.1 V»:
s A :31» 1.1...
31:. ale...) m .
4...! .r: In... :

. 2. .imnueu.
i. . , . uuahmmgdw .m&
(.5 $.61 . , 13%.”?! .4. . h
a .. WW Aw...»
. . 8h.......n. -Jﬂl

..:r...,1:.
a. .
i... z: y
.. .9. ..
can 3.3...
i z .
km.
5....
.. ..
van." 5:,
.
Inuitutlvuia
5.} u ..
. b .
a: .2 .

T... x 1...! 90343. A

.. Aunt.

4 km“ Ewm I
. 93.0.1!

. an... kt...» a
.rflii 5.. 3
. i . . 3.11%...553‘3569.
22.....1 . , 3.... . nun}. ‘ .
3n PE... . . . _ .3
.. 4. . ...m. . .

n .. 53!.
(”mm-2.. . “I“!
9%.“
Itf‘l'i
. 4
[£143
261....)

31

I: . nu...)
i....}..x.....,£...n s“ ..
. w. 1 . In"! .K >
- 11.3, 1.15 . . .5:
“31.1: 1......wbn.
3.2.... $13.?
.r iii:
if:

.15....

.21.. 42.";

1. 3.3:
I I. ‘I

. z
p .97;
. A NW?! ..
. : ,3

¢ Ho...” .44 UN:

 

Rh... .
~

.
:l . 7

 

Ill!

EH“

L LIBRARY
Qm’o Michigan State
"livenity 1

This is to certify that the
dissertation entitled

The X-ray Crystallographic Structures of the Angiogenesis

lnhibitor Angiostatin Bound to a Peptide from the Group A

Streptococcal Surface Protein PAM and the Metal-Bound
Conantokins Con-G and Con-T[K79amma]

presented by
Sara Elizabeth Cnudde

has been accepted towards fulﬁllment
of the requirements for the

Ph.D. degree in Biochemistry

 

 

W

Majo [/qPrifessor’ s Signature

 

Date

MSU is an aﬁirmative-action, equal-opportunity employer

-.-I-u-c--—-—..—-...- ,

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 Kthroj/AccsPrelelRC/DateDuemdd

THE X-RAY CRYSTALLOGRAPI—IIC STRUCTURES OF THE ANGIOGENESIS
INHIBITOR ANGIOSTATIN BOUND TO A PEPTIDE FROM THE GROUP A
STREPTOCOCCAL SURFACE PROTEIN PAM AND THE METAL-BOUND
CONANTOKINS CON-G AND CON-T[K7GAMMA]
By

Sara Elizabeth Cnudde

A DISSERTATION
Submitted to
Michigan State University

in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Biochemistry

2007

ABSTRACT
THE X-RAY CRYSTALLOGRAPHIC STRUCTURES OF THE ANGIOGENESIS
INHIBITOR ANGIOSTATIN BOUND TO A PEPTIDE FROM THE GROUP A
STREPTOCOCCAL SURFACE PROTEIN PAM AND THE METAL-BOUND
CONANTOKINS CON-G AND CON-T[K7GAMMA]
By

Sara Elizabeth Cnudde

Angiostatin is a fragment of plasminogen encompassing the ﬁrst three
kringle domains. Plasminogen is the zymogen of plasmin that is known to bind C-
terrninal lysine residues in ﬁbrin through the kringle domain lysine binding site. The
ﬁrst Structure of a multi-kringle containing compound bound to a ligand was not
done until the structure of the an giostatin/VEK-30 complex was determined and
reﬁned to 2.3 A resolution that of which is described herein. It provides a model of
the interaction between plasminogen and Streptococcal-derived pathogenic proteins
during infection. VEK-3O contains a through-space isostere for C-terminal lysine,
wherein Arg and Glu side chains, separated by one helical turn, bind within the
bipolar angiostatin kringle 2 (K2)-domain lysine-binding site. VEK-3O also makes
several contacts with K2 residues that exist outside of the canonical LBS and are not
conserved among the other plasminogen kringles, thus providing a molecular basis
for the selectivity of VEK-3O for K2. The structure also shows that plasminogen
kringle domains undergo signiﬁcant structural rearrangement relative to one another,
and reveals dimerization between two molecules of angiostatin/VEK-3O related by

crystallographic symmetry. This dimerization, which only exists in the crystal

structure, is consistent with the parallel coiled-coil full-length PAM dimer expected
from sequence similarities and homology modeling.

The challenge of generating a small, unstructured peptide capable of metal
ion-triggered helix formation and self-association has been satisﬁed in nature with
the peptide conantokin-G (con-G). Con-G antagonizes the N-methyl-D-aspartate
(NMDA) receptors. Con-G is 17 residues and contains ﬁve y-carboxyglumates
(Gla). A variety of metals can promote a conformational change from random coil
to a helix, but only Ca2+ allows for the formation of a dimeric con-G complex. From
these data, we proposed a model for the complex in which antiparallel con-G strands
are stabilized solely through Ca2+-bridging of Gla headgroups within the helix-helix
interface. This model represents a heretofore unknown motif which we deﬁne as the
“metallo-zipper.” A second member from the conantokin family, conantokin-T (con-
T), shares some sequence identity to con-G. However, several primary and
secondary structural differences exist between con-G and con-T, speciﬁcially at
position 7, which is occupied by Lys in con-T and Gla in con-G. While con-T does
not undergo Ca2+-induced self-assembly, replacing the Lys with a Gla (con-T[K7y])
allows it to form an antiparallel helix dimer in the presence of Ca2+. However the
dimer interface is substantially different from con-G.

We were able to further understand metal-dependant dimerization and helix
stabilization by determining the structure of the peptide bound to different metals,
such as Cd2+ and Mg”. X-ray structures have been determined of Ca2+-complexed

con-G and con-T[K7y] at high resolution (1.2 A for con-G and 1.6 A for con-

T[K7y]) as well as MgZI/Cdz‘Y complexed con-T[K7y] at 1.2 A.

To my husband, Jeremy

iv

ACKNOWLEDGEMENTS

I would like to thank my research advisor Dr. James Geiger for all of his
guidance, encouragement and support throughout the course of my graduate study. I
sincerely thank him for all of the invaluable experiences that I have had in his lab. I
would also like to thank our collaborator Dr. Frank Castellino and Dr. Mary Prorok
for providing the proteins and peptides that were used in both the angiostatin and
conantokin studies. I would also like to thank them for helping to write and review
the published papers. I would also like to thank my biochemistry advisor Dr. Bill
Henry for all of his help and support during my oral and defense. I am in gratitude
to my committee members as well Dr. Michael Garavito, Dr. John Wang, and Dr.
Hongao Yan.

I would like to acknowledge and thank all former and current members of the
Geiger lab, Dr. Stacy Hovde, Dr. Marta Abad, Dr. Xiangshu Jin, Dr. Soheila
Vaezeslami, Dr. Andrej Hanzlowsky, Paul Booth, Mike Wolf, Erika Mathes, Aimee
Brooks, Lei Feng, Xaofei, Susie, Tyra, Blanka, and Dorothy Tappenden. It has been
a pleasure to know you and work with you. A time in my life I will never forget.

No lab will ever be like the Geiger lab. I would also like to give a special thanks to
Dr. Stacy Hovde. Your help and friendship has meant a great deal to me. Stacy
thank you for teaching me to knit. I would also like to thank Dr. Jorge Rios and Dr.
Raghuvir Ami for their help learning crystallography and reﬁnement techniques.

Thank you everyone and I wish you good luck in your scientiﬁc careers.

I would also like to thank my parents who both pushed me to get an
education and provided financial support as well as unconditional love.

Last but certainly not least I would like to thank my husband. If not for your
unconditional love and continuing support I might not have ﬁnished. I am forever

grateful. I love you.

vi

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ......................................................................................... v

TABLE OF CONTENTS .......................................................................................... vii

LIST OF TABLES ..................................................................................................... ix

LIST OF FIGURES ...................................................................................................... x

ABBREVIATIONS .................................................................................................. xix
Chapter I: INTRODUCTION

1.1 Introduction to the Angiostatin/VEK-BO Complex Structure .................... 1

1.1.1 Plasminogen ............................................................................. 1

1.1.2 PAM ......................................................................................... 5

1.1.3 Angiostatin ............................................................................... 6

1.1.4 Physiological Importance of the Structure ............................. 20

1.2 Introduction to Metal Bound Structures of con-G and con-T[K7y] ........ 23

1.2.1 Protein Design ........................................................................ 23

1.2.2 Overview of Cone Snails and Conotoxins ............................. 25

1.2.3 con-G ...................................................................................... 27

1.2.4 con-T ...................................................................................... 28

1.2.5 NMDA Receptor .................................................................... 31

1.2.6 Structure-Activity Relationships ............................................ 35

1.2.7 Relevance of the project ......................................................... 39

1.3 Literature Cited ........................................................................................ 40

Chapter II: X-RAY STRUCTURE DETERMINATION

2.1 Structure Determination from X—ray Diffraction Data ........................... 53
2.1.1 Theory .................................................................................... 53
2.1.2 Overview of Crystallography ................................................. 62
2.2 Angiostatin/VEK-3O Complex ............................................................... 63
2.2.1 Crystallization and Data Collection of P6122 Structure ......... 63
2.2.2 Structure Determination and Reﬁnement P6122 Structure ..... 68
2.2.3 Crystallization and Data Collection of P6. Structure ............. 72
2.2.4 Structure Determination and Reﬁnement P6. Structure ......... 73
2.3 Ca2+/con-G .............................................................................................. 80
2.3.1 Crystallization and Data Collection ....................................... 80
2.3.2 Structure Determination and Reﬁnement ............................... 81
2.4 Ca2+lcon-T[K7y] ..................................................................................... 87
2.4.1 Crystallization and Data Collection ....................................... 87

vii

2.5

2.6

2.4.2 Structure Determination and Reﬁnement ............................... 88

Cd2+/Mg2+/con-T[K7y] ........................................................................... 92
2.5.1 Crystallization and Data Collection ....................................... 92
2.5.2 Structure Determination and Reﬁnement ............................... 93

Literature Cited ....................................................................................... 98

Chapter III: THE THREE DIMENSIONAL STRUCTURE OF THE

ANGIOSTATIN/VEK-30 COMPLEX

3.1 Overall Structure of Angiostatin/VEK-30 Complex ............................ 101
3.2 Interactions Between VEK-30 and Angiostatin ................................... 102
3.3 Kringle Domain Rotation ..................................................................... 106
3.4 The Kringle 3 Domain .......................................................................... 110
3.5 Dimerization ......................................................................................... 113
3.6 Kringle 2 Domain Specificity ............................................................... 118
3.7 Further Studies of VEK-30 ................................................................... 122
3.8 Angiostatin Binding to Proteins ........................................................... 127
3.9 Discussion ............................................................................................ 131
3.10 Literature Cited ..................................................................................... 133

Chapter IV: THE THREE DIMENSIONAL STRUCTURES OF CA2+-BOUND

CON-GAND CON-T[K77], AND CD2+IMGZ+ICON-T[K7y]

4.1 Overall Structure of Ca2+lcon-G and Ca2+/con-T[K7y] ...................... 136
4.2 The Structure of Ca2+lcon-G ................................................................ 137
4.3 The Structure of Ca2+lcon-T[K7y] ........................................................ 140
4.4 Discussion ............................................................................................ 149
4.5 Mutational Studies of the con-G and con-T Peptides ........................... 151
4.6 The con-G and con-T Structures and NMDA Receptor Binding ......... 152
4.7 Comparison of the Crystal Structures to the NMR Structures ............. 153
4.8 Conclusions From the Calcium-Bound Structures ............................... 156
4.9 The Overall Structure of Cd2+/Mg2+/con-T[K7y] ................................. 157
4.10 Comparison between Ca2+-con-T[K7y] and Cd2+fMg2+-con-T[K7y] 161
4.11 Literature Cited ..................................................................................... 166
APPENDIX
Appendix 5.1 Scalepack output ﬁle of Angiostatin/VEK-30 P6122 ........................ 169
Appendix 5.2 Scalepack output ﬁle of An iostatin/VEK-30 P61 ............................ 170
Appendix 5.3 Scalepack output ﬁle of Ca +Icon-G .................................................. 171
Appendix 5.4 Scalepack output ﬁle of Ca2+lcon-T[K7y] ......................................... 172
Appendix 5.5 Scalepack output ﬁle of Cd2+/Mg2+/con-T[K7y] ............................... 173

viii

LIST OF TABLES

CHAPT ER I: INTRODUCTION
Table 1.1 Metal ion effects secondary structure and Ma.pp of con-G and analogs ..... 29
Table 1.2 Metal ion effects secondary structure and Ma.pp of con—T and analogs ...... 31

Table 1.3 Effects of side chains on the NMDA receptor antagonist and
conformational properties of con-G .......................................................................... 37

Table 1.4 Effects of side chains on the NMDA receptor antagonist and
conformational properties of con—T ........................................................................... 38
CHAPT ER II: X-RAY STRUCTURE DETERMINATION

Table 2.1 Crystal parameters for the angiostatin/VEK-3O crystals in the space
groups P6122 and P61 ................................................................................................. 66

Table 2.2 Data statistics for the an giostatin/VEK-30 diffraction data collection ..... 67

Table 2.3 Reﬁnement statistics for the angiostatin/VEK—30 complex ...................... 67
Table 2.4 Crystal parameters for the metal bound conantokin structures ................. 81
Table 2.5 Data statistics for the metal bound conantokin structures ......................... 82
Table 2.6 Reﬁnement statistics for the metal bound conantokin structures .............. 82

CHAPTER III: THE THREE DIMENSIONAL STRUCTURE OF THE
ANGIOSTATIN/VEK-30 COMPLEX

Table 3.1 Interactions between angiostatin and VEK-30 ........................................ 107

LIST OF FIGURES

Images in this dissertation are presented in color.
CHAPTER I: INTRODUCTION

Figure 1-1 Schematic representation of the structure of human plasminogen ............ 1

Figure 1-2 The structure of K5 (PDB ID SHPG). The disulﬁde bonds are
colored red .................................................................................................................... 2

Figure 1-3 Structure of Kl-EACA (PDB ID ICEA). K1 is colored green and the
residues encompassing the LBS are shown in atom color (nitrogen, blue; oxygen, red
and carbon, green) and EACA is colored green and by atom color. Side chains are
labeled using plasminogen numbering ......................................................................... 4

Figure 1-4 Primary sequence of VEK-30. The underlined residues are the
direct repeat ................................................................................................................. 6

Figure 1-5 X-ray crystallographic structure of K2/VEK-30 (PDB ID 115K). K2 is
colored green and the VEK-3O peptide is colored magenta. Side chains are labeled
using plasrrrinogen numbering ...................................................................................... 7

Figure1-6 The overall structure of angiostatin (PDB ID IKIO). The interkringle
disulﬁde connecting K2 and K3 (C169-C297) is shown in red. The green
spheres are bicine molecules ...................................................................................... 10

Figure 1-7 The left molecule is bicine and the right molecule is EACA. Both
are colored by atom .................................................................................................... 11

Figure 1-8 Space-ﬁlling view of angiostatin. The LBS in each of the three
kringles is colored red. All other atoms are blue ....................................................... 11

Figure 1-9 Interaction of the angiostatin Kl LBS with bicine. K1 residues are
colored green and the bicine molecule is colored magenta. Residues are labeled
with plasminogen numbering and atom coloring ....................................................... 14

Figure 1-10 Interaction of the angiostatin K2 with bicine. K2 is colored green
and bicine is colored magenta. Residues are labeled with plasminogen numbering
and atom coloring ...................................................................................................... 14

Figure 1-11 Overlay of the angiostatin K2 onto the K2/VEK-30 structure.
Residues are labeled with plasminogen numbering and atom coloring ..................... 14

Figure 1-12 Interaction between angiostatin K3 and bicine. K3 is colored
green and bicine is colored magenta. Residues are labeled with plasminogen
numbering ................................................................................................................... 15

Figure 1-13 Overlay of K1/EACA onto angiostatin K3. K3 is colored blue
and K1 is colored orange. EACA is colored pink. Residues are labeled with
plasminogen numbering ............................................................................................. 16

Figure 1-14 Close view of the section of (XVB3 integrin that harbors the
residues involved in possible angiostatin binding. The [33 subunit is shown
in red and the residues are shown in green with atom coloring ................................. 21

Figure 1-15 Top view of 0:433 integrin with angiostatin docked onto a helix in

the [33 subunit assuming the conformation seen in the K2/VEK-3O structure. OLVBg
integrin is shown in silver, angostatin in blue, residues in angiostatin binding are
colored magenta, residues in the B3 subunit responsible for binding are in green while
the residues in the 0tV subunit are shown in yellow .................................................... 21

Figure 1-16 (a) Close view of the section of the alpha subunit of F 1-ATPase

that harbors residues for possible angiostatin binding. (b) Structure of bovine
mitochondrial Fl-ATPase (131). (c) Enlarged view of the beta subunit Fl-ATPase
that harbors residues for possible angiostatin binding ............................................... 22

Figure 1-17 The primary sequences of con-G, con-T, and con-T[K7y]. The
oc-helical heptad repeat assignments are shown above ............................................... 30

Figure 1-18 Helical wheel representation of the cross-sectional heptad

repeat of the antiparallel dimer of con-G in the presence of Ca2+. For the left

chain, the N-terminus is closest to the viewer and for the right chain the C-terrrrinus
is closest to the viewer. Gla7 and Gla14 of the left chain occupy position a, while
Gla3 and GlalOoccupy position d. Gla chelation pairs include Gla3-Glal4’,
Gla7-Gla10’, GlalO-Gla7’, Gla14-Gla3’. The lighter arrows are at the bottom of the
wheel, progressively thickening towards the top ....................................................... 32

Figure 1-19 Helical wheel representation of the cross-sectional heptad

repeat of the antiparallel dimer of con-T[K7y] in the presence of Ca“.

For the left chain, the N-terminus is closest to the viewer and for the right

chain the C-terminus is closest to the viewer. Gla7 and Gla14 of the left chain
occupy position a, while Gla3 and GlalO occupy position (1. Gla chelation pairs
include Gla3-Gla14’, G1a7-G1a10’, GlalO-Gla7’, Glal4-Gla3’. The lighter arrows
are at the bottom of the wheel, progressively thickening towards the top ................. 32

CHAPTER II: X-RAY STRUCTURE DETERMINATION

Figure 2-1 The hanging drop vapor diffusion method for crystallizing proteins ...... 65

xi

Figure 2-2 A non-single crystal of the angiostatin/VEK-30 complex grown
in 20% PEG 8000/0.1 M potassium dihydrogen phosphate ....................................... 65

Figure 2-3 A single crystal of the angiostatin/VEK-30 complex grown in
20% PEG 8000/01 M potassium dihydrogen phosphate/5% 1,4-dioxane ................ 65

Figure 2-4 The overall structure of the an giostatin/VEK-30 complex.
Angiostatin is shown in red and VEK-30 is shown in blue. One molecule of dioxane
is shown in green with two different conformations .................................................. 66

Figure 2-5 Ramachandran plot of angiostatin in the angiostatin/VEK-30
complex in the P6.22 space group ............................................................................. 70

Figure 2-6 Ramachandran plot of VEK-30 in the angiostatinNEK-30
complex in the P6.22 space group ............................................................................. 71

Figure 2-7 An example of the 2Fo-Fc map contoured at lo of the angiostatinNEK-
30 complex in the P6 .22 space group ........................................................................ 72

Figure 2-8 An example of the 2Fo-Fc map contoured at 1.20 of the
angiostatin/VEK-30 complex P6. structure. The map is centered on the interkringle

K2-K3 disulﬁde bond (C169-C297) ........................................................................... 75
Figure 2-9 Ramachandran plot of angiostatin molecule A in the

angiostatin/VEK-3O complex P6. structure ............................................................... 76
Figure 2-10 Ramachandran plot of angiostatin molecule B in the
angiostatin/VEK-30 complex P6. structure ............................................................... 77
Figure 2-11 Ramachandran plot of VEK-30 molecule A in the

angiostatin/VEK-3O complex P6. structure ............................................................... 78
Figure 2-12 Ramachandran plot of VEK-30 molecule B in the

angiostatin/VEK-30 complex P6. structure ............................................................... 79
Figure 2-13 Crystal of Ca2+lcon-G ............................................................................ 80

Figure 2—14 Overall structure of the con-G antiparallel dimer with Fo-Fc

map calculated in the absence of calcium and contoured at 80'. The gray molecule is
related by crystallographic two-fold symmetry that forms the dimer. The side-chains
colored green (E2, 74, L5, Q9, 112) have been shown to decrease potency of the
NMDA receptor when mutated to Ala by at least 100-fold. The side-chains colored
yellow (GI, y3, L11, R13) have been shown to decrease the potency of the NMDA
receptor when mutated to Ala by at least 10-60-fold. Ca3 is positioned directly on

xii

the crystallographic two-fold axis. Ca2b and Ca3b are crystallographically related to
Ca2 and Ca3. The peptide side-chains are colored by atom type .............................. 85

Figure 2-15 Ramachandran plot of the Ca2+/con-G structure ................................... 86

Figure 2-16 The overall structure of the antiparallel con-T[K7y] dimer with a Fo-Fc
map calculated in the absence of calcium and contoured at 86. The side-chains
colored green (E2, y4, Y5) have been shown to decrease potency of the NMDA
receptor when mutated to Ala by at least lOO-fold. The side-chains colored

yellow (G1, y3, M8, L12) have been shown to decrease the potencyof the NMDA
receptor when mutated to Ala by at least 10-60-fold. In all cases, the peptide side-
chains are colored by atom type ................................................................................. 89

Figure 2-17 Ramachandran plot of helix A in the Ca2+lcon-T[K7y] structure ......... 90
Figure 2-18 Rarrrachandran plot of helix B in the Ca2+lcon-T[K7y] structure .......... 91

Figure 2-19 The overall structure of Cd2+/Mg2+/con-T[K7y] with an

anomalous difference map contoured at 80'. The four Cd2+ are shown as blue
spheres and the Mg2+ is shown as a magenta sphere. The side-chains are colored by
atom ............................................................................................................................ 94

Figure 2-20 The overall structure of Cd2+lMg2+lcon-T[K7y] with a For,

map calculated in the absence of metal cations and contoured at 56. The four Cd2+
are shown as blue spheres and the Mg2+ is shown as a magenta sphere. The side-
chains are colored by atom ......................................................................................... 96

Figure 2-21 Ramachandran plot of the Cd2+/Mg2+/con-T[K7y] ................................ 97

CHAPTER III: THE THREE DINIENSIONAL STRUCTURE OF THE
ANGIOSTATIN/VEK-30 COMPLEX

Figure 3-1 Interactions between the angiostatin K2 LBS and VEK-30.

R101 and E104 are spaced by almost one helical turn and form the pseudo-lysine
residue. Angiostatin is colored green and VEK-30 is shown in magenta. K2
residues are labeled using plasminogen numbering and VEK—30 residues are labeled
with PAM numbering. All atoms are colored by atom type (nitrogen, blue; oxygen,
red) ............................................................................................................................ 104

Figure 3-2 A n-cation interaction occurs between the guanidino group of VEK-30
R101 and angiostatin K2 LBS residue W235. Angiostatin is shown in green and
VEK-30 is shown in magenta. All atoms are colored by atom type (nitrogen, blue;
oxygen, red). Angiostatin residues are labeled with plasminogen numbering.
VEK-30 residues are labeled with PAM numbering ................................................ 105

xiii

Figure 3-3 Overlay of the K2 domains of the unliganded angiostatin and
angiostatin/VEK-30 complex structures. The structure of angiostatin is shown in
blue and the structure of the angiostatin/VEK-30 complex is shown in red
(angiostatin) and green (VEK-30). H114 of angiostatin is colored yellow in both
structures. Kl of the angiostatin/VEK-30 complex undergoes a 481° rotation
relative to the angiostatin structure .......................................................................... 108

Figure 3-4 Residues responsible for interdomain rotation as determined by
the program DynDom ............................................................................................... 109

Figure 3-5 Overlay of the K2 of angiostatin with K2 of the

angiostatin/VEK-30 complex from the P6. crystal form where 3 residues of
angiostatin K3 are ordered. The P6. structure is shown in red and the structure of
angiostatin is shown in blue. The residues are labeled with Pg numbering. The
inter-kringle disulﬁde bond in the P6. complex structure has rotated 3.3A from its
position in the angiostatin structure .......................................................................... 111

Figure 3-6 An overlay of residues Pro296-Ly5298 of the VEK-30-bound

with the corresponding residues in the crystal structure of free angiostatin.
Angiostatin in the angiostatin/VEK-30 complex is colored magenta, K3 of
angiostatin is shown in green, and the angiostatin/VEK-30 symmetry related
molecule is shown in cyan ........................................................................................ 112

Figure 3-7 Dimerization of the angiostatin/VEK-30 complex. One molecule
has angiostatin colored blue and VEK—30 colored red, while the other molecule has
angiostatin colored magenta and VEK-30 colored yellow ....................................... 114

Figure 3-8 Dimerization of the K2/VEK-30 structure. K2 is shown in green
and VEK-30 is colored magenta .............................................................................. 114

Figure 3-9 Overlay of the K2/VEK-30 dimer onto the angiostatin/VEK-30

dimer. The angiostatin/VEK-30 dimer is shown in red and the K2NEK-30 dimer is
shown in blue. The Cor positions of the dimeric structure of K2/VEK-30
superimpose well with the two angiostatin/VEK-30 complexes that are related by
crystallographic 2-fold symmetry (rmsd for all atoms ~0.4 A) ................................ 115

Figure 3-10 Residues involved in dimerization. VEK-30 residues are shown

in green and angiostatin K2 residues are shown in yellow. The symmetry-related
molecule of angiostatin K2 is shown in cyan with its VEK-30 residue shown in
magenta. The atoms are colored by type. The water molecules are designated as W1
and the symmetry-related waters W2a and W2b. All residues are labeled with Pg
and PAM numbering ................................................................................................ 117

Figure 3-11 A sequence alignment of all ﬁve Pg kringle domains. Yellow
highlighted residues are conserved ........................................................................... 120

xiv

Figure 3-12 An overlay of angiostatin K2 onto angiostatin/VEK-30 K2.

Residues from an giostatin alone are colored magenta, residues from the
Angiostatin/VEK—30 complex are colored green, and VEK-30 residues are colored
yellow. The atoms are colored by type. All residues are labeled with Pg and PAM
numbering. D219 is ﬂipped out of the LBS, resulting in a tight salt bridge contact
with R220 ................................................................................................................. 121

Figure 3-13 The primary sequences of VEK-30 (top) and VEK-32L (bottom).
They are labeled with PAM numbering above. The green box indicates the direct
repeat sequence ......................................................................................................... 122

Figure 3—14 A model showing that two angiostatin K23 can bind to a longer

PAM peptide. The ﬁgure was made by overlaying the direct repeat sequences from
the angiostatin/VEK-30 structure. The angiostatin/VEK-3O structure is colored
green and the second molecule magenta .................................................................. 123

Figure 3-15 A model showing that two angiostatin K23 can bind to a longer

PAM peptide but disrupts the dimer. The ﬁgure was made by overlaying the direct
repeat sequences from the an giostatin/VEK-30 structure. The angiostatin/VEK-30
structure is colored green and the dimer colored cyan whereas the second molecule
magenta .................................................................................................................... 124

Figure 3-16 Gel ﬁltration chromatogram results from Superdex 200 10/300 GL column.
A) The dotted line is Biorad molecular weight standards (670 kDa thyroglobulin,

158 kDa bovinegamma-globulin, 44 kDa chicken ovalbumin, 17 kDa equine
myoglobin, and 1.3 kDa Vitamin B12). B) Angiostatin/VEK-32L complex
approximately 33 kDa. C) Unbound angiostatin approximately 30 kDa. All
experiments were performed at a ﬂow rate of 1 mL/min and the fractions were

5 mL ......................................................................................................................... 126

Figure 3-17 Endostatin modeled into the angiostatin K2-K3 cavity. This was

done by overlaying the helices of endostatin and VEK-30. Angiostatin is colored
blue and endostatin is colored orange. R158 and D161 of endostatin are colored
green with atom coloring .......................................................................................... 129

Figure 3-18 The residues possibly involved in endostatin/angiostatin binding are
shown. Endostatin was modeled into the angiostatin K2-K3 cavity by overlaying the
helices of endostatin and VEK-30. Angiostatin is colored green and endostatin is
colored lavender. The residues are colored by atom ............................................... 130

CHAPTER IV: THE THREE DIMENSIONAL STRUCTURES or CA“-
BOUND CON-GAND CON-T[K77], AND CD2+IMG2*/CON-T[K77]

Figure 4-1 Ca2+ coordination in con-G. (Top left) CalA (magenta) coordinates

with 73 and 77 and waters W30 and W38 (blue). (Top right) Ca2B (magenta)
coordinates with 73 and Q6 of helix A and 714 and 710 of the crystallographically-

XV

related helix B forming the dimer. Q6 is displayed in blue. (Bottom) Ca3 (gray)
lies on the crystallographic twofold and coordinates with 77 and 710 of each helix as
well as waters W34 and W34b (blue), a symmetry-related water molecule. The
symmetry-related helix forming the dimer is colored yellow. In all cases, the peptide
side-chain atoms are colored by atom type, except as otherwise indicated ............. 138

Figure 4-2 Comparison of the con-T[K77] helices. Overlay of the two
con-T[K77] helices show their identity .................................................................... 141

Figure 4—3 Comparison of the con-T[K77] and con-G helices. (Top) Overlay of the
con-T[K77] (magenta) and con-G(green) helices demonstrates that these helices are
nearly identical to each other. Blue spheres represent Ca2+ ions of con-G and green
spheres designate Ca2+ ions of con-T[K77]. (Bottom) The boxed area is magniﬁed
and includes waters (red spheres). In all cases, the peptide side—chain hetero-atoms
are colored by atom type .......................................................................................... 143

Figure 4—4 Ca2+ coordination in the con-T[K77] dimer. (a) CalA (magenta)
coordinates with )8 and )7 of helix A and y14 of helix B, as well as with with one
water molecule, W71 (red sphere). Ca2B (magenta) coordinates with 77 and 710 of
helix A and yl4 and 710 of helix B, along with water molecule W79 (red sphere).

