. 2:3:
,‘3 ~ 153:“
#5215: mf‘ixvx 2‘-‘~-! :
« .r-‘b xixziér‘fié" ~--«
If~£tt€zé{.%~£§§g
‘ ‘t I . ‘.I‘

a\

L. A «763.1. . ~ 1..
‘r” dfimfi‘fi‘ “
- ~ \-
w '5. ‘ . . :fi'fik‘ifl‘g‘l
t .3' TA“ 1. — ”-3
. éi‘u‘afizfl'f‘ég“ 3
h l- 1'" .

. ;_. a

’ ‘EERHW
war-:1;
I}

’ .
Kati-«é
u if
. ‘9
.,.
..

I“: l
_ V‘
‘ h'.“ '1
rail-'05:?
L7 13.355-
ayflv ”

5
4.
‘
.'

- I .‘T .1”
$3 11;" . . '
I I,'- 9' k
. f— v 1"

"
i:
[A

yr

- .3".
, ‘ Wu.
'3“ :7. 3;; if;
}

I

..

~.
- , _, _ , . r r _ ‘ . .‘ v. ;‘_:;:‘(:-a ' .‘Jt
- "-1me fl
, .

«N; {3-}. -~~n.5‘- ...:.
J‘ra s :3; '2.“ vi“
I“ ‘ ‘3?

. ix . .g
gawk: ”3%;
“m 3,3: " wt”??? ,5

w. ; 3"
fi~ s‘

.1. v

[u mill—v
C

h. ‘ Vi": '

.N . Q; "VI:
,‘E‘ia‘i‘wrféat ~
, . ‘ - 1*; ~51
23’3"} {.‘l~1:"‘l’ 1“}?
‘A ‘ u" 0:
~ ”Wig-255%.;
~Ah ‘
ark: 5 ‘
. - a .
31.

A
v,
’- If
5"
M‘v‘
”.‘q-v ' ’
.( “r53,
.

itfié

-3';' a
3‘

I
§
"1

'§

x
g: .'
J

,.

$55
. "P
F"
{at c

‘ 1:
1.1.
‘13.! 1.

%k',';‘ékw
$3323.: 2
L_ A

M" £5“ +

v k 7
if?
5‘.

u
- 99"”‘1'
1,21". 3..
_|~:,. {I ..
1??? -’ '

0....
w

J ”v.34:
'Wflu‘h‘,.-
W 9... W _ —",v
' «'JQ I “'15:, o
n-

--~ -
- . ~

n.-

n

 

‘ililiiiillliii

 

This is to certify that the

dissertation entitled

THE STRUCTURE OF A COMPLEX OF RECCNBINANT HIRUDIN
AND HUMAN oz-THRQMBIN AT 2.3 A RESOLUTION

presented by

Timothy John Rydel

has been accepted towards fulfillment
of the requirements for

Ph . D . degree in Chemistry

 

Gm Tuséfi

 

Major professor

Datew

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

k

F

 

PLACE IN RETURN BOX to
TO AVOID FINES return on

     

L

manna-r

Michigan state

University

fl

or betore due due.

DATE DUE DATE DUE DATE DUE

Tl

 

remove this checkout from your record.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

'l

 

 

 

 

 

 

 

 

 

 

J

 

 

 

 

 

 

 

 

 

J

 

MSU Is An Affirmative Action/Equal Opportunity Institution

_H—

cWMG-nt

‘

 

 

THE STRUCTURE OF A COMPLEX OF RECOMBINANT HIRUDIN
AND HUMAN «PTHROMBIN AT 2.3 A RESOLUTION

BY

Timothy John Rydel

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemistry

1991

ABSTRACT

THE STRUCTURE OF A COMPLEX OF RECOMBINANT HIRUDIN
AND HUMAN a-THROHBIN AT 2.3 A RESOLUTION

By
Timothy John Rydel

Crystals of a complex of the recombinant hirudin, rHVZ-K47, and
human a-thrombin were obtained via vapor diffusion from hanging drops.
The hirudin-thrombin complex crystallizes in the tetragonal crystal
system, space group P43212, with a = b = 90.54 A, c = 132.04 A and one
molecule in the asymmetric unit.

The structure was solved by Patterson search techniques using
PPACK-human arthrombin as a model and a 2.8 A resolution diffractometer
intensity data set. The rotational search was conducted in Patterson
space using 8.0 to 3.0 A resolution data and the SEARCH routine in the
PROTEIN package of macromolecular crystallographic programs. The
translational search also used 8.0 to 3.0 A resolution data, and was
conducted with programs written by Lattman and modified by Deisenhofer
and Huber. The R-value of the initial solution was 0.39, but it reduced
to 0.34 when the rotational and translational parameters were refined
with the rigid body refinement program TRAREF.

The initial structure was improved using 2.3 A resolution FAST area
detector data, the energy restraint-crystallographic refinement program
EREF, and graphics interventions. After six iterative stages of
model—building and refinement, the hirudin-thrombin complex (includes
207 water molecules) had an R-value of 0.193 for 21,960 reflections
between 7.0 and 2.3 A resolution.

The complex was further refined using PROFFT, the fast Fourier

.Ff‘
.‘4

 

our

i

.i..\

0'1.

ch
ch

in

fig

g.‘

"r

0‘

 

 

transform version of the stereochemically-restrained least squares
refinement program PROLSQ. After five iterative stages of
model-building and refinement, the complex (includes 265 water
molecules) had excellent overall stereochemistry and an R-value of 0.173
for 21,056 reflections between 7.0 and 2.3 A resolution.

Hirudin consists of an N-terminal globular domain with a
three-cystine core and an extended C-terminal tail. The N-terminal
tripeptide of hirudin interacts in a novel manner in the active site
region of thrombin. The C—terminal tail of hirudin makes many
electrostatic as well as hydrophobic interactions in the anion-binding
exosite of thrombin. In all, hirudin engages in 216 contacts of less
than 4.0 A with thrombin. This abundance of interactions is likely the
source of the high affinity and specificity of hirudin.

The overall folding of the thrombin structure in the complex is
very similar to that of PPACK-thrombin, however many significant main
chain and side chain differences exist between the two. Most of the
differences can be accounted for in terms of four categories of factors

or influences.

To my parents,
Arlene and Joseph Rydel,
and to my wife,
Odette,
for their encouragement, love, and support.

iv

ulq er I c .

G. . . w. . G.
. .. .fi .2 . . a... c ..
.x. 5b. a . Au» v. v.
.1 . ,. , v . .1. C.

 

 

 

a:

 

p u

ACKNOVLEDGEHENTS

To Professor Alexander Tulinsky I extend my sincere gratitude for
his constant enthusiasm, guidance, interest, and support throughout all
aspects of this work. I could not have asked for a better scientific
mentor, and it has been a wonderful learning experience to be a part of
his laboratory. I am also very grateful to him for allowing me to
return to Michigan State to continue work toward a Ph.D. degree in his
group.

To Dr. K.G. Ravichandran, a former post—doc of the laboratory, I
extend my sincere thanks for his important contributions to the
structure solution of the hirudin—thrombin complex. In particular, I am
grateful to him for his assistance with the model-building and
refinement work on the complex conducted at the Max-Planck-Institut fur
Biochemie in Martinsried, Germany, and for his friendship during two
exciting and exhausting months spent overseas.

