V' \\\\\‘
.‘X‘Vs‘ 3“?»
‘ “ ‘5“ "V“:
\\. 30'.»" A\?«\‘ \A
‘i:.d§" \ .. ‘\

 
      
   
 
   
     
      
  
 

  
 

   
 

  

s 2 "(£93 3944‘. _
wavy any 3:;
‘ 0‘ K 3“» o‘r‘.
:5‘ r"§:’ \ O‘o" ‘ ’Q‘.
. " D'A‘1r.\:' $9" ‘ F‘
‘2»“14 «an «

 

  

Figure 7.9. Stereoview of the electron density corresponding to F PAM in the thrombin
complex. Basket contour at 10.

164

97 97
99 195 99 195

2,174 21 {1,1396 2174 2 £33516

Figure 7.10. Stereoview of FPAM bound in active site of thrombin. F PAM in bold;
hydrogen bonds, broken.

165

 

Figure 7.11. Stereoview of the comparison of FPAM (bold) and PPACK in their

thrombin complexes.

166

C. FPAM—Thrombin Interaction

The active site of thrombin, which displays a preference for arginyl and lysyl
sidechains, has the form of an elongated channel and is defined by peptide segments
Tyr60A-Trp60D, Arg97—Leu99, Thr172—Arg175, Cylel-Gly196 and Ser214—Glu217
(Figure 7.10). The density observed for the catalytic triad indicates an intermediate
hemiketal is formed between the carbonyl group of Arg16 and Ser195 OG (2.2 A) (Fig-
ure 7.10). Both FPAM and PPACK being chloromethylketone derivatives have the
same contacts at this region. The 31 specificity pocket of the F PAM complex is occu-
pied with an arginyl group with geometry similar to that of arginine in PPACK— [99],
hirulog 1— [98], and 6—homoarginine in hirulog 3—thrombin [100]: the guanidinium
group of the Arg16 forms a doubly hydrogen bonded salt—bridge with the carboxyl
oxygens of Asp189 (2.5 A and 3.1 A)(Figure 7.10). Moreover, the guanidinium group
makes a close contact with the carbonyl oxygen of Gly219 (3.1 A) and the main
.chain nitrogen atom of Arg16 may be involved in a hydrogen bond with the carbonyl
oxygen of Ser214 (3.1 A). The sidechain of Val5 in the S2 subsite is buried within
a hydrophobic cage that is the apolar binding site of the thrombin (Figure 7.10).
The valyl group makes hydrophobic contacts with sidechains of Tyr60A, Trp60D and
Leu99 (3.7 A, 3.8 A, 4.3 A respectively) and occupies a spatial region similar to
Pro of PPACK in PPACK—thrombin complex [99]. The P3 interaction observed in
this structure is due primarily to an anti—parallel 6—sheet hydrogen bond between the
amide nitrogen of G1y14 and the carbonyl group of Gly2l6 (2.4 A) that appears to
be important in positioning the bicyclic ring and is different from that in PPACK-
thrombin. The bicyclic ring corresponding to a 6—bend, which was presumed to be
at the 11-12 position of F PA [94], has a (S,S) conformation according to the chirality
at carbon atoms CH and CB (Figure 7.6) and is located in the region bordered by

residues of Tyr60A, Trp60D, Leu99, Trp215 and Glu217 (Figure 7.10). The ring, al-

167

though not aromatic, forms an end—to—face contact with the indole ring of Trp215 and
produces an aromatic-like interaction that is common in proteins [45]. In addition
to the stacking interaction, the bicyclic ring is also stabilized by other hydrophobic
contacts formed with sidechains of Tyr60A, Trp60D and Leu99. The N-terminal of
the FPAM concludes with two phenyl groups that are also located near the non-polar
region of the S2, S3 subsites. The Phal ring makes a good van der Waals contact
with the sidechain of Ilel74 (2.6 A) and shows density for its phenyl ring. However,
Ben2 on the other side has no significant interaction with thrombin except for loose
hydrophobic contacts with residues of Tyr60A and Pro60C, and is most likely the

reason why the density is not as well-defined as that of Phal (Figures 7.9 and 7.10).