(b) Ca2A (magenta) coordinates with le and 714 of helix A and y10 and y7 of helix
B. Ca2A also coordinates with two water molecules, W37 and W68 (red spheres).
(c) CalB (magenta) coordinates with yl4 of helix A and )7 and 73 of helix B. CalB
(magenta) also coordinates with one water molecule, W44 (red sphere). (d) The salt
bridge interactions in con-T[K77]. All distances are shown in Angstroms. In each
case, the peptide side-chain hetero-atoms are colored by atom type. Helix A is
always displayed on the left whereas Helix B is displayed on the right .................. 144

Figure 4-5 The con-T[K77] dimer interface is signiﬁcantly different from

con-G. A water network exists at the Ca2+ldimer interface ofCa2+/con-T[K77]. All
distances are shown in Angstroms. The Ca2+ ions are shown as blue spheres and the
waters are shown as red spheres. The top four Gla residues belong to Helix A and
the bottom four Gla residues belong to Helix B. The peptide side-chain hetero-
atoms are colored by atom ........................................................................................ 145

Figure 4-6 The con-T[K77] dimer interface is signiﬁcantly different from con-G.
The fundamental differences between the Ca2+lcon-G structure and Ca2+lcon-

T[K77] structure are shown in a schematic wiring diagram. Each dotted line
represents a coordination between one Gla carboxylate and a Ca2+. Helix A is
colored green and Helix B is colored yellow. The black spheres are Ca2+ ............. 147

Figure 4-7 The con-T[K77] dimer interface is signiﬁcantly different from

con-G dimer. Overlay of the con-G and con-T[K77] structures showing difference in
helix alignment. The con-G dimer is colored green and its Ca2+ ions colored red.

xvi

The con-T[K77] dimer is colored magenta and its Ca2+ ions colored blue. In all
cases, peptide side-chain hetero-atoms are colored by atom .................................... 148

Figure 4-8 Overlay of the NMR and X-ray crystal structures of Ca2+lcon-G.
Cyan, the X-ray structure of Ca2+/con-G; green, a low energy NMR structure; white,
Ca2+. In all cases, peptide side-chain hetero-atoms are colored by atom type ........ 154

Figure 4-9 Cadmium coordination at the calcium interface. (a) Cadmium Cd2

is six—coordinate. The overall structure of Cd2+/Mg2+-con-T[K77] is shown in green.
Cd2 is colored as a blue sphere and the water W20 is colored as a red sphere. A
crystallographic-related oc-helix molecule is colored yellow. (b) Cadmium
coordination at Cd3 is tetrahedrally coordinated. Cd3 sits directly on a
crystallographic two-fold axis and coordinates bidendately with both 74 and Y7 of the
structure and its symmetry-related molecule. Distances to each oxygen atom range
from about 2.3-2.6A. The structure is shown in green and the symmetry-related
molecule is colored yellow. Cd3 is shown as a blue sphere. (c) Cadmium
coordination at Cd4 is octahedrally coordinated. Cd4 coordinates with 710 of the
structure (green) and a crystallography-related molecule (magenta). Cd4 also
coordinates with waters W22a and W51a and the symmetry-related waters W22b
and W51b. Cd4 is partially colored as a blue sphere and all waters are colored as red
spheres. All distances are measured in Angstroms. The side-chains are colored by
atom .......................................................................................................................... 159

Figure 4—10 Magnesium coordination of Mg] is octahedrally coordinated. Mgl
coordinates bindentately with 710 and 714. Mgl also coordinateswith two waters
W24 and W23. W23 has two different conformations shown as W23a and W23b.
Mgl is colored as a magenta sphere and the water molecules are shown as red
spheres. The side-chains are colored by atom. All distances are shown in
Angstroms ................................................................................................................ 160

Figure 4-11 On the opposite side of the calcium interface cadmium coordination

at Cd] is six coordinate. Cdl coordinates with E16 and K19 of the crystallographic
symmetry-related molecule. Cdl also coordinates with G1 and two waters W18 and
W19. All distances are shown in Angstroms. The structure is colored green and a
symmetry-related molecule is colored magenta. Cdl is colored as a blue sphere and
the waters are shown as red spheres. All side-chains are colored by atom ............. 160

Figure 4-12 Gla residues 3, 4, and 7 are able to move relative to the speciﬁc

metal cation. Overlay of the Cd2+/l\/Ig2+—con-T[K77] structure (magenta) onto the
Ca2+—con—T[K77] dimer (blue) structure. Cd2+ are shown as magenta spheres, Mg2+
is shown as a green sphere, and Ca2+ are blue spheres. The side-chains are colored
by atom ..................................................................................................................... 162

Figure 4-13 An overlay of the Cd2+/Mg2+-con-T[K77] structure onto the

Ca2+-con-T[K77] dimer. The Ca2+-con-T[K77] dimeric structure is colored blue and
the Cd2+/Mg2+-con-T[K77] structure is colored magenta. Mgl is shown as a green

xvii

sphere whereas Cal is shown as a blue sphere. The distances shown are all in
Angstroms. The side-chains are colored by atom .................................................. 164

Figure 4-14 A closer view of the Mg2+ site and the 73-77 chelation pair.

All distances are shown in Angstroms. All side-chains are colored by atom. The
magenta atom is Mgl and the blue atoms are Cd2+ .................................................. 165

xviii

LIST OF ABBREVIATIONS

A — alanine

Ala - alanine

apoA - apolipoprotein A
Arg - arginine

Asn - asparagine

Asp — aspartic acid

bicine — N,N Bis(2-hydroxyethyl) glycine

C - cysteine

Cys - cysteine

CC - correlation coefﬁcient

C terminal — carboxy terminus

CHO cells — Chinese hamster ovary cells
con-G - conantokin G

con-T - conantokin T

D — aspartic acid

E — glutarrric acid
EACA — e-aminocaproic acid

F — phenylalanine

Fcalc -— calculated structure factors

FH — structure factor contribution of the derivative
Fhk. — structure factor for a reflection labeled hkl
Fobs — observed structure factors

Fm - structure factor for protein plus derivative
FAK - focal adhesion kinase

G —glycine

Gla - 7-carboxyglutamate

Gln - glutamine

Glu - glutamic acid

glu-Pg — full length plasminogen
Gly - glycine

H -— histidine
hepes — N-[2-hydroxyethyl] piperazine-N’-[ethane sulfonic acid]
His - histidine

I — isoleucine

xix

I — intensity
150 - isoleucine

K — lysine
Kg — kringle
L — leucine

LBS —- lysine binding site
Leu - leucine

LLG - log likelihood gain
Lys - lysine

M — methionine

Met - methionine

MPD -— 2-Methyl-2,4—pentanediol
mL — milliliter

min - minute

N - asparagines

NOE —- nuclear overhauser effect

N terminal - amino terminal

NTD — N terminal domain

NMDA receptor - N-methyl-D-aspartate receptor

P — proline

PAM - plasminogen binding group A Streptococcal M-like surface protein
PEG - polyethylene glycol

Phe - phenylalanine

PP - phasing power

Pm — plasmin

Pg — plasminogen

PGK - phosphoglycerate kinase

Pro - proline

PSA — prostate speciﬁc antigen

Q — glutamine
R — arginine
rrnsd - root mean square deviation

S — serine

SAD — single wavelength anomalous dispersion

SBC - structural biology center

SDS-PAGE - sodium dodecyl sulfate polyacrylamide gel electrophoresis
Ser - serine

SeMet — selenium methionine

SIR - single isomorphous replacement

XX

SK - streptokinase

T — threonine

Thr - threonine

TM - transmembrane

tPA — tissue type plasminogen activator

Trp - tryptophan
Tyr - tyrosine

uPA —- urinary type plasminogen activator

V — valine
Val - valine

W — tryptophan
WT — wild type

Y - tyrosine

xxi

Images in this thesis/dissertation are presented in color.

Chapter I

Introduction

1.1 Introduction to the Angiostatin/VEK-3O Complex Structure
1.1.1 Plasminogen

A critical reaction in the generation of the ﬁbrinolytic response is the production
of the serine protease plasmin (Pm) from the activation of the zymogen, plasminogen
(Pg). Pm catalyzes the proteolysis of the ﬁbrin network, resulting in the dissolution of
blood clots. Conversion of Pg to Pm results from cleavage of the Arg561-Va1562
peptide bond by tissue-type plasminogen activator (tPA) or urinary-type plasminogen
activator (uPA). Cleavage at this site in Pg results in the formation of two-chain Pm,
which is composed of a heavy chain and light chain linked by two disulﬁde bonds. The
heavy chain consists of the N-terminal domain (NTD) followed by ﬁve consecutive
homologous triple disulﬁde-bonded kringle (Kg or K) domains. The light chain

possesses the C-terminal serine protease catalytic unit (Figure 1-1)7.
W

m \
1‘ I I

Wu:

0 9
09¢

Figure 1-1 Schematic representation of the structure of human plasminogen.

Kringle domains are found in a variety of proteins including Pg8, prothrombing,
the unactivated form of thrombin, tPAlo and uPAll and apolipoprotein A (apoA)12.
Kringle modules have been shown to be protein recognition modules in virtually all
cases where a function has been identiﬁed. Kringle domains are compact structures that
are held together by three critical disulﬁde bonds. An example of the overall kringle
domain structure is shown in Figure 1-2. The ﬁve kringle domains of plasminogen
share signiﬁcant sequence and structural homology. A superposition of the Pg kringle

domains yields an rmsd value of about 0.40 A.

 

Figure 1-2 The structure of plasminogen K5 (PDB ID 5HPG)3. The
disulﬁde bonds are colored red.

Many of these kringles display an afﬁnity for C-terminal lysine residues on proteins, and
for small molecules such as epsilon-amino caproic acid (EACA) that mimic C-terminal
lysines'3'15. Their binding modes for lysine-like ligands have been extensively studied
both structurally and by site-directed mutagenesis. The speciﬁc region responsible for
ligand-binding is the lysine-binding site (LBS). Structures of several kringle domains
bound to EACA are known and deﬁne the LBS as bipolar with a cationic and anionic
center that stabilizes the carboxy] and amino group of a C-terminal lysine residue2'3‘16’l7.
The LBS is deﬁned by residues R115, R153, D137 and D139 in K1; R234, D219 and
E221 in K2; R290, R324, D309 and H317 in K3; D411, D413 and R426 in K4; D516,
D518 and R530 in K52. Between the two charged regions of the binding site is a
hydrophobic region consisting of two aromatic residues that act as forceps for the
intervening hydrophobic methylene chain of the C-terminal lysine. Figure 1-3 shows
K1 bound to EACA. Comparison of all kringle structures known to date shows that all
the structures are quite similar with root mean square deviations of no more than 1-2 A.
Pg-Kl, -K4, and -K5 all show reasonably high afﬁnity for EACA while Pg-K2 has
signiﬁcantly lower afﬁnity for EACA. Pg-K3 displays no afﬁnity for any of the C-
terminal lysine mimics”. However, speciﬁc and high afﬁnity interactions between Pg
kringle domains and proteins lacking C-terminal lysines have also been identiﬁed.

. . -7
These include tetranectln,19 “1

which binds to Pg—K4, the streptococcal surface protein
PAM (Pg-binding group A streptococcal l\_/I_-like protein),6’22'24 which binds to Pg-K2,
DANCE (developmental arteries and neural crest epidermal growth factor), which binds

to Kringle IV type 2 domains of apo(a)25, factor Va, which binds to K1 and K2 of

 

Y156

Figure 1-3 LBS of Kl-EACA (PDB ID lCEA)2. K1 is green and the
residues are colored by atom (nitrogen, blue; oxygen, red; carbon,
green) and EACA is colored magenta. Side chains are labeled using
plasminogen numbering.

26’27, and the NG2 proteoglycan, which binds to angiostatinzs’zg. A ligand

prothrombin
for Pg-K3 has yet to be identiﬁed. The ﬁve kringle domains that make up most of the
N-terminus of Pg are the most studied examples of kringle domains.

The interaction between the C-terrrrinal lysine of ﬁbrin and the lysine binding
kringles of Pm or Pg serve to localize Pg to ﬁbrin, thereby promoting proteolytic
dissolution of the ﬁbrin knot by Pm30. The ﬁve Pg kringle domains have an additional
regulatory function involving a dramatic conformational change31‘32. The full-length Pg
(glu-Pg) exists as a tightly compact structure in the presence of Cl' and is relatively

6.31.33-35

inactive toward activation by Pg activators . Upon EACA binding this structure

becomes signiﬁcantly less compact and far more susceptible to activation“. This
compact conformation depends, at least in-part, on interactions between Pg-K5 and the
N-terminal 77-residue domain (NTD), since recombinant Pg bearing LBS mutations in
KS or N-terrninal lysine substitutions exist only in the extended conformation and show
no activation in response to EACA”. The LBS's of K1 and K4 also play an important
role in the maintenance of the closed conformation, though the target for these binding
sites is less clear. Therefore the ﬁve kringle domains of Pg must be capable of
signiﬁcant structural rearrangement relative to one another.
1.1.2 PAM

The group A streptococcal surface protein PAM, a 43 kDa member of the M-
protein family, binds Pg-K2 with high afﬁnity”. M and M-like proteins account for
several interactions between group A streptococci and plasma proteins and are known to

38.39

act as virulence factors by inhibiting phagocytosis In fact, PAM is required for

infection by several streptococcal strains'w‘“.

It acts by localizing Pm to the bacterial
surface, inhibiting fibrin encapsulation during infection. Pm then catalyzes both extra-
cellular matrix and ﬁbrin degeneration, thwarting bacterial encapsulation during
infection. M and M-like proteins are highly related structurally, consisting of a
continuous OL—helix encompassing most of the structure, with a membrane-binding
domain on the N—terminus. The continuous (Jr—helices then dimerize, forming extended
parallel coiled-coil structures that extend tens of angstroms from the cell surface. These
M-protein protrusions are easily visualized in BM projections”. A region of PAM,

spanning amino acids 91 - 116, contains two direct repeat sequences and is responsible

for Pg binding by PAM. VEK-30, an oc-helical peptide derived from residues 85-113 of

PAM and containing the ﬁrst and most of the second direct repeat, possesses a high
afﬁnity binding site for Pg-K2 (Kd = 460 nM),6’23‘24'38 even though it does not contain a
C-terminal lysine residue. Figure 1-4 shows the primary sequence of VEK-30.
VEK-30:
85 95 105

VEKLTADAELQRLKNERHEEAELERLKSEY

Figure 1-4 Primary sequence of VEK-30. The underlined residues are the direct repeat.

VEK-30 specifically binds Pg-K2, having no measurable afﬁnity for any of the
other isolated Pg kringle domains and displays almost identical afﬁnity for K2 compared
to the full length protein, indicating that it contains most of the PAM protein's interface
with K2334. The crystal structure of Pg-mK2 (Figure 1-5), mutated to contain an
upregulated (an order of magnitude higher) lysine binding site (C169G/EZZlD/L237Y),
bound to VEK-30 demonstrates a novel kringle LBS interaction, where the C-terminal
lysine is mimicked by an argininyl and glutamyl side-chain residues displaced by almost
one helical turn". This arrangement of residues is named a “pseudo-lysine” moiety.
This structure shows that kringle domains can make interactions with a variety of
bipolar protein ligands and enlarges signiﬁcantly the possible targets for kringle
interaction.

1.1.3 Angiostatin
Angiostatin is an internal fragment of Pg containing the ﬁrst three or four kringle

modules43’44.

It was one of the ﬁrst angiogenesis inhibitors to be identiﬁed.
Angiogenesis is the sprouting of blood vessels from pre-existing capillary beds and is

essential for development and wound healing. In 1994, Folkman and coworkers ﬁrst

identiﬁed the plasminogen fragment angiostatin containing K1-4 through its anti-tumor
effects in mice“. Angiogenesis is critical for the growth of most solid tumors, as a

blood supply must be recruited to stimulate signiﬁcant growth45'46.

For this reason,
intense interest has been focused on angiogenesis inhibitors for use as potential anti-

cancer agents.

 

Figure 1-5 X-ray crystallographic structure of mK2NEK-30 (PDB ID 115K)6. mK2
is colored green and the VEK-30 peptide is colored magenta. mK2 side chains are
labeled using plasminogen numbering and VEK-30 pseudo-lysine labeled using
PAM numbering.

Although the function of angiostatin in angiogenesis inhibition is uncertain, agents
containing Pg-Kl-3, -K1-4, and -K1-5 Show potent anti-angiogenic and/or anti-tumor
growth activity in animal models. These fragments, as well as individual kringle
modules, are also inhibitory toward endothelial cell migration and/or proliferation in
vitro. Later studies showed, however, that angiostatin corresponding to Kl-3 engenders
all the determinants responsible for maximal inhibition of cell proliferation and

motility“.

The Production of Angiostatin

Until recently, the extracellular component involved in disulﬁde reduction of Pm
in Chinese hamster ovary cells (CHO cells) or HT1080 cells remained elusive. Hogg
and coworkers discovered that the Pm disulﬁde bonds were reduced by a reductase
secreted by CHO or HT 1080 cells and that this disulﬁde bond reduction triggered the
proteolysis of Pm generating an giostatin48. They believed that Pm reduction is the ﬁrst
step in the formation of angiostatin fragments consisting of kringle domains 1-3, 1-4,
and 1-41/248‘5 2. The proteolysis of Pm is thought to occur in three stages and a model
was proposed by Hogg and coworkers48'52. The model proposes that the ﬁrst step
involves the reduction of disulﬁde bonds in KS by a Pm reductase. The Pm reductase
requires reduced glutathione or cysteine as cofactors. After disulﬁde reduction,
proteolysis occurs at peptide bonds in KS by a serine proteinase producing K1-4I/2.
Finally, the fragment kringle domain consisting of Kl-4'I2 is cleaved by various matrix

metalloproteinases to produce either kringle fragments 1-3 or 1-453'58. The Pm reductase

involved in Pm processing has been shown to be phosphoglycerate kinase (PGK), a
glycolytic enzymesz. High levels of PGK have been observed in plasma of mice bearing
ﬁbrosarcoma tumors and treatment with recombinant PGK inhibits tumor growth
between 50-70%, depending on the tumor type. The reduction of Pm by PGK is thought
to occur independently of thiolsso.

Other molecules exist as well that mediate the cleavage of plasminogen to
angiostatin. The proteolytic activity of elastase catalyzes the production of angiostatin

. . . 4 . 9
1n Lewrs lung carcrnoma 35.

Tumor cells are believed to up regulate elastase
production. In human prostate carcinoma, Pm generates angiostatin in the presence of
free sulfhydral donors“). Prostate speciﬁc antigen (PSA) has also been shown to cleave
plasminogen to form angiostatin in vitro‘“. Most recently PSA has been shown to be an
inhibitor of angiogensiséz. Thioredoxin and protein disulﬁde isomerase are also thought
to be involved in the generation of angiostatin“.
The Structure of Angiostatin

The X-ray crystallographic structure of angiostatin was previously determined to
a resolution of 1.75 A63. As shown in Figure 1-6, the three kringle domains of
angiostatin come together to form a triangular bowl-like structure. An inter-kringle
disulﬁde bond between K2 residue C169 and K3 residue C297 contributes signiﬁcantly
to the relative orientation of K2 and K3, while Kl's position is somewhat constrained by
the short three-residue inter-kringle peptide linking K1 with K2. However, ﬁxing the
orientation of K2 and K3 by disulphide linkage is not required for angiostatin's activity

as disruption of the linkage has little effect on its antiproliferative activity“. Numerous

interactions between the inter-kringle peptides and the kringles appear to stabilize the

arrangement seen in the structure. However, there are no direct interactions between any
of the three kringle domains, which raises the possibility that signiﬁcant motion may
occur between the three domains. This motion is limited somewhat by the inter-kringle
disulﬁde bond between K2 and K3.

The crystal structure of angiostatin provides important insights regarding the
varying binding site preferences of the three kringle LBS's because each of the three
LBS's is bound to a molecule of bicine (N ,N Bis (2-hydroxyethyl) glycine) in the
structure63 (Figure 1—6 and 1-7). Interestingly, the LBS of K1, which does have high
afﬁnity for C-terminal lysine residues, is located on the opposite face of the molecule

relative to K2

 

Figure 1-6 The overall structure of angiostatin (PDB II) IKIO). The
interkringle disulﬁde connecting K2 and K3 (C169-C297) is shown in
red. The green spheres are bicine molecules.

10

 

Bicine EACA

Figure 1-7 The left molecule is bicine and the right molecule is EACA. Both are
colored by atom.

Front Back

Kl

        

    

K2 K3

K3

Figure 1-8 Space-ﬁlling view of angiostatin. The LBS in each of the three
kringles is colored red. All other atoms are blue.

and K3 (Figure 1-8). Since the structures of K1 bound to EACA2 and K2 bound to
VEK-306 are known, detailed comparisons can be made between the structures. The Kl
LBS (Figure 1-9) binds bicine in an orientation reminiscent of its interaction with
EACA, the carboxylate of the bicine interacting with the positive end of the LBS
deﬁned by two positively charged residues and the hydrophobic chains of bicine
tracking through the V-shaped cleft formed by two aromatic side chains (W144 and
Y154). One of the bicine hydroxyl groups then makes a hydrogen bond with one of the
aspartates (D137) that deﬁnes the negative end of the bipolar binding site. The situation
is similar in the K2 LBS (Figure 1-10), with a bicine similarly oriented in the LBS.
However, D219, one of the two acidic residues that deﬁne the negative end of the K2
LBS, is ﬂipped out of the binding site of K2 and instead makes a tight salt bridge with
R220, a residue that is not conserved in other kringle domains. The altered
conformation of D219 may explain the lower affinity for EACA that K2 displays when
compared to other lysine-binding kringle domains. In contrast, D219 is found to be
flipped into the LBS in the structure of K2 bound to the VEK-30 helix. D219 is ﬂipped
into the active site in this structure, recapitulating the LBS while R220 does interact with
VEK-30. In fact, if the K2 structure of angiostatin is overlayed onto the K2NEK—30
complex structure (Figure 1-11), a collision occurs between the D219/R220 salt bridge
and Q95 and L94 of VEK-30. We surmise that binding of angiostatin to VEK-30
induces this conformational "switch" recapitulating the LBS of K2 by sterically
interfering with this salt bridge and interacting with D219 and R220.

Although the K3 LBS (Figure 1-12) is also bound by bicine, the binding is

completely different. This is because the K3 LBS is structurally distinct from all other

12

kringle domain LBS's known. One of the two acidic residues that deﬁne the negative
end of the LBS is mutated to lysine in K3. This lysine (K311) not only neutralizes the
negative end of the LBS, it also tracks across the hydrophobic cleft through the middle
of the LBS occluding half of the LBS. Overlaying the K3 LBS over the K1/EACA
structure shows that a signiﬁcant collision would occur between K311 and EACA in the
K3 LBS (Figure 1-13), explaining its complete lack of afﬁnity for EACA or C-terminal
lysines. However, bicine is still able to bind in the LBS, but it does so by rotating
almost 90 degrees relative to its orientation in K1 and K2, inserting its carboxylate
group into the remaining binding pocket, which is now essentially positively charged.
This interaction may mimic the interaction of K2 with carboxylate side chains from
other proteins. Taken together, this indicates that the cleft between K2 and K3 may act
by binding to opposite sides of an angiostatin ligand, with the K2 side interacting with
some bipolar surface possibly reminiscent of VEK-30's arrangement of glutamate and
arginine residues one helical turn away from one another, while the K3 side of the cleft
would interact with a protruding carboxylate sidechain. Together, the two kringles K2
and K3 may act like "molecular forceps" speciﬁcally interacting with a ligand on two
sides. The role of K1 in this binding model is not clear as its LBS is on the opposite

face of angiostatin relative to the K2 and K3 LBS's (Figure 1-8).

13

R117
Y154
F118
Y156 \ D137

D139

Figure 1-9 Interaction of the angiostatin Kl LBS with bicine. K1 residues are
colored green and the bicine molecule is colored magenta. Residues are labeled
with plasminogen numbering and atom coloring.

4 D219
E221 ‘ ‘

Figure 1-10 Interaction of the
angiostatin K2 with bicine. K2 is
colored green and bicine is colored
magenta. Residues are labeled with
plasminogen numbering and atom
coloring.

 

Figure l-11 Overlay of the
angiostatin K2 onto the mK2/VEK-
30 structure. Residues are labeled
with plasminogen numbering and
atom coloring.

 

 

Figure 1-12 Interaction between angiostatin K3 and bicine. K3 is colored green
and bicine is colored magenta. Residues are labeled with plasminogen
numbering.

15

 

Figure 1-13 Overlay of K1/EACA onto angiostatin K3. K3 is colored blue
and K1 is colored orange. EACA is colored lavender. Residues are labeled
with plasminogen numbering.

How Does Angiostatin Inhibit Angiogenesis

Blood vessels can be formed by two different processes, vasculogenesis and
angiogenesis. Vasculogenesis is the formation of new blood vessels when there are no
pre-existing ones. Speciﬁcally, vasculogenesis involves the formation of new blood
vessels through the differentiation of precursor cells into endothelial cells, forming a
vascular network“. Vasculogenesis occurs during embryonic development. However,
angiogenesis is the process by which blood vessels form from pre-existing capillary
beds“. Angiogenesis is triggered by a stimulus that forces endothelial dormant cells
into the cell cycle. This process starts with basement membrane degradation in

endothelial cells and the formation of the lumen. Simultaneously, the endothelial cells

change morphology, proliferate, migrate, form microtubes, and sprout new capillaries.
This process is highly regulated involving muliple controls that can be either turned on
or off. Both activators and inhibitors of angiogenesis are often found together in tissues
that are both active in angiogenesis and those that are quiescent leading to the notion of
an "angiogenic switch," where the switch is "off" when angiogenic inhibitors
predominate and "on" when the balance is tilted toward activators“.

Angiogenesis is essential for development and wound healing. While it is
intimately involved in the development of the vascular system, angiogenesis is for the
most part not active in adults. There are, however, several pathologies that appear to
require aberrant angiogenesis. These include rheumatoid arthritis, diabetic retinopathy,
cardiovascular diseases, and tumorigenesis”. In rheumatoid arthritis, capillaries invade
the joints and destroy cartilage. In diabetic retinopathy, new blood vessels invade the
retina and cause blindness. Ocular vascularization is the most common cause of
blindness. Tumor growth and metastasis depend on angiogenesis as well. Avascular
tumors can not grow beyond 2-3 mm3 without recruiting its own blood supply. But once
the tumor is vascularized, the tumor grows rapidly45‘46. Not only do the blood vessels
within the tumor promote tumor growth, it also serves as a portal for tumor cells to enter
the blood stream and metastasize.

Once it was determined that tumor growth and metastasis were angiogenesis
dependent processes, it was proposed that blocking angiogenesis was a strategy for
cancer treatment. Antiangiogenic therapy is a form of cancer treatment that uses
angiogenesis inhibitors that speciﬁcally stop new blood vessel growth, which in turn

starves the tumor. Unlike conventional chemotherapy, antiangiogenic therapy attacks

17

only new blood vessel cells selectively without affecting normal vasculature and does
not kill cancer cells directly. Chemotherapy kills all cells which leads to severe side
effects. In order to keep cancers from regrowing, patients may need to take
angiogenesis inhibitors for the rest of their lives. Currently, more than 300 angiogenesis
inhibitors, including angiostatin, have been discovered so far. Angiostatin has been
shown to inhibit tumor growth and metastasis dissemination in animal models“.
Angiostatin causes no side effects, toxicity, or weight loss and is currently in phase II
clinical trials at the Indiana University School of Medicine.

In Spite of the intense interest in angiogenesis inhibitors, little is known about the
precise mechanism of action of angiostatin or indeed of most of the other angiogenesis
inhibitors, but several potential targets of action have been identiﬁed. Pizzo and
coworkers have discovered that angiostatin binds directly to the mitochondrial F(1)-F(0)
ATP synthase and further that the ATPase can be found on the surface of endothelial
cells“. They have also shown that angiostatin is a potent inhibitor of both the cell-
surface and puriﬁed mitochondrial forms of the enzyme“. The authors speculate that
endothelial cell-surface F(1)-F(0) ATP synthase plays an important role in the
maintenance of intracellular pH in the acidic tumor cell environment68'69. Angiostatin's
role is to inhibit the proton pump, allowing the intracellular pH to drop, triggering
ap0ptotic events in endothelial cells69’70.

Alternatively Takada and coworkers have found that angiostatin speciﬁcally
binds to the angiogenesis activator Oth3 integrin on the surface of CHO cells and bovine

arterial endothelial cells (BAE cells) in an EACA dependent fashion". Further, they

have shown that Pm, in contrast to Pg, is also a ligand for OLVB3 integrin and that it

18

induced endothelial cell migration”. Angiostatin, an RGD-containing peptide (a known
ligand and inhibitor of (XVB3) and a serine protease inhibitor all effectively blocked Pm-
induced cell migration in these assays”. The mechanistic conclusion to be drawn from
these studies is that localization of Pm's protease domain to the integrin on the cellular
surface is necessary for its activity. Angiostatin, and potentially many of the other Pg
kringle domains, would then block this association. These results are quite intriguing as

(XVB3 integrin is a known angiogenesis activator, though it is not absolutely required for

angiogenesis because avB3-knockout cells still show all the hallmarks of angiogenesis”.

A third potential target for angiostatin has also been identiﬁed. Angiomotin was
originally identiﬁed in a yeast two hybrid screen using a placental cDNA library".
Paradoxically, angiomotin does not appear to be a cell surface receptor nor does it
contain a signal sequence for secretion, but nevertheless appears to bind angiostatin on
the surface of endothelial cells. Angiomotin is thought to consist of conserved coiled-

75

coil and PDZ binding domains . Cells that contained angiomotin were able to both

bind and internalize angiostatin leading to an induction of focal adhesion kinase (FAK)
activity”. FAK regulates cell motility and adhesion-dependent cell survival and FAK
plaques are involved in proton transport76’77. Independently, angiostatin was previously
shown to activate FAK activity possibly resulting in the inhibition of migration and
apoptosis78. However, little else is known regarding the function of angiomotin.
Interestingly, an exposed helix with a positive and negative residue separated by
approximately one helical turn is found on the B3 subunit of (XVB3 integrin, where lysine

and glutamate are found on the exposed Side of the helix shown in Figure 1-1479.

Simply docking angiostatin onto this helix, assuming a similar conformation and

19

interaction to that seen in the K2/PAM-30 structure indicates that angiostatin could bind
in this position and would signiﬁcantly occlude the integrin's binding site (Figure 1-15).
However, without further data this postulate remains highly speculative. It is also of
interest that OLVB3 integrin and angiostatin complexation is potentially inhibited by
concentrations of EACA sufﬁcient to fully saturate both K1 and K2 LBS's71. A similar
K2 binding site can also be found on the exposed C-terminal helices of both or and B
subunits of Fl-ATPase encompassing residues K496 and E499 or K472 and E475 in the
or and [3 subunits respectively (Figure 1-16)80. Unfortunately, the importance of the
LBS's for binding of angiostatin to either angiomotin or F(1)-F(O) ATP synthase has yet
to be determined by EACA competition experiments.
1.1.4 Physiological Importance of the Structure
To further understand kringle domain function and kringle domain speciﬁcity,
angiostatin bound to the VEK-30 peptide was crystallized and its three-dimensional
structure determined to a resolution of 2.3 A. The angiostatinNEK-30 structure is the
ﬁrst example of an interaction between a multiple kringle-containing protein and a
biologically relevant ligand to be visualized at atomic resolution. Furthermore, as the
only existing structure of an angiostatin/ligand complex, it provides a model at atomic
resolution to test the genesis of anti-angiogenic activity following angiostatin/1i gand
binding. Structures of other multi-kringle domain complexes will be required to
understand how these domains cooperatively interact and undergo conformational
rearrangement in order to provide a better understanding of plasminogen structure and

function. Plasminogen is an important serine protease that binds to the surface of

bacteria. An interesting example resides with ﬂesh-eating bacteria. In order for the

20

 

Figure 1-14 Close view of the section of OIVB3 integrin that harbors the residues
involved in possible angiostatin binding. The B3 subunit is shown in red and the
residues are shown in green with atom coloring.

 

Figure 1-15 Top view of (XVB3 integrin with angiostatin docked onto a helix in the
83 subunit assuming the conformation seen in the mK2NEK-30 structure. 0!.ng
integrin is silver, angostatin blue, residues in angiostatin binding are colored
magenta, residues in the B3 subunit responsible for binding are green while the
residues in the av subunit are shown in yellow.

21

a)

e)

 

Figure 1—16 (a) Close view of the section of the alpha subunit of Fl-ATPase that
harbors residues for possible angiostatin binding. (b) Structure of bovine mitochondrial
Fl-ATPase (131). (c) Enlarged view of the beta subunit Fl-ATPase that harbors
residues for possible angiostatin binding.

bacterial cell to invade the extracellular matrix of human skin, it must bind an abundant
serine protease, plasminogen. The structure of the angiostatin/VEK-30 complex will
provide further insight into plasminogen function during bacterial infection. A

mechanism for plasminogen function during bacterial infection is proposed herein.