I have also benefitted a great deal from being a part of a very
friendly, instructive and supportive research group. I would especially
like to thank Drs. Vasili Carperos, Anne Mulichak, Pappan Padmanabhan,
Eva Skrzypczak-Jankun, and Manuel Soriano-Garcia for their assistance
and helpful discussions.

It is also a pleasure for me acknowledge Dr. Carolyn Roitsch of
Transgene, S.A., in Strasbourg, France for supplying us with ample
amounts of recombinant hirudin, and for her genuine enthusiasm and

interest in the project. The generous financial support of Transgene,

S.A., is also gratefully acknowledged. Thanks is also extended to Dr.
John V. Fenton II of the New York State Department of Health in Albany,
NY for supplying us with large quantities of human a-thrombin.

It is an honor and a privilege for me to thank Professor Robert
Huber of the Max—Planck-Institut fur Biochemie in Martinsried, Germany
for his role in this project. The invitation to come to his laboratory
to collaborate on the structure solution of the complex, and his advice
and guidance during my visit are greatly appreciated. I would also like
to thank Drs. Wolfram Bode, Albrecht Messerschmidt, and Milton Stubbs,
and Ms. Monika Schneider - all members of Professor Huber’s laboratory
- for their valuable advice and assistance during my stay.

Prior to returning to Michigan State to work on this project. I was
very fortunate to have worked in two industrial protein crystallography
laboratories where I was given the opportunity to learn and develop new
skills in both biochemistry and crystallography. To Drs. Cele
Abad-Zapatero and John Erickson of Abbott Laboratories, and Drs. Howard
Einspahr and Keith Vatenpaugh of The Upjohn Company, I am thankful for
these opportunities and for their encouragement of my returning to
Michigan State to continue work toward a Ph.D. degree.

I am fortunate to have had so many supportive family members and
friends behind me in this effort. I am especially indebted to my
wonderful wife Odette, and to my parents, my brother and sisters, and my
Grandma for their continual love and encouragement.

Support by the National Institutes of Health is gratefully

acknowledged.

vi

CHAPTER

II

III

IV

VI

VII

LIST OF TABLES

LIST OF FIGURES .

INTRODUCTION

A. Thrombin .

B. Hirudin

EXPERIMENTAL

A. Sample Preparation . . .

8. Sample Molecular Veight Estimation .

C. Crystallization . . . .

D. Crystal Characterization . . . . . . .

E. 2. 8 A Resolution Diffractometer Intensity
Data Collection . .

F. 2. 3 A Resolution Area Detector Data Collection .

DATA REDUCTION

A. Converting the 2. 8 A Resolution Diffractometer
Intensity Data to Structure Amplitudes .

8. Post- Data Collection Processing of the 2. 3 A

TABLE OF CONTENTS

Resolution Area Detector Structure Factors .

MOLECULAR REPLACEMENT CALCULATIONS . . . . . . . .

Introduction . . . . . .
The Rotation Function Calculations . . . . .
The Translation Function Calculations .
The Initial Hirudin- Thrombin Electron Density
Map Calculation . . . . . . . . . . . . .

MODEL-BUILDING AND EREF REFINEMENT

MODEL-BUILDING AND PROLSQ REFINEMENT

RESULTS AND DISCUSSION

A.
B.
C.

The Hirudin Structure . . . . . . .
The Hirudin- Thrombin Interaction . . . . .
Crystal Packing . . . . .

vii

PAGE

ix

. xiii

62
68
76
76
79
81
84
89

95

120

120
139
178

 

 

eh.‘
-

-

.p—"

«0"

i...

 

 

CHAPTER PAGE

VIII A COMPARISON OF THE THROMBIN STRUCTURE IN THE
HIRUDIN—THROMBIN COMPLEX AND IN PPACK—THROMBIN . . . . . 185
LIST OF REFERENCES . . . . . . . . . . . . . . . . . . . 218

viii

TABLE

10

11

12

13
14

LIST OF TABLES

The Primary Sequence of Human a—Thrombin. The amino

acid type is indicated using the standard single
letter code. The numbering of residues is based on
topological similarities with chymotrypsinogen. A
number followed by an alphabetic character indicates
an insertion. CP represents a cis-proline.
(Numbering taken from [6].) . . . . . . . . . . . .

Summary of Crystal Data of rHV2-K47—Human
a-Thrombin e e e e e e e e e e e e e e

The Average Cell Parameters of the HTNZ (3.6-2.8 A)

and of the HTN3 (2.8 A) Crystals. Cell lengths in in
units, angles in degrees . . . . . . . . . . . .

Reflection Statistics on the Hirudin-Thrombin

FAST Data Set Evaluated by MADNES and Corrected using
ABSCOR O O O O O O O O O O O O O O O O O O O O O O 0

Results of the Translation Search . . .
Course of Model-Building and EREF Refinement

Final Model Parameters of the BREE-refined
rHVZ—K47 Human a—Thrombin Structure . .

Summary of Model-Building and PROLSQ Refinement .

Summary of Hirudin- Thrombin PROLSQ Refinement
Restraint Parameters . . . . . . . . . .

Final Model Parameters of the PROLSQ—refined
Hirudin-Thrombin Complex . . . . . . . . .

The Refactor vs. Resolution for the Final
Hirudin-Thrombin Model Parameters .

Average Individual B-factor and Average Occupancy
Factor for the Hirudin-Thrombin Model Parameters

Secondary Structural Elements of Hirudin
Sulfur— Sulfur Distances (A) in Hirudin. A '*'

indicates a disufide bond . . . . .

ix

PAGE

34

45

74
85
91

92

. 100

. 102

. 103

. 110

. 112
. 123

. 125

TABLE PAGE

15 Intramolecular Hydrogen Bonds in Hirudin. Hydrogen

atoms are assigned geometrically idealized positions.

Donor atoms are denoted 'D' and acceptor atoms ’A'. . . . 130
16 Dihedral Angles and Dihedral Energies of the

Disulfide Bridges in Hirudin. A ’*' indicates the
energy value was excluded from the calculation of the
mean . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

17 Intermolecular Hydrogen Bonds in the Hirudin-Thrombin
Complex. Hydrogen atoms are assigned geometrically
idealized positions. Donor atoms are denoted 'D'
and acceptor atoms 'A'. A protonated amino terminus
is designated ’N+' and an unprotonated amino terminus
is designated ’N' for Ilel'. A '*' indicates the
mean and standard deviation were calculated using
values for an unprotonated Ilel' amino terminus . . . . . . 145

18 Hirudin-Water Molecule Hydrogen Bonds in
the Complex. Hydrogen atoms are assigned
geometrically idealized positions. Donor atoms are
denoted ’D', acceptor atoms 'A’ . . . . . . . . . . . . . . 151

19 Solvent-Bridged Hirudin-Thrombin Hydrogen Bond
Interactions. The protein atoms involved in the
interaction are represented as 'A' and ’B', and the
mediating water 'V’ . . . . . . . . . . . . . . . . . . . . 152

20 Solvent-Bridged Intramolecular Hydrogen Bond
Interactions in Hirudin. The protein atoms involved
in the interaction are represented as 'A' and 'B',
and the mediating water 'V' . . . . . . . . . . . . . . . . 152

21 Hirudin a—Thrombin Interactions. Numbers are
intermolecular contacts < 4.0 A; parentheses indicate
ion pairs; a dash followed by a number is a hydrogen
bond; hydrophobic or neutral contacts are underlined
in bold . . . . . . . . . . . . . . . . . . . . . . . . . . 169

22 Solvent Accessible Surface Area of Thrombin Residues
Alone and in Complex with Hirudin. Abbreviations
used: TASA: total atom surface accessibility; MASA:
main chain atom surface accessibility; SASA= side

chain atom surface accessibility . . . . . . . . . . . . . 173
23 Solvent Accessible Surface Area of Hirudin

Residues Alone and in Complex with Thrombin.