D. Comparison of FPAM Related Structures

There are five FPA derived structures that are relevant in any comprehensive com-
parison of thrombin in its bound state: (1) the FPA bovine thrombin complex [95],
(2) the chloromethyl ketone of F PA alkylating His57 of the active site [96], (3) the
solution NMR structure of F PA bound to thrombin [94], (4) the present FPAM—
thrombin structure and (5) the modeled FPAM—thrombin complex structure [92].
The structures of F PA in (1) and (2) are practically identical with rmsA=0.6 A
and only 0.3 A if side chains of Leu9 and Glull are omitted (Figure 7.12). The
<15, 11), conformational angles of the P1-P2—P3 residues of (1), (2), (3), and (5) are
all in close agreement with the conformation of PPACK-thrombin, except for 1/23 of
PPACK which is negative due to the D-Phe enantiomer [99]. In addition, the con-
formational angles of the Pl—P2—P3 residues of (1), (2), (3), and (5) are also similar
to that of Prol3-Cysl4—Ly315 of bovine pancreatic trypsin inhibitor (BPTI) bound
in the BPTI-trypsin complex [101]. The P1—P2-P3 residues of F PA of the NMR
structure bound to thrombin [94] are not in agreement with the foregoing, especially

with respect to the 162 and 663, 163 angles. This lack of agreement was first noticed

168

Figure 7.12. Stereoview of the comparison of FPA (bold) and FPA-chloromethyl
ketone in their thrombin complexes.

169

when attempts to dock and model the N MR structure in the active site of thrombin

by placing Arg16 in the 31 subsite, failed because of massive collisions of FPA with

the enzyme [92]. When the BPTI trypsin bound structure was used as a template to
reorient the P1—P2—P3 residues of the N MR structure, an excellent fit of these in the
active site of thrombin was achieved. In the same work, similar rationalizations were
employed to model FPAM in the active site.

The structure of FPAM bound to thrombin is compared with that of FPA in
the thrombin complex [95] in Figure 7.13, from which it will be seen that the first
generation F PA mimetic does not correspond to FPA in two important aspects. The
first aspect is that a peptide insertion that follows the bicyclic system could place

the 6—turn of the mimetic more optimally with respect to the turn of FPA. The

initial positioning of the turn in FPAM was based on the NMR position (between

Gly11 and Va115). The second aspect is in the conformation of the Phal moiety:

rather than reversing to interact with Va115 as in FPA, the FPAM molecule assumes

a more or less extended conformation in the thrombin complex (Figures 7.10 and

7.13). Both of these shortcomings can be easily rectified with some additional design

features and synthesis (insertion of a peptide, reconforming the bicyclic system).

Another notable difference from the F PA complex is that Gly14 N of FPAM makes a

hydrogen bond with Gly216 0 thereby altering the conformation somewhat between

the P2fP3 positions. The most important aspect that emerges, however, is that the

F PAM—thrombin complex displays yet another binding mode in the active site of

thrombin [102] bringing the total to four. The other three are: (1) FPA/substrate—
Iike, (2) N -——terrninal hirudin—like [93, 43] and (3) argatroban—like [103].

170

 

Figure 7. 1 3. Stereoview of the comparison of F PAM (bold) and FPA in their thrombin

complexes .