22

1.2 Introduction to the Metal Bound Structures of con-G and con-T[K77]
1.2.1 Protein Design

Knowledge of the interactions that stabilize local protein structure is critical for a
global understanding of protein folding. To this end, much effort has been invested in
resolving the physical and chemical principles that direct the formation of the a—helix,
the most widespread secondary structural motif, and its supersecondary and tertiary
structural derivates, which include coiled-coils and helix bundles. Short (=3 20 residues),
linear peptides corresponding to structured segments of naturally occurring proteins
would clearly be instructive in this regard, owing to the ease with which their primary
sequences can be manipulated and their structures interpreted. However, excepting a

few notable examples“82

, when examined in isolation in aqueous milieu, such
abbreviated protein segments rarely adopt the conformations observed within the context
of the full-length protein. The de novo peptide/protein design ﬁeld has largely emerged
to overcome this problem. Within this discipline, considerable success has been
achieved in the design of peptides that are sufficiently short to be synthetically practical,
yet capable of adopting stable helices that can further organize into higher order
complexes. Among the various strategies that have been employed for directing the
formation of monomeric and/or interacting helices, the most tractable and effective
approach involves the use of metal ions. For instance, helix nucleation and the attendant
folding of unstructured or partially structured peptides can be readily facilitated through
metal ion bridging of appropriately spaced natural and unnatural metal-binding amino
acid383'85. In these particular cases, introduction of the metal induces 80-90% helicity in

83.84

a l7-mer and over 80% helicity in an ll-residue peptidegs, depending on the metal

23

ion. Similarly, bridging interactions have been exploited to stabilize contacts between
two or more helical strands. Many of these studies have focused on the metal-assisted
assembly of prefolded de novo structures via pendant metal ligands, namely porphyrin
and bipyridyl derivativesS6'87.
A more daunting challenge in de novo construction of higher order structures
involves the metal-assisted organization of random coils into helix-helix assemblies.
This has been accomplished by induction of a triple-stranded coiled-coil following
transition metal ion binding to His or Cys residues contained within a four heptad repeat
peptide of a random structure88‘89. Further, design of Zn2+-chelating sites into a model
four-helix bundle resulted in a highly stabilized, ordered structure compared to the apo-
variant90. A particularly dramatic example of metal ion-induced folding was the design
of two disulﬁde-linked 35-mers that underwent a random coil to coiled-coil transition in
the presence of various metal ions”. The molecular interactions pivoting this self-
organization is interstrand metal ion bridging between 7-carboxyglutamate (Gla)
residues occupying positions e and g in the a—helical heptad repeat. In all of these
examples, the metal ion is crucial to initiating strand association, but the predominant
forces driving and stabilizing complex formation rely on interstrand hydrophobic
contacts. In some of these studies, covalent bonds between the helix-forming strands are

necessary for complex stabilization”91

. Hence, the metal-ion triggered association of
peptide chains in these instances cannot be regarded as unqualiﬁedly autonomous.
Additionally, structural characterization of such complexes has proven elusive, raising

questions as to the precise nature of stabilizing interhelix interactions.

24

A substantial body of evidence has recently been analyzed to suggest that the
challenge of generating a small, unstructured peptide capable of metal ion-triggered
helix formation and self-association has been satisﬁed by natures. This peptide is known
as conantokin-G (con-G) and is described further in this thesis.

1.2.2 Overview of Cone Snails and Conotoxins

The cone snails (genus Conus) are venomous marine mollusks that use small
structured peptide toxins (conotoxins) for prey capture, defense, and competitor
deterrence”. Each of the approximately 500 different Conus species can express about
100 different conotoxins. In order to efﬁciently impose the desired physiological effect
on the prey, multiple conopeptides act together in a synergistic fashion. Cone snails can
be deadly to humans, speciﬁcally, the species Conus geographus. The recognition that
this snail could kill people led to the initial scientiﬁc investigation of cone snail venoms.

There are two broad divisions of cone snail venom components: disulfide-rich
conopeptides and non-disulﬁde rich conopeptides. Conopeptides are usually small
containing 12-30 amino acids unlike other polypeptidic toxins from other venoms 40-80
amino acids in length. Conopeptides have a high frequency of posttranslationally
modiﬁed residues including hydroxyproline, D-amino acids”, sulfated-tyrosine94, 6-Br-
tryptophan”, O-glycosylated serine or threonine96, and 7-carboxy-glutamic acid”.

The molecular targets of the individual components of Conus venom are diverse,
including G protein-coupled receptors and neurotransmitter transporters. However, the
majority of the Conus venom components that have been presently characterized target

either ligand-gated ion channels or voltage-gated ion channels. These peptides are able

25

to discriminate between molecular isoforrns within a certain ion channel family. This
unprecedented selectivity is an important tool for deﬁning ion channel function.

The voltage-gated ion channel family consists of membrane-bound proteins
activated by changes in the transmembrane voltage. Their most important physiological
role involves the generation, shaping, and transduction of the cells’ electrical signals.
They are generally divided into Na+, K+, and Ca++ channels. Upon activation to the open
state, voltage-gated ion channels undergo a conformational change which allows cations
through the pore of the channel. An additional conformational change to a closed state
causes deactivation and thereby a nonconducing state. Three different conopeptide
families are known to target voltage—gated sodium channels: u-conotoxins, 5-
conotoxins, and pO-conotoxins. lc-conotoxins target K channels and (o-conotoxins
target Ca channels.

The ligand-gated ion channels are membrane-bound proteins that mediate fast
synaptic transmission. One major group of ligand-gated ion channels are those activated
by acetylcholine, serotonin, GABA, or glycine. These all belong to the same gene
superfamily. a-conotoxins and w-conotoxins are known to target the nicotinic receptors.
The other gene superfamily of ligand-gated ion channels is the glutamate receptors,
including the N—methyl-D-aspartate (NMDA) and non-NMDA (kainite/AMPA)
receptors. The conantokins are an unusual class of Conus peptides containing a high
content of 7—carboxyglutamate (Gla) residues and no Cys residues and are antagonists of

the NMDA receptors. A third family of ligand-gated ion channels is the ATP receptors.

26

1.2.3 con-G

The peptide, conantokin-G (con-G), is a component of the venom of the
predatory marine snail, Conus geographusgg. It was originally puriﬁed using a
behavioral assay. The peptide put young rrrice to sleep elucidating the nickname
“sleeper peptide” but made older mice hyperactivegg’loo. In mammals, con-G
antagonizes the N-methyl-D-aspartate (NMDA) subtype of ionotropic glutamate
receptors, the dysfunction of which is linked to numerous chronic and acute
neuropathologies such as acute neuronal cell death after ischemic stroke and Ca2+-
related neurodegeneration in Parkinson’s and Alzheimer’s diseasesm'. Con-G is 17
arrrino acids in length, contains no disulﬁdes, and contains ﬁve Gla (7) residues at
positions 3, 4, 7, 10 and 14. Figure 1-16 shows the primary sequence of con-G. Of
these, only Gla4 is an absolute requirement for NMDA receptor activity'oz. Gla is
synthesized from glutamate by a vitamin K-dependent 7-glutamyl carboxylase. The
discovery of Gla in con-G established for the ﬁrst time the presence of Gla in
invertebrates103 . Because of the absence of disulﬁdes in this peptide, it is the Gla
residues that provide the structural framework for forming a helical conformation
through an interaction with a variety of metal cations. Table 1.1 displays the metal ion
effects on the secondary structure and Malpp of con-G and con-G analogs”. While a
variety of di- and trivalent metals such as Ca2+, Mgz”, and Zn2+ can promote a
conformational change from random coil to a helix content of 50% and higher, only Ca2+
allows for the formation of a dimeric con-G complex with a stoichiometry of 2-3

molecules per mole of peptides. The Ca2+-induced dimerization of con-G was also

27

shown to be dependant on pH and calcium concentrations. The measured binding
constant, Kd, of con-G and calcium is 2.8 mMm. In the absence of metal cations, con-
G has virtually no secondary structure. The calculated apo monomeric molecular weight
of con-G is 2265 Da. Interestingly, metal cations such as Mg2+ bind tighter to con-G but
do not induce metal-mediated dimerization. Thiol-disulﬁde rearrangement experiments
with Cys-containing con-G variants are congruous with an antiparallel alignment of
peptide chains in the Ca2+-containing dimer, and Ala-replacement studies have
conﬁrmed the participation of Gla residues 3, 7, 10 and 14 in the dimerization event.
From these data, a model for the complex in which antiparallel con-G strands are
stabilized was proposed solely through Ca2+-bridging of Gla headgroups within the
helix-helix interface (Figure 1-18). This model represents a heretofore unknown motif
among natural] y-occurrin g proteins or peptides.
1.2.4 con-T

After the discovery of con-G, another member of the conantokin family
conantokin-T (con-T), from the marine snail Conus tulipa was isolated105 . Con-T shares
sequence identity with con-G at eight sequence positions, including four positions
occupied by Gla residues. For con-T, Gla residues exist at positions 3, 4, 10, and 14.
However, unlike con—G, con-T at position seven contains a Lys residue instead of a Gla.
Also, con-T is a longer peptide containing 21 residues whereas con—G only has 17
residues. Figure 1-16 shows the primary sequence of con-T. The two peptides also
have pronounced secondary structural characteristics. As stated above, con-G exhibits
rrrinimal structure in its metal-free form, yet adopts an end-to-end helix dimer structure

in the presence of divalent metal cationss. Table 1.2 shows the metal ion effects on the

28

Table 1.1 Metal ion effects secondary structure and Mapp of con-G and analogss. The percentage
of helicity was derived from CD measurements, the Mapp was derived from ultracentrifugation

experiments. and Kd was derived from ITC experiments.

 

 

Number
Peptide Metal Ion % Helix Mapp (Da) Kd (M) of
Molecules
Con-G None <2 2550
Ca2+ 50 3950 2800 23
Mg“ 72 2700 46 2
Zn2+ 68 3110 0.2 3
Mn2+ 67 2875 3.9 2
Ba~+ 7 4200
Sr’+ 11 3480
Con-G[73A] None 5 2330
Ca2+ 31 2530
Con-G[74A] None <2 2480
Ca2+ 36 3120
Con-G[77A] None 5 2480
Ca2+ 24 2650
Con-G[77K] None 10 2470
Ca2+ 32 2450
Con-G[710A] None 7 2410
Ca2+ 14 2310
Con-G[714A] None 10 2110
Ca2+ 27 2160

 

29

secondary structure of con-T and con-T analogs. Similarly, con-T also adopts a full 0t-
helical conformation in the presence of metal cationsm‘m'm. However, this change is
less dramatic since con-T already has high helical content in the absence of metal
cations. Another large difference exists between con-G and con-T. In the presence of
Ca2+, con-G undergoes calcium induced self-assembly forming an antiparallel dimeric
structure whereas con-T does not4'5. However, previous studies have shown that
replacing the Lys at position 7 with a Gla (con-T[K77]) allows the peptide to form an
antiparallel helix dimer in the presence of calcium". Interestingly, neither con-G nor
con—T[K77] forms a dimeric structure in the presence of Mgz” even though both peptides
have a higher afﬁnity for Mg2+ than Ca2+1'4’5. Con-T also has a lO-fold greater afﬁnity
for Ca2+ than con-G4. This is consistent with the notion of a pre-formed cation binding
site in con-T versus an induced cation binding site in con-G. A model for the Ca2+-
bridged con-T[K77] dimer, similar to that proposed for con—G (Figure 1-18), is presented

in Figure 1-19.

bcdefgabcdefgabcdefga

con-G G—E-7-7-L—Q— 7- N—Q—7—L—I— R-7-K-S- N-NH2
con-T G—E-7-7-Y-Q-K-M—L—7—N—L—R—7—A—E-V-K-K-N-A—NH2
con-T[K77] G~E—7—7—Y-Q— 7-M-L-7-N-L-R-7-A-E-V-K-K-N-A-NH2

Figure 1-17 The primary sequences of con-G, con-T, and con-T[K77]. The
a-helical heptad repeat assignments are shown above.

30

1.2.5 NMDA receptor

The N-methyl-D-aspartate (N MDA) receptor is a subclass of ionotropic
glutamate-dependent receptors that is widely distributed in the mammalian CNS'OB. The
NMDA receptor forms a cell membrane ion channel that controls the ﬂux of Ca2+ and
Na2+ into nerve cells. The receptor is activated only if glycine and glutamate bind in
concert and polyamines assist in the opening or closing of the ion channel, depending on

concentration. Protons, Zn2+, and redox agents further modulate the opening of the

Table 1.2 Metal ion effects secondary structure and Mapp of con-T and analogs".
The percentage of helicity was derived from CD measurements, the Mapp was
derived from ultracentrifugation experiments, and Kd was derived from ITC
experiments.

 

 

- Number
Peptide Metal Ion % Helix Mapp (Da) Kd (mM) Mogums

Con-T None 55 3 160

Ca2+ 72 3240 428 1

Mg2+ 82 3290 10 1
Con-T[K77] None 12 3390

Ca2+ 67 6660 120 2-3

Mg2+ 74 3330 18 1-2

Zn2+ 72 3750

Mn” 70 3700

13a2+ 50 10080

Sr2+ 52 7060

 

31

 

Lll

 

Figure 1-18 Helical wheel representation of the cross-sectional heptad repeat of the
antiparallel dimer of con-G in the presence of Ca”. For the left chain, the N-terminus
is closest to the viewer and for the right chain the C-terminus is closest to the viewer.
Gla7 and Gla14 of the left chain occupy position a, while Gla3 and GlalO occupy
position (1. Gla chelation pairs include Gla3-Gla14’, Gla7-Gla10’, GlalO-Gla7’,
Gla14-Gla3’. The lighter arrows are at the bottom of the wheel, progressively
thickening towards the top.

N20

 

 

N11
K18

Figure l-l9 Helical wheel representation of the cross-sectional heptad repeat of the
antiparallel dimer of con-T[K77] in the presence of Ca2”. For the left chain, the
N-terminus is closest to the viewer and for the right chain the C-terminus is closest
to the viewer. Gla7 and Gla14 of the left chain occupy position a, while G133 and
GlalO occupy position d. Gla chelation pairs include Gla3-Glal4’, Gla7-Gla10’,
GlalO-Gla7’, Glal4-Gla3’. The lighter arrows are at the bottom of the wheel,
progressively thickening towards the top.

32

NMDA receptor. The general features of the native form of the NMDA receptor are as
follows: a) a voltage dependent channel block by Mgz”; b) glycine and glutamate
coactivation; c) allosteric agonism by polyamines such as spermine and spermidine; d)
voltage-insensitive and voltage-sensitive allosteric blocks by Zn“; e) when the ion
channel is open high Ca2+ permeability; f) proton allosteric antagonism, g) sensitivity to
oxidizing and reducing agents; and h) phosphorylation regulationlm. The opening of the
ion channel is believed to be due to voltage-dependent relief of the Mg2+ channel block
which therefore allows Ca2+ to enter the cells'm‘m. When a stroke occurs, a detrimental
inﬂux of calcium initiates a cascade of events that will eventually lead to neuronal cell
death. Therefore, elucidating the effects of the mechanisms of the NMDA antagonists
and developing potent molecules to attenuate Ca2+ inﬂux are the focus of intense
research.

The action of con-G and presumably con-T as non-competitive antagonists of the
NMDA receptor has been attributed to the inhibition of the positive modulatory effects
of polyamines. They are the only known naturally occurring peptide ligands that
speciﬁcally antagonize this receptor. These peptides speciﬁcally inhibit polyamine-
stimulated dizolcipine (MK-801) binding in rat brain membranes with an ICso of about
500 anom 10, attenuate the NMDA-stimulated rise in neuronal intracellular Ca2+ (IC50 =
2 11M)”, decrease the NMDA-stimulated cGMP levels in rat cerebellar granule neurons
(ICso = 77477 nM)l ‘ 1, and inhibit glutamate-induced neurotoxicity1 1°. The inhibitory
mechanism for which con-G and con-T work involve their inhibition of the positive

effector roles of spermine1 10, glutamate and glycinem‘1 13.

33

The glutamate class of receptors are believed to be tetrarners consisting of two
types of subunits, NR1 and NR2”. So far, only one gene has been found to encode NR1
subunits, but several splice variants have been discovered. However, the four NR2
subunits, NR2A, NR2B, NR2C, and NR2D, are encoded by different genes. Based on
homology considerations with other receptors, a membrane binding model of NR1 has
been developed’m’n". This model includes an 18-residue signal peptide, followed by a
large N-terminal extracellular portion (543-residues) containing ten potential

115,116

glycosylation sites and regions that inﬂuence Zn2+ and spermine function and

”7" 18 and possibly glutamate binding sites1 '9. However, there is some

glycine
speculation that the glutamate binding site resides in the NR2 subunitlzo. Downstream
of this extracellular region, a transmembrane (TM) domain of about l9-residues is
present followed by a l9-residue linker and a second transmembrane-like domain, TM2.
TM2 is thought to exist as a hairpin structure and also contains Asn616, a residue
responsible for full expression of the Mg”, Zn”, and MK-801 blocks as well as Ca2+
permeability12 ”23 . Next, a relatively short linker region exists followed by a third TM
domain that spans the membrane. A large 163-residue linker exists after TM3 and is

thought to contain a region of the glycine binding site”"’125

and a site that partially
controls responses to spermine and protonsm. Finally, a fourth TM domain is present
along with an intracellular C-terminal region that contains sites that become
phosphorylatedm. It has been suggested that NRlA is the prevalent mRNA in adult rat

128

brain . NR2 only has about 20% homology with NRlA and is known to be longer

than NR1 101. Assuming similar domain distributions of the NR1 and NR2 subunits, the

34

major difference resides in the C-terminal domain length. NR2 homomers are inactive
with each other and must be co-assembled with NR1 for active channels to result.

Activation of the NMDA receptor ion channel leads to an increase in
intracellular Ca2+ concentration'm. This has signiﬁcant ramiﬁcations resulting in many
neuropathologies related to the NMDA receptor. One result of the Ca2+ inﬂux is the
activation of the Ca2+lcalmodulin-dependent protein kinase II (CaMKII)ml. It is thought
that this inﬂux of Ca2+ following NMDA activation results in a sustained CaMKII
activation which is known to lead to nitric acid production. It is thought that this may be
in-part responsible for neurodegenerative effects.
1.2.6 Structure-Activity Relationships

The mechanism by which conantokins inhibit ion ﬂow in NMDA receptors
appears complex. Small linear peptides like the conantokins may be systematically
altered easily. This makes the conantokins prime targets for structure-activity studies.
The neurochemical activity of these peptides is usually studied through concentration
effects on spermine-enhanced [3H]MK-801 binding to rat neuronal membranes. Single
Ala-replacement studies were also performed in order to characterize the functional and
structural contributions of the side chains in the conantokinsl. Table 1.3 shows the
effects of side chains on the NMDA receptor antagonist and conformational properties
of con-G. Substitutions by Ala at Glu2, Gla4, Leu5, Gln9, and Ile12 resulted in ZOO-fold
increases in the ICSO value associated with [3H]MK-801 binding. Glyl, Gla3, Leul 1,
and Arg13 exhibited inhibitory potencies diminished by 8-20-fold. The remainder of the
substituted side chain variants did not Show any signiﬁcant deviations in activity versus

the wild-type con-G. A similar experiment was performed on con-T. These results are

35

displayed in Table 1.4. The most drastic deviations occurred for Ala replacements at
Glu2, Gla3, Gla4, and Tyr5. Glyl, Met8, and Leul2 displayed intermediate effects.
However, Gln6, Asnl l, and Glal4 actually amplified the potency slightly. Position 7
occupied by Gla in con-G and Lys in con-T has no bearing on the activity of
conantokins when replaced with Ala. Moreover, con-G[E7K] and con-T[K7C] have 1C5.)
values in the [3H]MK-801 binding assay that are comparable to the wild-type. The
replacement of GlalO or Gla14 does not alter con-G nor con-T receptor activity.
Replacing either GlalO or Glal4 with Ala reduces con-G and con-T helicity in the
presence of Ca2+. GlalO and Gla14 have been Shown to be important for maintaining
metal binding. Since these residues are not essential for activity, metal cation binding to
these residues as well as the increased helicity cannot be a factor in NMDA receptor
binding. This is also consistent with studies using con-T where the 1C5.) values to
[3H]MK-801 binding were unchanged by the addition of 1.5 mM Ca2+”). It is also
interesting to note that the introduction of a helix-breaking Pro residue at position 7 in
con-G completely eliminated activity'og. This supports the importance of a helical
structure.

It is possible that con-G and con-T interact with different NMDA receptor
subunit combinations. It has been shown that con-G selectively antagonizes the
NRlA/NRZB receptor combination without inﬂuencing receptors containing NRlA and
NR2A subunitsm. Also electrophysiological studies using recombinant receptors
consisting of NRla, NRlb, NR2A, and NR2B subunits indicate that the subunit
selectivity of con-G may be derived from the presence of Ieu5 while in con-T Tyr5

confers broad selectivityl 3'.

36

Table 1.3 Effects of side chains on the NMDA receptor antagonist and conformational
properties of con-G].

 

 

Peptide 1C5.) (11M) Apo % Helix Ca2+ % Helix
Con-G[G1A] 6.5 0 32
Con-G[E2A] >100 2 42
Con-G[ EBA] 9.6 5 31
Con-G[C4A] >100 0 36
Con-G[L5A] >100 0 35
Con-G[Q6A] 0.5 0 35
Con-G[ :7 A] 0.1 5 24
Con-G[N8A] 0.5 2 44
Con-G[Q9A] >100 0 36
Con-G[C10A] 2.0 7 14
Con-G[L11A] 5.9 0 37
Con-G[Il2A] >100 0 28
Con-G[R13A] 3.4 0 15
Con-G[ E14A] 0.2 11 21
Con-G 0.5 2 50
Con-G [1-15] 1.2 O 27
Con-G [1-14] 5.9 0 17
Con-G [1-13] 2.9 0 4
Con-G [1-12] >100 O 6
Con-G [l-ll] >100 0 5

 

37

Table 1.4 Effects of side chains on the NMDA receptor antagonist and conformational
properties of con-Tl.

 

 

Peptide 1C5.) (uM) Apo % Helix Ca2+ % Helix
Con-T[GIA] 4.2 53 61
Con-T[E2A] 38.7 52 65
Con-T[CBA] 13.5 59 75
Con-T[ 31A] >60 22 47
Con-T[YSA] >100 53 86
Con-T[Q6A] 0.17 50 60
Con-T[K7A] 1.6 41 65
Con-T[M8A] 5.8 35 51.
Con-T[L9A] 0.60 26 39
Con—T[ DOA] 0.71 56 22
Con-T[Nl 1A] 0.17 51 58
Con-T[L12A] 5.0 39 50
Con-T[R13A] 0.43 22 36
Con-T[ [NA] 0.20 51 35
Con-T 0.40 55 82
Con-T[l-l7] 1.5 46 57
Con—T[l-l3] 2.7 46 57
Con-T[l-l 1] 6.9 15 8
Con-T[1-9] 9.2 10 4
Con—T[1-8] 51.3 2 2
Con-T[l-6] >100 2 2

 

38

1.2.7 Relevance of the project

In the realm of protein design, the forced oligomerization of engineered helices
has potential uses in a variety of biotechnological applications, including afﬁnity
puriﬁcation, hydrogel formation, biosensor development, and is of general interest for
expanding our knowledge of the forces that govern molecular self-organizations. For
these applications, a reversible control of the association event through a stimulus-
sensitive trigger, such as temperature or pH, is desirable. In this contextual framework,
metal ions have been employed to direct the self-assembly of designed helices.
However, work describing metal ion induced association of designed peptides has been
limited until the discovery of the naturally occurring peptide con-G. In previous metal
bound complexes, metal ions initiate strand association but rely mostly on interstrand
hydrophobic contacts. However, structural characterization of these complexes has
proven to be elusive, raising many questions as to the precise nature of stabilizing
interhelix interactions. In order to further understand how metal ions induce self-
association and helix stabilization, the crystal structure of con-G and con-T[K7C] in the
presence of Ca2+ has been determined as well as the structure of con—T[K7 C] in the
presence of Cd2+ and Mgz". The structures of these complexes are further described

herein.

39

1.3 Literature Cited

1.

10.

Prorok, M. & Castellino, F.J. Structure-function relationships of the NMDA
receptor antagonist conantokin peptides. Curr Drug Targets 2, 313-22 (2001).

Mathews, II, Vanderhoff—Hanaver, P., Castellino, F. J ., and Tulinsky, A. Crystal
structures of the recombinant kringle 1 domain of human plasminogen in
complexes with the ligands epsilon-aminocaproic acid and trans-4-
(aminomethyl)cyclohexane-l-carboxylic Acid. Biochemistry 35, 2567-76 (1996).

Chang, Y., Mochalkin, I., McCance, S. G., Cheng, B., Tulinsky, A., and
Castellino, F. J. Structure and ligand binding determinants of the recombinant
kringle 5 domain of human plasminogen. Biochemistry 37, 3258—71 (1998).

Dai, Q., Castellino, F.J. & Prorok, M. A single amino acid replacement results in
the Ca2+—induced self-assembly of a helical conantokin-based peptide.
Biochemistry 43, 13225-32 (2004).

Dai, Q., Prorok, M. & Castellino, E]. A new mechanism for metal ion-assisted
interchain helix assembly in a naturally occurring peptide mediated by optimally
spaced gamma-carboxyglutamic acid residues. J Mol Biol 336, 731-44 (2004).

Rios-Steiner, J.L., Schenone, M., Mochalkin, I., Tulinsky, A., and Castellino, F.
J. Structure and binding determinants of the recombinant kringle-2 domain of

human plasminogen to an internal peptide from a group A Streptococcal surface
protein. J Mol Biol 308, 705-19 (2001).

Castellino, F.J. & Ploplis, V.A. Structure and function of the
plasminogen/plasmin system. Thromb Haemost 93, 647-54 (2005).

Sottrup-Jensen, L., Claeys, H., Zajdel, M., Petersen, TE, and Magnusson, S.
Prog Chem F ibrinolysis Thrombolysis 3, 191-209 (1978).

Magnusson, S., Petersen, T.E., Sottrup-Jensen, L., and Claeys, H. in Proteases
and Biological Control, 123-149 (Cold Spring Harbor Laboratory, Cold Spring
Harbor, NY, 1975).

Pennica, D. et al. Cloning and expression of human tissue-type plasminogen
activator cDNA in E. coli. Nature 301, 214-21 (1983).

40

ll.

12.

l3.

14.

15.

16.

17.

18.

19.

20.

Steffens, G.J., Gunzler, W.A., Otting, F., Frankus, E. & Flohe, L. The complete
amino acid sequence of low molecular mass urokinase from human urine. Hoppe
Seylers Z Physiol Chem 363, 1043-58 (1982).

McLean, J.W. et al. cDNA sequence of human apolipoprotein(a) is homologous
to plasminogen. Nature 330, 132-7 (1987).

Urano, T., Deserrano, V. S., Chibber, B. A. K., and Castellino, F. J. The Control
of the Urokinase-Catalyzed Activation of Human Glutamic-Acid l-Plasminogen
by Positive and Negative Effectors. J Biol Chem 262, 15959-15964 (1987).

Urano, T., Chibber, B.A.K. & Castellino, E]. The Reciprocal Effects of Epsilon-
Aminohexanoic Acid and Chloride-Ion on the Activation of Human
(Glul]Plasminogen by Human Urokinase. J Biol Chem 84, 4031-4034 (1987).

Winn, E.S., Hu, S. P., Hochschwender, S. M., and Laursen, R. A. Studies on the
lysine-binding sites of human plasminogen. The effect of ligand structure on the
binding of lysine analogs to plasrrrinogen. EurJ Biochem 104, 579-86 (1980).

McCance, S.G., Menhart, N., and Castellino, F. J. Arrrino-Acid-Residues of the
Kringle-4 and Kringle-5 Domains of Human Plasminogen That Stabilize Their

Interactions with Omega- Amino Acid Ligands. J Biol Chem 269, 32405-32410
(1994).

Wu, T.P., Padmanabhan, K., Tulinsky, A., and Mulichak, A. M. The reﬁned
structure of the epsilon-aminocaproic acid complex of human plasminogen
kringle 4. Biochemistry 30, 10589-94 (1991).

Geiger, J.H. & Cnudde, S.E. What the structure of angiostatin may tell us about
its mechanism of action. J Thromb Haemost 2, 23-34 (2004).

Clemmensen, 1., Petersen, L.C. & Kluft, C. Puriﬁcation and characterization of a
novel, oligomeric, plasminogen kringle 4 binding protein from human plasma:
tetranectin. EurJ Biochem 156, 327-33 (1986).

Graversen, J .H., Lorentsen, R. H., J acobsen, C., Moestrup, S. K., Sigurskjold, B.
W., Thogersen, H. C., and Etzerodt, M. The plasminogen binding site of the C-
type lectin tetranectin is located in the carbohydrate recognition domain, and
binding is sensitive to both calcium and lysine. J Biol Chem 273, 29241-6
(1998).

41

21.

23.

24.

25.

26.

27.

28.

29.

30.

Graversen, J H Sigurskjold, B. W., Thogersen, H. C., and Etzerodt, M.
Tetranectin-binding site on plasminogen kringle 4 involves the lysine-binding

pocket and at least one additional arrrino acid residue. Biochemistry 39, 7414-9
(2000).

Schenone, M.M., Warder, S. E., Martin, J. A., Prorok, M., and Castellino, F. J.
An internal histidine residue from the bacterial surface protein, PAM, mediates

its binding to the kringle-2 domain of human plasminogen. J Pept Res 56, 438-
45 (2000).

Wistedt, A.C., Ringdahl, U., Muller-Ester], W., and Sjobring, U. Identiﬁcation of
a plasminogen-binding motif in PAM, a bacterial surface protein. Mol Microbiol
18, 569-78 (1995).

Wistedt, A.C., Kotarsky, H., Marti, D., Ringdahl, U., Castellino, F. J ., Schaller,
J ., and Sjobring, U. Kringle 2 mediates high afﬁnity binding of plasminogen to
an internal sequence in streptococcal surface protein PAM. J Biol Chem 273,
24420-4 (1998).

Kapetanopoulos, A. et a1. Direct interaction of the extracellular matrix protein
DANCE with apolipoprotein(a) mediated by the kringle IV -type 2 domain. Mol
Genet Genomics 267, 440-6 (2002).

Deguchi, H., Takeya, H., Gabazza, E.C., Nishioka, J. & Suzuki, K. Prothrombin
kringle 1 domain interacts with factor Va during the assembly of prothrombinase
complex. Biochem J 321 ( Pt 3), 729-35 (1997).

Kotkow, K.J., Deitcher, S.R., Furie, B. & Furie, BC. The second kringle domain
of prothrombin promotes factor Va-mediated prothrombin activation by
prothrombinase. J Biol Chem 270, 4551-7 (1995).

Chekenya, M. et al. NG2 proteoglycan promotes angiogenesis-dependent tumor
growth in CNS by sequestering angiostatin. Faseb J 16, 586-8 (2002).

Goretzki, L., Lombardo, C.R. & Stallcup, W.B. Binding of the NG2
proteoglycan to kringle domains modulates the functional properties of
angiostatin and plasmin(ogen). J Biol Chem 275, 28625-33 (2000).

Suenson, E. & Thorsen, S. Secondary-Site Binding of Glu-Plasmin, Lys-Plasmin
and Miniplasrrrin to Fibrin. Biochem J 197, 619-628 (1981).

42

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

Christensen, U. & Molgaard, L. Positive co-operative binding at two weak
lysine-binding sites governs the Glu-plasrrrinogen conformational change.
Biochem J 285 ( Pt 2), 419-25 (1992).

Takada, Y., Urano, T. & Takada, A. Conformational change of plasminogen:
effects of N-terrninal peptides of Glu-plasminogen. Thromb Res 70, 151-9
(1993).

Mangel, W.F., Lin, B., and Ramakrishnan, V. Characterization of an Extremely
Large, Li gand-Induced Conformational Change in Plasrrrinogen. Science 248,
69-73 (1990).

Marshall, J .M., Brown, A. J., and Ponting, C. P. Conformational Studies of
Human Plasminogen and Plasminogen Fragments - Evidence for a Novel 3rd
Conformation of Plasrrrinogen. Biochemistry 33, 3599-3606 (1994).

McCance, S.G. & Castellino, F.J. Contributions of individual kringle domains
toward maintenance of the chloride-induced tight conformation of human
glutamic acid-1 plasrrrinogen. Biochemistry 34, 9581-6 (1995).

Lin, L.F., Houng, A.Y. & Reed, G.L. Epsilon arrrino caproic acid inhibits
streptokinase-plasminogen activator complex formation and substrate binding
through kringle-dependent mechanisms. Biochemistry 39, 4740-4745 (2000).