Abbreviations used are as defined in Table 22 . . . . . . . 173
24 Percent Loss of Hirudin Solvent Surface

Accessibility Due to Complexation with Thrombin.
Results presented with regard to hirudin residue

TABLE

25

26

27

28

ranges. Abbreviations used are as defined in Table
22 O I O O I 0 O I O O 0 O O O O O O O O O

Residue Surface Accessibility of Free Hirudin.
Conformation of hirudin as it appears in the complex.
Abbreviations used: SASAt= theoretical side chain
atom surface accessibility. All other abbreviations
used are as defined in Table 22. All surface
accessibility results reported in A2. SASA/SASAt
ratios indicate how the actual side chain
accessibility compares to the maximum possible.
TASA/SASAt ratios greater than one indicate that that
there is at least some main chain accessibility.
These ratios are not given for glycines as glycine
has no non-hydrogen side chain atoms. A '*'
preceding a line indicates that the values given are
susceptible to error due to disorder in or near these
residues. A '+' preceding a line indicates a residue
which has at least one direct contact of less than
4.0 A with thrombin . . . . . .

Hydrogen Bonds Between Symmetry-Related Molecules.

Hydrogen atoms are assigned geometrically idealized
positions. Donor atoms are denoted 'D' and acceptor
atoms ’A'. The symmetry molecule number refers to
the symmetry equivalent position of the
symmetry-related molecule in the interaction, as
given below. The symmetry-related molecule residue
number is preceded by a ’*' . . . . . . . . . . . . .

Results of an Optimal Superposition of the Thrombin

Structures of the Complex and PPACK-Thrombin [6]
using the Program MOLFIT. Fitting based on N, CA, G
Atoms. Results expressed in A units as root mean
square deviations (r.m.s.d.). Numbers in the '{}'
brackets represent the number of atom pairs
contributing to the r.m.s.d . . . . . . . . . . . .

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6]
due to Residues Poorly Defined in Electron Density or
Tentatively Placed in Either Structure, or due to
Alternate Interpretations of the Density Maps.
Optimal superpositioning of thrombin molecules as in
Table 27. Abbreviations: HT s hirudin-thrombin
complex; PT . PPACK-Thrombin; MCA r.m.s.d. = main
chain atom average root mean square deviation; SCA
r.m.s.d. a side chain atom average r.m.s.d.; S =
surface; AS 2 active site; conf. = conformation;
rot. = rotation. Deviations > la on main chain (0.45
A) and/or side chain (1.43 A) atoms are listed

xi

PAGE

. 174

. 176

. 182

. 186

. 191

TABLE

29

30

31

32

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6] due
to Crystal Packing Interactions in the Complex.
Optimal superpositioning of thrombin molecules as in
Table 27. Abbreviations: as = aromatic stacking; hb

=hydrogen bond; hip: hydrogen— bonding ion pair; hc
a hydrophobic contact; pc = polar contact; sympos =
symmetry molecule positions (see Table 26), wmi=
water mediated interaction. All other abbreviations
are defined in Table 28. Residues in brackets [] are
not involved in crystal contacts but have significant
conformational differences which could be due to
packing interactions of nearby residues. A '*'
denotes the average MCA and SCA r.m.s.d. for
residues Asn143-Gln151. Table 32 givess a complete
listing of the MCA and SCA r. m. s. d. for the
residues. Deviations > 1o on main chain (0. 45 A)
and/or side chain (1. 43 A) atoms listed .

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6] due
to the Interaction with Hirudin. Optimal
superpositioning of the thrombin molecules as in
Table 27. All abbreviations are defined in Tables 28
and 29. Residues in brackets [] do not interact with
hirudin, but have significant conformational
differences which could be due to hirudin
interactions of nearby residues. Deviations > lo on
main chain (0. 45 A) and/or side chain .(1. 43 A) atoms
listed . . . . . . .

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6]
due to Surface Residue Conformational Variability.
Optimal superpositioning of the thrombin molecules as
in Table 27. Abbreviations: fs 2 free sidechain;
ident. . identical. All other abbreviations defined
in Tables 28 and 29. Deviations > 1o on main chain
(0.45 A) and/or side chain (1.43 A) atoms listed

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6] in
the Vicinity of the 149A-149E Insertion Loop.

Optimal superpositioning of the thrombin molecules as

in Table 27. Abbreviations: r.m.s.d. = root mean
square deviation; MCA r.m.s.d. = average main chain
atom r.m.s.d.; SCA r.m.s.d. = average side chain

atom r.m.s.d. All residues listed have deviations >
1o on main chain (0.45 A) and/or side chain atoms

(1.43 A)

xii

PAGE

192

. 193

194

196

 

w
sssss

 

 

FIGURE

LIST OF FIGURES

The Prothrombinase Complex. Prothrombin (PT)

is composed of fragment 1 (F1), fragment 2 (F2), and
prethrombin 2 (Pre-2); Factor Va . Va; factor Xa =
Xa; Ca+2 ions are represented by filled circles
binding to y—carboxyglutamic acid residues

The Products of PT Activation. CHO represents

a carbohydrate side chain; "a", "b", and "c"
represent the sites of cleavage of bovine and human
PT; an additional cleavage, "b'" occurs in human PT
(taken from [3])

The Polysaccharide of Human a—Thrombin

[8] and its Components: N-Acetylneuraminic Acid
(NeuNAc), N-Acetylglucosamine (GlcNAc), Mannose
(Man), Galactose (Gal), and Fucose (Fuc) . . .

Schematic Drawing of the Fibrinogen Molecule [9].

The symmetric molecule is composed of a dimeric
central domain containing the amino—termini of all
six chains (a,B,Y), two connecting coiled coils, two
terminal domains, and two Aa polar appendages.
Fibrinopeptides A and B (EPA and FPB) are at the
amino-terminal ends of the a and B chains. Four
carbohydrate clusters (CHO) occur, and are located on
each 7 chain near the central domain and on the B
chains of each terminal domain. Primary
cross-linking sites (XL) can be found near the
carboxy-termini of the 7 chain and in the Au polar
appendages . . . . . .