171
7 .4 Discussion

A hallmark of the enzyme thrombin is its remarkable specificity. This is due to
the inherently deep and constricted binding site of thrombin and a requirement of
its substrates to adopt a specific conformation to productively bind in the active
site. Limited proteolysis by trypsin—like serine proteases plays an important role
in coagulation/hemostatis [104] and complements activation [105]. The potential
to specifically intervene in these processes to ameliorate a number of disease states
is significant and well recognized. However, controlled and selective interference is
critical to the success of the strategy that is made considerably more difficult by
the high degree of sequence homology within this family of proteases, which generally
contain a trypsin-like core with insertions which modify specificity and are responsible

for interaction with additional macromolecular components [106].
Of the 181 Arg/Lys—Xaa sequences in fibrinogen [107], only two bonds are cleaved
by thrombin. Experimental rationalization of this was provided through N MR investi-
gations of the complex between FPA and bovine thrombin [94]. A striking feature that
emerged in this study is the cluster of nonpolar residues (Phe8, Leu9, and Vall5) that
were apparently brought into close proximity by a reverse turn. Reverse turns have
been implicated in enhancing the specificity of proteolytic processing of prohormones,
zymogens and viral proteins [108, 109, 110]. The role of secondary structural elements
in proteolysis has been investigated through the incorporation of peptidomemtic pros-
thetic units [91 , 111, 112] and recently a model was proposed for the bound structure
of F PA [92]. The most critical feature of the model involved the reorientation of the
P1 to P3 residues of the NMR derived FPA structure to coincide in alignment with
the active site conformation of BPTI, which is believed to represent a canonical loop
proteolysis substrate mimic [101, 113, 114]. This proved to be very effective in that

the reoriented FPA model then fit well into the thrombin active site and satisfied all

172

of the previously reported NOE data [94]. Hybrid mimetic substrates based on the
foregoing were found to have similar kinetic parameters to those of FPA1_52, and thus
believed to effectively mimic the bound conformation of natural substrates.

We anticipated that because the FPAM inhibitor was designed around a specific
natural substrate for thrombin, it would exhibit a high degree of selectivity. This
is indeed the case, in that FPAM exhibits a degree of specificity similar to that of
PPACK, a well known selective inhibitor of thrombin. Utilization of the canonical
loop motif [113] of a natural proteinaceous inhibitor as a lead for designing reverse turn
peptidomimetic inhibitors may provide a general strategy for introducing specificity
into an inhibitor.

Although the model for the bound structure of F PAM is consistent with the ob-
served crystallographic structure [95], particularly in the orientation of the P1 to P3
sites, not surprisingly there are some significant differences. Most striking is the fact
that the hydrophobic pocket, formed by the 60 insertion loop of thrombin and residues
Leu99, Ilel74 and Trp215, is not fully occupied by the N—benzyl group back—tracking
to Va115 as anticipated, but rather by the bicyclic 6—turn template in a similar man-
ner to the dansyl group of DAPA—thrombin [102] and related molecules [103]. This is
due to the relatively extended structure of F PAM and is yet another example of the
intriguing dichotomy that exists between the specificity and promiscuity of throm-
bin [102]. A peptide insertion between Gly14 and Rng3 (Figure 7.6) and an alternate
conformation or stereochemistry in the C—7 ring of the bicyclic 6—turn prosthetic
unit could conceivably match the bound F PA conformation with considerable fidelity
(Figure 7.13).

Information garnered from these investigations is being utilizied in the design and
synthesis of novel nonpeptidic thrombin inhibitors by Kahn and his collaborators.
This stepwise process, starting with natural protease substrates or inhibitors and

culiminating in truly nonpeptide inhibitors will generate new structures that should

 

173

maintain the specificity that nature has so elegantly and carefully crafted.

APPENDICES

APPENDIX A

Amino Acid Shorthand Used in
the Thesis.

 

 

 

 

Amino Acid Three-letter one-letter
abbreviation symbol
Alanine Ala A
Arginine Arg R
Asparagine Asn N
Aspartic acid Asp D
Cysteine Cys C
Glutamine Gln Q
Glutamic acid Glu E
’ Glycine Gly G
Histidine His H
Isoleucine Ile I
Leucine Leu L
Lysine Lys K
Methionine Met M
Phenylalanine Phe F
Proline Pro P
Serine Ser S
Threonine Thr T
Tryptophan Trp W
Tyrosine Tyr Y
Valine Val V
174

 

BIBLIOGRAPHY

BIBLIOGRAPHY

[1] B. Wiman and D. Collen Nature, vol. 272, pp. 549—550, 1978.
[2] B. Wiman and P. Wallen Thromb. Res, vol. 1, pp. 213—222, 1977.