Cockell, C.S., Marshall, J .M., Dawson, K.M., Cederholm-Williams, S.A. &
Ponting, C.P. Evidence that the conformation of unliganded human plasrrrinogen

is maintained via an intramolecular interaction between the lysine-binding site of
kringle 5 and the N-terminal peptide. Biochem J 333 (Pt 1), 99-105 (1998).

Berge, A. & Sjobring, U. PAM, a Novel Plasminogen-binding Protein from
Streptococcus pyogenes. J Biol Chem 268, 25417-25424 (1993).

Fischetti, V.A. Streptococcal M protein: molecular design and biological
behavior. Clin Microbiol Rev 2, 285-314 (1989).

Sanderson-Smith, M. et a1. Divergence in the plasminogen-binding group A
streptococcal M protein family: functional conservation of binding site and
potential role for immune selection of variants. J Biol Chem 281, 3217-26
(2006).

43

41.

43.

45.

46.

47.

48.

49.

50.

Boxrud, RD, and RE. Bock. Streptokinase binds preferentially to the extended
conformation of plasminogen through lysine binding site and catalytic domain
interactions. Biochemistry 39, 13974-13981 (2000).

Phillips, G.J., Flicker, R, Cohen, C., Manjula, B. & Fischetti, V.A. Streptococcal
M protein: alpha-helical coiled-coil structure and arrangement on the cell
surface. Proc Natl Acad Sci U S A 78, 4689-93 (1981).

O'Reilly, M.S., Holmgren, L., Shing, Y., Chen, C., Rosenthal, R. A., Moses, M.,
Lane, W. S., Cao, Y., Sage, E. H., and Folkman, J. Angiostatin: a novel

angiogenesis inhibitor that mediates the suppression of metastases by a Lewis
lung carcinoma. Cell 79, 315-28 (1994).

O'Reilly, M.S., Holmgren, L., Shing, Y., Chen, C., Rosenthal, R. A., Cao, Y.,
Moses, M., Lane, W. S., Sage, E. H., and Folkman, J. Angiostatin: a circulating
endothelial cell inhibitor that suppresses angiogenesis and tumor growth. Cold
Spring Harb Symp Quant Biol 59, 471-82 (1994).

Folkman, J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nat
Med 1, 27-31 (1995).

Folkman, J. & Shing, Y. Angiogenesis. J Biol Chem 267, 10931-4 (1992).

MacDonald, N.J., Murad, A. C., Fogler, W. E., Lu, Y. Y., and Sim, B. K. L. The
tumor-suppressing activity of angiostatin protein resides within kringles l to 3.
Biochem Biophys Res Commun 264, 469—477 (1999).

Stathakis, P., Fitzgerald, M., Matthias, L.J., Chesterman, C.N. & Hogg, P.J.
Generation of angiostatin by reduction and proteolysis of plasmin. Catalysis by a
plasrrrin reductase secreted by cultured cells. J Biol Chem 272, 20641-5 (1997).

Stathakis, P. et al. Angiostatin formation involves disulﬁde bond reduction and
proteolysis in kringle 5 of plasmin. J Biol Chem 274, 8910-6 (1999).

Lay, A.J., Jiang, X.M., Daly, E., Sun, L. & Hogg, P.J. Plasmin reduction by
phosphoglycerate kinase is a thiol-independent process. J Biol Chem 277, 9062-8
(2002).

51.

52.

53.

54.

55.

56.

57.

58.

59.

60.

Lay, A.J. & Hogg, P.J. Measurement of reduction of disulﬁde bonds in plasmin
by phosphoglycerate kinase. Methods Enzymol 348, 87-92 (2002).

Lay, A.J. et al. Phosphoglycerate kinase acts in tumour angiogenesis as a
disulphide reductase. Nature 408, 869-73 (2000).

Dong, Z.Y., Kumar, R., Yang, X.L. & Fidler, I.J. Macrophage-derived
metalloelastase is responsible for the generation of angiostatin in Lewis lung
carcinoma. Cell 88, 801-810 (1997).

Patterson, B.C. & Sang, Q.X.A. Angiostatin-converting enzyme activities of
human matrilysin (MMP-7) and gelatinase B type IV collagenase (MMP-9). J
Biol Chem 272, 28823-28825 (1997).

Lijnen, H.R., Ugwu, F., Bini, A. & Collen, D. Generation of an angiostatin-like
fragment from plasminogen by stromelysin-l (MMP-3). Biochemistry 37, 4699-
4702 (1998).

Cornelius, L.A. et al. Matrix metalloproteinases generate angiostatin: Effects on
neovascularization. J. Immunol. 161, 6845-6852 (1998).

O'Reilly, M.S., Wiederschain, D., Stetler—Stevenson, W.G., Folkman, J. &
Moses, M.A. Regulation of angiostatin production by matrix metalloproteinase-2
in a model of concomitant resistance. J Biol Chem 274, 29568-29571 (1999).

Gonzalez-Gronow, M., Grenett, H.E., Weber, M.R., Gawdi, G. & Pizzo, S.V.
Interaction of plasrrrinogen with dipeptidyl peptidase IV initiates a signal
transduction mechanism which regulates expression of matrix metalloproteinase-
9 by prostate cancer cells. Biochem J. 355, 397-407 (2001).

Cao, Y.H., Ji, R. W., Davidson, D., Schaller, J ., Marti, D., Sohndel, S.,
McCance, S. G., Oreilly, M. S., Llinas, M., and Folkman, J. Kringle domains of
human angiostatin - Characterization of the anti-proliferative activity on
endothelial cells. J Biol Chem 271, 29461-29467 (1996).

Gately, S. et al. The mechanism of cancer-mediated conversion of plasrrrinogen
to the angiogenesis inhibitor angiostatin. Proc. Natl. Acad. Sci. U. S. A. 94,
10868-10872 (1997).

45

61.

62.

63.

65.

66.

67.

68.

69.

70.

71.

Heidtrnann, H.H. et al. Generation of an giostatin-like fragments from
plasminogen by prostate-speciﬁc antigen. Br J Cancer 81, 1269-73 (1999).

Fortier, A.H. et al. Recombinant prostate Speciﬁc antigen inhibits angiogenesis in
vitro and in vivo. Prostate 56, 212-219 (2003).

Abad, MC. et al. The X-ray crystallographic structure of the angiogenesis
inhibitor angiostatin. J Mol Biol 318, 1009-17 (2002).

Lee, H., Kim, H. K., Lee, J. H., You, W. K., Chung, S. 1., Chang, S. 1., Park, M.
H., Hong, Y. K., and Joe, Y. A. Disruption of interkringle disulﬁde bond of
plasminogen kringle 1-3 changes the lysine binding capability of kringle 2, but
not its antiangiogenic activity. Arch Biochem Biophys 375, 359-63 (2000).

Risau, W. Differentiation of endothelium. Faseb J 9, 926-933 (1995).

Pepper, M.S. Extracellular proteolysis and angiogenesis. Thromb Haemost 86,
346-55 (2001).

O'Reilly, M.S., Holmgren, L., Chen, C. & Folkman, J. Angiostatin induces and
sustains dormancy of human primary tumors in mice. Nat Med 2, 689-92 (1996).

Moser, T.L. et al. Angiostatin binds ATP synthase on the surface of human
endothelial cells. Proc Natl Acad Sci U S A 96, 2811-6 (1999).

Moser, T.L. et al. Endothelial cell surface Fl-FO ATP synthase is active in ATP
synthesis and is inhibited by angiostatin. Proc Natl Acad Sci U S A 98, 6656-61
(2001).

Moser, T.L., Stack, M.S., Wahl, M.L. & Pizzo, S.V. The mechanism of action of
angiostatin: can you teach an old dog new tricks? Thromb Haemost 87, 394-401
(2002).

Tarui, T., Miles, L.A. & Takada, Y. Speciﬁc interaction of angiostatin with
integrin alpha(v)beta(3) in endothelial cells. J Biol Chem 276, 39562-8 (2001).

46

72.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

Tarui, T., Majumdar, M., Miles, L.A., Ruf, W. & Takada, Y. Plasmin-induced
migration of endothelial cells. A potential target for the anti-angiogenic action of
angiostatin. J Biol Chem 277, 33564—70 (2002).

Reynolds, L.E. et al. Enhanced pathological angiogenesis in mice lacking beta3
integrin or beta3 and betaS integrins. Nat Med 8, 27-34 (2002).

Troyanovsky, B., Levchenko, T., Mansson, G., Matvijenko, O. & Holmgren, L.
Angiomotin: an angiostatin binding protein that regulates endothelial cell
migration and tube formation. J Cell Biol 152, 1247-54 (2001).

Bratt, A. et al. Angiomotin belongs to a novel protein family with conserved
coiled-coil and PDZ binding domains. Gene 298, 69-77 (2002).

Akasaka, T., van Leeuwen, R.L., Yoshinaga, I.G., Mihm, M.C., Jr. & Byers,
H.R. Focal adhesion kinase (p125FAK) expression correlates with motility of
human melanoma cell lines. J Invest Dermatol 105, 104-8 (1995).

Plopper, G.E., McNamee, H.P., Dike, L.E., Bojanowski, K. & Ingber, D.E.
Convergence of integrin and growth factor receptor signaling pathways within
the focal adhesion complex. Mol Biol Cell 6, 1349-65 (1995).

Claesson-Welsh, L. et al. Angiostatin induces endothelial cell apoptosis and
activation of focal adhesion kinase independently of the integrin-binding motif
RGD. Proc Natl Acad Sci U S A 95, 5579-83 (1998).

Xiong, JP. et al. Crystal structure of the extracellular segment of integrin alpha
Vbeta3. Science 294, 339-45 (2001).

Abrahams, J.P., Leslie, A.G., Lutter, R. & Walker, J .E. Structure at 2.8 A
resolution of Fl-ATPase from bovine heart mitochondria. Nature 370, 621-8
(1994).

Brown, J .E. & Klee, W.A. Helix-coil transition of the isolated amino terminus of
ribonuclease. Biochemistry 10, 470-6 (1971).

Siedlecka, M., Goch, G., chhart, A., Sticht, H. & Bierzyski, A. Alpha-helix
nucleation by a calcium-binding peptide loop. Proc Natl Acad Sci U S A 96, 903-
8 (1999).

47

83.

84.

85.

86.

87.

88.

89.

90.

91.

93.

Ghadiri, M.R. & Choi, C. Secondary structure nucleation in peptides. Transition
metal ion stabilized alpha-helices. J Am Chem Soc 112, 1630-2 (1990).

Ghadiri, M.R. & Femholz, A.K. Peptide architecture. Design of stable .alpha.-
helical meta110peptides via a novel exchange-inert ruthenium(III) complex . J Am
Chem Soc 112, 9633-5 (1990).

Ruan, F., Chen, Y. & Hopkins, P.B. Metal ion-enhanced helicity in synthetic

peptides containing unnatural, metal-ligating residues. J Am Chem Soc 112,
9403-4 (1990).

Ghosh, D. & Pecoraro, V.L. Understanding metalloprotein folding using a de
novo design strategy. Inorg Chem 43, 7902-15 (2004).

Doerr, A.J. & McLendon, G.L. Design, folding, and activities of metal-
assembled coiled coil proteins. Inorg Chem 43, 7916-25 (2004).

Li, X.Q. et al. Soft metal ions, Cd(11) and Hg(H), induce triple-stranded alpha-
helical assembly and folding of a de novo designed peptide in their trigonal
geometries. Protein Sci 9, 1327-33 (2000).

Kiyokawa, T. et al. Binding of Cu(II) or Zn(11) in a de novo designed triple-
stranded alpha-helical coiled-coil toward a prototype for a metalloenzyme. J Pept
Res 63, 347-53 (2004).

Handel, T., Williams, S.A. & Degrado, W.F. Metal ion-dependent modulation of
the dynamics of a designed protein. Science 261, 879-85 (1993).

Kohn, W.D., Kay, C.M., Sykes, B.D. & Hodges, R.S. Metal Ion Induced Folding
of a de Novo Designed Coiled-Coil Peptide. J Am Chem Soc 120, 1124-32
(1998).

Terlau, H. & Olivera, B.M. Conus venoms: a rich source of novel ion channel-
targeted peptides. Physiol Rev 84, 41-68 (2004).

Jimenez, E.C., Olivera, B.M., Gray, W.R. & Cruz, L.J. Contryhan is a D-
tryptophan-containing Conus peptide. J Biol Chem 281, 28002-28005 (1996).

48

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

Loughnan, M. et al. Ot-Conotoxin, EpI, a novel sulfated peptide from Conus
episcopatus that selectively targets neuronal nicotinic acetylcholine receptors. J
Biol Chem 273, 15667-15674 (1998).

Jimenez, RB. et al. Bromocontryphan: post-translational bromination of
tryptophan. Biochemistry 36, 989-994 ( 1997).

Craig, A.G. et al. An O-glycosylated neuroexcitatory Conus peptide.
Biochemistry 37, 16019-16025 (1998).

McIntosh, J .M., Olivera, B.M., Cruz, L.J. & Gray, W.R. y-Carboxy-glutamate in
a neuroactive toxin. J Biol Chem 259, 14343-14346 (1984).

McIntosh, J ., Olivera, B.M., Cruz, L. & Gray, W. Garnma-carboxyglutamate in a
neuroactive toxin. J Biol Chem 259, 14343-6 (1984).

Olivera, B.M. et al. A sleep-inducing peptide from Conus geographus venom.
Toxicon 23(1985).

Olivera, B.M., Cruz, L.J. & Yoshikarni, D. Effects of Conus peptides on the
behavior or mice. Curr Opin Neurobiol 9, 772-777 (1999).

Castellino, F.J. & Prorok, M. Conantokins: inhibitors of ion ﬂow through the N -
methyl-D-aspartate receptor channels. Curr Drug Targets 1, 219-35 (2000).

Warder, S.E., Bland], T., Klein, R.C., Castellino, F.J. & Prorok, M. Amino acid
determinants for NMDA receptor inhibition by conantokin-T. J Neurochem 77,
812-22 (2001).

Mcintosh, J.M., Olivera, B.M., Cruz, L.J. & Gray, W.R. 7-carboxy-glutamate in
a neuroactive toxin. J Biol Chem 259, 14343-14346 (1984).

Prorok, M., Warder, S.E., Blandl, T. & Castellino, P.J. Calcium binding
properties of synthetic gamma-carboxyglutamic acid-containing marine cone

snail "sleeper" peptides, conantokin-G and conantokin-T. Biochemistry 35,
16528-34 (1996).

49

105.

106.

107.

108.

109.

110.

111.

112.

113.

114.

Haack, J .A. et al. Conantokin-T. A gamma-carboxyglutamate containing peptide
with N -methyl-d-aspartate antagonist activity. J Biol Chem 265, 6025-9 (1990).

Skjaerbaek, N., Nielsen, K.J., Lewis, R.J., Alewood, P. & Craik, D.J.
Determination of the solution structures of conantokin—G and conantokin-T by
CD and NMR spectroscopy. J Biol Chem 272, 2291-9 (1997).

Warder, S.E. et al. The roles of individual gamma-carboxyglutamate residues in
the solution structure and cation-dependent properties of conantokin-T. J Biol
Chem 273, 7512-22 (1998).

Blandl, T., Prorok, M. & Castellino, F .J . NMDA-receptor antagonist
requirements in conantokin-G. FEBS Lett 435, 257-62 (1998).

Ault, 8., Evans, R.H., Francis, A.A., Oakes, D.J. & Watkins, J .C. Selective
depression of excitatory amino acid induced depolarizations by magnesium ions
in isolated spinal cord preparations. J Physiol 307, 413-428 (1980).

Skolnick, P., Boje, K., Miller, R., Pennington, M. & Maccecchini, M.L.
Noncompetitive inhibition of N-methyl-D-aspartate by conantokin-G: evidence
for an allosteric interaction at polyamine sites. J Neurochem 59, 1516-21 (1992).

Mena, E.E. et al. Conantokin-G: a novel peptide antagonist to the N-methyl-D-
aspartic acid (NMDA) receptor. Neurosci Lett 118, 241-4 (1990).

Hammerland, L.G., Olivera, B.M. & Yoshikami, D. Conantokin-G selectively
inhibits N-methyl-D-aspartate-induced currents in Xenopus oocytes injected with
mouse brain mRN A. Eur J Pharmacol 226, 239-44 (1992).

Huang, C.C., Lyu, P.C., Lin, C.H. & Hsu, K.S. Conantokin-T selectively
antagonizes N-methyl-D-aspartate-evoked responses in rat hippocampal slice.
Toxicon 35, 355-63 (1997).

Wo, Z.G. & Oswald, R.E. Transmembrane topology of two kainate receptor
subunits revealed by N-glycosylation. Proc Natl Acad Sci U S A 91, 7154-7158
(1994).

50

115.

116.

117.

118.

119.

120.

121.

122.

123.

124.

Williams, K., Kashiwagi, K. & Fukuchi, J. An acidic amino acid in the N-

methyl-D-aspartate receptor that is important for spermine stimulation. Mol
Pharmacol 48, 1087-1098 (1995).

Zheng, X., Zhang, L., Durand, G.M., Bennett, M.V. & Zukin, R.S. Mutagenesis
rescues spermine and Zn2+ potentiation of recombinant NMDA receptors.
Neuron 12, 811-818 (1994).

Kuryatov, A., Laube, B., Betz, H. & Kuhse, J. Mutational analysis of the
glycine-binding site of the NMDA receptor: structural similarity with bacterial
amino acid-binding proteins. Neuron 12, 1291-1300 (1994).

Laube, B., Kuryatov, A., Kuhse, J. & Betz, H. Glycine-glutamate interactions at
the NMDA receptor: role of cysteine residues. F EBS Lett 335, 331-334 (1993).

Stem-Bach, Y. et al. Agonist selectivity of glutamate receptors is speciﬁed by
two domains structurally related to bacterial amino acid-binding proteins.
Neuron 13, 1345-1357 (1994).

Kendrick, S.J., Lynch, D.R. & Pritchett, D.B. Characterization of glutamate
binding sites in receptors assembled from transfected NMDA receptor subunits.
J Neurochem 67, 608-616 (1996).

Sakurada, K., Masu, M. & Nakanishi, S. Alteration of Ca2+ permeability and
sensitivity to Mg2+ and channel blockers by a single amino acid substitution in
the N-methyl-D-aspartate receptor. J Biol Chem 268(1993).

Mori, H., Masaki, H., Yamakura, T. & Mishina, M. Identiﬁcation by
mutagenesis of a Mg(2+)-block site of the NMDA receptor channel. Nature 358,
673-675 (1992).

Kawajiri, S. & Dingledine, R. Multiple structural determinants of voltage-
dependent magnesium block in recombinant NMDA receptors.
Neuropharmacology 32, 1203-1211 (1993).

Wood, M.W., VanDongen, H.M. & VanDongen, A.M. An alanine residue in the
M3-M4 linker lines the glycine binding pocket of the N -methyl-D-aspartate
receptor. J Biol Chem 272, 3532-3537 (1997).

51

125.

126.

127.

128.

129.

130.

131.

Hirai, H., Kirsch, J ., Laube, B., Betz, H. & Kuhse, J. The glycine binding site of
the N-methyl-D-aspartate receptor subunit NR1: identiﬁcation of novel

determinants of co-agonist potentiation in the extracellular M3-M4 loop region.
Proc Natl Acad Sci U S A 93, 6031-6036 (1996).

Kashiwagi, K., Fukuchi, J., Chao, J ., Igarashi, K. & Williams, K. An aspartate
residue in the extracellular loop of the N-methyl-D-aspartate receptor controls
sensitivity to spermine and protons. Mol Pharmacol 49, 1131-1141 (1996).

Tingley, W.G., Roche, K.W., Thompson, A.K. & Huganir, R.L. Regulation of
NMDA receptor phosphorylation by alternative splicing of the C-terminal
domain. Nature 364, 70-73 (1993).

Sugihara, H., Moriyoshi, K., Ishii, T., Masu, M. & Nakanishi, S. Structures and
properties of seven isoforms of the NMDA receptor generated by alternative
splicing. Biochem Biophys Res Commun 185, 826-832 (1992).

Nielsen, K.J. et al. Structure-activity studies of conantokins as human N-methyl-
D-aspartate receptor modulators. J Med Chem 42, 415-26 (1999).

Donevan, S.D. & McCabe, R.T. Conantokin G is an NR2B-selective competitive
antagonist of N -methyl-D-aspartate receptors. Mol Pharmacol 58, 614-23
(2000).

Klein, R.C., Prorok, M., Galdzicki, Z. & Castellino, E]. The amino acid residue
at sequence position 5 in the conantokin peptides partially governs subunit-

selective antagonism of recombinant N-methyl-D-aspartate receptors. J Biol
Chem 276, 26860-7 (2001).

52

Chapter II

X-ray Structure Determination

2.1 Structure Determination from X-ray Diffraction Data
2.1.1 Theory

When a crystal is exposed to X-rays, constructive interferences between rays
scattered from successive planes in the crystal will only take place if the path
differences between the rays are equal to an integral number of wavelengths. This is

known as Bragg’s law:

2d sin19 = M (2.1)

The Bragg equation gives the condition for diffraction so that if a crystal is
rotated in a beam of X-rays, the scattering pattern is a series of intensity maxima. In
a crystal, electrons in atoms are the scatters, and each atom has a different
effectiveness as a scatter. Consequently, when an experiment is carried out, a set of
diffraction maxima of different intensities are observed. The crystal is rotated to
obtain the scattering intensity at various angles. The scattering intensity depends on
the scattering effectiveness of the individual atoms and the phase of the wave from

each scattering source, known as individual structure factor ( ﬁ' ). The structure

factor, F(hkl) , for each plane (hkl) can be deﬁned as the sum of the structure factors

for individual atoms, fj, times a phase factor, Ot(hkl), for each atom. In other

words, the structure factor can be represented by its amplitude and phase:

53

where, F(hkl) is the amplitude and Ot(hkl) is the phase.

When the diffracted X-ray is recorded, all information on the phase is lost
and only a measurement of the intensity of the diffracted beam is recorded. The

intensity in each Spot of the diffraction pattern is given by:

[(hld) =[1=(h1c1)]2 (2.3)

The electron density, p(r) , is a function of the coordinates of the scattering

centers (the atoms) and has a maximum around the position of each atom. It is
desirable to convert the measured structure factors into atomic coordinates. This is
done by taking the Fourier transform of equation 2.2. In this case, the Fourier
transform takes the Structure factors, which are functions of the electron density, and

inverts the functional dependence so that the electron density is expressed as a

p(r) = Lmrmmonspace F (s)e'2”"r’sdvs (2.4)

function of the structure factors:

where, st is a small unit of volume in diffraction space. The integration can be
replaced by a summation since F (s) is not continuous and is non-zero only at the

reciprocal lattice points. Therefore:

where, p(xyz) is the electron density at any point x,y,z and F(hkl)is the amplitude,

which is proportional to the square root of the measured intensity of each reﬂection,

labeled hkl.

54

1 —7ti vc+y+z
p(xyz)=;,-ZZZF(h/d)ez (’ " ’l

h k l
1 - . _ . . .
=_ F hkl 1a(hIJ) 27n(hrc+k_\v+lu) (25)
V;;;( A e

The problem is that only the intensity, which is the square of the amplitude,
can be directly derived from the measured intensity of the diffracted beam. Since
only intensities but not phases are measured in the recorded diffraction pattern it is
impossible to determine the electron density, and therefore a structure, directly from
a recorded diffraction pattern. This is a major problem in crystallography and is
referred to as the “phase problem”]‘2.

There are few methods by which the phase problem can be solved. (1) The

Patterson summation: This is a Fourier summation based on the experimentally
observed [F (hkl )T. It is basically a vector map of the structure and is applied for

relatively small molecules. (2) Direct methods: In this method mathematical
relationships between the reﬂections can be used to provide phase information. (3)
Heavy atom isomorphous replacement: In this method a heavy atom is introduced to
a structure to provide phase information. (4) Anomalous scattering: In this method
phase information is obtained from the information contained in the scattering by an
atom, when its natural absorption frequency is close to the frequency of the incident
radiation. (5) Molecular replacement method: A known structure will be used to
ﬁnd the phase of the unknown structure. However molecular replacement is limited
because to have a good chance of success in ﬁnding a solution, the search and target

molecules must have a reasonable sequence identity (>25%). Likewise, having data

55

with high completeness can be crucial. Most of the time, but not always, molecular
replacement seems to be relatively easier than the other methods and is the ﬁrst
choice in solving the phase problem.

Since many crystal structures of kringles have already been solved including
the multiple kringle containing compound angiostatin3, we could easily use
molecular replacement to determine the initial phases of the angiostatinNEK-30
complex structure using angiostatin and/or its individual kringle components as
models. Since the structures of metal bound conantokins-G and -T[K77] are known
to adopt a-helical conformations“, any a-helix can be used as a model. This will be
further discussed in this chapter.

Generally there are two steps in molecular replacement and these are known
as the rotation and translation functions.

Rogtion Function

The rotation function should allow the orientation of the search molecule,
which produces a maximal overlap with the target structure to be determined in the
absence of any phases for the unknown structure. To do this, it compares the
Patterson self-vectors of the known and unknown structures at different orientations
of the search model. It should be noted that Patterson functions can be calculated
from the amplitudes only and using Patterson space so that the translation vector is
irrelevant, since all intramolcular vectors are shifted to the origin. The rotation
function is usually calculated as a function of Eulerian angles, 0t, [3 and 7. The
molecule is placed in an orthogonal coordinate system with the axis of highest

symmetry along Z (about Ct) to reduce the amount of computation.

56

mum Function

Having determined the angles, 0t, B and 7, from the rotation search the
rotation matrix can be used and applied to the coordinates of the search molecule.
The shift vector, which is required to position the search molecule correctly relative
to the symmetry elements of the space group, can be determined by one of a number
of translation searches. Patterson methods can be used to measure the overlap of the
target Patterson cross-vectors with those calculated for the oriented search molecule
as it ranges through the target cell. The simpler way to solve the translation problem
is the reliability factor (R-factor) search. It involves the calculation of an R-factor as
the search molecule and its symmetry mates are moved through the unit cell of the
target crystal. The correct position should give the lowest R-factor deﬁned as:

Z |Fobs| —|Fcal”

R: (2-6)

ZlFobsl

where, Fobs and Fcal are the observed and calculated structure factors, respectively.

 

 

Basically when a crystal structure is known, values of F(hkl)can be calculated and a
way to test the correctness of the structure is how well the calculated values of
F(hkl ) agree with the observed ones. Any random collection of atoms in the cell has

been shown to result in an R-factor of 83% and 59% for centric and acentric space
groups, respectively. Therefore any model that gives R-factors approaching these
values is just a little better than a collection of atoms randomly placed in a cell. An

R-factor around 45% tells us that the solution is not hopeless but major changes are

57

needed to ﬁt the model. An R-factor of ~35% is likely to be a correct solution. An
R-factor of 25% and below indicates that the model is most probably correct except
for small (1.0 A) atomic Shifts and changes in temperature factors (B-factors).

Other parameters such as the Correlation Coefﬁcient (CC) can be used to

measure the agreement between the Fobs and Fcal as the search model is moved

around. CC can be expressed as:
Z( 2 ) (|Fa..|2 - 2) (2.7)

[El 2)’2( 2- 2)];

The CC runs from —1 (perfect inverse correlation), through 0 (no correlation),

 

 

2

Fobs Fobs Fcal

 

 

 

 

 

 

 

 

2

Fobs F obs Fcal Fcal

 

 

 

 

 

 

 

 

to +1 (perfect correlation). It has the advantages of being almost independent of

scaling between Fobs and Fear , and is much more sensitive than the R-factor in the
region where the R-factor approaches its random limits. Conversely, the coefﬁcient
approaches 1.0 closely as the R-factor goes below 20%, and thus becomes of limited
value6.

If successful, a preliminary model of the target structure will be obtained by
correctly orienting and positioning the search molecule in the target cell.
Subsequently, this model can be reﬁned and optimized by rigid body reﬁnement,
restrained reﬁnements, as well as manual reﬁtting of the models.

A number of programs exist for performing molecular replacement. These
include AMoRe7, MolRep within CCP48, CaspRg, X-PLOR'O, and Phaser”. All of

these programs except Phaserll output R-factors and Correlation Coefﬁcients.

58

Instead Phaserll outputs Z-scores and the log-likelihood gain (LLG). The likelihood
is the probability that the data would have been measured, given the model, so it
allows us to compare how well different models agree with the data. The LLG is the
difference between the likelihood of the model and the likelihood calculated from a
Wilson distribution, so it measures how much better the data can be predicted with
your model than with a random distribution of the same atoms. The LLG allows us
to compare different models against the same data set, but the LLG values for
different data sets should not be compared with each other. If the best LLG is
negative, then the model is worse than a collection of random atoms. The LLG
should always be positive, and it should increase as the solution progresses. By
default, Phaser11 selects solutions over 75% of the difference between the top
solution and the mean. Ideally, only the number of solutions you are expecting
should be selected by this criterion, but if the signal-to-noise of your search is low,
there will be noise peaks in this selection also. For a translation function, the correct
solution will generally have a Z-score (number of standard deviations above the
mean value) over 5 and be well separated from the rest of the solutions. For a
rotation function, the correct solution may be in the list with a Z-score under 4, and
will not be found until a translation function is performed and picks out the correct
solution.

Phaser11 proved to be the only molecular replacement program capable of
providing a correct solution for the metal bound structure of conantokin-G. This is
because the other programs rely on a spherically shaped protein whereas in the case

of conantokin-G, the protein was not spherically Shaped but helical. So, the

59

intermolecular vectors are extremely limited whereas the intramolecular vectors are
not.

As mentioned above, heavy atom isomorphous replacement is another
method of solving the phase problem. In this technique, differences in the
diffraction pattern are measured after the introduction of a heavy atom. In order to
be able to record these differences reliably, the atom or group of atoms must have
many electrons. If hard ions such as lanthanides or actinides are used as derivatives,
the interaction between the protein and the heavy atom is usually ionic. On the other
hand, soft ions such as mercury, gold, platinum, etc. tend to react with sulfurs on
cysteines, deprotonated nitrogen on histidines or even with the sulfur from
methionine. These interactions are covalent, so they tend to bind more speciﬁcally.
For this technique to work, it is important that the protein molecules within the
crystal be identically bound to the heavy atom compound with essentially no change
in the structure of the native crystal lattice.

The coordinates and the diffraction pattern of the heavy atom alone can be
determined by calculating the difference between the diffraction patterns of the
native and heavy atom replaced crystal. This is because the contributions from every
atom to a reﬂection are combined in an additive way. With the diffraction pattern of
just the heavy atom, a small number of atoms exist and the structure is easier to
solve. This technique was used in order to provide phases for K3 in the
angiostatin/VEK-30 complex structure. This technique was also used in order to
determine the structure of conantokin-G bound to Ca2+. This will be further

discussed below in this chapter.

60

Anomalous dispersion is another method mentioned above that was tried in
order to determine the structure of Ca2+ bound conantokin-G and the Cd2+/l\/Ig2+/con-
T[K77] structure. Anomalous dispersion involves the use of anomalous scatterers in
the protein. In many cases, the sulfur in methionine is substituted with selenium by
expressing the protein in the presence of SeMet. Advantage is taken of the
anomalous differences by irradiating the crystal with X-ray radiation at the
absorption edge of the scatterer (0.98 A for Se). However, in the case of the
conantokin structures, this method was used in order to locate Ca2+ sites in the
conantokin-G structure and Cd2+ sites in the Cd21/Mg2+/con-T[K77] structure. This
will be further described below.