The Central Bioregulatory Functions of

asThrombin in Hemostasis. Coagulation factor
zymogens are identified by roman numerals; an "a"
following a numeral indicates an activated enzyme.
Primes and double primes are used to indicate active
and inactive forms of factors of V and VIII.
Abbreviations: F1.2 = prethrombin fragment 1.2; FpA
. fibrinopeptide A; FpB fibrinopeptide B; PL =
phospholipid; and TPN thromboplastin (taken from
[18]) . . . . . . . . . . . . . . . . . . .

Amino Acid Sequences of the Three Natural
Hirudin Variants: HV1, HV2, and HV3. The boxes

xiii

PAGE

., .yo-F
' e

‘I
..V'.-~

‘Q

.1

- A

«i.

'_.1
(I,

 

FIGURE PAGE

delineate regions of homology and the stars (*) refer
to variant residues. HV3 contains an additional
residue at position 63. Y’ indicates a sulfated
tyrosine (taken from [66]) . . . . . . . . . . . . . . . . 14

7 The Disulfide Bond Connectivity Patterns of

Hirudin and Epidermal Growth Factor [67] . . . . . . . . . 16
8 The Sequence of Recombinant Hirudin Variant

2-Lysine 47 (rHVZ-K47) . . . . . . . . . . . . . . . . . . 18
9 Stereo Drawing of Nine Optimally Superimposed

DISMAN Distance Geometry N, CA, G Structures from the
2D NMR Constraint Data of Haruyama and Vuthrich [80].
Disorder exists in the residue ranges 1-3, 31-36 and
49-65 (taken from [80]) . . . . . . . . . . . . . . . . . . 19

10 Human apThrombin and its Proteolytic Derivatives.
The numbering of residues is taken from Table 1. The
A-chain and B-chain of thrombin are designated by "A"
and "B", respectively. CHO represents the
polysaccharide linked to nitrogen ND2 of Asn(N)60G.
The location of the catalytic triad residues is noted
(857, D102, and $195). Excised proteolytic fragments
are denoted by broken lines . . . . . . . . . . . . . . . . 21

11 100Log(fo100) vs. ZT Gel Concentration. The
slope for each line gives the retardation coefficient
(Kr) for the corresponding protein oligomer.
Abbreviations: L(1ysozyme), CA(carbonic anhydrase),
CEA(chicken egg albumin), BSA-M(bovine serum
albumin-monomer), BSA—D(bovine serum albumin),
HT(hirudin-thrombin) . . . . . . . . . . . . . . . . . . . 29

12 Log(-Kr) vs. Log(Molecular Weight). Linear
regression best fit line to the standard protein
retardation coefficient data. Abbreviations used
have the same meaning as in Figure 11. The
extrapolated molecular weight estimate for the
hirudin-thrombin complex is 38.5 kD . . . . . . . . . . . . 31

13 Photograph of an rHVZ—K47-Human a-Thrombin
Crystal. Crystal size is approximately 1.0 x 0.4 x
0.4 mm. Note the trace of the seed in the center of

the crystal . . . . . . . . . . . . . . . . . . . . . . . . 33
14 hkO Precession Photograph of a Hirudin-Thrombin

Crystal. The resolution limit at the edge of the

photo is 3.7 A. (mu angle = 12.0°) . . . . . . . 35
15 Morphology of an Idealized Hirudin-Thrombin

Crystal . . . . . . . . . . . . . . . . . . 38

xiv

«A

hr-

4‘

LL

‘1

54

(I

 

FIGURE

16

17

18

19

20

21

22

23

24

25

26

27

28

A Hirudin-Thrombin Crystal Mounted in a
Capillary for X-ray Crystallographic Analysis .

The Goniostat of a Four-Circle
Diffractometer [111]

Axial Intensity Distributions of the
Recombinant Hirudin-Human a—Thrombin Crystals .

The w—Profile of (6,16,17) Taken Before
the 3-D HTNZ Data Collection. Peak width at
half-height approximately 0.18° . . . . . . . .

Intensity Absorption Plot of (0,0,8)

Taken Before the HTN3 3.56 A Resolution Data
Collection. 1(6) in total counts/measurement. The
45° 6 range employed in the data collection
indicated by "D.C." . . . . . .

Drawing of an Enraf—Nonius Diffractometer
with a Conventional Detector [116]

A Flow Chart of Operations in a MADNES Area
Detector Data Collection . . . . . . . .

Dependence of X- -ray Decay Correction with
Scattering Angle for the HTNZ Hirudin-Thrombin
(3.6-2. 8) A Resolution Data Collection . . . .

The <|Fobs|2> versus <29) Distribution for

the 2. 8 A Resolution Hirudin- Thrombin Diffractometer
Data Set . . . . . . . . . . . . . . . . .
Plot of Sin weighting factor V versus X

where V-Il(X)/IO(X) and XxZIFobs(h)||Fcal(h)|/Iu
(taken from [137]). 11(X) and 10(X) are first- and
zero-order modified Bessel functions, respectively.
Iu is an estimate of the contribution from the

unknown structure, or that which is not accounted for
in the madel D O O O O O O O O O O O O O

Stereoview of the First Unit of Thrombin

Carbohydrate Modeled into Map Density. Displayed in
bold is N-acetylglucosamine (NAGl) linked to the side
chain atom ND2 of Asn60G. Map density is from the
final 2Fo-Fc map used for model-building; contours at
0.8a. Atomic positions are from the final
coordinates . . . . . . . . . . . . . . . . . . . . .

Histogram of w Angles in the Hirudin-Thrombin
Complex . . . . . . . . . . . . . . . .

The Ramachandran Plot of Hirudin-Thrombin.
Only non-glycine residue angles are displayed .

XV

PAGE

39

41

43

48

50

52

55

65

69

88

98

. 107

. 108

at.

at

a

Re
~44

lhu.

)4-

1114

 

FIGURE PAGE

29 Plot of R-factor vs. Resolution for
Hirudin-Thrombin. Large filled circles represent the
data points of the complex. The small filled circles
represent structure factor amplitude discrepancy data
due to specified average coordinate errors (Ar)
according to Luzzati [148] . . . . . . . . . . . . . . . . 111

30 Plot of the Average Individual B-factor
of Main and Side Chain Atoms for the Protein Atoms in
the Complex. The <Bi> for main and side chain atoms
is represented by '+' and 'X' symbols, respectively.
Main chain atom <Bi> values connected by a line. The
break in <Bi> in hirudin corresponds to the
completely undefined segment Ser32'-Lys35'. Bold
arrows point to main and side chain (Bi) values of

Cys residues engaged in disulfide linkages . 114
31 Histogram of the Individual B-factors of the

Water Molecules in the Hirudin-Thrombin Structure.

. 117

32 Histogram of the Occupancy Factors of the

Water Molecules in the Hirudin—Thrombin Structure.