[3] G. Markus, J. L. DePasquale, and F. C. Wissler J. Biol. Chem., vol. 253,
pp. 727—732, 1978.

[4] G. Markus, R. L. Priore, and F. C. Wissler J. Biol. Chem., vol. 254, pp. 1211—
1216, 1979.

[5] G. Markus, J. L. Evers, and G. H. Hobika J. Biol. Chem., vol. 253, pp. 733—739,
1978.

[6] S. Magnusson, T. E. Petersen, L. Sottrup-Jensen, and H. Claeys in Proteases
and Biological Control (E. Reich, D. B. Rifkin, and E. ShawD, eds.), pp. 123—
149, Cold Spring Harbor Laboratory, Cold Spring, NY, 1975.

[7] L. Sottrup-Jensen, H. Claeys, M. Zajdal, T. E. Peetersen, and S. Magnusson
Prog. Chem. Fibrinolysis Thrombolysis, vol. 3, pp. 191—209, 1978.

[8] C. T. Esmon and C. M. Jackson J. Biol. Chem., vol. 249, pp. 7791—7797, 1974.

[9] P. G. Lerch, E. E. Rickli, W. Lergier, and D. Gillessen Eur.J.Biochem., vol. 107,
pp. 7—13, 1980.

[10] L. Patthy Cell, vol. 41, pp. 756—663, 1985.

[11] G. J. Steffens, W. A. Gunzler, F. Otting, E. Frankus, and L. Flohe Hoppe—
Seyler’s Z. Physiol. Chem., vol. 363, pp. 1043-1058, 1982.

[12] B. A. McMullen and K. Fujikawa J.Biol.Chem., vol. 260, pp. 5328-5340, 1985.

[13] D. Pennica, W. E. Holmes, W. J. Kogr, R. N. Harkins, G. A. Vehar, C. A. Ward,
W. F. Bennett, E. Yelverton, P. H. Seeberg, H. L. Hrynober, D. V. Goeddel,
and D. Collen Nature, vol. 301, pp. 214—221, 1983.

175

176

[14] T. Nakaamura, T. Nishizawa, M. Hagiya, T. Seki, M. Shimonishi, A. Sugimura,
K. Tashiro, and S. Shimizu Nature, vol. 342, pp. 440—443, 1989.

[15] J. W. Mclean, J. E. Tomlison, W.-J. K., D. L. Eaton, E. Y. Chen, G. M. Fless,
A. M. Scanu, and R. M. Lawn Nature, vol. 330, pp. 132-137, 1987.

[16] A. Varadi and L. Patthy Biochem. Biophys. Res. Commun., vol. 103, pp. 97—
102,1981.

[17] M. Trexler, Z. Valli, and L. Patthy J. Biol. Chem., vol. 257, pp. 7401—7406,
1982.

[18] A. J. van Zonneveld, H. Veerman, and H. Pannekoek Proc.Nat.Acad.Sci. U.S.A.,
vol. 86, pp. 4670—4674, 1986.

[19] E. S. Winn, S. P. Hu, S. M. Hochschwender, and R. A. Laursen Eur.J.Biochem.,
vol. 104, pp. 579-586, 1980.

[20] C. L. Young, W. C. Barker, C. M. Tomaselli, and M. O. Dayhoff in Atlas
of Protein Sequece and Structure (M. O. Dayhoff, ed.), pp. 73—93, National
Biomedical Research Foundation, Silver Spring, Maryland, 1978.

[21] A. Tulinsky, C. H. Park, and E. Jankun J. Mol. Biol, vol. 202, pp. 885—901,
1988.