Structure Reﬁnement

Once one has overcome the phase problem and a solution is found for the
structure, 2Fobs—Fcal and Fobs-Fcal Fourier maps are calculated, and the atomic
positions are taken as the locations of the electron density function. The advent of
high-speed computers has led to widespread use of the method of least squares,

which automatically adjusts the parameters so as to minimize some functions such as
2(Fobs- Fcal)2. Each residue will be ﬁtted manually in the structure until the bias

introduced by the starting model is reduced considerably. A correct structure Should
have a satisfactory R-factor with no major unexplained discrepancies. A satisfactory
value for the R-factor depends on the resolution of the data. The higher the
resolution the lower the R-factor must be. Before reﬁning the model, a fraction of
the reﬂections, usually 5-10%, are randomly chosen and put aside for cross-

validation. This set of randomly selected reﬂections is called the test set. All the

61

ﬁtting of the residues and reﬁnement will be done on the working set (95% or 90%
of the data). Then two separate R-factors are calculated. The R-factor calculated for
the working set is called “Rm...” or “Rm.” and that of the test set is called “Rfrec”. As
the Rm. decreases upon model ﬁtting, the Rﬁce should concurrently decrease.
Generally wae is higher than Rm“, however they should not differ by more than 6-
7%'~6.

Values of Rwork and R.rec are measures of the agreement between the values of the

observed structure factors, given by these equations:

Rwork: M (2.8)

ZlFobsl

Rm: leFobsl -|Fcal” (2.9)

2 I Fobsl

2.1.2 Overview of Crystallography

The ﬁrst step in every structure determination is the production of single and
well diffracting crystals. For this purpose it is essential to have access to large
quantities of highly pure material. The protein is set up for crystallization using
sparse matrices of precipitating solutions that have been proven successful in
crystallizing other proteins. The most common method for macromolecular
crystallization is the hanging drop vapor diffusion method (Figure 2-1). In this
method, a drop containing a mixture of protein and precipitating solution is
equilibrated over a reservoir containing the precipitating solution. The protein is
slowly precipitated over time to where the molecules adopt identical orientations in

order to form a three dimensional array of molecules that are held together by non-

62

covalent interactions. A well-formed crystalline lattice aids in better diffracting
crystals. The crystallization process involves setting up thousands of
protein/precipitant drops as well as monitoring of each of these drops. When
monitoring these drops, it is important to look for precipitation behavior, relative
solubility, and the appearance of crystals. Based on the observations from the initial
screens, new optimized screens can be made. This can turn into an iterative process
that will hopefully yield crystals suitable for X-ray diffraction data collection. This
Strategy was followed in the crystallization of the angiostatinNEK-30 complex and
metal bound structures of con-G and con-T[K77].
2.2 AngiostatinNEK-30 Complex
2.2.1 Crystallization and Data Collection of the P6122 Structure

The human angiostatin mutant N289E (this mutant lacks N-linked
glycosylation) containing K1-3 was expressed in Pichia pastoris and puriﬁed as
previously described”. The puriﬁed protein was provided by EntreMed Inc.
(Rockville, MD) and was buffer exchanged into saline buffer (0.15 M NaCl/lOO mM
Tris pH 7.5) and concentrated to 15 mg/ml. The VEK-30 peptide was synthesized
and puriﬁed as previously described and provided by our collaborators Frank

Castellino and co-workers13

. An angiostatin/VEK-30 complex was made using a
ﬁve molar excess of VEK-30 compared to that of angiostatin. The protein was
extensively screened for crystallization by using the hanging drop vapor diffusion
method (Figure 2-1). The search for well diffracting crystals was done by using

several crystallization screens at two different temperatures, 25°C and 4°C. The best

crystals of the angiostatin/VEK-30 complex were grown at 25°C with a precipitant

63

solution containing 20% PEG 8000/0.1 M potassium dihydrogen phosphate after
optimizing the crystallization condition from Hampton Research Crystal Screen #42
(20% PEG 8000/0.05 M potassium dihydrogen phosphate). However, these crystals
were not single (Figure 2-2). In order to grow Single crystals of the
angiostatin/VEK-30 complex, multiple crystal screen additives were tried. The best
single crystals were grown in 20% PEG 8000/0.1 M potassium dihydrogen
phosphate/5% 1,4-dioxane. Diamond-like crystals appeared in 1 day and continued
to increase in size for 2-3 weeks (Figure 2-3).

The crystals were brieﬂy soaked in a cryoprotectant solution of 22% PEG
8000/0.l M potassium dihydrogen phosphate/5% 1,4-dioxane/30% glycerol at 298 K
and ﬂash frozen by immersion in liquid N2. Data were collected at the Advanced
Photon Source IMCA-CAT 17-11) at Argonne National Laboratory to a resolution of
2.0 A and data were processed and scaled using the I-IKL suite of programs”.
Crystals were in the hexagonal space group P6.22 with unit cell parameters a = b =
58.4 A and c = 391.0 A. The crystal to detector distance was 270 mm and 200° of
data were collected with an oscillation range of 05°. Assuming one molecule of
angiostatin/VEK-30 complex per asymmetric unit, the crystal volume per protein
mass is 2.9, which corresponds to approximately 57.5% solvent content in the
crystal. This value is within the range observed for protein crystals. The crystal
parameters of the angiostatin/VEK-30 crystal are listed in Table 2.1. A synchrotron
X-ray diffraction data set to a resolution of 2.0 A, with an overall 1/0 of 45.12, was

obtained. The data was 80.2% complete with an Rmerge of 9.1% for 28,416 unique

64

Protein + Reservoir Cover slip

Solution \

/ U
Grease ; H20

 

 

Reservoir

 

 

 

Figure 2-1 The hanging drop vapor diffusion method for
crystallizing proteins.

 

Figure 2-2 A non-single crystal of the
angiostatin/VEK-30 complex grown in 20%
PEG 8000/0.1 M potassium dihydrogen
phosphate.

 

Figure 2—3 A single crystal of the
angiostatin/VEK-30 complex grown in 20%
PEG 8000/0.1 M potassium dihydrogen
phosphate/5% 1,4-dioxane.

65

Table 2.1 Crystal parameters for the angiostatin/VEK-30 crystals in the space groups

 

 

Crystal Form Hexagonal Hexagonal
Space Group P6122 P61
a=b=58.4Ac=391.0A a=b=58.8Ac=389.2A
UM Ge" 0 = 13 = 900° 7 = 120.0° a = I3 = 900° 7 = 120.0°
Solvent Content 575% 574%
Mo]. Per Asymmetric l 2

Unit

 

Figure 2-4 The overall structure of the angiostatin/VEK-30 complex. Angiostatin
is shown in red and VEK-3O is shown in blue. One molecule of dioxane is shown
in green with two different conformations.

66

Table 2.2 Data statistics for the angiostatin/VEK-30 diffraction data collection21

 

 

Space Group P6.22 P6.
Wavelength, A 1.0000 1.0719
Resolution, A 2.0 (2.07-2.00) 3.0 (3.20-3.04)
Completeness, % 80.2 (78.7) 78.5 (97.1)
I/o 45.12 (4.8) 12.38 (3.02)
Rmerge, % 9.1 (30.8) 8.4 (31.2)
Unique Reﬂections 28,416 18,600
Measured Reﬂections 763,409 281,739

 

aThe parentheses denote those values for the last resolution shell.

Table 2.3 Reﬁnement statistics for the angiostatin/VEK-30 complex

 

 

Space Group P6.22 P6.
Rm... 20.85 % 20.17 %
R... 25.45 % 29.56 %

Resolution 152.3 A 203.0 A

rmsd Bonds 0.01 A 0.02 A

rmsd Angles 158° 259°

Mean B value 25.6 A2 35.5 A2

 

67

reﬂections from a total of 763,409 measured reﬂections. Detailed data collection
statistics are found in Table 2.2.

2.2.2 Structure Determination and Refinement of the P6.22 Structure
The structure was solved by molecular replacement using the AMoRe7 program and
the structures of K1 and K2 from human angiostatin as search models (PDB id
lKIO)7. A translation search with K2 gave one solution and a translation search
with Kl after ﬁxing the K2 translation solution also yielded one solution. The
crystal packing of K2 and K1 was consistent with the location of K2 and K1 in the
structure. This solution had a correlation factor of 41.9 and an R value of 44.2%,
after rigid body reﬁnement. Fixing the positions of angiostatin K1 and K2 and
calculating an electron density map revealed density corresponding to the VEK-3O
helix. The same electron density map revealed density connecting K1 and K2. Even
after ﬁxing K1, K2, and VEK-30, rotation and translation searches performed using
human angiostatin K3 as a search model (PDB id IKIO3) were unsuccessful. After
reﬁnement of K1, K2, and VEK-30, an electron density map revealed density for
two residues in the K2-K3 linker peptide, residues T244 and T245, but no electron
density was ever seen past residue T245. No electron density was seen for C297 of
K3 at the inter-kringle K2-K3 disulﬁde bond. When C297 was built into the
Structure, negative density was calculated for residue C297 and an increase in R and
Rm were observed. This indicated that residues 246 to 333 encompassing K3 are
disordered in this structure. CNSlo reﬁnements of the angiostatin/VEK-30 structure
yielded extremely high B-factors averaging 68 A2. In order to ﬁnish reﬁning the

structure and lower the average B-factors, TLS restrained reﬁnements were carried

68

out within the CCP48 programs suite. Three TLS groups were defined using K1 in
group 1, K2 in group 2, and VEK-30 in group 3. The refinement parameters are
shown in Table 2.3. The ﬁnal model (Figure 2-4) includes 165 residues from
angiostatin (containing amino acids 81-245), 23 residues from VEK-30 (containing
amino acids 88-110), and 321 water molecules. In addition to the water molecules,
electron density for a single dioxane molecule, the additive used in crystallization,
was located in two different conformations. A Ramachandran diagram is a plot of 0
(angle between N and C01) and (p (angle between C and Ca). According to geometry
and steric restrictions the (I) and (p angles must be within certain values. A11 residues
with the exception of glycines and prolines, that don’t have a side chain, must lie
within these allowed regions. The red region is the most favored, bright yellow is
the additionally allowed region, dull yellow is the generously allowed region, and
white is the disallowed region. The Ramachandran plot of angiostatin in the
complex structure contained 138 non-glycine, non-proline residues (80.4%) in the
most favored regions and 15.9% in the additionally allowed regions with C169 and
E165 in the disallowed region (Figure 2-5). The Ramachandran plot of VEK-30 in
the complex contained 100% in the most favored regions (Figure 2-6). An example

of the 2Fo-FC map contoured at 10' is shown in Figure 2-7.

69

 

 

 

 

 

o—i
J-
LII
3
,—
U
U
C
6

-180 4’35 -90 45

Figure 2-5 Ramachandran plot of angiostatin in the angiostatin/VEK-30
complex in the P6.22 space group. Phi (degrees) is x and Psi (Degrees) is y.

70

 

-135-

 

 

 

 

4.8) -135 —90

Figure 2-6 Ramachandran plot of VEK-30 in the angiostatin/VEK-30 complex
in the P6.22 space group. Phi (degrees) is x and Psi (Degrees) is y.

71

   

  
  
 

I: \ fl’.’ A?!
‘u 411
t - V-
[A 3.1V "=5. 9.5! M,
:1,

  
  

" 0.1.,

   

  

Figure 2-7 An example of the 2Fo-Fc map contoured at 16 of the angiostatin/VEK-30
complex in the P6.22 space group.

2.2.3 Crystallization and Data Collection of the P6. Structure

In order to ﬁnd K3, the heavy atom isomorphous replacement was used.
Heavy atom soaks were performed on angiostatin/VEK-30 complex crystals grown
in the native crystal condition using 5 mM Pt(C5H5N)2C12 (cis-dichloro-
bis(pyridine)—platinum11). After soaking for 24 hr, the crystal was back-soaked in its
cryoprotectant solution. The crystal was ﬂash frozen by immersion in liquid N2.
Data were collected at the DND-CAT S-ID beamline at the APS to a resolution of
2.8 A. The crystal to detector distance was 200 mm and 70° of data were collected
with an oscillation of 05°. The data were processed and scaled using The HKL
suite15 in the space group P6. with unit cell parameters a = b = 58.8 A and c = 389.2
A. Assuming two molecules of angiostatin/VEK-30 complex per asymmetric unit,

the crystal volume per protein mass is 2.9, which corresponds to approximately

72

57.4% solvent content in the crystal. This value is within the range observed for
protein crystals. The crystal parameters of the angiostatin/VEK-30 P6. crystal are
listed in Table 2.1. A synchrotron X-ray diffraction data set to a resolution of 3.0 A,
with an overall I/o of 12.38, was obtained. The data was 78.5% complete with an
Rmerge of 8.4% for 18,600 unique reﬂections from a total of 281,739 measured
reﬂections. Detailed data statistics are tabulated in Table 2.2.
2.2.4 Structure Determination and Refinement of the P6. Structure
Automated heavy atom Patterson searches using the program SOLVE16
failed to locate heavy atom positions. Since the space group changed from the
native crystal, molecular replacement was performed in order to ﬁnd K3. Molecular
replacement using MolRep with CCP48 and the P6 .22 structure as a model yielded
two translation solutions with a correlation of 46.6 and 73.6 and an R of 43.2% and
31.2%, respectivelyg. These two solutions represented the two molecules of K1, K2
and VEK-30 in the asymmetric unit of the P6. crystal form. Molecular replacement
using K3 of angiostatin (PDB id IKIO) as a model produced no solution. An
electron density map was calculated and density for residues T244 and T245 of the
K2-K3 linker was seen as in the P6.22 structure. No density was seen after residue
T245. However, density was seen at the inter-kringle K2-K3 disulﬁde bond (C169-
C297) for C297 of K3 for only one of the molecules in the asymmetric unit (Figure
2-8). Residues P296 and K298 were then built into corresponding density in the K3
disulﬁde region. However, no density was seen beyond K298 or before P296.
Further reﬁnement of the structure was carried out using TLS restrained reﬁnement

in CCP48 similar to that of the P6.22 structure. The ﬁnal model includes 332

73

residues from angiostatin (containing amino acids 81-245 of two molecules A and B
and 296-298 of molecule A), and 48 residues from VEK-30 (containing amino acids
88-110). The Ramachandran plots of the P6. structure are shown in Figure 2-9 and
2-10. Molecule A contained 139 non-glycine, non-proline residues (59%) in the
most favored regions and 36% in the additionally allowed regions with E165, C169,
and K212 in the disallowed region (Figure 2-9). Molecule B contained 137 non-
glycine, non-proline residues (55.5%) in the most favored regions and 38.7% in the
additionally allowed regions with E130, C169, and K212 in the disallowed region
(Figure 2-10). The Ramachandran plots of the corresponding VEK-30 peptides are
shown in Figure 2-11 and 2-12. Molecule A contained 22 non-glycine, non-proline
residues (59.1%) in the most favored regions and 31.8% in the additionally allowed
regions with T89 in the disallowed region (Figure 2-11). Molecule B contained 22
non-glycine, non-proline residues (72.7%) in the most favored regions and 22.7% in
the additionally allowed regions (Figure 2-12). Reﬁnement statistics are shown in
Table 2.3. All model building was done using TURBO FRODOl7 and the
reﬁnement and map calculations were carried out using CNS10 and CCP48’18. The

details of this structure will be discussed in Chapter H1.

74

 

Figure 2-8 An example of the 2Fo-Fc map contoured at 1.26 of the angiostatin/VEK-
30 complex P6. structure. The map is centered on the interkringle K2-K3 disulﬁde
bond (C169-C297).

75

 

 

 

 

 

Figure 2-9 Ramachandran plot of angiostatin molecule A in the angiostatinNEK-3O
complex P6. structure. Phi (degrees) is x and Psi (Degrees) is y.

 

-135-'

 

ﬂ“ .'! A
-1&) -135 -90

 

Figure 2-10 Ramachandran plot of angiostatin molecule B in the angiostatin/VEK-30
complex P6. structure. Phi (degrees) is x and Psi (Degrees) is y.

77

 

 

 

 

-180 —135 —90 45 0 45 90 135 180

Figure 2-11 Ramachandran plot of VEK-30 molecule A in the angiostatinNEK-30
complex P6. structure. Phi (degrees) is x and Psi (Degrees) is y.

 

 

 

 

 

-180 -135 -90 45 0 4'5 90 135 180

Figure 2-12 Ramachandran plot of VEK-30 molecule B in the angiostatin/VEK-30
complex P6. structure. Phi (degrees) is x and Psi (Degrees) is y.

2.3 Ca2+/con-G
2.3.1 Crystallization and Data Collection

The con-G peptide was synthesized, puriﬁed, and characterized as previously
described5 and provided by Frank Castellino and co-workers in lyophilized form.
The lyopholized solid of con-G was dissolved in 50 mM CaCl2 and 100 mM Tris-
HCl, pH 8, to a concentration of 10 mg/ml. Previous experiments had shown that
the concentration of Ca2+ and pH were important in forming a dimeric structure of
Ca2+/con-G5. The search for well diffracting crystals was performed using several
crystallization screens at two different temperatures, 25°C and 4°C. The best
crystals of con-G were grown at 25°C by the hanging drop vapor diffusion method in
35% v/v 1,4-dioxane from the Hampton Research Crystal Screen 11 #4. Crystals ﬁrst
appeared within 24 hr and were highly variable in
size (Figure 2-13).

Crystals of con-G were brieﬂy soaked in a
cryoprotectant solution containing 35% 1,4-dioxane

and 30% MPD at 25°C and ﬂash frozen by

 

immersion in liquid N2. X-ray diffraction data were
Figure 2-13 Crystal of Ca2+/con-G.

collected at the Advanced Photon Source COM-CAT
32-ID at the Argonne National Laboratory to a resolution of 1.2 A and were

processed and scaled using the HKL suite of programs"'5

in the tetragonal
spacegroup P4222 with unit cell parameters a = b = 29.3 A and c = 46.9 A. The

crystal to detector distance was 60 mm, and 200° of data were collected with an

80

Table 2.4 Crystal parameters for the metal bound conantokin structures

 

 

C212+Icon-G Ca2+/con-T[K77] Cd21/Mg2+/con-
T [K77]
Crystal Form Tetragonal Cubic Hexagonal
Space Group P4222 P4332 P6322

Unit Cell (a,b,c) 29.3, 29.3, 46.9 A 89.0, 89.0, 89.0 57.1, 57.1, 32.8 A

(01, [3, v) 90.0, 90.0, 900° A 90.0, 90.0, 120.0°
90.0, 90.0, 900°

Solvent Content 46% 71.3% 43.8%

Mol. Per Asymu 1 2 1

 

oscillation of 1°. The crystal parameters of con-G are listed in Table 2.4. Assuming
one molecule of con-G per asymmetric unit, the crystal volume per protein mass is
2.3, which corresponds to approximately 46% solvent. A synchrotron X-ray
diffraction data set to a resolution of 1.2 A, with an overall I/o of 20.91, was
obtained. The X-ray diffraction data was 97.9% complete with an Rmerge of 7.9% for
12,427 unique reﬂections from a total of 266,169 measured reﬂections. Detailed
data collection statistics are found in Table 2.5.
2.3.2 Structure Determination and Refinement

In order to determine the structure of Ca2+/con-G molecular replacement
using the program MolRep within CCP48 was utilized. Many different models were
tried in order to obtain a correct solution. Models using known NMR structures of
con-G were not helpful in molecular replacement attempts as well as polyalanine

helices of many different lengths from previously determined crystal structures. The

81

Table 2.5 Data statistics for the metal bound conantokin structuresa

Ca2+/con-G Ca2+/con-T[K77] Cd2+/Mg2+/con-T[K77]

 

Wavelength, A 1.0000 1.0000 1.0332
Resolution, A 1.2 (1.29-1.24) 1.6 (1.66-1.60) 1.2 (129-127)
Completeness, % 97.9 (97.7) 99.5 (99.5) 84.2 (62.7)
I/o 20.91 (1.31) 21.31 (1.88) 29.34 (2.70)
Rmerge, % 7.9 9.8 6.7
Unique Reﬂections 12,427 16,539 10,347
Measured Reﬂections 266,169 215,476 132,076

 

aThe parentheses denote the values for the last resolution shell

Table 2.6 Reﬁnement statistics for the metal bound conantokin structures

 

Ca2+/con-G Ca2+/con-T[K77] Cd2+/Mg2+/con-T[K77]

 

Rwork, % 11.99 13.35 11.03

Rfree, % 16.15 18.63 14.40
Resolution, A 8-1.25 10—1.7 8-1.3
rmsd Bond Length, A 0.01 0.01 0.01
rmsd Angle Length, ° 0.02 0.03 0.02
Mean B-factors, A2 28.6 34.6 19.1

 

same was true for the programs AMoRe7 and CaspRg. In order to determine the
structure, data sets were collected near the Ca2+ edge (5000 kEV and 6000 kEV) in
order to determine the positions of Ca2+ sites. However, the program SOLVE16
failed to ﬁnd any sites.

The isomorphous replacement method using natively grown crystals soaked
with BaCl2 was also done. Previous experiments had shown that con-G in the
presence of Ba2+ formed a dimeric structures. Ba2+ has more electrons than Ca2+ and
the Ba2+ edge is easier to achieve than Ca2+. Soaks were performed on natively
grown Ca2+/con-G crystals. Different concentrations as well as different time spans
were tried. Crystals that showed no evidence of degradation were taken to the
synchrotron. A number of data sets were collected and compared with the native for
intensity differences and isomorphism. However, the program SOLVEl6 failed to
ﬁnd any Ba2+ sites.

The structure of Ca2+/con-G was ﬁnally solved by molecular replacement
using the recently developed program, Phaser”, and a 14-mer polyalanine helix as a
model. For con-G, brute force rotation and translation functions were performed,
followed by reﬁnement. The translation Z-score and LLG were 6.83 and 48.46,
respectively indicating a correct solution. The LLG score after reﬁnement improved
to 63.82 further implicating that a correct solution had been obtained. Phaserll
proved to be the only molecular replacement program capable of locating a correct
solution, as many other programs such as AMoRe7, MolRep within CCP48, and

CaspR9 failed. Examination of the crystal packing of the solution revealed that no

83

collisions occur with any of the symmetry-related molecules further supporting that
the molecular replacement solution is correct.

A rigid body reﬁnement was performed using Refmac5'9 within the CCP4
suite of programsg’zo, producing Rwork 53.7% and Rm... 54.6%. Side-chains were built
into the structure from the electron density map. Subsequent reﬁnements of all 17
residues of con-G led to an Rwork of 35.9% and RM 45.4%. In order to determine
the positions of the Ca2+ ions, a Fo-Fc map was calculated and contoured at 80
(Figure 2-14). This map showed that three calcium ions were present within the
structure. After these ions were added, a reﬁnement was performed where the R“...
and Rﬁ-ee dropped to 27.16% and 32.29%, respectively. Ar‘p/warplg’21 was then run in
order to add water molecules. Further cycles of reﬁnement using the RefmacS19
program in CCP48’20 produced a ﬁnal Rm... of 18.22% and R38, 21.37%. In order to
ﬁnish reﬁning the structure, the program SI-IELXL-9722 was employed. Further
reﬁnement and additional water molecules were added, as well as adding side-chain
secondary conformations of residues L5, Q9, and 816. SI-IELXL produced a ﬁnal
Rm... and Rm of 11.99% and 16.15% respectively. The ﬁnal model includes all 17
residues of con-G, 3 Ca2+, and 68 water molecules. Electron density was also seen
for the C-terminal NH2 atom. The Ramachandran plot of the structure contained 15
non-glycine, non-proline residues with 100% in the most favored region (Figure 2-
15). Reﬁnement statistics are shown in Table 2.6. All model building was done

using TURBO FRODO”.

84

 

Figure 2-14. Overall structure of the con-G antiparallel dimer with Fo-Fc map
calculated in the absence of calcium and contoured at 86. The gray molecule is
related by crystallographic two-fold symmetry that forms the dimer. The side-
chains colored green (E2, 74, L5, Q9, 112) have been Shown to decrease potency of
the NIvaA receptor when mutated to Ala by at least 100-fold. The side-chains
colored yellow (G1, 73, L11, R13) have been shown to decrease the potency of the
NMDA receptor when mutated to Ala by at least 10-60-fold. Ca3 is positioned
directly on the crystallographic two-fold axis. CalA and Ca2A are
crystallographically related to CalB and Ca2B, respectively. The peptide side-
chains are colored by atom type.

85

 

 

 

 

 

30

Figure 2-15 Ramachandran plot of the Ca2+/con-G structure.
Phi (Degrees) is x and Psi (Degrees) is y.

86

2.4 Ca2*lcon-T[K7y]
2.4.1 Crystallization and Data Collection

The con-T[K77] peptide was synthesized, puriﬁed, and characterized as
previously described5 and provided by Frank Castellino and co-workers in a
lyophilized form. The lyopholized solid of con-T[K77] was dissolved in 50 mM
CaCl2 and 100 mM Tris-HCl, pH 8, to a concentration of 10 mg/ml. Previous
experiments had shown that the concentration of Ca2+ and pH were important in
forming a dimeric structure of Ca2+/con-T[K77]4. The search for well diffracting
crystals was performed using several crystallization screens at two different
temperatures, 25°C and 4°C. The best crystals of con-T[K77] were grown at 4°C by
the hanging drop vapor diffusion method in 3 M (NH4)2SO4 and 0.1 M NaOAc, pH
5.5. Crystals ﬁrst appeared after 1 year.

Crystals of con-T[K77] were brieﬂy soaked in a cryoprotectant solution
containing 4 M sodium dihydrogen phosphate, pH 5.5/ 20% glycerol at 4° C, and
ﬂash frozen by immersion in liquid N2. The salt component was changed in order to
cryoprotect the crystal since the glycerol concentration would not be high enough in
the original crystal condition from above. Data were collected at the Advanced
Photon Source COM-CAT 32-ID at the Argonne National Laboratory to a resolution
of 1.6 A, and data were processed and scaled using AUTOMAR in the cubic
spacegroup P4332 with unit cell parameters a = b = c = 89.0 A. The crystal to
detector distance was 100 mm and 40° of data were collected with an oscillation of

1°. The crystal parameters of con-T[K77] are listed in Table 2.4. Assuming two

87

molecules of con-T[K77] per asymmetric unit, the crystal volume per protein mass is
4.3, which corresponds to approximately 71.3% solvent in the crystal. A
synchrotron X-ray diffraction data set to a resolution of 1.6 A, with an overall I/o of
21.31, was obtained. The X-ray diffraction data was 99.5% complete with an Rmerge
of 9.8% for 16,539 unique reﬂections from a total of 215,476 measured reﬂections.
Detailed data statistics are found in Table 2.5.
2.4.2 Structure Determination and Refinement

The structure of Ca2+/con-T[K77] was solved by molecular replacement
using the program, Phaser“, and a 14-mer polyalanine helix as a model. Automated
searches using were performed in order to ﬁnd the correct solutions. The translation
Z-score and LLG for the ﬁrst molecule were 8.6 and 57, respectively, and 19.1 and
216 for the second molecule indicating that two distinct correct solutions had been
found. Examination of the crystal packing of the solution revealed that no collisions
occur with any of the symmetry-related molecules, further supporting that the
molecular replacement solution is correct.

A rigid body reﬁnement was performed as with con-G, producing Rwork
52.5% and Rﬁ-ee 56.5%. After building in side-chains and adding Ca2+ and water
molecules, an Rwork of 20.36% and Rfree 23.08% was obtained. All model building
was done using the program TURBO FRODO17 and reﬁnement and map
calculations were carried out using CCP48. In order to ﬁnalize the reﬁnement of the
structure, SI-IELXL-9722 was employed, producing a ﬁnal Rwork and Rﬁ-ec of 13.35%
and 18.63%, respectively. The ﬁnal model includes all 21 residues of con-T[K77], 4

Ca2+, and 142 water molecules (Figure 2-16). No alternative amino acid Side-chain

88

conﬁrmations are observed for any of the residues. The NH2 moiety of the C-
terrninal amide was not seen in the electron density map of either con-T[K77] helix.
As with con-G, the positions of the Ca2+ ions was determined by calculating a Fo-Fc.

map in the absence of calcium and contoured at 86 (Figure 2-16).

 

4

.

  
 

5

Av Y4

G1

  
    

Figure 2-16. The overall structure of the antiparallel con-T[K77] dimer with a
Fo-Fc map calculated in the absence of calcium and contoured at 86. The
side-chains colored green (E2, 74, Y5) have been shown to decrease potency of
the NIVHDA receptor when mutated to Ala by at least lOO-fold. The side-chains
colored yellow (G1, 73, M8, L12) have been shown to decrease the potency of
the NMDA receptor when mutated to Ala by at least 10-60-fold. In all cases,
the peptide side-chains are colored by atom type.

89

The Ramachandran plot of helix A of the structure contained 19 non-glycine, non-
proline residues (94.7%) in the most favored region and 5.3% in additional allowed
region (Figure 2-17). The Ramachandran plot of helix B of the structure contained
19 non-glycine, non-proline residues 100% in the most favored region (Figure 2-18).

Reﬁnement statistics are shown in Table 2.6.

 

 

 

 

Figure 2-17 Ramachandran plot of helix A in the Ca2+/con-T[K77]
structure. Phi (Degrees) is x and Psi (Degrees) is y.

90

 

 

 

 

-180 -135 —90 45 0 4'5 90 135 180

Figure 2-18 Ramachandran plot of helix B in the Ca2+/con-T[K77]
structure. Phi (Degrees) is x and Psi (Degrees) is y.

2.5 Cd2*lMgz*Icon-T[K7v]
2.5.1 Crystallization and Data Collection

The con-T[K77] peptide was synthesized, puriﬁed, and characterized as
previously described5 and provided by Frank Castellino and co-workers in a
lyophilized form. The lyopholized solid of con-T[K77] was dissolved in 50 mM
MgC12 and 100 mM Tris-HCl pH 8 to a concentration of 10 mg/ml. The search for
well diffracting crystals was performed using several crystallization screens at two
different temperatures, 25°C and 4°C. The best crystals were grown at 25°C by the
hanging drop vapor diffusion method in 1 M sodium acetate, 0.1 M Hepes pH 7.8,
and 0.05 M CdSOa from Hampton Research Crystal Screen 11 #34. The crystals
appeared after 3 days. The crystals were brieﬂy soaked in a cryoprotectant solution
containing 1 M sodium acetate, 0.1 M Hepes pH 7.8, 0.05 M CdSOa, and 30%
glycerol at 25°C and ﬂash frozen by immersion in liquid nitrogen. X-ray diffraction
data was collected at the Advanced Photon Source SBC 19-BM at Argonne National
Laboratory to a resolution of 1.2 A and data were processed and scaled using the
HKL suite of programs to the hexagonal space group P6322 with unit cell parameters
a = b = 57.1 A and c = 32.8 A. The crystal to detector distance was 125 mm and
120° of data were collected with an oscillation of 1°. The crystal parameters are
listed in Table 2.4. Assuming one molecule of con-T[K77] per asymmetric unit, the
crystal volume per protein mass is 2.2, which corresponds to approximately 43.8%
solvent in the crystal. A synchrotron X-ray diffraction data set to a resolution of 1.2

A, with an overall 1/6 of 29.34, was obtained. The X-ray diffraction data was 84.2%

92

complete with an Rmerge of 6.7% for 10,347 unique reﬂections from a total of
132,076 measured reﬂections. Detailed data statistics are found in Table 2.5.
2.5.2 Structure Determination and Refinement

The structure was solved by molecular replacement using the program
Phaser“ and the Ca2+/con-T[K77] (PDB id 2DPR) as a model. The rotational and
translational Z-scores were 9.4 and 10.4, and the LLG was 150 indicating a correct
solution. Examination of the crystal packing of the solution revealed that no
collisions occur with any of the symmetry-related molecules further supporting that
the molecular replacement solution is correct.