All water molecules in the last occupancy shell, 1.0

- 1.1, have an occupancy of 1.0 . . . . . . 118

33 Stereoview of the Folding of Hirudin in the
Complex. N, CA, C atoms only; hirudin disulfides in
bold; disordered or poorly defined regions indicated
by dashed lines . . . . . . . . . . . . . . . . . . . . . . 121

34 The Sequence of Recombinant Hirudin Variant
2-Lysine 47 (rHV2-K47). Isolated loop designated ’A'
and loop segments comprising the two interconnected
loops designated '8', 'C', and 'D'. . . . . . . . . . . . . 122

35 Stereoview of the Cystine Disulfide Core
of Hirudin. Disulfide linkages in bold . . . . . . . . . . 124

36 Stereoview of the Folding of NH2-terminal
Domain of Hirudin in the Complex. N, CA, C atoms
only; designations ’A', 'B', 'C', and 'D' correspond
to those of Figure 34; disorder indicated by dashed
lines . . . . . . . . . . . . . . . . . . . . . . . . . 127
37 Stereoview of the 81 Antiparallel B—Strand
in Hirudin. Hydrogen bonds indicated by dashed
lines . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

38 Stereoview of the 82 Antiparallel B—Strand
in Hirudin. Hydrogen bonds indicated by dashed

lines . . . . . . . . . . 129

xvi

.,.vw~fl
b

. I
Ao-V“~

q,_- p
‘U

.4

46

47

 

FIGURE PAGE

39 Stereoview of the Intramolecular Lys47'
Hydrogen Bonds in the Hirudin N-terminal Domain. CA
structure of residues 1'-48' including disulfides
(disulfide linkages in bold), and hydrogen bonding
residues (bold). The disorder in residues 32'—35’ as
well as the hydrogen bonds are indicated by dashed
lines . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

40 Stereoview of the Polyproline II Helix
in Hirudin. N, CA, C atoms indicated in bold . . . . . . . 133

41 Stereoview of the Disulfides in the Crystal
Complex and in the NMR Hirudin Structure [79]. NMR
hirudin disulfide side chains indicated by dashed
lines; disulfide linkages in bold . . . . . . . . . . . . . 136

42 Stereoview of the T4 Type III Helical
Reverse Turn in Hirudin. Hydrogen bonds indicated by
dashed lines; main chain atoms in bold; Gln65’ also
included . . . . . . . . . . . . . . . . . . . . . . . . . 140

43 Stereoview of the CA Structure of the
Hirudin-Thrombin Complex. Hirudin CA and disulfides
of hirudin and thrombin in bold. N- and C—terminals
of thrombin designated . . . . . . . . . . . . . . . . . . 141

44 Stereoview of PPACK in the Active Site
Region of Thrombin. PPACK, from the PPACK-thrombin
structure [6], optimally superimposed into thrombin
active site based on N, CA, C fitting of
PPACK-thrombin and hirudin- thrombin complex thrombin
structures; PPACK in bold . . . . . . . . . . . . . . . . 143

45 Stereoview of the Interaction of the Hirudin
N-terminal Tripeptide in the Active Site of Thrombin.
Hirudin N-terminal tripeptide in bold; hydrogen bonds
of the N-terminal nitrogen indicated by dashed lines.
. . . . ..... . . . . . . . . . . . . . . . . . . . . . 144

46 Stereoview of Hydrophobic Contacts of the
N-terminal Hirudin Domain with Thrombin. Hirudin
residues in bold. Residues in figure: Ile174,
Pro6OC, Leu13', Va121', and Pro46'. . . . . . . . . . . . . 148

47 Stereoview of Electrostatic Interactions of the
N-terminal Hirudin Domain with Thrombin. Hirudin
residues in bold. Hydrogen bonds indicated by dashed
lines, and ion pairs by +/- . . . . . . . . . . . . . . ._. 149

48 Stereoview of a Hirudin-Thrombin Water Mediated
Interaction. Hirudin residues in bold. Hydrogen
bonds indicated by dashed lines, ion pairs by +/—,
the water molecule by * . . . . . . . . 153

xvii

4..
:4

[Ho
pKv

FIGURE PAGE

49 Stereoview of Hirudin Pro46’-Lys47’-Pro48' and the
Hirudin N-Terminus—Thrombin Active Site Interaction.
Hirudin residues in bold. Hydrogen bonds indicated
by dashed lines, ion pairs by +/-, and water

molecules by * . 154
50 Charged Residues of the Anion-Binding Exosite

Region of Thrombin. Thrombin CA structure only with

thrombin disulfides and charged anion-binding exosite

residues in bold . . . . . . . . . . . . . . . . . . . . . 156
51 Basic Residues of the Postulated Heparin

Binding Site of Thrombin. Thrombin CA structure only

with thrombin disulfides and basic residues of

heparin binding site in bold . . 157

52 Hirudin(Glu49'-Ser50’-Hi551')-Thrombin
Interactions. Hirudin residues in bold. Hydrogen
bonds indicated by dashed lines, ion pairs by +/~,
and water molecules by * . . . . . . . . . . . . . . . . . 159

53 Electrostatic Hirudin(Asp55'-Gln65')-
Thrombin Interactions. Hirudin residues in bold.
Hydrogen bonds indicated by dashed lines, ion pairs
by +/-, and water molecules by * . . . . . . . . . . . . . 160

54 Hirudin Phe56' in a Hydrophobic Thrombin
Cavity. Phe56' in bold . . . . . . . . . . . . . . . . . . 162

55 Hirudin-Thrombin Interactions near the
Type III Helical Reverse Turn of Hirudin. Hirudin
residues in bold, with N, CA, G of hirudin bolder.
Ion pairs indicated by +/— . . . . . . . . . . . . . . . . 164

56 Space-filling Drawing of the rHV2-K47-Human
apThrombin Complex. Hirudin in sky blue . . . . . . . . . 168

57 The Sequence of Human asThrombin.

Insertion residues with respect to the sequence of
chymotrypsinogen [6] are graphically indicated as
protrusions from the linear sequence chain.

Disulfide linkage connections are indicated, and

insertion residues possessing contacts of < 4.0

with hirudin in the hirudin-thrombi complex are

circled . . . . . . . . . . . . . . . . . . . . . . . . . . 171

58 Stereoview of a Hydrophobic Crystal Packing
Interaction at a Two—Fold Symmetry Axis. Top:
Molecules involved are at symmetry positions 1 and 5
(relation 42, text). Disulfides and hirudin in bold.
Two-fold axis indicated. The hydrophobic
interactions occur in the circled region and are
detailed below. Bottom: Nonpolar/aromatic residues

xviii

“'9‘."

C-’d

‘ II
\L')

\7\

'74.

(7\
‘AJ

 

FIGURE

59

60

61

62

63

64

which comprise the hydrophobic cluster. Groups
involved: Leu14G, Tyr14J, Ile14K, Leu129C, Tyr134,
Pr0204, and Phe204A (all bold) and the corresponding
symmetry- -equivalent groups . . . . . . . .

Steroview of an Electrostatic Crystal Packing

Interaction at a Two-Fold Symmetry Axis. Top:
Molecules involved are at symmetry positions 1 and 6
(relation 42, text). Hirudin and all disulfides are
in bold. Two—fold axis indicated. The electrostatic
crystal packing interactions are within the circled
region and are detailed below. Bottom: The double
hydrogen—bonding ion pairs involved in the crystal
contacts. Dashed lines indicate hydrogen bonds;
hirudin glutamate residues are in bold. Two-fold
symmetry axis offset approximately 30 degrees from
the perpendicular to the plane of the paper .