[22] T. P. Seshadri, A. Tulinsky, E. Skrzypczak-Jankun, and C. H. Park J. Mol.
Biol, vol. 220, pp. 481-494, 1991.

[23] A. M. Mulichak, A. Tulinsky, and K. G. Ravichrandran Biochemistry, vol. 30,
pp. 10576—10588, 1991.

[24] H. W. Wyckoff, M. Doscher, D. Tsernoglou, T. Inagami, L. N. Johnson, K. D.
Hardman, N. M. Allewell, D. M. Kelly, and F. M. Richards J.Mol.Biol., vol. 27,
pp. 563—578, 1967.

[25] A. C. T. North, D. C. Philips, and F. S. Mathews Acta Cryst., vol. 24A, pp. 351—
359, 1968.

[26] A. J. Howard, G. L. Gilliland, B. C.‘ Finzel, T. L. Poulous, D. H. Olhendorf,
and F. R. Salemme J. Applied. Cryst., vol. 20, pp. 383—387, 1987.

[27] M. G. Rossmann and D. M. Blow Acta Cryst., vol. 15, pp. 24—31, 1962.

177

[28] W. Steigemann, Dissertation. PhD thesis, Technische Unicersitat, Munchen,
1974.

[29] M. Fujinaga and R. J. Read J.Appl.Cryst., vol. 20, pp. 517-521, 1987.

[30] W. A. Hendrickson and J. H. Konnert in Biomolecular Structure, Function,
Conformation and Evolution (R. Srinivasan, ed.), pp. 43 - 57, Oxford, Per-
magon, 1980.

[31] B. C. Finzel J.Appl.Cryst., vol. 20, pp. 53—55, 1987.

[32] T. A. Jones in Computational Crystallography (D. Sayre, ed.), pp. 303—317,
Oxford, Clarendon Press, 1982.

[33] B. N. Violand and F. J. Castellino J. Biol. Chem., vol. 251, pp. 3906—3912,
1976.

[34] L. Sottrup-Jensen, H. Claeys, M. Zajdel, T. E. Petersen, and S. Magnusson
Prog. Chem. Fibrinolysis Thrombolysis, vol. 3, pp. 139-209, 1977.

[35] S. M. Hochschwender and R. A. Laursen J. Biol. Chem., vol. 256, pp. 11166—
11171, 1981.

[36] V. Ramesh, A. M. Petros, M. Llinas, A. Tulinsky, and C. H. Park J. Mol. Biol,
vol. 198, pp. 481—498, 1987.

[37] M. Andrew, V. Ramesh, and M. Llinas Biochemistry, vol. 28, pp. 1368—1376,
1989.

[38] A. Tulinsky, C. H. Park, B. Mao, and M. Llinas Protein, vol. 3, pp. 85—96, 1988.

[39] A. M. Mulichak and A. Tulinsky Blood Coagulation and Fibrinolysis, vol. 1,
pp. 673—679, 1990.

[40] T. P. Wu, K. Padmanabhan, A. Tulinsky, and A. M. Mulichak Biochemistry,
vol. 30, pp. 10589—10594, 1991.

[41] A. M. Mulichak, Dissertation. PhD thesis, Michigan State Univeraity, East
Lansing, Michigan, 1991.

[42] A. M. Mulichak, C. H. Park, A. Tulinsky, and M. Llinas J.Biol.Chem., vol. 264,
pp. 1922—1923, 1989.

178
[43] T. J. Rydel, A. Tulinsky, W. Bode, and R. Huber J. Mol. Biol., vol. 221,
pp. 583—601, 1991. '

[44] M. Trexler, L. Banyai, L. Patthy, N. D. Pluck, and R. J. P. Williams
Eur.J.Biochem., vol. 152, pp. 439-446, 1985.

[45] S. K. Burley and G. A. Petsko Science, vol. 229, pp. 23—28, 1985.