A rigid body reﬁnement was performed using RefmacS '9 within the CCP48
suite of programs producing Rm... and Rfm of 53.2%. An anomalous difference map
was calculated using CCP48 in order to determine which sites were Cd2+ and which
sites were Mg2+ since Cd2+ existed in the crystallization condition (Figure 2-19). A
total of four Cd2+ sites were seen in the electron density of the anomalous difference
map at peaks of 246 (Cd3), 146(Cd1), and 116 (Cd2 and Cd4). The program
SOLVE16 was used to determine actual Cd2+ site occupancy. Only Cd3 is 100%
occupied whereas Cdl is 80% occupied and Cd2 and Cd4 are both 50% occupied.
We were unable to determine whether these partially occupied Cd2+ sites were also
partially occupied by magnesium since magnesium does not have an anomalous
signal. One Mg2+ site was located from the Fo-Fc map contoured at 86 (Figure 2-
20). Figure 2-21 also shows electron density for the Cd2+ sites. Subsequent
reﬁnements using Refrnacs19 within CCP420 of all 21 residues of con-T[K77] after

adding 4 Cd“, 1 Mg2+, and 28 waters produced an Rwork of 18.61%

93

10 Cd4
1111 Q
OMgl

E16

 

0

K19

Figure 2-19 The overall structure of Cd2+lMg2+lcon-T[K77] with an anomalous
difference map contoured at 86. The four Cd2+ are shown as blue spheres and
the Mg2+ is shown as a magenta sphere. The side-chains are colored by atom.

94

and Rﬁce 20.47%. In order to ﬁnish reﬁning the structure, the program SHELXL-
9722 was employed. Further refinement was done by adding secondary side-chain
conformations of Met8 and Val 17 and adding more waters. SHELXL produced a
ﬁnal Rwork and Rfrce of 11.03% and 14.40% respectively. The ﬁnal model includes
all 21 residues of con-T[K77], 4 Cd2+, 1 Mg2+ and 61 water molecules. The NH2
moiety of the C-terminal amide was seen in the electron density map of the con-
T[K77] helix. The Ramachandran plot of the structure contained 19 non-glycine,
non-proline residues (100%) in the most favored region (Figure 2-21). Reﬁnement

statistics are shown in Table 2.6.

95

 

Figure 2-20 The overall structure of Cd2+/Mg2+/con-T[K77] with a Fo-Fc map
calculated in the absence of metal cations and contoured at 56. The four Cdz“
are shown as blue spheres and the Mg2+ is shown as a magenta sphere. The
side-chains are colored by atom.

96

 

 

 

 

 

480 -135 -90 45 o 45 90 135 180

Figure 2—21 Ramachandran plot of the Cd2+/Mg2+/con-T[K77].
Phi (Degrees) is x and Psi (Degrees) is y.

97

2.6

to

10.

Literature Cited

McPherson, A. Introduction to Macromolecular Crystallography, John
Wiley & Sons Inc (Hoboken, NJ, 2003).

Hammes, G.G. Spectroscopy for the Biological Sciences, John Wiley & Sons
Inc (Hoboken, NJ, 2005).

Abad, M.C. et al. The X-ray crystallographic structure of the angiogenesis
inhibitor angiostatin. J Mol Biol 318, 1009-17 (2002).

Dai, Q., Castellino, P.J. & Prorok, M. A single amino acid replacement
results in the Ca2+-induced self-assembly of a helical conantokin-based
peptide. Biochemistry 43, 13225-32 (2004).

Dai, Q., Prorok, M. & Castellino, R]. A new mechanism for metal ion-
assisted interchain helix assembly in a naturally occurring peptide mediated
by optimally Spaced gamma-carboxyglutamic acid residues. J Mol Biol 336,
73 1-44 (2004).

Stout, G.H. & Jensen, L.H. X-ray Structure Determination, John Wiley &
Sons Inc (New York, N. Y., 1989).

Navaza, J. AMoRe, an automated program for molecular replacement. Acta
Crystallog. sect. A 50, 157-163 (1994).

Collaborative Computational Project, N. The CCP4 Suite: Programs for
Protein Crystallography. Acta Cryst D50, 760-763 (1994).

Claude, J.-B., Suhre, K., Notredame, C., Claverie, J .-M. & Abergel, C.
CaspR: a web-server for automated molecular replacement using homology
modelling. Nucleic Acids Research 32, W606-W609 (2004).

Brunger, A.T. X-PLOR, version 3.1, a system for X-ray Crystallography and
NMR, (New Haven, CT, 1992).

98

11.

12.

13.

14.

15.

l6.

17.

18.

19.

20.

McCoy, A.J., Grosse-Kunstleve, R.W., Storoni, L.C. & Read, R.J.
Likelihood-enhanced fast translation functions. Acta C ryst D61, 458-64
(2005).

Shepard, S.R., Boucher, R., Johnston, J ., Boemer, R., Koch, G., and Madsen,
J. Large-scale puriﬁcation of recombinant human angiostatin. Protein Expr
Purzf20, 216-227 (2000).

Rios-Steiner, J .L., Schenone, M., Mochalkin, 1., Tulinsky, A., and Castellino,
F. J. Structure and binding determinants of the recombinant kringle-2 domain

of human plasminogen to an internal peptide from a group A Streptococcal
surface protein. J Mol Biol 308, 705-19 (2001).

Otwinowski, Z. Oscillation data reduction program, 56-62 (SERC
Daresbury Laboratory, Daresbury, UK, 1993).

Otwinowski, Z.a.M., W. Processing of x-ray diffraction data collected in
oscillation mode. Methods Enzymol 276, 307-326 (1997).

Terwilliger, T.C. & Berendzen, J. Automated MAD and MIR structure
solution. Acta Crystallographica D55, 849-861 (1999).

Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Improved methods
for building protein models in electron density maps and the location of
errors in these models. Acta Cryst A47, 110-119 (1991).

Brunger, A.T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., and
Grosse-Kuntsleve, R. W. Crystallography & NMR System: a new software
suite for macromolecular structure determination. Acta Crystallog. sect. D
54, 905-921 (1998).

Murshudov, G.N., Vagin, A.A. & Dodson, E.J. Reﬁnement of
Macromolecular Structures by the Maximum-Likelihood Method. Acta Cryst
D53, 240-55 (1997).

COLLABORATIVE COMPUTATIONAL PROJECT, N. The CCP4 Suite:
Programs for Protein Crystallography. Acta Cryst D50, 760-3 (1994).

99

IQ
Ix)

Perrakis, A., Morris, R.J.H. & Lamzin, V.S. Automated protein model
building combined with iterative structure reﬁnement. Nature Struct Biol,
458-463 (1999).

Sheldrick, GM. & Schneider, T.R. SHELXL: High resolution reﬁnement. in
Methods in EnzymOIOgy, Vol. 277 (eds. Sweet, R.M. & Carter Jr, CW.) 319-
43 (Academic Press, Orlando, Florida, 1997).

100

Chapter III

The Three Dimensional Structure of AngiostatinNEK-30

The angiostatin/VEK-30 complex structure has been solved and further
described in this chapter (Cnudde SE, et al. 2006. Biochemistry. 45(37): 11052-60).
The methods used in order to determine this structure have been thoroughly described in
Chapter II. The complex of angiostatinNEK-30 was crystallized and its structure
determined in two different space groups, P6.22 and P6.. The P6. structure shows the
interkringle K2-K3 disulﬁde and two residues on either side of the disulﬁde region in
K3. This and other comparisons discussed herein prove that plasminogen kringle
domains are capable of signiﬁcant structural rearrangement relative to one another. This
provides insight into the mechanism of plasminogen during streptococcal infection and
further characterizes plasminogen’s importance as a critical serine protease for bacteria.
Also, VEK-30 binds speciﬁcally to angiostatin K2 through a pseudo-lysine moiety. This
pseudo-lysine motif seen in VEK-30 may help to identify unknown angiostatin ligands.
3.1 Overall Structure of AngiostatinNEK-30 Complex

The resolved portion of the overall structure of the angiostatin/VEK-30 complex
is shown in Figure 2-4. In comparison to the unbound form of angiostatin (Figure 1-6
shows the overall structure of angiostatin), no electron density was detected for any of
the residues of Pg-K3 (residues C256-C333) in the angiostatin/VEK-3O complex.
Additionally, electron density is absent for most of the residues in the K2-K3
interkringle peptide (residues P246-Q255). However, the remaining residues of the Pg-

KI and Pg-K2 domains are well ordered and exist in a relatively extended orientation

101

with essentially no interactions between them. There are also very few interactions
between either Pg-Kl or Pg-K2 and the linker peptide that connects the two domains.
Electron density for the VEK-30 peptide is absent for N-terrninal residues Val85-Lys87
and C-terminal residues K111-Y114 (Figure 1-4 shows the primary sequence of VEK-
30). However, of the remaining 23 residues, 20 correspond to about 5 turns of a well-
deﬁned or-helix that is approximately 30 A long (Figure 2-4).

3.2 Interactions Between VEK-30 and Angiostatin

Figure 3-1 and 3-2 shows the interactions occurring between VEK-30 and angiostatin.
Table 3.1 lists distances involved in the interaction between VEK-3O and angiostatin K2.
VEK-30 interacts with K2 of angiostatin from VEK-30 residues Glu93 to Glu104. This
region encompasses most of the ﬁrst direct repeat and the ﬁrst residue of the second
(Figure 1-4 shows the VEK-30 primary sequence). Most of these interactions occur
between a single face of the helix and the K2 motif. One side of the helix consisting of
residues K98-E104 makes numerous contacts with the angiostatin K2 LBS. Hydrogen
bond and salt bridge electrostatic interactions occur between angiostatin K2 and VEK-
30, both within and outside of the K2 LBS. There are no interactions between
angiostatin K1 and VEK-30. The total number of contacts between angiostatin K2 and
VEK-30 with a distance less than 3.8 A is about 61. The LBS of the K2 domain
contains consensus anionic (Asp219 and Glu221) and cationic (Arg234) centers (Figure
3-2 and 3-3). VEK-30 forms a pseudo-lysine by inserting the residues Arg101 and
Glu104 located on one face of the a-helix into the LBS of angiostatin K2. The principal
interactions that occur between the anionic loci of the K2 LBS and VEK-30 involve

residues Asp219 and Glu221 of K2 and Lys98, Arg101, and Hile2 of VEK-30.

102

Critical salt bridge interactions occur between angiostatin K2 LBS residue Asp219 and
VEK-30 residues Arg101 and Lys98. In addition, Glu221 makes a tight (distance 2.57
A) salt bridge contact with VEK-30 residue Arg101, and also makes contacts with
His102 of VEK-30. Figure 3-2 shows a rt-cation interaction. This interaction occurs
between the guanidino group of Arg101 of VEK30 and one of the two K2 Trp residues
(Trp235) that make up the hydrophobic portion of the LBS (distances shown in Table
3.1). At the cationic Site, a salt bridge exists between residues Arg234 of K2 and
Glu104 of VEK-30. An interaction between residues Arg220 of angiostatin and Asp91
of VEK-30 is also observed. VEK-30 Arg101 has proved to be vital for angiostatin K2
binding, since mutating this residue to Ala results in no measurable afﬁnity for K22.
Mutating VEK-30 residues Lys98 and HiS102 to Ala and Glu104 to Gln results in a
decreased affinity for the angiostatin K2 domain. These results suggest that the overall
binding relies on interactions mediated by VEK-30 residues Lys98, Arg101, Hile2 and
are consistent with our crystallographic results. There are also signiﬁcant interactions
between VEK-30 and K2 outside the LBS. A salt bridge interaction occurs between the
angiostatin K2 residue Lys204 and VEK-30 residue Glu93 (Figure 3-1). This
interaction was also ﬁrst observed in the structure of the isolated K2NEK-30 (Pg-K2)
complexz. It is unknown whether this exosite interaction is critical for binding. The
exosite salt bridge interaction between angiostatin K2 residue Ly8204 and VEK-30
residue Glu93 is further stabilized by hydrogen bonds and hydrophobic contacts of
VEK-30 residues Leu94, Leu97, and Lys98 and angiostatin K2 residues Tyr200,

Phe205, Asp219, and Arg220.

103

R234

 

E104
W235
H102 D Y200
K204
E221 K98
D219 E93
R220

2 D91

Figure 3-1 Interactions between the angiostatin K2 LBS and VEK-30. R101 and
E104 are Spaced by almost one helical turn and form the pseudo-lysine residue.
Angiostatin is colored green and VEK-30 is shown in magenta. K2 residues are
labeled using plasminogen numbering and VEK-30 residues are labeled with
PAM numbering. All atoms are colored by atom type (nitrogen, blue; oxygen,
red).

W225 w235
3.77A

101

 

D219

Y200

Figure 3—2 A rt-cation interaction occurs between the guanidino group of
VEK-30 R101 and angiostatin K2 LBS residue W235. Angiostatin is
shown in green and VEK-30 is shown in magenta. All atoms are colored by
atom type (nitrogen, blue; oxygen, red). Angiostatin residues are labeled
with plasminogen numbering. VEK-30 residues are labeled with PAM
numbering.

105

Most of the interactions discussed above are also seen in the structure of a mutated Plg
K2 domain (C169G, E221D, L237Y) bound to VEK-302. However, two of the mutated
residues, D221 and Y237, made interactions with VEK-30 in the LBS, calling into
question whether the wild-type LBS would provide a similar interface. Our structure
conﬁrms that in Spite of these mutations, most of the interface is similar in the two
structures.
3.3 Kringle Domain Rotation

Comparison of the Structures of unbound angiostatin to the VEK-30 bound form
reveals that Kl has rotated signiﬁcantly from its unbound conformation when the K2
regions are overlaid (Figure 3-3). Kl of the angiostatin/VEK-30 complex rotates 48.1°
and translates about 0.5 A from its position in the unbound angiostatin crystal structure
(relative motion determined using the program DynDoml). Residues encompassing the
angiostatin K1-K2 linker peptide are the bending residues primarily responsible for such
a large rotation. More speciﬁcally, the changes in psi- and phi-dihedral angles of
residues Glu163 and Glu165 contribute signiﬁcantly to most of the interdomain rotation
and are likely responsible for the large rotation of the angiostatin K1 domain seen in the
angiostatin/VEK-30 structure (Figure 3-4). Cys166 also contributes to the rotation as
well. VEK-30 binding does not appear to be directly responsible for the rotation of K1
since there are no interactions between K1 and VEK-30. However, crystal packing
dictates the K1 position in each Structure indicating relative motion in solution.
Angiostatin K1 clashes into a symmetry-related molecule of angiostatin/VEK-30 when

both K2 regions are overlaid. The same phenomenon is seen

106

Table 3.1 Interactions between angiostatin and VEK-30

 

 

Angiostatin VEK-30 P6.22 (A)
Y200-OH L97-O 3.36
Y200-OH K98-N 3.19
Y200-OH R101-NH1 3.20
K204-NZ 1393-0131 3.61
K204-NZ 1593052 3.48

D219-OD1 R101-NH1 3.69

D219-OD2 R101-NH1 2.58

0219.0 K98-NZ 2.66

R220-NH1 D91-OD2 2.97
13221-0131 R101-NH2 2.57
E221-0E1 H102-NE2 3.48

W225-NE1 R101-NH1 3.67

W225-CE2 R101-NH1 3.77

W225-CH2 R101-NE 3.80

W225-CH2 R101-CD 3.80

W225-CZ2 R101-NH1 3.75

W235-CD2 R101-NH2 3.43

W235-CE3 R101-NH2 3.41

W235-CZ3 R101-NH2 3.43

W235-CH2 R101-NH2 3.45

w235-c22 R101-NH2 3.48

R234-NH1 E104-0E1 3.15

W235-NE1 R101-O 3.15

 

107

 

when K2 of the angiostatin/VEK-30 complex is overlaid onto angiostatin K2. This
motion demonstrates that the kringle domains of angiostatin are not rigidly positioned as
previously thought but are in fact mobile relative to each other. The LBS of Pg-Kl and
Pg-K4 are known to play important roles in the maintenance of the closed conformation
of Pg3 although the involvement of the NTD with these binding sites is unclear. The
fact that kringle domains of Pg are capable of signiﬁcant structural rearrangement
relative to one another indicates that transitions between the open and closed Pg

conformations may also involve signiﬁcant motion of K1-K3.

 

Figure 3-3 Overlay of the K2 domains of angiostatin and angiostatin/VEK30.
The structure of angiostatin is Shown in blue and the angiostatin/VEK-30
complex is shown in red (angiostatin) and green (VEK-30). H114 of angiostatin
is colored yellow in both structures.

108

ILE-159
LEU-160
GIN-161

 

6110-163
GLU-164

  

61.04165

 

 

 

 

G’s-1'56
I I
-200 0 200 400
Percentage“)

Figure 3-4 Residues responsible for interdomain rotation as determined by the
program DynDoml.

109

3.4 The Kringle 3 Domain

A second, P6 . crystal form of the angiostatinNEK-30 complex was produced by
soaking crystals grown in the ﬁrst, P6.22, form with Pt(II)-derivatizing molecules. This
second crystal form preserves most of the crystal packing of the P6.22 form, although it
loses the 2-fold axis perpendicular to the six-fold axis. This results in a crystal form
with two molecules in the asymmetric unit, instead of one. Though the structure of the
complex is quite similar, several important differences are seen between the P6. and
P6.22 crystal forms of the angiostatin/VEK-30 complex. In one molecule of the P6.
structure, the inter-kringle disulﬁde bond is ordered. Residues Pr0296, Cys297, and
Lys298 of K3 are seen in the electron density though no other residues of K3 were
identiﬁed (Figure 2-8). This shows that the K2-K3 interkringle disulﬁde bond remains
intact in these structures, even though K3 appears to be disordered in the structures.
Overlaying K2 of angiostatin with K2 of the angiostatin/VEK-30 complex in the P6.
structure displays a significant motion of about 3.3 A at the inter-kringle disulﬁde bond
(Figure 3-5). When residues Pro296-Lys298 of the VEK-30-bound form are overlaid
with the corresponding residues in the crystal structure of free angiostatin, it can be seen
that angiostatin K3 encroaches upon a crystallographic symmetry related molecule of
K2 (Figure 3-6). This indicates that there may be some structural differences within the
angiostatin K3 domain to avoid steric clashes with a crystallographically related
molecule. It should be noted that the conformation of the tripeptide alone ﬁts in either

structure. The motion of angiostatin K3 is limited due to the inter-kringle disulﬁde bond

110

 

Figure 3-5 Overlay of the K2 of angiostatin with K2 of the angiostatinNEK-
30 complex from the P6. crystal form where 3 residues of angiostatin K3 are
ordered. The P6. structure is shown in red and the structure of angiostatin is
shown in blue. The residues are labeled with Pg numbering. The inter-
kringle disulﬁde bond in the P6. complex structure has rotated 3.3A from its
position in the angiostatin structure.

111

a}; d3
~V 5 ,. 431’ ..
it.“ v - ﬁﬁl’ 043;“:

V‘s \, P‘ v“ '
4t a
‘rﬁ ’ ’ ' '*‘ g:‘.‘
I
N .
‘

Figure 3—6 An overlay of residues Pr0296-Ly5298 of the VEK—30-
bound with the corresponding residues in the crystal structure of
free angiostatin. Angiostatin in the angiostatin/VEK-30 complex
is colored magenta, K3 of angiostatin is shown in green, and the
angiostatin/VEK30 symmetry related molecule is shown in cyan.

112

between residues Cysl69 of K2 and Cy8297 of K3 (Figure 1-6). It has been suggested
that the disruption of the K2-K3 inter-kringle disulﬁde bond is required for maximum
angiogenic inhibition". However, the angiostatin double mutant (C169S, C297S), which
eliminates the inter-kringle disulﬁde bond, has little effect on angiogenic activity, but
resulted in the loss of EACA binding by K2 leading to the supposition that lysine
binding by K2 was unimportant for anti-angiogenic activitys. However, this loss of
EACA binding is not in agreement with the binding of a series of 0,00 —amino acids and
VEK-30 to the C169G mutant of K2,6 and the observation that the C169D/C297R
double mutant retains the chloride- and EACA-induced hydrodynamic properties of
wild-type plasminogen3. Similar conclusions regarding the irrelevance of lysine binding
to angiostatin were drawn from comparisons of lysine-binding afﬁnity and anti-
angiogenic potency7.
3.5 Dimerization

Inspection of the crystal packing indicated that dimerization between two
angiostatin/VEK-30 complex molecules occurs along a crystallographic 2-fold axis
(Figure 3-7). The same dimerization was seen in the K2/VEK-30 structure between the
two molecules of K2NEK-30 in the asymmetric unit2 (Figure 3-8). This indicates that
dimerization is not an artifact of crystal packing since there is obviously no relationship
between the crystal packing in the K2/VEK-30 complex and the angiostatin/VEK-3O
complex. Figure 3-9 shows that the Cor positions of the dimeric structure of K2/VEK-30
superimpose well with the two angiostatin/VEK-30 complexes that are related by

crystallographic 2-fold symmetry (rmsd for all atoms ~0.4 A). Dimerization results in

113

 

Figure 3-7 Dimerization of the angiostatin/VEK-30 complex. One molecule
has angiostatin colored blue and VEK-30 colored red, while the other molecule
has angiostatin colored magenta and VEK-30 colored yellow.

 

Figure 3—8 Dimerization of the K2/VEK-30 structurez. K2 is shown in
green and VEK-30 is colored magenta.

114

C
.V " K1 K1 i.
«in; 1

o 4 l
-. ,
“it" 1‘}
K2 “)9": " K2

0 ' ‘

'

VEK-30

Figure 3-9 Overlay of the K2/VEK-30 dimer onto the angiostatin/VEK-30
dimer. The angiostatin/VEK30 dimer is shown in red and the K2/VEK-30
dimer is shown in blue. The C01 positions of the dimeric structure of K2NEK-
30 superimpose well with the two angiostatin/VEK-30 complexes that are
related by crystallographic 2-fold symmetry (rmsd for all atoms ~0.4 A).

115

two molecules of (rt—helical VEK-30 packing parallel and side-by-side in the center of
the dimer, and two K2 domains located on either side of the parallel helices.
Interestingly, full-length PAM is predicted to homo-dimerize as a coiled-coil that
extends on either side of the VEK-30 regions. In fact, several homology modeling
programs predict the PAM structure based on the extended coiled-coil structure of
tropomyosing’g. The only region of PAM (between amino acids 60-310) that is not well-
ﬁtted by the coiled-coil structural prediction is the Pg-binding direct repeat region that
encompasses the VEK-30 peptide (W. Wedemeyer, unpublished results). This is
consistent with the structure of the angiostatin/VEK-30 dimer, in that the two helices do
not form a classical coiled coil, though they are parallel and stacked side-to-side.
Numerous contacts between the two molecules at the dimerization interface are
observed. As shown in Figure 3-10, water-mediated interactions occur at the
dimerization interface between angiostatin K2 residue Gln193 of the symmetry-related
molecule and VEK-30 Asn99. Another water-mediated interaction occurs at angiostatin
K204 and VEK-30 Glu94. Numerous hydrogen bonds also play a role at the
dimerization interface. Hydrogen bonds occur between VEK-30 residue Glu103 and
angiostatin K2 symmetry-related molecule residues His196 and Ala197. Another
hydrogen bond occurs between VEK-30 residue E100 and angiostatin symmetry related
molecule Arg234, the cationic site within the K2 LBS. The calculated total buried
surface area of a dimer of angiostatin/VEK30 is 1627 A2 suggesting a relatively strong
interaction. Furthermore, because a dimeric structure is exhibited by both K2NEK-30

and angiostatin/VEK-30, it is plausible that such higher order structures may exist in

116

 

Figure 3-10 Residues involved in dimerization. VEK-30 residues are
shown in green and angiostatin K2 residues are shown in yellow. The
symmetry-related molecule of angiostatin K2 is shown in cyan with its
VEK-30 residue shown in magenta. The atoms are colored by type. The
water molecules are designated as W1 and the symmetry-related waters
W2a and W2b. All residues are labeled with Pg and PAM numbering.

117

solution. To address this possibility, sedimentation equilibrium analysis was conducted
on angiostatin in the absence and presence of VEK-30 by our collaborators Frank
Castellino and co-workers. The apparent molecular weight of angiostatin (at a
concentration of 13 11M) was determined to be 27,7001100 (calculated sequence-based
weight = 29,000). In the presence of a 5-fold molar excess of VEK-30, (allowing for
virtually all angiostatin to exist in VEK-30-bound form assuming a K. of 460 nM for the
angiostatin/VEK30 interaction6), an apparent molecular weight of 28,4002800 was
obtained. These data fail to support a model of VEK-30-mediated dimerization of
angiostatin at angiostatin concentrations that are physiologically feasible based on
circulating plasma levels (ca., 2 11M) of the Pg parent"). However, this does not rule out
the possibility of a PAM-induced angiostatin (or Pg) dimer on the bacterial surface,
where high effective concentrations of both binding partners can be encountered and
where full-length PAM is strongly predicted to exist as a parallel coiled-coil dimer, as
are all proteins in the M protein family8’9.
3.6 Kringle 2 Domain Specificity

The Pg kringles have high sequence and structural homology. However, the Pg
kringles are very different with respect to afﬁnity for C-terminal lysine mimics and in
their biological functions in proliferation and migration assays. A closer examination of
the residues of K2 involved in VEK-30 interactions reveals that many of these residues
are not conserved in Pg kringle domains. Figure 3-11 shows a sequence alignment for
all ﬁve Pg kringles. For instance Gln193 is responsible for mediating interactions at the
angiostatin/VEK-30 dimerization interface. In K1 and K4, this sequence position is

occupied by Thr and Met, respectively. Ala197 of K2 is not conserved in Pg kringle

118

domains and also plays a role at the dimerization interface (K1, K4, and K5 all have Arg
residues while K3 has a Thr). K2 residue Lys204 is also not conserved and is involved
in mediating interactions at the dimerization interface, as well as directly interacting
with VEK-30 through an exosite region. Focusing on residues within the LBS, Y200 is
not conserved, where K1 and K5 have Phe and K3 and K4 have Arg and Lys,
respectively. Tyr200 of K2 makes numerous interactions with VEK-30. Speciﬁcally,
the hydroxyl group of Tyr200 makes a hydrogen bond with ArglOl-NHI. Finally,
Arg220 makes contacts with VEK-30, and this site is also not conserved in Pg kringle
domains. Speciﬁcally, Arg220 forms a salt bridge with Asp219 in the angiostatin
structure (Figure l-9). Because of the interaction with Arg220, Asp219 is ﬂipped out of
the LBS and is incapable of interacting with the C-terminal group of EACA, possibly
explaining the poor EACA binding afﬁnity of K27‘H'l4. However, VEK-30 abrogates
the salt bridge interaction between Arg220 and Asp219 so that Asp219 ﬂips into the
LBS. This rearrangement recapitulates the canonical LBS architecture, permitting
extensive LBSNEK-30 interactions (Figure 3-12). Mouse Pg has substantially lower
afﬁnity for PAM relative to human Pg, although the two domains are 86% identical in
sequence. Arg220 is the only residue that both interacts, either directly or indirectly,
with VEK-30 and is not identically conserved in mouse Pg. This strongly implicates
Arg220 as a residue that is both important for binding and critical to the species-
speciﬁcity of PAM. In summary, many of the K2 residues that interact with VEK-30
are not conserved among Pg-Kl, -K3, -K4 and -K5, likely explaining why only human
Pg-K2 has afﬁnity for VEK-30. The sequence analysis results identify potential targets

for further mutagenesis studies. For example, it would be interesting to learn whether

119

166 IQ damn ECQDMBQSP Harris-YIPSK

K3 am GWAVTVSG-I Tar-um? HITE—R’I'FBJ

(358) K4 K639115911? I-RI-O—KTIEN
(462) K5 04mm G<RATIVIGT PCQJIWQEP I-RI-ISIFI'PEI’
41 — 60 — - 80

(123) K1 mm W RYDKIIIZLEC
(205) IQ mm mm— Han-Tm niELcnIPnc
E29 K3 moms»! mumm- PlvornNsQI maps:
39 K4 mm camc— HllanDPsv mm:

— Residues in vicinity of LBS
Magenta — Residues interacting with VEK-30

Figure 3-11 A sequence alignment of all ﬁve Pg kringle domains. Yellow
highlighted residues are conserved.

mutating K1 residue N138 to R (same R220 in K2) and determine if this mutated K1
binds VEK-30 or mutate the mouse K2 to contain R220 to see if it now binds VEK-30.
These studies would help characterize kringle domain Speciﬁcity and identify why only

plasminogen K2 binds VEK-30. These mutagenesis studies may also help to identify

120

L94

D91

0219 ‘

 

E221

 

Figure 3-12 An overlay of angiostatin K2 onto angiostatin/VEK30 K2.
Residues from angiostatin alone are colored magenta, residues from the
AngiostatinNEK-3O complex are colored green, and VEK-30 residues are
colored yellow. The atoms are colored by type. All residues are labeled
with Pg and PAM numbering. D219 is ﬂipped out of the LBS, resulting
in a tight salt bridge contact with R220.

121

other potential kringle domain ligands. Ligands for plasminogen K1 and K3 have not
been identiﬁed.
3.7 Further Studies of VEK-3O

As previously mentioned in Chapter I, a region of PAM, spanning amino acids
91 - 116, contains two direct repeat sequences and is responsible for Pg binding by
PAM. VEK-30, an (rt-helical peptide derived from residues 85-113 of PAM and
containing the ﬁrst and most of the second direct repeat, possesses a high afﬁnity
binding site for Pg-K2 (K. = 460 nM),2"5"7 even though it does not contain a C-terminal
lysine residue. This led us to explore whether a peptide of PAM containing two full
direct repeats (VEK-32L) might bind two molecules of K2 since it appears that PAM
may have two distinct Pg binding regions from the analysis of the PAM sequence.
Figure 3-13 shows both primary sequences of VEK-30 and the wild-type sequence
further named VEK-32L. A model was made showing that binding of two K2 domains
to a longer peptide of PAM was possible (Figure 3-14). However, Figure 3-15 shows
that two angiostatin K2 binding to a longer peptide of PAM would disrupt the dimer
seen in the angiostatin/VEK-30 and mK2/VEK-30 crystal structures. Gel ﬁltration and
sedimentation velocity experiments using analytical ultracentrifugation were performed

to determine whether this postulate is true.

85 95 105 114
VEK—30 VEKLTADAELQRLKNERHEEAELERLKSEY

VEK-SZL VEKLTADAELQRLKNERHEEAELERLKSERHE
E

 

Figure 3-13 The primary sequences of VEK-30 (top) and VEK-32L
(bottom). They are labeled with PAM numbering above. The green
box indicates the direct repeat sequence.

122

 

Figure 3-14 A model showing that two angiostatin K2s can bind to a longer
PAM peptide. The ﬁgure was made by overlaying the direct repeat sequences
from the angiostatin/VEK-30 structure. The an giostatin/VEK-30 structure is
colored green and the second molecule magenta.

123

K149" K1

K‘~:ff{ ‘2‘): K1 '63:}
‘11 r . ‘
any;

”1‘

K2 10'4“;
1'
00)

Figure 3-15 A model showing that two angiostatin K2s can bind to a longer
PAM peptide but disrupts the dimer. The ﬁgure was made by overlaying the
direct repeat sequences from the angiostatin/VEK-30 structure. The
angiostatin/VEK-30 Structure is colored green and the dimer colored cyan
whereas the second molecule magenta.

’ K2

VEK-30

124

Gel filtration was ﬁrst performed to determine whether two molecules of
angiostatin interact with a PAM peptide containing both direct repeats. Gel ﬁltration
analysis was performed on unbound angiostatin (~29 kDa) and on the angiostatin/VEK-
32L (~33 kDa) complex using a Superdex 200 10/300 GL column. Different sample
volumes as well as varying concentrations were tried. The results showed that only one
molecule of angiostatin was bound to the VEK-32L peptide (Figure 3-16). However, it
was unclear whether the complex dissociated over the column since the VEK-32L
peptide cannot be seen using SDS-PAGE.

Sedimentation velocity experiments were then performed using a Beckman XL-I
analytical ultracentrifuge Operated in absorbance mode (280 nm) at 20°C. Samples were
rotated at speeds of 50,000 rpm and 45,000 rpm. For experiments conducted on
unbound angiostatin, angiostatin dissolved in saline buffer (0.15 M NaCl and 100 mM
Tris, pH 7.5) at concentrations of 0.5 mM and 1 mM were used. For experiments
conducted on an giostatin/VEK-32L, a ﬁve molar excess of VEK-32L was used in the
saline buffer previously used with unbound angiostatin. The data was analyzed through
programs designed by W. Wedemeyer. The results showed that only one molecule of
angiostatin binds to VEK-32L. However, a binding constant of VEK-32L for
angiostatin was not determined. Because of this, it was unclear whether angiostatin was

in fact binding to the VEK-32L peptide.