Stereoview of a Crystal Packing Interaction

Involving the Thrombin Lysl45-Gly150 Undecapeptide
Loop. The crystal contact involves molecules at
symmetry positions 1 and 7 (relation 42, text).
Residues involved: Lysl45, Trp148, Thr149, Asn149B,
Va1149C, Gly#1D, A1a#lB, Leu#123, Lys#235, Gln#244,
and Phe#245. N, CA, C atoms and residues of the
Lysl45-G1y150 loop engaged in hydrogen bond
interactions in bold; hydrogen bonds indicated by
dashed lines . . . . . . . . . . . . . . . . .

Stereoview of the Optimally Superimposed

Thrombin CA Structures of the Complex and
PPACK-Thrombin. Superpositioning corresponds to the
MOLFIT N, CA, C fitting results of Table 27.
Residues of the complex in bold . . . . . . . . . . .

Average R.M.S. Deviation of Main Chain Atoms

versus Residue Number for the Superimposed Thrombin
Molecules of the Complex and PPACK-Thrombin.
Superpositioning as in Figure 61. The horizontal
lines represent lo and 2a deviations. Most of
residues with deviations > lo are identified

Average R.M.S. Deviation of Side Chain Atoms
versus Residue Number for the Superimposed Thrombin
Molecules of the Complex and PPACK-Thrombin.
Superpositioning as in Figure 61. The horizontal
lines represent lo and 20 deviations. Many of the
residues with deviations > Is are identified

Stereoview of the Superimposed Residues

LysZ36-Phe245 of the Complex and PPACK-Thrombin.
Superpositioning as in Figure 61. Residues of the
complex in bold. Symmetry molecule interactions of

xix

PAGE

. 179

. 181

. 184

. 187

. 189

. 190

—' P."
. l

god"

4'.
(l)

I 1'»
‘LJ

‘rx

 

 

 

FIGURE

65

66

67

68

69

70

71

the complex displayed; symmetry molecules indicated
by a '#' preceding the residue number. Hydrogen
bonds indicated by dashed lines . . . .

Stereoview of the Superimposed Residues
Glu146-Gly150 of the Complex and PPACK-Thrombin.
Superpositioning as in Figure 61. Residues of the
complex in bold . . . . . . . . . . . .

Stereoview of the Superimposed Residues

Glu146-Gly150 of PPACK-Thrombin and the Complex,
including the Hirudin N-Terminal Pentapeptide.
Superpositioning as in Figure 61. Residues of
PPACK-thrombin in bold. Thrombin catalytic triad of
the complex displayed . . . . . . . . . . . . .

Stereoview of the Superimposed Residues

LysllO-Prolll of the Complex and PPACK-Thrombin.
Superpositioning as in Figure 61. Residues of the
complex in bold. Peptide bond preceding Prolll is
cis- in PPACK—thrombin, trans- in the complex .

Stereoview of the Superimposed Residues

LysZOZ-Arg206 of the Complex and PPACK-Thrombin.
Superpositioning as in Figure 61. Residues of the
complex in bold . . . . . . . . . . . . . .

Stereoview of Superimposed Active Site

Residues of PPACK-Thrombin and the Complex, including
the Hirudin N-Terminal Tripeptide. Residues of the
complex in bold. Superpositioning as in Figure 61.
Hydrogen bonds indicated by dashed lines . . . . .

Stereoview of the Superimposed Anion-Binding

Exosite Residues Arg75-Lys81 of the Complex and
PPACK-Thrombin. Additional residues of the complex
(thrombin Ala113, hirudin G1u57'-Glu58', and the
water molecule V467 ) included. Residues of the
complex in bold. Superpositioning as in Figure 61.
Hydrogen bonds indicated by dashed lines . . . . . .

Stereoview of Residues Leu41 and Phe60H-Leu64

of the Complex and PPACK-Thrombin. Residues of the
complex in bold. Superpostioning as in Figure 61.
Hydrogen bonds indicated by dashed lines . . . . .

XX

PAGE

. 198

. 200

. 202

. 205

. 207

208

. 211

213

Ffl’fifl
nun '

‘Hv—

ucgh'

5-,..- .

1-4»0

sit;

fetid

~‘HUI

 

 

I INTRODUCTION

A. Thrombin

aPThrombin is a glycoprotein serine proteinase [1] which is
generated in the penultimate step of the blood coagulation cascade. In
this step, prothrombin (PT), a zymogen of molecular weight 74,000, is
’activated' by limited proteolysis in the prothrombinase complex [2] to
form thrombin. The complex, schematically represented in Figure 1, is
composed of PT substrate, the proteolytic enzyme Factor Xa, the
membrane—bound cofactor Factor Va, the phospholipid surface, and Ca+2
ions. A representation of the prothrombin activation products in bovine
and human systems is shown in Figure 2 [3]. During bovine PT
activation, Factor Xa cleaves PT at Arg-Thr and Arg-Ile bonds, sites "b"
and "c" respectively in Figure 2, to produce PT fragment 1-2 and
dethrombin. In the human system, a-thrombin additionally cleaves at
sites "a” and "b'" to produce PT fragment 1, PT fragment 2, and
arthrombin. The result is that the A chain of human thrombin is 13
residues shorter than the A chain of bovine thrombin.

Human arthrombin, the thrombin form used in this study, is a two
chain molecule which possesses a molecular weight of approximately
36,600 [4,5]. The sequence of human aethrombin is given in Table 1 [6].
The A chain contains 36 residues and is covalently linked to the B chain
through a disulfide (1-122). The B chain is 259 residues long and
possesses 3 intrachain disulfide linkages (42-58, 168-182, 191-220).
Thrombin is a serine proteinase by virtue of the B chain, which is
highly homologous to other coagulation/fibrinolytic serine proteinases
(Factor IXa, Factor Xa, protein C, urokinase, tissue plasminogen

activator, and plasmin) and is approximately 50% homologous with the

Figure 1.

PROTHROMBIN DIMER

 

The Prothrombinase Complex. Prothrombin (PT) is composed

of fragment 1 (F1), fragment 2 (F2), and prethrombin 2
(Pre-Z); Factor Va . Va; factor Xa s Xa; Ca+2 ions are
represented by filled circles binding to y-carboxyglutamic
acid residues.

 

 

 

 

 

Prothrombin
a b' c
cao cno I : one
I
AL‘. 1 I ;SER jTHR ' ILE—-—L—-5ER
V
L‘ Prothrombin ‘L‘
"7 fragment 1 '1‘

 

Prethrombin l———]

‘47Prothrombin‘

1‘
l'

f 2 71‘ Prethrombin 2———o]
ragmenr

In——-Prothromb in fragment l-Z—o‘

le-A :4: 3 —'
L—sr-s ,,

o-Thromb in

Figure 2. The Products of PT Activation. CHO represents a carbohydrate
side chain; "a", "b", and "c" represent the sites of cleavage

of bovine and human PT; an additional cleavage, "b'" occurs
in human PT (taken from [3]).

u
._.v"

m}
N145

 

Table 1.