[46] M. Llinas, A. D. Marco, S. M. Hochschwender, and R. A. Laursen Eur. J.
Biochem., vol. 135, pp. 379—391, 1983.

[47] A. Motta, R. A. Laursen, N. Rajan, and M. Llinas J. Biol. Chem., vol. 261,
pp. 13684-13692, 1986.

[48] J. R. Powell, J. M. Beals, and F. J. Castellino Arch. Biochem. Biophys., vol. 248,
pp. 390-400, 1986.

[49] M. Trexler and L. Patthy Proc. Natl. Acad. Sci. U. S. A., vol. 80, pp. 2457—2461,
1983.

[50] A. Motta, R. A. Laursen, M. Llinas, A. Tulinsky, and C. H. Park Biochemistry,
vol. 26, pp. 3827—3836, 1987.

[51] Z. Vali and L. Patthy J.Biol.Chem., vol. 259, pp. 13690-13694, 1984.

[52] J. C. W. Carter and C. W. Carter J. Biol. Chem., vol. 254, pp. 12219-12223,
1979.

[53] A. T. Brunger, “X—plor manual (version 2.1),” 1990.
[54] E. E. Lattman Acta Cryst., vol. B28, pp. 1065-1068, 1972.

[55] A. M. de Vos, M. H. Ultsch, R. F. Keeley, K. Padmanabhan, and A. Tulinsky
Biochemistry, vol. 31, pp. 270—279, 1992.

[56] R. K. Arni, K. Padmanabhan, K. P. Padmanabhan, T. P. Wu, and A. Tulinsky
Biochemistry, vol. 32, 1993.

[57] X. Li, R. A. G. Smith, and C. M. Dobson Biochemistry, vol. 31, pp. 9562—9571,
1992.

[58] T. Ny, F. Elgh, and B. Lund Proc. Natl. Acad. Sci. U.S.A., vol. 81, pp. 5355—
5359, 1984.

179
[59] G. Pohl, M. Kallstrom, N. Bergsdorf, P. Wallen, and H. Jorncall Biochemtry,
vol. 23, pp. 3701—3707, 1984.

[60] P.Wallen, N. Bergsdorf, and M. Ranby Biochim. Biophys. Acta, vol. 719,
pp. 318-328, 1982.

[61] D. C. Rijken, M. Hoylaerts, and D. Collen J. Biol. Chem., vol. 257, pp. 2920-
2925, 1982.

[62] A. J. V. Zonneveld, H. Veerman, and H. Pannekoek Proc. Natl. Acad. Sci.
U.S.A., vol. 83, pp. 4670-4674, 1986.

[63] G. A. W. de munk, M. P. M. Caspers, G. T. C. Chang, P. H. Pouwels, B. E.
Engervalk, and J. H. Verheijen Biochemistry, vol. 28, pp. 7318-7325, 1989.

[64] E. J. D. Weening-Verhoeff, P. H. A. Quax, R. T. van Leeuwen, E. F. Rehberg,
K. R. Marotti, and J. H. Verheijen Protein Engineering, vol. 4, pp. 191-198,
1990.

[65] I. L. Byeon and M. Llinas J. Mol. Biol., vol. 222, pp. 1035—1051, 1991.

[66] I. L. Byeon, R. F. Kelley, and M. Llinas Eur. J. Biochem., vol. 197, pp. 155—165,
1991.

[67] I. L. Byeon, R. F. Kelley, and M. Llinas Biochemistry, vol. 28, pp. 9350—9360,
1989.

[68] J. W. I. Fenton Ann. N. Y. Acad. Sci., vol. 370, pp. 468-495, 1981.
[69] L. J. Berliner Mol. Cell. Biochem., vol. 61, pp. 159—172, 1984.

[70] K. G. Mann and R. L. Lundblad in Hemostasis and Thrombosis: Basic Princi-
ples and Clinical Practice (R. W. Coleman, J. Hirsh, V. J. Marder, and E. W.
Salzman, eds.), pp. 148—161, J. B. Lippincott Co., Philadelphia, 1987.