125

 

:i’
.1;
{3

 

 

 

 

 

 

 

 

 

Figure 3-16 Gel ﬁltration chromatogram results from Superdex 200
10/300 GL column. A) The dotted line is Biorad molecular weight
standards (670 kDa thyroglobulin, 158 kDa bovine gamma-globulin, 44
kDa chicken ovalbumin, 17 kDa equine myoglobin, and 1.3 kDa Vitamin
B 12). B) AngiostatinNEK-32L complex approximately 33 kDa. C)
Unbound angiostatin approximately 30 kDa. All experiments were
performed at a ﬂow rate of 1 mIJmin and the fractions were 5 mL.

126

3.8 Angiostatin Binding to Protein Domains
It has been determined that kringle domains not only bind six carbon zwitterions

-2
2’18 0. The structures of

such as C-terminal lysines and EACA, but also protein domains
the VEK-30 bound forms of K2 and angiostatin presented a model for protein binding at
the surface of bacteria. The VEK-30 peptide has an arginine and a glutamate separated
by almost one turn of the helix. This resembles a C-terminal lysine with a positive
(R101) and a negative (E104) end that interacts with the K2 anionic and cationic
residues in the LBS. This arrangement of residues in VEK-30 has been called a
“pseudo-lysine.” Known angiostatin ligands such as 01.03 integrin and F(1)-F(0) ATP
synthase contain similar “pseudo-lysine” moieties as seen in VEK-30 (Figure 1-13 and
1-15, respectively). This led us to explore other possible molecules relevant to
angiostatin’s mode of action. The X-ray structure of the angiogenesis inhibitor
endostatin reveals that it contains an a-helix with the RGAD sequence”. R158 and
D161 appear to form a pseudo lysyl site similar to the “pseudo-lysine” within the VEK-
30 peptide (Figure 3-18). Figure 3-17 shows endostatin modeled in the angiostatin
cavity between K2 and K3. Endostatin ﬁlls the cavity with very few collisions.
Moreover, inspection of the angiostatin K3 LBS shows that endostatin’s E272 may
possibly interact with the K3 residues R290 and R324 (Figure 3-18). This validates the
previously suggested observation that K3 is suited to bind short carboxylate ligands.
Although there is no biochemical data indicating that an giostatin binds endostatin, there

have been observations of an increase in tumor reduction when both inhibitors are given

. . . . . 7
1n comblnatlon wrth cancer patrentsz“.

127

Mameros and Olsen proposed a mechanism to explain the role of endostatin as
an angiogenesis inhibitor”. This mechanism is based on the binding afﬁnity of the
endostatin domain of collagen XVIII. Collagen XVIII is involved in the activation of
cell migration by interactions with extracellular components through its endostatin
domain. It is proposed that endostatin binding competes for the binding of matrix
components, inhibiting endothelial cell migration and ultimately angiogenesis. We
propose that in the same way, an giostatin interacts with the endostatin domain of
collagen XVIII by the interactions previously described thus providing an inhibitory

mechanism of angiostatin.

128

Figure 3—17 Endostatin modeled into the angiostatin K2-K3 cavity.
This was done by overlaying the helices of endostatin and VEK-30.
An giostatin is colored blue and endostatin is colored orange. R158
and D161 of endostatin are colored green with atom coloring.

129

 

Figure 3-18 The residues possibly involved in endostatin/angiostatin
binding are shown. Endostatin was modeled into the angiostatin K2-
K3 cavity by overlaying the helices of endostatin and VEK-30.
Angiostatin is colored green and endostatin is colored lavender. The
residues are colored by atom.

130

3.9 Discussion

The angiostatin/VEK-30 structure is the only interaction between a multiple
kringle-containing protein and a ligand to be characterized, and is the sole model for
angiostatin complexation with its cognate protein-binding partners. Interestingly, other
known angiostatin protein ligands, 01.53 integrin and F(l)-F(0) ATP synthase, both
contain a possible pseudo-lysine arrangement similar to that of VEK-307 (Figure 1-13
and 1-15, respectively). However, data deﬁning the interface between angiostatin and
these targets has not yet been reported.

The complex between VEK-30 and angiostatin suggests a possible mechanism
for Pg activation during group A. streptococci infection. SK is secreted by group A
streptococci and activates human Pg. It does so by binding to the catalytic domain of
Pg, causing a conformational change in Pg within the complex that results in formation
of an active site in the bound Pg (SK-Pg’). This SK-Pg’ complex activates other
molecules of Pg or is converted to SK-Pm, another potent Pg activator. Previous studies
have Shown that both SK and PAM contribute to virulence during group A streptococci
impetigo“. Speciﬁcally, “molecular cooperation” between SK and PAM occurs,
resulting in the bacterial acquisition of the host Pm so that the pathogen is able to invade
host tissues”. Previous experiments suggest that streptokinase activates PAM-bound Pg
and also preferentially binds the open form of Pg26. Our structure of the
angiostatin/VEK-30 complex suggests that the PAM region might induce the open
conformation of Pg on the cell surface by altering the relative orientation of the Pg
kringle domains and by steric interference between its extended helical structure and Pg.

Although VEK-30-mediated dimerization was not observed under physiologically-

131

relevant solution conditions, this does not rule out dimerization of Pg by full-length
PAM on the bacterial cell surface. This dimerization may also promote a Pg
conformation that is more amenable to SK binding and would also bring two molecules
of Pg together on the cell surface. SK binding to one Pg could then result in proximity-

accelerated activation of the second.

132

3.8

Literature Cited

Hayward, S. & Berendsen, H.J.C. Systematic Analysis of Domain Motions in
Proteins from Conformational Change; New Results on Citrate Synthase and T4
Lysozyme. Proteins, Structure, Function, and Genetics 30, 144(1998).

Rios-Steiner, J .L., Schenone, M., Mochalkin, 1., Tulinsky, A., and Castellino, F.
J. Structure and binding determinants of the recombinant kringle-2 domain of

human plasminogen to an internal peptide from a group A Streptococcal surface
protein. J Mol Biol 308, 705-19 (2001).

McCance, S.G. & Castellino, P.J. Contributions of individual kringle domains
toward maintenance of the chloride-induced tight conformation of human
glutarrric acid-1 plasrrrinogen. Biochemistry 34, 9581-6 (1995).

Cao, Y., Ji, R. W., Davidson, D., Schaller, J., Marti, D., Sohndel, S., McCance,
S. G., O'Reilly, M. S., Llinas, M., and Folkman, J. Kringle domains of human

angiostatin. Characterization of the anti-proliferative activity on endothelial cells.
J Biol Chem 271, 29461-7 (1996).

Lee, H., Kim, H. K., Lee, J. H., You, W. K., Chung, S. 1., Chang, S. 1., Park, M.
H., Hong, Y. K., and Joe, Y. A. Disruption of interkringle disulﬁde bond of
plasminogen kringle 1-3 changes the lysine binding capability of kringle 2, but
not its antiangiogenic activity. Arch Biochem Biophys 375, 359-63 (2000).

Nilsen, S.L., Prorok, M. & Castellino, F.J. Enhancement through mutagenesis of
the binding of the isolated kringle 2 domain of human plasminogen to omega-
amino acid ligands and to an internal sequence of a Streptococcal surface
protein. J Biol Chem 274, 22380-6 (1999).

Geiger, J.H. & Cnudde, S.E. What the structure of angiostatin may tell us about
its mechanism of action. J Thromb Haemost 2, 23-34 (2004).

Jones, D. TI-IREADER: Protein Sequence Threading by Double Dynamic
Programming. in Computational Methods in Molecular Biology, Vol. Chapter 13
(eds. Salzberg, S., Searls, D. & Kasif, S.) (Elsevier Science, 1998).

Whitby, F. & Phillips Jr., G. Crystal structure of tropomyosin at 7 Angstroms
resolution. Proteins 38, 49-59 (2000).

133

10.

ll.

12.

l3.

14.

15.

16.

17.

18.

19.

20.

Collen, D. & Lijnen, H.R. Basic and clinical aspects of ﬁbrinolysis and
thrombolysis. Blood 78, 3114-3124 ( 1991).

Winn, E.S., Hu, S. P., Hochschwender, S. M., and Laursen, R. A. Studies on the
lysine-binding sites of human plasrrrinogen. The effect of ligand structure on the
binding of lysine analogs to plasminogen. Eur J Biochem 104, 579-86 (1980).

Urano, T., Chibber, B.A.K. & Castellino, R]. The Reciprocal Effects of Epsilon-
Aminohexanoic Acid and Chloride-Ion on the Activation of Human
(Glul]Plasminogen by Human Urokinase. J Biol Chem 84, 4031-4034 (1987).

Urano, T., Deserrano, V. S., Chibber, B. A. K., and Castellino, F. J. The Control
of the Urokinase-Catalyzed Activation of Human Glutamic-Acid l-Plasminogen
by Positive and Negative Effectors. J Biol Chem 262, 15959-15964 (1987).

Abad, M.C. et al. The X-ray crystallographic structure of the angiogenesis
inhibitor angiostatin. J Mol Biol 318, 1009-17 (2002).

Wistedt, A.C., Ringdahl, U., Muller-Ester], W., and Sjobring, U. Identiﬁcation of
a plasminogen-binding motif in PAM, a bacterial surface protein. Mol Microbiol
18, 569-78 (1995).

Wistedt, A.C., Kotarsky, H., Marti, D., Ringdahl, U., Castellino, F. J., Schaller,
J ., and Sjobring, U. Kringle 2 mediates high afﬁnity binding of plasminogen to
an internal sequence in streptococcal surface protein PAM. J Biol Chem 273,
24420-4 (1998).

Berge, A. & Sjobring, U. PAM, a Novel Plasrrrinogen-binding Protein from
Streptococcus pyogenes. J Biol Chem 268, 25417-25424 (1993).

Moser, T.L. et al. Angiostatin binds ATP synthase on the surface of human
endothelial cells. Proc Natl Acad Sci U S A 96, 2811-6 (1999).

Troyanovsky, B., Levchenko, T., Mansson, G., Matvijenko, O. & Holmgren, L.
Angiomotin: an angiostatin binding protein that regulates endothelial cell
migration and tube formation. J Cell Biol 152, 1247-54 (2001).

Tarui, T., Miles, L.A. & Takada, Y. Speciﬁc interaction of angiostatin with
integrin alpha(v)beta(3) in endothelial cells. J Biol Chem 276, 39562-8 (2001).

134

23.

24.

25.

26.

Hohenester, E., Sasaki, T., Olsen, B.R. & Timpl, R. Crystal structure of the
angiogenesis inhibitor endostatin at 1.5A resolution. EMBO J 17, 1656-1664
(1998).

Yokoyama, Y., Dhanabal, M., Grifﬁoen, A.W., Sukhatme, V.P. &
Ramakrishnan, V. Synergy between angiostatin and endostatin: inhibition of
ovarian cancer growth. Cancer Res 60, 2190-2196 (2000).

Mameros, A.G. & Olsen, B.R. Matrix Biol 20, 337-345 (2001).

Svensson, M.D., Sjobring, U., Luo, F. & Bessen, D.E. Roles of the plasrrrinogen
activator streptokinase and the plasminogen-associated M protein in an
experimental model for streptococcal impetigo. Microbiology 148, 3933-3945
(2002).

Ringdahl, U. et al. Molecular co-operation between protein PAM and
streptokinase for plasmin acquisition by Streptococcus pyogenes. J Biol Chem
272, 6424-6230 (1998).

Boxrud, RD, and PE. Bock. Streptokinase binds preferentially to the extended
conformation of plasminogen through lysine binding site and catalytic domain
interactions. Biochemistry 39, 13974-13981 (2000).

135

Chapter IV
The Three Dimensional Structures of Calcium-Bound con-G

and con-T[K77], and Cd2*lMgz*/con-T[K7v]

The physiological importance of the metal bound structures of con-G and
con-T[K77] has been discussed in Chapter I. In order to get a better understanding
of conantokin helix self-assembly, calcium-bound con-G and con-T[K77] (Cnudde,
SE et al. 2007. JAm Chem Soc. 129(6): 1586-93) as well as Cd2+/Mg2+/con-T[K7v]
were crystallized and their structures determined. Structure determination including
the crystallization conditions, data collection and reﬁnement statistics are discussed
in Chapter II. The structural characterization of these complexes is the ﬁrst example
of helix stabilization and self-association that is mediated by metal ion coordination
and electrostatics with absolutely no component due to the hydrophobic effect. The
details are discussed herein.

4.1 Overall Structure of Ca2+/con-G and Ca2+/con-T[K7y]

The X-ray crystal structures of the Ca2+/con-G and Ca2+/con-T[K77] have
been determined at high resolution. Their overall structures are shown in Figure 2-
13 and 2-15, respectively. The con-G a-helix is approximately 26 A in length. A
total of three Ca2+, one Cl", and 68 water molecules were included in the model.
Alternative side-chain conformations exist for con-G residues, Leu5, Gln9, and
Serl6. The con-T[K77] or-helix is approximately 30 A in length with a total of four
Ca2+ and 142 water molecules. No alternative Side-chain conformations exist in the

Ca2+/con-T[K77] structure.

136

4.2 The Structure of Ca2+/con-G

In the presence of Ca2+, con-G forms a helical antiparallel dimeric structure
(Figure 2-13). Four of the Gla residues (3,7,10,14) are present on one face of the
helix, speciﬁcally the Ca2+ coordination interface, while Gla4 resides on the opposite
side of the helix and does not coordinate Ca2+. A crystallographic two-fold axis
passes through the central Ca2+ ion, Ca3, relating the two helices to form a dimer.
Therefore, the con-G dimer contains a total of 5 Ca2+, two associated with each of
the crystallographically equivalent helices, and one shared between the two that is
bisected by a crystallographic two-fold axis. The two helices are further referred to
as Helix A and Helix B, for the unique and crystallographically equivalent helix,
respectively. Ca2+ ions are further designated as CalA, Ca2A, CalB, Ca2B and Ca3,
for those associated with Helix A, Helix B, and shared, respectively. CalA and
CalB, and Ca2A and Ca2B, are related by crystallographic symmetry, as are Helix
A and Helix B. CalA is coordinated by both carboxylates of two Gla residues, Gla3
and Gla7, accounting for a total of four ligands to CalA (Figure 4-1 top left).
Similarly, Ca2A is chelated by Gla10 and Gla14, with all four carboxylates of the
two Gla residues acting as ligands (Figure 4-1 top right). Both CalA and Ca2A have
octahedral geometry with the two carboxylates of a single Gla residue occupying
coordination sites separated by about 90°. The central Ca2+ ion, Ca3, is also
coordinated by two Gla residues, Gla7 and Gla10, but each Gla residue donates only
one carboxylate (Figure 4-1 bottom). Therefore, Helix A donates a total of only two
ligands to Ca3. There is an identical arrangement of chelated Ca2+ ions on Helix B,

as required by the crystallographic symmetry relating the two helices. It is clear that

137

 

Figure 4-1 Ca2+ coordination in con-G. (Top left) CalA (magenta) coordinates with
73 and 77 and waters W30 and W38 (blue). (Top right) Ca2B (magenta) coordinates
with 73 and Q6 of helix A and 714 and 710 of the crystallographically-related helix B
forming the dimer. Q6 is displayed in blue. (Bottom) Ca3 (gray) lies on the
crystallographic twofold and coordinates with 77 and 710 of each helix as well as
waters W34 and W34b (blue), a symmetry-related water molecule. The symmetry-
related helix forming the dimer is colored yellow. In all cases, the peptide side-chain
atoms are colored by atom type, except as otherwise indicated.

138

the helical conformation is stabilized by the metal-mediated interactions of Gla
residues located one helical turn apart. Metal chelation enforces the positioning of
all four Gla residues and, by nature of the Ca3 chelation of Gla7 and Gla10, enforces
the location of all four Gla residues on a single helical face. A similar chelation
arrangement is likely for other metal ions, such as Mg2+, which also leads to a helical
conformation in con-G. Absent metal ions, the electrostatic repulsion caused by the
concentration of negative charges from the Gla residues on a single helical face
would destabilize a helical conformation, consistent with all biophysical
measurements that Show that con-G is not helical in the absence of metal ions 1'5.
The dimerization interface between Helix A and Helix B is deﬁned almost
exclusively by intermolecular Ca2+ coordination between the two helices. There are
no direct contacts between residues of the two helices, although there are some
water-mediated interactions (vide infra). Three of the 5 Ca2+, Ca2A, Ca2B, and
Ca3, are involved in intermolecular coordination. Helix B donates two ligands to
Ca2A, the Side-chain carbonyl oxygen of Gln6 and a single carboxylate group of
Gla3 (Figure 4-1 top right). This creates an octahedral coordination geometry about
Gla3, with two of the six ligands originating from Helix B. Since Ca3 resides on a
crystallographic two-fold axis that relates Helix A with Helix B, the coordination of
Ca3 to Helix B is identical to that of Ca3 to Helix A. This results in four
carboxylate ligands, one each from Gla7 and Gla10 from each of the two helices
acting as ligands for Ca3. The other two coordination sites of Ca3 are occupied by

water molecules to complete the octahedral geometry. The other two Ca2+, CalA

and CalB, do not participate in interhelical contacts (Figure 4-1 top left), though

139

mutational studies indicate that their ligands, Gla3 and Gla7, are both important for
Ca2+-mediated helix dimerization 6. This is most likely due to the induction of a
Stable monomeric helix upon occupation of the CalA/CalB site formed by Gla3 and
Gla7.

4.3 The Structure of Ca2*lcon-T[K77]

The overall structure of Ca2+/con-T[K77] is shown in Figure 2-15. Unlike
con-G, there is no crystallographic two-fold symmetry relating the helices of the
con-T[K77] dimer, even though the two con-T[K77] helices are virtually identical
(overlay in Figure 4-2). Con-G contained one molecule in the asymmetric unit
whereas the con-T variant con-T[K77] contains two molecules in the asymmetric
unit. Also, the Ca2+ coordination within each con-T[K77] helix is also identical.
There are a total of four Ca2+ bound to the con-T[K77] helix, CalA and Ca2A, which
are predominantly associated with Helix A, and CalB and Ca2B, which are
associated with Helix B of the dimer. As occurs with CalA and Helix A of con-G,
CalA coordinates Helix A of con-T[K77], via a tetravalent interaction between all of
the carboxylate residues of Gla3 and Gla7. The analogous scheme exists for Ca2A,
which chelates all four carboxylates of Gla10 and Gla14 of Helix A, resulting in an
essentially identical interaction to that seen for Ca2A of con-G. CalB and Ca2B of
con-T[K77] are correspondingly identical to CalB and Ca2B of con-G. Taken
together, it is clear that when a single helix is viewed in isolation, the con-G and

con-T[K77] structures are very similar, except that Ca3 appears to be missing in the

con-T[K77] structure (Figure 4-3 top and 4-3 bottom).

140

N-terminal

  
  

C-terminal

Figure 4—2 Comparison of the con-T[K77] helices. Overlay of the two con-
T[K77] helices show their identity.

141

However, the dimerization interface of con-T[K77] is not similar to that of con-G.
Given the structural similarity between con-G and con-T [K77], it was surprising to
see a difference in the dimer interface between the two structures. While both
peptides yield antiparallel helical dimers, and both are largely stabilized by Ca2+-
mediated chelation, the relative orientation of the helices as well as the Ca2+
coordination in the interface, is radically different in the con-T variant (overlays in
Figure 4-3).

The two helices of con-T[K77] are almost parallel and all four Ca2+ are
involved in the dimerization interface of con-T[K77]. As seen in Figure 4-4 a,
CalA, chelated by Gla3 and Gla7 of Helix A, coordinates one carboxylate of Gla14
of Helix B, while Ca2A, chelated by Gla10 and Gla14 of Helix A, coordinates one
carboxylate each of Gla7 and GlalO of Helix B (Figure 4-4 b). The coordination
sphere of CalB is identical to that of CalA (Figure 4-4 c), and the coordination
sphere of Ca2B (Figure 4-4 a) is identical to that of Ca2A, except that the residues of
Helix A are transposed for the residues of Helix B. The result of this chelation
pattern is that the two helices are staggered by approximately one helical turn along
the colinear helix axes and the Gla residues account for all of the Ca2+ chelation in
con-T[K77]. In addition to the Ca2+ interactions, a short salt bridge exists between
Arg13 of Helix A and Gla3 of Helix B and an identical salt bridge is formed between
Arg13 of Helix B and Gla3 of Helix A to complete the symmetric interaction (Figure

44 d).

142

 

Figure 4-3 Comparison of the con-T[K77] and con-G helices. (Top) Overlay of the
con-T[K77] (magenta) and con-G(green) helices demonstrates that these helices are
nearly identical to each other. Blue spheres represent Ca2+ ions of con-G and green
spheres designate Ca2+ ions of con-T[K77]. (Bottom) The boxed area is magniﬁed
and includes waters (red spheres). In all cases, the peptide side-chain hetero-atoms are
colored by atom type.

143

 

Figure 4-4 Ca2+ coordination in the con-T[K77] dimer. (a) CalA (magenta)
coordinates with )8 and 77 of helix A and y14 of helix B, as well as with with one water
molecule, W71 (red sphere). Ca2B (magenta) coordinates with 77 and y10 of helix A
and 714 and y10 of helix B, along with water molecule W79 (red sphere). (b) Ca2A
(magenta) coordinates with y10 and y14 of helix A and 7’10 and y7 of helix B. Ca2A
also coordinates with two water molecules, W37 and W68 (red spheres). (c) CalB
(magenta) coordinates with y14 of helix A and V7 and 73 of helix B. CalB (magenta)
also coordinates with one water molecule, W44 (red sphere). (d) The salt bridge
interactions in con-T[K77]. All distances are shown in Angstroms. In each case, the
peptide side-chain hetero-atoms are colored by atom type. Helix A is always displayed
on the left whereas Helix B is displayed on the right.

A further stabilization of the con-T[K77] dimeric structure involves a spine
of water molecules that runs between the Ca2+, (Figure 4-5), all of which are

chelated by Ca2+. This continuous spine of hydration is not seen in the con—G

structure.
.2 62
232. 5.5.8 29W442 ZOEYMSE‘?” {0:17 A ;Y7 Helix A
. ,.....8. In ‘239 :,:2;44 22:.W75::382B 3.14: CalA

   

7:243". ~'W79
2.49 -'.""'."".w71

 
 

73
Helix B
710 714

Figure 4-5 The con-T[K77] dimer interface is signiﬁcantly different from con-G. A water
network exists at the Ca2+/dimer interface of Ca2+/con-T[K77]. All distances are shown in
Angstroms. The Ca2+ ions are shown as blue spheres and the waters are shown as red
spheres. The top four Gla residues belong to Helix A and the bottom four Gla residues
belong to Helix B. The peptide side-chain hetero-atoms are colored by atom.

145

The fundamental differences between the dimer interface in con-G versus con-
T[K77] are illustrated by the schematic "wiring diagram" presented in Figure 4-6.
First, there is a clear correspondence between CalA, CalB, Ca2A and Ca2B in all of
the structures, since they have identical chelation from one of the two helices. For
example, all Cal-type ions are chelated to Gla3 and Gla7 of a single helix by both of
their carboxylate moieties, while Ca2-type ions are chelated to GlalO and Gla14 of a
single helix by both of their carboxylate moieties. In addition, all four helices
coordinate a single Ca2+ via G137 and GlalO in bidentate fashion, using only one
carboxylate from each Gla. Therefore, if any of the four helices are viewed in
isolation, the Ca2+ coordination appears to be identical. The dramatic difference
occurs because each Ca2+ interacts differently with the adjacent helix in the two
structures. For example, in con-T[K77] all Cal-type ions chelate with a single
carboxylate of Glal4 on the partner helix, thus creating a symmetric interaction. On
the other hand, in con-G Cal-type ions do not chelate with their partner helix.
Furthermore, the Ca2-type ions of con-T [K77] chelate Gla7 and GlalO of their
partner helix, while in con-G, the Ca2-type ions chelate the side-chain of Gln6, while
Ca3 coordinates Gla7 and Gla10 residues from both helices. The result of this
coordination difference is to cause the con-G helices to be twisted out of plane to
accommodate the octahedral coordination about Ca3, while the helices in con-
T[K77] are parallel, but are shifted along the helix axis by one turn to accommodate

the chelation of both Cal and Ca2 (Figure 4-7).

146

   

con-T[K77]

Figure 4-6 The con-T[K77] dimer interface is signiﬁcantly different from con-G.
The fundamental differences between the Ca2+/con-G structure and Ca2+/con-
T[K77] structure are shown in a schematic wiring diagram. Each dotted line
represents a coordination between one Gla carboxylate and a Ca2+. Helix A is
colored green and Helix B is colored yellow. The black spheres are Ca“.

147

    

Helix A C 2+ Helix B
3

Interface

Figure 4-7 The con-T[K77] dimer interface is signiﬁcantly different from con-
G dimer. Overlay of the con-G and con-T[K77] structures showing difference
in helix alignment. The con-G dimer is colored green and its Ca2+ ions colored
red. The con-T[K77] dimer is colored magenta and its Ca2+ ions colored blue.
In all cases, peptide side-chain hetero-atoms are colored by atom.

148

4.4 Discussion

While both con-G and the con-T variant (con-T[K77]) provide an identical
Gla pattern that allows dimerization in the presence of Ca2+ in an antiparallel
orientation and would therefore be expected to yield similar Ca2+-bridged dimeric
interfaces, this was not the case. There is a substantial difference in dimerization
and Ca2+ afﬁnity between these two peptides. In fact, con-T[K77] exhibits a 10-fold
higher association constant for dimerization and 10-fold higher afﬁnity for Ca2+ than
con-G 7. This is consistent with the present structures as there are four Ca2+ in the
dimer interface of con-T[K77] and only three in con-G. There are also more
bridging interactions in con-T[K77], with six cross-helix ligand interactions
compared to only four in con-G (these are Ca2+-mediated interactions where the
metal ion is assumed to be bound to one of the helices and coordination sites to the
partner helix are counted). There are also two tight salt-bridge interactions
(involving Arg13 and Gla3) between helices in con-T[K77] that are not seen in con-
G. A spine of water molecules exists that span all four Ca2+ in con-T[K77]. Each of
the six water molecules that constitute this spine are coordinated by one of the four
Ca2+ ions. This spine of hydration is absent in the con-G structure, at least partly due
to the fact that the Ca2+ ions in con-G have fewer coordinated water molecules.
Taken together, these observations are consistent with the higher binding and
dimerization afﬁnity of con-T[K77] compared to that of con-G.

Our structure is also consistent with the fact that native con-T binds Ca2+

without dimerization 6. The absence of Gla7 destabilizes, if not eliminates, the CalA

149

and CalB binding sites, leading to loss of two helix bridging interactions to Gla14.
Gla7 is also involved in helix bridging interactions to CalA and Ca2B. Therefore,
elimination of this residue leads to probable loss of three of the helix bridging
interactions seen in the structure. It is interesting to note that metal-free native con-
T has considerable helical content as opposed to both con-G and con-T[K77]7. This
is probably due to two factors. First, electrostatic repulsions between the four Gla
residues on one side of a helix almost certainly lead to destabilization. Gla7
contributes signiﬁcantly to this destabilization because it resides in the center of the
helix. Replacement of Gla7 with a positively charged side-chain (e.g., Lys), would
signiﬁcantly reduce this electrostatic repulsion and may possibly add electrostatic
stabilization by forming salt bridging interactions with both Gla3 and Gla10.
Consistent with this is the fact that the con-G[77K] mutant is signiﬁcantly more
helical in the absence of metal ions than is native con-G 8. The con-T peptide is also
four residues longer than con-G, and all of these residues are helical in the crystal
structure. It is well acknowledged that longer helices are more stable than shorter
helices.

Less clear is what causes the different dimerization interface of con-T[K77]
relative to con-G. Although eight amino acids differ between the two peptides, all of
the residues that are involved in Ca2+ binding and dimerization are conserved. The
overlay of the con-G and con-T[K77] helices shows that most of the residues in the
interface have identical conformations in the two structures. The most signiﬁcant
difference is in the conformation of Gln6, which is ﬂipped into the interface and

chelates Ca2B (or Ca2A for the B helix) in con-G, but is ﬂipped out of the interface

150

in con-T[K77] and interacts with Tyr5, a residue that is not conserved in the two
peptides. Hence, the conformation of Gln6 appears to be critical to both dimer
interfaces. As one of the cross-helical Ca2+ ligands in the con-G interface, the
ﬂipped-out Gln6 conformation would compromise this mode of Ca2+ coordination.
On the other hand, in con-T[K77], the ﬂipped-in conformation of Gln6 would
sterically clash with Arg13 from the other helix and prevent the formation of the salt
bridge between Arg13 and Gla3 in con-T[K77]. Overlaying the two helices indicates
that the Gln6 ﬂipped-out conformation is opposed in con-G both by the
conformation of Gln9, which is also not conserved in con-T, and the resulting water
structure around Gln9 and Glu2. The ﬂipped-out conformation would bring the
Gln6 side-chain oxygen atom within 3 A of the Gln9 oxygen atom and within 2 A of
two very well-ordered water molecules that form a bridge between Glu2 and Gln9,
all of which represent electrostratic and steric clashes. The side-chain oxygens of
both Gln6 and Gln9 are well-deﬁned, since the Gln6 side-chain nitrogen must
interact with Glu2, and the Gln9 nitrogen must interact with the main-chain carbonyl
of Leu5. Therefore the dimer interface is deﬁned primarily by the metal chelation of
the four critical Gla residues, Gla 3, 7, 10 and 14, but is also strongly affected by
subtle, indirect effects on the conformation of other side-chains as well.
4.5 Mutational Studies of the con-G and con-T Peptides

The systematic mutational studies of both con-G and con-T 8'9 have shed
much light on our structural results. For example, deletion of residues from the C-
terminus of either the con-T or con-G helices leads to reduced helicity. This is not

surprising given that longer helices are usually found to be more Stable than shorter

151

ones and our structure demonstrates that an end-to-end helix exists for both peptides.
Further, individual replacement of Gla residues 3, 7, 10 or 14 leads to signiﬁcantly
reduced helicity for con-G 8, consistent with our observation that chelation of Ca2+ to
these residues deﬁnes and stabilizes the helical peptide structure. Though
replacement of many of the other residues also leads to some loss of helicity, it is
particularly interesting to note that in con-G, the R13A variant manifests a nearly
complete loss of helicity in the presence of Ca2+. In the case of con-T, the R13A
replacement likewise diminishes the helix content of both Ca2+'bound and metal-free
peptide forms 8‘9. In our structure, Arg13 makes a tight salt bridge with Gla10 of the
same helix in both con-G and con-T[K77], and in the latter makes an additional salt
bridge with Gla3 of the adjoining helix. Thus, it appears that Arg13 is important for
stabilization of the dimer form, through neutralization of the considerable negative
charge within the helix-helix interface and also participates in intrapeptide
electrostatic contacts that stabilize the monomeric form of apo con-T.
4.6 The con-G and con-T Structures and NMDA Receptor Binding
Those residues found by alanine replacement to be important for NMDA
receptor binding or antagonism define one face of both the con-G and con-T helices
(Figure 2-13 and 2-15) 9. The N-termini of the peptides are especially critical for
biological activity, consistent with the fact that the N-termini of con-T and con-G are
identical through the first four residues. Gla4 is particularly important, since
mutation of this residue, even conservatively to Asp or Glu, results in no detectable
NMDA receptor activity 9. The fact that these residues are localized to one face of

the helix strongly indicates that both peptides bind the receptor in a helical

152

conformation, even though replacements that signiﬁcantly destabilize the con-G or
con-T helix do not always have a signiﬁcant effect on NMDA receptor binding (e.g.,
710A). However, for both con-G and con-T, the 73A replacement results in the
reduction of 1C5.) values by at least an order of magnitude. Our hypothesis is that
NMDA receptor binding is mostly dependent on the maintenance of the helical
conformation in the N-terminal portion of the helix. Alanine replacement of 73
results in helix fraying at the N-terminus, possibly through loss of the stabilizing
effect that the negatively charged Gla residue imparts to the helix macro-dipole at
the N-terminus thereby reducing peptide afﬁnity for the receptor. In further support
of the helical hypothesis is con-G[77P], which is incapable of helix formation and is
completely inactive at the NMDA receptor 8. On the other hand, the C-terminal part
of the helix is less critical for the interaction and therefore loss of helicity in this
region is less deleterious to NMDA receptor binding.
4.7 Comparison of the Crystal Structures to the NMR Structures
Several NMR-based structures of both metal-bound and apo forms of con-T

d 2'5‘10. Most relevant to this discussion are the

and con-G have been detennine
NMR-derived structures of Ca2+-complexed con-G 25”. While the helical
backbones of these structures are very similar to the X-ray structure determined
herein (Figure 4-8), neither of the NMR-derived structures predicts the dimerization
interface. This is consistent with our data, which indicate that there are no
interhelical proton distances closer than 5 A. This would obviate detection of the

dimeric species by 2D-NMR, since NOEs do not typically evolve from proton-

proton distances with the mixing time employed in the afore-referenced studies.