The Primary Sequence of Human a-Thrombin.
is indicated using

numbering
with chymotrypsinogen.
character

cis-proline.

the

residues

indicates

The amino acid type

(Numbering taken from [6].)

Human u-Thrombin A-chain

T1H
L3
313
8141

316
R4
014
Y14J

61?
P5
K14A
114K

513
L6
T148
014L

Human ot-‘l'hrombin B-Chain

116
M26
x36
845
A55
0603
V56
Y76
L85
was
1104
:11:
p124
0131
w141

N149!

0156
P166
1176
K185
C191
M201
0209
C220
T229
0239

V17
$27
S36A
L46
A56
K60?
R67
377
386
N96
L105
8115
0125
A132
6142

V1496

V157
V167
T177
P186
3192
K202
M210
D221
H230
K240

318
P28
CP37
147
H57
N606
168
R773
K87
R97
M106
0116
3126
6133
N143

61490

V158
C168
0178

0186A

6193
S203
6211

R221A

V231
V241

619
N29
038
848
C58
F603
669
N78
188
397A
K107
1117
3127
1134
L144

K1493

N159
K169
N179

3186B

D194
P204
1212
D222
F232
1242

610
F7
314C
614M

520
030
E39
D49
L59
T601
K70
179
189
N98
L108
1118
T128
K135
K145
6150
L160
D170
M180

61860

$195

F204A

V213
6223
R233
0243

standard single letter code.
based on topological similarities
A number followed by an alphabetic
insertion. CP represents
31C A18 013 C1 62
88 K9 K10 811 L12
R140 3143 L143 L146 E14H
R15
D21 A22 B23 124 625
V31 M32 L33 F34 R35
L40 L41 C42 G43 A44
R50 w51 V52 L53 T54
L60 160A P608 P606 W600
361 N62 D63 L64 L65
H71 872 R73 T74 R75
380 K81 182 583 M84
190 R91 P92 R93 194
L99 0100 R101 0102 1103
K109 K110 P111 V112 A113
H119 P120 V121 C122 L123
A129 A129A. $1298 L129C L130
6136 R137 V138 T139 6140
3146 T147 N148 T149 A149A
0151 P152 8153 V154 L155
P161 1162 V163 8164 R165
$171 T172 R173 1174 R175
P181 C182 A183 6184 Y184A
K1860 R187 6188 0189 A190
6196 6197 P198 P199 V200
N2048 N205 R206 W207 1208
8214 N215 6216 8217 6219
K224 1225 6226 F227 Y228
L234 K235 K236 W237 1238
0244 P245 6246 B247

4

. ~Ai~
Vb:
-»-n¢

 

. _ cu. I.
.5 .. ._ ”n a. 1.. A... .. fie ... cc 2. .. a. .6. 4. v. u . .
. n. .u E r. . . .. a Q. 2‘ ‘ y. .HJ . y. Y. A.
. . _ . . . . C A. p. E y: .1 .5. [NC 63 s a ax E
.o. .6. Ce _ n. I. .. we a. . . n. 0. Ce 5-. .7. ?~ A». n. a:
. . a” _.. . ~ J . .2. C . .. .t p . rJ s... pf n0. 5». e... 2.

digestive serine proteinases trypsin, chymotrypsin, and elastase [7].
The 8 chain of thrombin contains the three invariant catalytic site
residues — Hi557, Asp102, and Ser195 (see Table 1) - commonly called the
catalytic triad. Finally, the thrombin B chain also contains a single
biantennary chain carbohydrate which is linked to Asn60G; the structure
of the carbohydrate moiety is given in Figure 3 [8].

The ultimate step of blood coagulation is also the primary function
of thrombin: the conversion of fibrinogen to fibrin. Fibrin forms the
network which encapsulates the hemostatic plug or blood clot during
wound healing. Fibrinogen is a large, dimeric six chain molecule with a
molecular weight of approximately 340,000 and a overall length of about
450 A [9]. A schematic drawing of fibrinogen is shown in Figure 4 [9].

Fibrin assembly proceeds in two stages [10]. In the first stage,
fibrin monomers polymerize end-to-end to form fibrin I polymers or
protofibrils. In the second stage, protofibrils associate laterally to
form fibers. The first stage of fibrin assembly is initiated by the
cleavage of two Aa chains of fibrinogen at 16Arg—17Gly with the
subsequent release of 2 molecules of fibrinopeptide A or FPA [11]. FPA
release leads to an unfolding of a polymerization domain in the
NHZ-terminal portion of fibrinogen and exposes a complementary portion
of the 7 chain, which exists in intact fibrinogen [12]. This
FPA-deficient fibrinogen molecule is known as the fibrin monomer.
Fibrin monomers polymerize via the available polymerization sites to
produce fibrin I polymers or protofibrils. The resulting fibrin I
polymers are comprised of fibrin monomers laid end-to-end with an
overlap equal to about 225 A or one—half the monomer length [13].

The second stage of fibrin assembly involves the lateral

association of protofibrils to form fibers. A fiber may contain over

NeuNAcaZ-6Galal-4GchAcal-2Manal Fuc a}
3 6
Han al-4GlcNAc31-4GlcNAc-l-N-Asn
6

leuMAcoZ-GGalIl-4GlcNAcal—2Menol

witffij b(§

ONHO OW HO
COOH
HOCHZJV OH GALACTOSE
N-ACEWLNEURAMINIC
ACID
CH20H H
H o \ OH H H0 H
o n- '
N—ACE TY LGLU COSAM l N E MANNOSE
H\ C//o
Hoin
HA...
.10..
Hot.
H3
L-Fucose

Figure 3. The Polysaccharide of Human a—Thrombin [8] and its Component
Hexoses.

 
 
 
  
   
 
    
   

   

 
 

  

1501 CENTRAL 1501
TERMINAL »——————‘1 Dow... .————————-4 Iggms
DOMNN (mwazwm
(uw 67.200) t_.———.
. wk
60A
FPS
caoru»
tuwzxm) SS mNG .J, :r “"-H
"L PROTEASE (“33:35} 160 A
‘7’ , sausmvs
5”55 coueo cou
(MW 39.100)

(W 42.300)

Figure 4. Schematic Drawing of the Fibrinogen Molecule [9].

The symmetric molecule is composed of a dimeric central
domain containing the amino—termini of all six chains
(a,8,y), two connecting coiled coils, two terminal domains,
and two Au polar appendages. Fibrinopeptides A and B (FPA
and FPB) are at the amino-terminal ends of the u and B
chains. Four carbohydrate clusters (CHO) occur, and are
located on each 7 chain near the central domain and on the B
chains of each terminal domain. Primary cross-linking sites
(XL) can be found near the carboxy-termini of the y chain and
in the Au polar appendages.

.v

v -

il.

‘8'

_.4

3.

3

PC

nb.