[71] M. E. Nesheim, L. S. Hibbard, P. B. Tracy, J. W. Bloom, K. H. Myrmel, and
K. G. Mann in The Regulation of Coagulation (K. G. Mann and F. B. Taylor,
eds.), pp. 145—149, Elsevier—North Holland, New York, 1980.

[72] R. J. Butkowski, J. Elion, M. R. Downing, and K. G. Mann J. Biol. Chem.,
vol. 252, pp. 494241957, 1977.

[73] A. R. Thompson, D. L. Enfield, L. H. Ericsson, M. E. Legaz, and J. W. F. II
Arch. Biochem. Biophys., vol. 178, pp. 356—367, 1977.

180
[74] W. Bode, I. Mayr, U. Baumann, R. Huber, S. R. Stone, and J. Hofsteenge
EMBO J., vol. 8, pp. 3467—3475, 1989.

[75] A. Henschen, F. Lottspeich, M. Kehl, and C. Southan Ann. N. Y. Acad. Sci.,
vol. 408, pp. 28—43, 1983.

[76] R. F. Doolittle Annu. Rev. Biochem., vol. 53, pp. 195-229, 1984.

[77] B. Blomback, B. Hessel, D. Hogg, and L. Therkildsen Nature, vol. 275, pp. 501—
505, 1978.

[78] E. F. Casassa J. Chem. Phys, vol. 23, pp. 596—597, 1955.
[79] R. R. Hantgan and J. Hermans J. Biol. Chem., vol. 254, pp. 11272—11281, 1979.

[80] R. R. Hantgan, W. E. Fowler, H. P. Erickson, and J. Hermans Thromb.
Haemost., vol. 44, pp. 119—125, 1980.

[81] P. H. Ehrlich, J. H. Sobel, Z. A. Moustafa, and R. E. Canfield Biochemistry,
vol. 22, pp. 4184—4192, 1983.

[82] R. Chen and R. F. Doolittle Proc. Natl. Acad. Sci. USA, vol. 66, pp. 472—479,
1970. i

[83] B. Blomback Ann. N. Y. Acad. Sci., vol. 484, pp. 120—123, 1986.
[84] H. A. Scheraga Ann. N. Y. Acad. Sci., vol. 485, pp. 124—133, 1986.

[85] Y. C. Meinwald, R. A. Martinelli, J. W. Nispen, and H. A. Scheraga Biochem-
istry, vol. 19, pp. 3820—3825, 1980.

[86] H. C. J. Marsh, Y. C. Meinwald, S. Lee, and H. A. Scheraga Biochemistry,
vol. 21, pp. 6167-6171, 1982.

[87] H. C. J. Marsh, Y. C. Meinwald, T. W. Thannhauser, and H. A. Scheraga
Biochemistry, vol. 22, pp. 4170—4174, 1983.

[88] B. Blomback, M. Blomback, P. Olsson, L. Svendsen, and G. Aberg Scand. J.
Clin. Lab. Invest, vol. 24, pp. 59-64, 1969.

[89] I. D. Rae and H. A. Scheraga Int. J. Rept. Protein Res., vol. 13, pp. 304—314,
1979.

[90] F. Ni, Y. Konishi, R. B. Frazier, and H. A. Scheraga Biochemistry, vol. 28,
pp. 3082—3094, 1989.

181

[91] M. Kahn, S. Wilke, B. Chen, and K. Fujita J. Am. Chem. Soc., vol. 110,
pp.1638—1639,1988.

[92] H. Nakanishi, R. A. Chrusciel, R. Shen, S. Bertenshaw, M. E. Johnson, R. T. J.,
A. Tulinsky, and M. Kahn Proc. Natl. Acad. Sci. USA, vol. 89, pp. 1705—1709,
1992.