153

N-terrrrinal

112

 

C-terminal

Figure 4-8 Overlay of the NMR and X-ray crystal structures of Ca2+/con-G.
Cyan, the X-ray structure of Ca2+/con-G; green, a low energy NMR structure”;
white, Ca2+. In all cases, peptide side-chain hetero—atoms are colored by atom

154

While side-chain orientations on the Gla-rich face of the helix are similar between
the NMR and X-ray structures, profound differences are observed at the opposite
face and at the termini. These latter disparities likely reﬂect some of the
uncertainties involved in garnering NMR-based structural information in the more
ﬂexible regions of the peptide, wherein an ensemble of solution conformers must be
averaged. Furthermore, computational docking of Ca2+ into the structure of con-G 27
yields stoichiometry and placement that is different from that associated with the
crystal structure. A total of four bound Ca2+ are predicted from the NMR data
whereas three exist per monomer, in the crystal structure. While three of these Ca2+
ions form a similar coordination network in both structures, the NMR-derived
complex deviates from our structure in that it invokes Gla3 and Gla4 as ligands for a
fourth Ca2+. Also, the CaZ+-docked NMR structure does not reveal Ca2+ coordination
to the side-chain carbonyl of Gln6 of con-G.

It is difﬁcult to accurately predict metal positions in peptides and proteins
solely by using conventional NMR in tandem with modeling techniques. While
single crystal X-ray diffraction is a powerful approach for elucidating the subtle
details of metal ion binding, it is not without uncertainties caused by the structural
ramiﬁcations of crystal packing interactions, whereas dynamic solution structures
are without these constraints. Thus, both NMR and X—ray crystallography continue
to provide valuable information even though there are some differences using such

disparate approaches.

155

4.8 Conclusions From the Calcium-Bound Structures

Limited success has been obtained in the past in the design of short peptides
that adopt stable, structurally homogeneous Ot—helical conformations. Almost all
helices are stabilized by the hydrophobic effect, promoting the formation of a
hydrophobic helical face that is critical to stabilize the structure. It is also the
primary force that drives helix self-assembly as seen in coiled-coil multimers, e.g.,
the leucine zipper motif. A possible alternative to the hydrophobic effect is the use
of metal ion coordination to electrostatically stabilize the helical conformation.
However, to date, there are no examples of a peptide that is solely dependent on
metal ion chelation for both helix formation and self-assembly. Further, there were
no high resolution crystal structures of any of the metal-promoted helices, leaving
open the full nature of the metal coordination. Employing the con-G and con-
T[K77] peptides, we provide the ﬁrst example of helix stabilization and self-
assembly that is completely dependent on electrostatics and metal coordination, with
no component of the stabilization due to the hydrophobic effect. The unique
dispensation of Gla residues on one face of the helix, resulting in a regular
coordination of metal ions along this face, give rise to this effect, resulting in what
we call a “metallo-zipper motif”. In addition, the natural peptides form speciﬁc anti-
parallel helix dimers in the presence of Ca2+, and fold into monomeric helices in the
presence of Mg2+. Our structures of these peptides Show very clearly the nature of
the metal chelation that leads to helix formation and also elaborate the dimerization
interface in both peptides. The structures also demonstrate that while the metal

chelation is very similar in the two peptides, they surprisingly result in completely

156

different dimer interfaces. The dimer interface appears to be exquisitely sensitive to
amino acid changes that are not directly involved in the metal chelation that results
in dimerization. This is the ﬁrst example of a short peptide helix that is completely
stabilized by metal ion chelation. Helix formation depends only on the disposition
of the metal chelating Gla residues. The dimer interface consists exclusively of
metal coordination interactions in the case of the con-G dimer, while a salt bridge
interaction plays an important role in deﬁning the con-T[K77] interface. These
structures provide a fertile platform for the design of short peptide helices that are
comparatively insensitive to amino acid sequence in all but a few positions. The
structures also provide a much better understanding of the mechanism of metal-
dependent oligomerization.

4.9 The Overall Structure of Cd2*lMgz*lcon-T[K77]

The X-ray crystal structure of the Cd2+/Mg2+/con-T[K77] has been
determined at high resolution. The overall structure is shown in Figures 2-18 and 2-
19. The con-T[K77] or-helix is approximately 30 A in length with a total of 4 Cd“, 1
Mg2+, and 61 water molecules. Alternative side-chain conformations exist for
residues Met8 and Vall7. Electron density was also seen for the C-terminal NH2
atom.

2+Icon-G and Ca2+/con-

Much like other conantokin structures such as Ca
T[K77], Gla residues 3, 7, 10, and 14 reside on the same face of the helix. However,
three cadmiums are seen at the Gla interface. In Figure 4-9a, Cd2 is six-coordinate

chelating residues Gla3 and Gla7 of the helix as well as Gla3 of a crystallographic-

related symmetry molecule. Cd2 also coordinates with water W20. However, Cd2

157

is only 50% occupied. It is unclear as to whether this partially occupied Cd2+ is also
partially occupied by Mg2+ since magnesium does not have an anomalous signal
(Figure 2-18). Figure 4-9b shows that Cd3 is tetrahedral] y coordinated involving 74
and Y7 of the helix as well as 74 and 77 of a symmetry-related molecule. Also, Cd3
sits directly on a crystallographic two-fold axis. Unlike Cd2, Cd3 has 100%
occupancy. The Cd3 metal ion site is unique Since the previous crystal structures of
both con-G and con-T[K77] bound to Ca2+ do not have a metal ion coordinating 74
and 77. Figure 4-9c shows Cd4 coordination. Cd4 is octahedrally coordinated and
involves 710 of the helix and a symmetry-related 710. Cd4 also coordinates with two
waters W22a and W51a as well as their symmetry-related waters W22b and W51b.
Much like Cd3, Cd4 sits directly on a crystallographic two-fold. One Mg2+ is also
seen at this interface (Figure 4-10). Mgl is octahedrally coordinated involving
residues 710 and 714. Mgl makes bidentate contacts with both carboxylates of 710
and 714 as well as coordinating with waters W23 and W24. However, W23 exits in
two different conformations (W 23a and W23b) and Mgl makes contact with W23a
only since W23b is too far (2.53 A). Unlike the other Cd2+, Mgl does not coordinate
with a crystallographic symmetry-related molecule. The primary role of Mgl
involves Gla residue chelation of a single helix. On the opposite face of the helix,
Cdl coordinates with the helix and a symmetry-related molecule (Figure 4-11). Cdl
coordinates with both the nitrogen and oxygen atoms of G1 of the helix. Cdl also

coordinates with E16 and K19 of a crystallographic related molecule. Two waters

are also involved at this site. W18 and W19 both coordinate with Cdl. The Cd2+

158

 

Figure 4-9 Cadmium coordination at the calcium interface. (a) Cadmium Cd2 is six-
coordinate. The overall structure of Cd2+/Mg2+-con-T[K77] is shown in green. Cd2
is a blue sphere and the water W20 is a red sphere. A crystallographic-related
molecule is colored yellow. (b) Cadmium coordination at Cd3 is tetrahedrally
coordinated. Cd3 sits directly on a crystallographic two-fold axis and coordinates
both 74 and 77 of the structure and its s metry-related molecule. Distances to each
oxygen atom range from about 2.3-2.6 . The structure is shown in green and the
symmetry—related molecule is colored yellow. Cd3 is shown as a blue sphere. (c)
Cadmium coordination at Cd4 is nr‘ ’ J ally W.,. J' ‘ 4. Cd4 coordinates with
710 of the structure (green) and a crystallography-related molecule (magenta). Cd4
also coordinates with waters W22a and W51a and the symmetry-related waters W22b
and W51b. Cd4 is partially colored as a blue sphere and all waters are colored as red
spheres. All distances are measured in Angstroms. The side-chains are colored by
atom.

 

 

159

Figure 4-10. Magnesium coordination of
Mg] is octahedrally coordinated. Mgl
coordinates bindentately with 710 and 714.
Mgl also coordinates with two waters W24
and W23. W23 has two different
conformations shown as W23a and W23b.
Mgl is colored as a magenta sphere and the
water molecules are shown as red spheres.
The side-chains are colored by atom. All
distances are shown in Angstroms.

 

 

Figure 4-11. On the opposite side of the calcium interface cadmium coordination at
Cd] is six coordinate. Cdl coordinates with E16 and K19 of the crystallographic
symmetry-related molecule. Cdl also coordinates with G1 and two waters W18 and
W19. All distances are shown in Angstroms. The structure is colored green and a
symmetry-related molecule is colored magenta. Cdl is colored as a blue sphere and
the waters are shown as red spheres. All side-chains are colored by atom.

160

distances range from about 1.9 A — 2.5 A whereas the Mg2+ distance ranges from
about 1.8 A — 2.3 A. Cadmium is essential for crystallization since no crystals
grew in the absence of cadmium or decreased concentrations of cadmium. The
structure conﬁrms the need for cadmium during crystallization since three out of the
four cadmium sites reside on crystallographic axises proving that cadmium is crucial
to form a crystalline lattice.

4.10 Comparison between Ca2+-con-T[K77] and Cd2*lMgz*-con-T[K77]
An overlay of the Cd2+/1v1g2+-con-T[K7y] onto the Ca2+-con-TlK7y] dimer is shown
in Figure 4-12. Gla residues 3, 4, and 7 are move signiﬁcantly relative to the metal
ion binding sites whereas 710 and 714 are virtually in the same position. 74 moves
the most because of the 7477 Cd2+ chelation that exists only in the Cd2+/Mg2+-con-
T[K77] structure. Interestingly, K18 is also able to move signiﬁcantly. However,
this is due to crystal packing interactions. Also, Mgl is near one of the Ca2+ (CalB)

and Cd2 is near CalA. However Cd4 is not close to Ca2A and there is no Ca2+ near

Cd3.

161

 

Figure 4-12 Gla residues 3, 4, and 7 are able to move relative to the speciﬁc metal
cation. Overlay of the Cd2+/Mg2+-con-T[K77] structure (magenta) onto the Ca2+-con-
T[K77] dimer (blue) structure. Cd2+ are shown as magenta spheres, Mg2+ is shown as
a green sphere, and Ca2+ are blue spheres. The side-chains are colored by atom.

162

In order to understand why Mg2+ does not cause dimerization in either con-
T[K77] and con-G, we must take a closer look at the Mg2+ metal cation site. Figure
4-12 shows that the Mg2+ metal cation is relatively close to one of the Ca2+ metal
sites (CalB). Figure 4-13 shows an overlay of the Cd2+/Mg2+-con-T[K77] structure
onto the Ca2+-con-T[K77] dimer. The distance between Mgl and Cal is about
0.84A. Cal coordinates with 710 and 714 of one helix and 77 and 710 of the dimer
helix whereas Mgl only coordinates with 710 and 714. The distances for Mgl
coordination with the Ca2+ dimer helix residues 710 and 77 are 3.08A and 2.91A,
respectively. This distance is much too far for Mg2+. If you were to replace Cal
with Mgl, the chelation distances would still be too far for Mg2+. Mg2+ generally
chelates with a distances between 1.7-2.3 A whereas Ca2+ coordinates with distances
ranging from 2.2-2.6 A. This is expected since Mg2+ is a much smaller metal cation
than Ca2+. Another difference occurs in the coordination requirements. Ca2+ is not
as rigid in its requirements for coordination. For example, in the structures of Ca2+-
con-G and Ca2+-con-T[K77], Ca2+ chelation is six-coordinate and occasionally even
seven-coordinate. However, Mg2+ metal cations are most often octahedrally
coordinated. We postulate that the reason why Mg2+ does not form the dimer is
because Mg2+ is too small of a cation to make the larger distance contacts seen at the
calcium induced dimerization interface. Since Mg2+ coordinates with distances
closer than that of Ca2+, Mg2+ will not be able to reach critical Gla residues to
interact with the dimer. This explains why con-T[K77] and con-G do not dimerize in

the presence of Mg2+.

163

 

Figure 4-13 An overlay of the Cd2+/Mg2+-con-T[K77] structure onto the Ca2+-con-
T[K77] dimer. The Ca2+-con-T[K77] dimeric structure is colored blue and the
Cd2+/Mg2+-con-T[K77] structure is colored magenta. Mgl is shown as a green
sphere whereas Cal is shown as a blue sphere. The distances shown are all in
Angstroms. The side-chains are colored by atom.

Another issue that must be understood is why Mg2+ only chelates 710 and
714 and not any of the other Gla residues speciﬁcally 73 and 77. Figure 4-14 Shows
a closer view of the Mg2+ site and the 73-77 chelation pair. The structure shows that
R13 stabilizes the negative charge of the Gla residues (710 and 714) allowing the
Mg2+ to bind better. The 73-77 chelation pair is not as negative as the 710-714 pair
because of R13. The Mg2+ has also forced 710 and 714 to move closer to one
another (Figure 4-13) than from the Ca2+ bound con-T variant structure. This cannot
happen at the 73-77 pair since 74 has ﬂipped away to stabilize the negative charge

and chelate Cd2+.

 

Figure 4-14 A closer view of the Mg2+ site and the 73-77 chelation pair. All
distances are shown in Angstroms. All side—chains are colored by atom. The
magenta atom is Mgl and the blue atoms are Cd2+.

165

4.11

Ex)

Literature Cited

Prorok, M., Warder, S.E., Blandl, T. & Castellino, F.J. Calcium binding
properties of synthetic gamma-carboxyglutamic acid-containing marine cone

snail "sleeper" peptides, conantokin-G and conantokin-T. Biochemistry 35,
16528-34 (1996).

Skjaerbaek, N., Nielsen, K.J., Lewis, R.J., Alewood, P. & Craik, D.J.
Determination of the solution structures of conantokin-G and conantokin-T
by CD and NMR spectroscopy. J Biol Chem 272, 2291-9 (1997).

Rigby, A.C., Baleja, J.D., Furie, B.C. & Furie, B. Three-dimensional
structure of a gamma-carboxyglutamic acid-containing conotoxin,

conantokin G, from the marine snail Conus geographus: the metal-free
conformer. Biochemistry 36, 6906-14 (1997).

Rigby, A.C. et al. Role of gamma-carboxyglutamic acid in the calcium-
induced structural transition of conantokin G, a conotoxin from the marine
snail Conus geographus. Biochemistry 36, 15677-84 (1997).

Chen, Z. et al. Conformational changes in conantokin-G induced upon
binding of calcium and magnesium as revealed by NMR structural analysis. J
Biol Chem 273, 16248-58 (1998).

Dai, Q., Prorok, M. & Castellino, F]. A new mechanism for metal ion-
assisted interchain helix assembly in a naturally occurring peptide mediated
by optimally spaced gamma-carboxyglutamic acid residues. J Mol Biol 336,
73 1-44 (2004).

Dai, Q., Castellino, P.J. & Prorok, M. A single amino acid replacement
results in the Ca2+-induced self-assembly of a helical conantokin-based
peptide. Biochemistry 43, 13225-32 (2004).

Blandl, T., Prorok, M. & Castellino, P.J. NMDA-receptor antagonist
requirements in conantokin-G. FEBS Lett 435, 257-62 (1998).

Warder, S.E., Blandl, T., Klein, R.C., Castellino, F.J. & Prorok, M. Amino
acid determinants for NMDA receptor inhibition by conantokin-T. J
Neurochem 77, 812-22 (2001).

166

10. Warder, S.E. et al. The NMR solution structure of the NMDA receptor
antagonist, conantokin-T, in the absence of divalent metal ions. F EBS Lett
411, 19-26 (1997).

167

APPENDIX

168

 

Appendix 5.1 Scalepack output ﬁle of Angiostatin/VEK-30 P6.22

Shell

Lower Upper
limit 0 1 2

Limit
50.00
4.95
3.93
3.44
3.12
2.90
2.73
2.59
2.48
2.38
All hkl

I/Sigma in resolution shells:

% of reﬂections with 1/ Sigma less than

3 5 10 20 >20

4.95 0.9 1.4 1.6 1.9 2.1 4.7 19.4 66.4
3.93 0.6 1.0 1.2 1.4 1.6 3.0 14.0 80.4
3.4-4 0.9 1.0 1.3 1.6 2.4 4.8 18.2 79.4
3.12 0.5 1.0 1.8 2.2 3.5 8.4 23.1 76.9
2.90 0.7 1.2 2.2 3.1 6.8 17.4 38.0 62.0
2.73 1.1 2.8 4.6 6.6 11.8 26.1 50.6 49.4
2.59 2.0 5.1 8.8 13.2 23.6 45.5 73.2 26.7
2.48 1.6 9.2 19.7 30.2 44.8 66.8 84.7 13.4
2.38 3.2 13.2 26.7 38.3 53.5 70.4 82.1 5.6
2.30 5.2 23.1 40.9 51.7 64.4 73.6 79.4 2.5

1.6 5.7 10.5 14.5 20.7 31.1 47.2 47.2

Summary of reﬂections intensities and R-factors by shells
R linear = SUM ( ABS(I - <I>)) / SUM (I)
quuare=SUM((I-<I>) ** 2)/SUM(I ** 2)
Chi**2 =SUM((I-<I>) ** 2)/(Error ** 2 * N/(N—1)))
In all sums single measurements are excluded

Shell Lower Upper Average
limit Angstrom I error
50.00 4.95 6673.7 308.1
4.95 3.93 7495.9 274.7
3.93 3.44 5756.9 203.1
3.44 3.12 2377.0 66.7
3.12 2.90 1197.1 39.3
2.90 2.73 745.7 28.8
2.73 2.59 526.6 32.9
2.59 2.48 339.8 36.9
2.48 2.38 239.3 39.9
2.38 2.30 161.9 45.3

All reﬂections 2648.7

Average Norm. Linear Square
stat. Chi**2 R-fac R-fac
224.6 1.169 0.068 0.082
228.8 1.908 0.079 0.095
167.8 2.287 0.095 0.124
57.2 2.294 0.108 0.144
35.3 1.921 0.117 0.137
28.7 2.217 0.141 0.201
32.8 2.100 0.203 0.548
36.9 1.220 0.161 0.145
39.9 0.925 0.198 0.175
45.3 0.693 0.247 0.201

110.3 91.7 1.860 0.094 0.112

169

total
85.8
94.4
97.6
99.9
99.9
100.0
99.9
98.1
87.7
81.9
94.4

Appendix 5.2 Scalepack output ﬁle of AngiostatinNEK-30 P6 .

Shell I/Sigma in resolution shells:

Lower Upper % of reﬂections with I / Sigma less than
limit limit 0 1 2 3 5 10 20 >20
50.00 5.81 0.2 0.4 0.7 1.1 2.4 7.1 43.8 51.0
5.81 4.62 0.8 1.4 1.9 2.5 4.1 11.4 40.7 57.4
4.62 4.03 1.1 1.7 3.1 4.6 9.4 25.0 64.3 34.3
4.03 3.66 1.1 2.7 5.5 9.8 21.6 48.0 83.3 15.0
3.66 3.40 1.4 5.5 13.5 24.1 41.8 72.6 95.7 3.2
3.40 3.20 3.2 16.4 37.6 57.0 77.2 93.7 98.0 0.5
3.20 3.04 7.2 37.9 64.3 79.6 91.6 96.2 97.1 0.0
3.04 2.91 9.5 46.3 63.8 69.3 72.3 73.5 73.6 0.0
2.91 2.80 4.3 19.6 24.7 25.9 27.0 27.3 27.4 0.0
2.80 2.70 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Allhkl 2.9 13.1 21.4 27.2 34.6 45.3 62.3 16.3

total
94.8
98.1
98.6
98.3
98.9
98.6
97.1
73.6
27.4

0.0
78.5

Summary of reﬂections intensities and R-factors by shells
R linear = SUM ( ABS(I - <I>)) / SUM (I)
R square = SUM ( (I - <I>) ** 2) / SUM (I ** 2)
Chi**2 = SUM ( (I - <I>) ** 2) / (Error ** 2 * N/(N-1)) )
In all sums Single measurements are excluded

Shell Lower Upper Average Average Norm. Linear Square
limit Angstrom I error stat. Chi**2 R-fac R-fac
50.00 5.81 10401.5 514.7 347.8 1.087 0.065 0.077
5.81 4.62 5723.4 259.5 218.4 1.462 0.069 0.083
4.62 4.03 3866.7 220.2 195.8 1.220 0.081 0.093
4.03 3.66 2660.2 212.2 201.4 0.968 0.102 0.111
3.66 3.40 1532.3 206.0 201.8 0.577 0.129 0.139
3.40 3.20 671.6 198.0 198.0 0.329 0.211 0.210
3.20 3.04 346.9 208.4 208.4 0.254 0.320 0.323
3.04 2.91 197.3 224.6 224.6 0.205 0.408 0.388
2.91 2.80 167.5 227.9 227.9 0.208 0.445 0.372
All reﬂections 3159.9 255.1 224.4 0.837 0.084 0.083

170

Appendix 5.3 Scalepack output ﬁle of Ca2+/con-G

Shell I/Sigma in resolution shells:

Lower Upper % of of reﬂections with 1/ Sigma less than
limit limit 0 1 2 3 5 10 20 >20
50.00 2.80 0.0 0.1 0.1 0.2 0.5 1.9 12.4 77.1
2.80 2.22 0.0 0.1 0.3 0.4 1.5 4.0 12.2 87.7
2.22 1.94 0.3 0.4 1.0 1.5 3.0 8.2 18.5 81.3
1.94 1.76 0.4 1.7 4.1 5.6 9.8 18.2 37.6 62.2
1.76 1.64 1.9 3.9 7.2 10.7 19.0 31.3 57.1 42.6
1.64 1.54 3.0 7.0 12.5 18.1 27.4 47.6 75.5 24.1
1.54 1.46 3.5 10.1 18.9 26.4 41.6 62.7 86.2 13.3
1.46 1.40 4.4 15.3 28.3 0.2 57.0 77.8 92.2 7.1
1.40 1.35 8.9 27.8 43.6 55.3 71.5 87.2 97.7 1.6
1.35 1.30 8.7 32.1 53.2 66.9 79.2 92.7 98.5 0.5
All hkl 3.1 9.8 16.8 22.4 30.9 42.9 58.5 40.0

total
89.5
99.9
99.8
99.8
99.7
99.6
99.5
99.3
99.3
99.0
98.5

Summary of reﬂections intensities and R-factors by shells
R linear = SUM ( ABS(I - <I>)) / SUM (I)
quuare=SUM((I-<I>) ** 2)/SUM(I ** 2)
Chi**2 = SUM ( (I - <I>) ** 2)/(Error ** 2 * N/(N-l) ))
In all sums single measurements are excluded

 

Shell Lower Upper Average Average Norm. Linear Square
limit Angstrom I error stat. Chi**2 R-fac R-fac
50.00 2.80 84068.8 3227.9 1681.8 1.916 0.062 0.071
2.80 2.22 41083.1 1074.5 435.1 1.548 0.057 0.064
2.22 1.94 20749.8 504.1 285.4 1.764 0.063 0.069
1.94 1.76 7616.3 225.4 178.1 1.700 0.085 0.085
1.76 1.64 3596.7 149.8 140.0 1.494 0.111 0.110
1.64 1.54 2105.6 131.4 126.4 1.284 0.153 0.139
1.54 1.46 1387.0 125.3 122.5 1.129 0.206 0.180
1.46 1.40 866.5 119.3 118.0 0.779 0.274 0.207
1.40 1.35 493.6 117.1 116.7 0.756 0.444 0.383
1.35 1.30 360.9 117.5 117.2 0.654 0.535 0.423
All reﬂections 15851.8 563.1 323.6 1.285 0.071 0.069

171

Appendix 5.4 Scalepack output ﬁle of Ca2+/con-T[K7y]

Shell
Lower Upper
limit limit
40.00 3.45

3.45 2.74
2.74 2.39
2.39 2.17
2.17 2.02
2.02 1.90
1.90 1.80
1.80 1.72
1.72 1.66
1.66 1.60
All hkl

I/Sigma in resolution shells:
% of of reﬂections with 1/ Sigma less than

0
0.7
0.5
2.1
2.6
2.6
3.6
6.9

1
1.5
2.0
5.5
5.1
7.0

11.4
18.0

2

3

5

2.3 3.4 5.1

3.7
8.6
8.1
13.2
19.6
29.9

11.7 28.0 44.6
12.7 33.9 52.9
14.8 42.2 63.3
5.7 15.1 24.1

5.3
10.5
11.4
18.5
27.1
40.7
57.9
66.6
77.1
31.2

7.7

15.7
16.0
28.3
42.2
55.8
75.7
82.4
90.0
41.0

10

8.4

15.0
25.7
30.7
47.5
67.2
81.2
94.2
97.2
99.0

20

18.2
31.0
48.8
58.3
79.7
94.0
98.3
99.7
99.8

>20 total
77.3 95.5
69.0 100.0
51.1 99.9
41.7 100.0
20.3 100.0
6.0 100.0
1.6 99.9
0.3 100.0
0.1 99.9

100.0 00100.0
55.6 71.7 27.7 99.5

Summary of reﬂections intensities and R-factors by shells
R linear = SUM ( ABS(I - <I>)) / SUM (I)
quuare=SUM((I-<I>) ** 2)/SUM(I ** 2)
Chi**2 = SUM ( (I - <I>) ** 2)/(Error ** 2 * N/(N-1)) )
In all sums single measurements are excluded

Shell Lower Upper Average Average
limit Angstrom I error stat.
40.00 3.45 21439.2 642.3 327.1
3.45 2.74 12805.1 386.4 234.9
2.74 2.39 5409.5 218.8 169.6
2.39 2.17 4343.7 209.4 173.0
2.17 2.02 2563.9 182.8 165.9
2.02 1.90 1430.5 165.5 158.4
1.90 1.80 918.1 152.8 148.9
1.80 1.72 477.9 141.8 140.4
1.72 1.66 368.1 143.1 142.2
1.66 1.60 275.6 146.2 145.6
All reﬂections 5180.1 242.9 182.1

Norm. Linear Square
Chi**2 R-fac

R-fac

1.427
1.152
0.899
0.782
0.682
0.599

0.567
0.523
0.547
0.565
0.775

172

0.060
0.069
0.090
0.103
0.148
0.245
0.361
0.638
0.895
0.000
0.099

0.070
0.074
0.093
0.106
0.144
0.249
0.292
0.615
0.856
0.000
0.079

Appendix 5.5 Scalepack output ﬁle of Cd2+/Mg2+/con-T[K7y]

Shell I/Sigma in resolution shells:

Lower Upper % of of reﬂections with 1/ Sigma less than

limit limit 0 1 2 3 5 10 20 >20 total
40.00 3.26 0.5 1.0 1.3 1.8 3.5 6.1 13.7 83.9 97.5
3.26 2.59 0.4 0.4 1.1 1.5 3.3 6.6 14.7 85.1 99.8
2.59 2.26 0.6 1.5 2.8 3.4 5.2 8.9 16.4 83.4 99.8
2.26 2.05 0.4 1.3 2.7 4.0 6.0 10.9 22.5 77.5 100.0
2.05 1.90 0.6 1.1 2.7 4.4 7.0 13.3 30.0 70.0 100.0
1.90 1.79 2.3 4.8 6.5 8.8 11.3 18.5 36.9 63.1 100.0
1.79 1.70 1.8 5.3 7.6 9.2 13.5 22.1 46.3 53.7 100.0
1.70 1.63 0.8 2.6 5.9 8.5 14.0 24.0 52.8 47.2 100.0
1.63 1.57 1.9 4.3 7.6 10.7 14.8 31.3 58.8 41.2 100.0
1.57 1.51 0.4 4.5 8.6 11.5 20.0 37.9 67.5 32.5 100.0
1.51 1.46 1.2 5.1 9.9 16.4 29.1 53.0 82.2 17.8 100.0
1.46 1.42 2.4 10.9 19.6 25.5 39.0 63.0 91.7 8.3 100.0
1.42 1.39 3.9 14.1 26.8 35.5 48.2 75.2 94.7 5.3 100.0
1.39 1.35 5.6 20.6 37.5 47.6 61.7 84.1 95.6 2.8 98.4
1.35 1.32 5.4 23.4 39.9 48.4 65.7 83.3 92.1 0.6 92.7
1.32 1.29 3.5 24.0 37.6 48.8 63.6 77.6 81.3 0.4 81.7
1.29 1.27 4.9 22.2 34.6 43.6 53.4 60.1 62.5 0.2 62.7
1.27 1.24 2.8 12.4 19.8 24.4 30.5 34.6 34.8 0.0 34.8
1.24 1.22 0.2 1.4 2.4 4.8 6.2 8.2 8.5 0.0 8.5
1.22 1.20 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
All hkl 1.9 7.9 13.4 17.5 24.3 35.3 49.4 34.7 84.2

Summary of reﬂections intensities and R-factors by shells
R linear = SUM ( ABS(I - <I>)) / SUM (I)
quuare=SUM((I-<I>) ** 2)/SUM(I ** 2)
Chi**2 = SUM ( (I - <I>) ** 2)/(Error ** 2 * N/(N-l) ))
In all sums single measurements are excluded

173

Shell Lower Upper Average

limit
40.00
3.26
2.59
2.26
2.05
1.90
1.79
1.70
1.63
1.57
1.51
1.46
1.42
1.39
1.35
1.32
1.29
1.27
1.24

Angstrom I

3.26
2.59
2.26
2.05
1.90
1.79
1.70
1.63
1.57
1.51
1.46
1.42
1.39
1.35
1.32
1.29
1.27
1.24
1.22

30150.7
17688.5
12812.1

9338.9
7049.7
4685.8
3458.2
2799.6
2148.1
1888.8
1563.5
1099.2
966.0
768.7
620.2
660.7
550.4
591.3
1072.6

All reﬂections 6170.2

error
716.3
395.7
278.2
249.6
204.4
153.1
126.2
111.0
96.9
91.2
95.0
105.2
123.4
146.8
166.4
189.4
203.9
231.5
283.5
210.3

Average

stat.
525.2
300.3
220.4
177.7
150.4
122.7
105.0
94.1
84.9
80.9
86.2
99.2
117.9
142.6
163.5
187.1
202.3
230.9
282.0
173.1

Norm. Linear Square
Chi**2 R-fac

1.596
1.282
1.040
0.731
0.699
0.617
0.590
0.610
0.589
0.568
0.542
0.450
0.429
0.418
0.394
0.395
0.358
0.543
0.666
0.752

174

0.056
0.059
0.059
0.062
0.064
0.072
0.081
0.091
0.105
0.110
0.108
0.131
0.134
0.156
0.163
0.138
0.151
0.213
0.206

R-fac
0.069
0.064
0.067
0.070
0.066
0.078
0.084
0.088
0.108
0.095
0.069
0.1 15
0.109
0.125
0.169
0.127
0.166
0.259
0.224

0.067 0.068