[93] T. J. Rydel, K. G. Ravichandran, A. Tulinsky, W. Bode, R. Huber, C. Roitsch,
and I. Fenton, J. W. Science, vol. 249, pp. 277-280, 1990.

[94] F. Ni, Y. C. Meinwald, M. Vasquez, and H. A. Scheraga Biochemistry, vol. 28,
pp.3094-3105,1989.

[95] P. D. Martin, W. Robertson, D. Turk, R. Huber, W. Bode, and B. F. P. Edwards
J. Biol. Chem., vol. 267, pp. 7911-7930, 1992.

[96] M. T. Stubbs, H. Oschkinat, I. Mayr, R. Huber, H. Angliker, S. R. Stone, and
W. Bode Eur. J. Biochem., vol. 206, pp. 187—195, 1992.

[97] T. P. Wu, V. Yee, A. Tulinsky, R. A. Chrusciel, H. Nakanishi, R. Shen,
C. Priebe, and M. Kahn Protein Engineering. in press.

[98] E. Skrzypczak—Jankun, V. Carperos, K. G. Ravichandran, A. Tulinsky,
M. Westbrook, and J. M. Maraganore J. Mol. Biol., vol. 221, pp. 1379-1393,
1991.

[99] W. Bode, D. Turk, and A. Karshikov Prot. Sci., vol. 1, pp. 426—471, 1992.

[100] X. Qiu, K. Padmanabhan, V. E. Carperos, A. Tulinsky, T. Kline, J. M.
Maraganore, and J. W. Fenton II Biochemistry, vol. 31, pp. 11689—11697, 1992.

[101] R. Huber and W. Bode Acc. Chem. Res., vol. '11, pp. 114—122, 1978.

[102] A. Tulinsky and X. Qiu Blood Coagulation and Fibrinolysis, vol. 4, pp. 305—312,
1993.

[103] D. W. Banner and P. Hadvary J. Biol. Chem., vol. 266, pp. 20085-20093, 1992.
[104] E. W. Davie and K. Fujikawa Annu. Rev. Biochem., vol. 44, pp. 799-829, 1975.
[105] H. J. Muller-Eberhard Annu. Rev. Biochem., vol. 44, pp. 697—724, 1975.

[106] S. Magnusson, L. Sottrup-Jensen, T. E. Petersen, G. D. Wojciechowska, and
H. Claey in Miami Winter Symposia (D. W. Ribbons and K. Brew, eds.), vol. 11,
pp. 203—239, Academic Press, New York, N .Y., 1976.

182

[107] H. Blomback in Blood Clotting Enzymology (W. H. Seegers, ed.), pp. 143—215,
Academic New York, 1967.

[108] M. Rholam, P. Nicolas, and P. Cohen FEBS Lett., vol. 207, pp. 1-6, 1986.

[109] M. Rhloam, P. Cohen, P. Brakch, N. Paolillo, A. Scarrurin, and C. DiBello
Biochem. Biophys. Res. Commun., vol. 168, pp. 1066—1073, 1990.

[110] E. Bek and R. Berry Chemistry, vol. 29, pp. 178—183, 1990.

[111] M. Kahn, H. Nakanishi, R. A. Chrusciel, D. Fitzpatrick, and M. E. Johnson J.
Med. Chem., vol. 34, pp. 3395-3399, 1991.

[112] R. Shen, C. Proebe, C. Patel, L. Rubo, T. Su, M. Kahn, and R. Sugasawara
Tetrahedron Lett., vol. 33, pp. 3417-3420, 1992.

[113] M. Laskowski and I. Kato Annu. Rev. Biochem., vol. 49, pp. 593-626, 1980.

[114] R. Huber and R. W. Carrell Biochemistry, vol. 28, pp. 8951—8966, 1989.

 

 

MICHIGAN STATE UNIV. LIBRARIES
[Illlllllllll[Will]llllllllllllll[lllllllllll
31293008853289