11

n?-

v‘

 

 

100 protofibrils [14] and it is these fibers which join irregularly to
produce the gel-like mass characteristic of in vitro clotted fibrin
[15]. The lateral association of fibrin I polymers is accompanied by
thrombin cleavage of the BB chains at l4Arg-15Gly in these polymers and
the subsequent release of fibrinopeptide B or FPB. The FPB—deficient
fibrin I polymers are known as fibrin II polymers. Release of FPB
unfolds another unique polymerization site [11] and results in an
interaction of polymeric species and in polymers and fibers of increased
strand width. It has been demonstrated that although FPB release is not
necessary for fiber formation or stage II fibrin assembly, the release
of FPB results in more rapid fiber formation [10]. Once fibrin assembly
is complete, the network is covalently stabilized by the action of
Factor XIIIa (fibrin stabilizing factor), which intoduces y-glutamyl
s-lysyl cross-links in fibrin [16].

The role of thrombin in hemostasis is by no means limited to the
activation of fibrinogen to clottable fibrin. Thrombin has been called
the central bioregulatory enzyme in hemostasis [17]. Thrombin regulates
hemostasis on three different levels - the fluid or plasma level, the
blood cell or cellular level, and the blood vessel or vascular level
[18]. Moreover, thrombin accomplishes these tasks by acting as a
positive feedback activator of coagulation [7] as well as a negative
regulator [19]. The central bioregulatory functions of thrombin in
hemostasis are indicated in Figure 5 [18].

On the fluid or plasma level, thrombin proteolytically cleaves
several zymogens to produce active enzymes or stimulate activation [17].
Thrombin activates Factor VIII (antihemophilic factor) [20] and Factor V
(accelerator globulin) [21] via limited proteolysis which results in

positive feedback to the coagulation pathway. Thrombin also activates

   

‘
fiET

Q

 
   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
       
    
 
    
 
    

 

I
I
I
: Kuuuuxus‘vm ,
l I Rm GEN
XII 1:le F8 0
it ”hum“
( xuo ( )Xle -I'IIXo
I CLOT
A a-rIIRouem BARRIER
qu-/"a_°:::-.-.::- ‘ 1' F
I rfl‘ ,/ PA
. KININS vm' I PFOTEINCO . "mam“
PREKALUKREIN TPN .
.. In.
H4 PROTEIN C/ / (FlBRlN-Z)
Lrummtm mum XIII marl
l l (INSOLUBLE)
Pusumocrn __‘ L “9""
L l _ _.p..s..... .Tl c...
BARRIER
r
'55”: PLASM'NOGE" FIBRIN SPLIT PRODUCTS

 

ACT I V'ATORS

CWPLEMENT
SYSTEM

ACTIVATED
NEUTROPHILS/MONOCYTES

vtssrr VESSEL
WALL LUMEN

Figure 5. The Central Bioregulatory Functions of a—Thrombin

in Hemostasis. Coagulation factor zymogens are identified by
roman numerals; an "a" following a numeral indicates an
activated enzyme. Primes and double primes are used to
indicate active and inactive forms of factors of V and VIII.
Abbreviations: F1.2 = prothrombin fragment 1.2; FpA =
fibrinopeptide A; FpB = fibrinopeptide B; PL = phospholipid;
and TPN = thromboplastin (taken from [18]).

"..o)\v

._...rt

’A‘ L‘C

. v”-
.p In

untfinri
n

«usex'v

' A
".‘h-
’va-u
il.‘ _
'P_I-\I
o-.II.‘,‘
V.
. a ,.
p
7‘ dri

 

“"10{
u."“

’c‘ ,.

‘~“h

-.
5".
L‘.‘ .

 

 

"1
(D
r 1

 

 

10

Factor XIV (protein C) by proteolytic cleavage; Factor XIVa is believed
to be responsible for the inactivation of Factors Va and VIIIa [22,23];
in this capacity thrombin is no longer a procoagulant but rather an
anticoagulant. Thrombin cleavage of Factor XIII (fibrin stabilizing
factor) produces Factor XIIIa'b which becomes an active enzyme in the
presence of Ca+2 ions [24]. The conversion of fibrinogen to clottable
fibrin also occurs at this level.

The endogenous proteinase inhibitors antithrombin III, heparin
cofactor II, al-proteinase inhibitor, and a2-macroglobulin all inhibit
thrombin on the plasma level [25]. Antithrombin III is the primary
inhibitor of thrombin in plasma. Antithrombin III and heparin cofactor
II are unique in that they display accelerated thrombin inhibition in
the presence of heparin and other glycosaminoglycans [26,27]. The most
potent naturally occurring plasma-level inhibitor of thrombin is
hirudin, a 65-66 amino acid peptide isolated from the salivary glands of
the European medicinal leech Hirudo medicinalis [28].

On the cellular level, thrombin is an agonist for a variety of
cellular activities in many tissues and cell types [29].
Thrombin-cellular interactions stimulate aggregation (platelets),
chemotaxis (monocytes), contractility (smooth muscle), proliferation or
cell growth (fibroblasts, macrophages, splenocytes), prostaglandin
synthesis (platelets, endothelial cells, fibroblasts, neuronal cells),
and secretion (platelets, endothelial cells). The induction of platelet
aggregation and secretion is a very important procoagulant cellular
function of thrombin. On the other hand, a particulary important
anticoagulant thrombin-cellular interaction involves the thrombomodulin
receptor on endothelial cells. Thrombomodulin is a 74,000 dalton

protein which forms a stoichiometric 1:1 noncovalent complex of high

:;;€ni‘

.‘..n

Ru 'D‘
'3 kill

TYOTEIY

hhrqON

.II.-- J

thrive-I
,llb a A V
p
e
A
'HrA-n
ustg
I l

.
-njn'w
n - n

 

A
u)“
\

 

 

11

affinity (Kd = 0.48 nM) with thrombin. Moreover, thrombin is inhibited
by thrombomodulin and this interaction also stimulates activation of
protein C which can in turn inactivate factors Va and VIIIa. Thus, the
thrombin—thrombomodulin interaction is yet another means by which
thrombin negatively regulates coagulation.

Functions of thrombin on the vascular level also involve
thrombin-cellular interactions [17]. Thrombin interacts with the
endothelial cells of the blood vessel wall. This interaction results in
the synthesis of prostacyclin [30], prostaglandin [30] and
platelet—activating factor [31], as well as the release of von
Villebrand factor [32], and tissue plasminogen activator (tPA) and its
inhibitor [33].

The mechanism by which serine proteinases cleave peptide bonds can
be described as a general base-catalyzed nucleophilic attack on the
carbonyl carbon of the substrate by the hydroxyl oxygen of Ser195
[34,35]. The x-ray crystal structures of the digestive serine
proteinases trypsin [36], a—chymotrypsin [37], and elastase [38]
indicate that prior to peptide bond cleavage, the sidechain of an amino
acid on the carbonyl-side of the bond to be cleaved fits into a
'specificity pocket' which brings the carbonyl carbon near to Ser195.
Moreover, the nature of the binding pocket reflects the specificity of
the enzyme [39]. The specificity pocket of archymotrypsin is large and
hydrophobic; thus, a—chymotrypsin is well-suited to cleave peptide bonds
after tryptophan, phenylalanine, tyrosine, and leucine sidechains. In
trypsin, the specificity pocket fis similar in size but it has an
aSpartate (Asp189) at the base of the specificity pocket as opposed to a
serine. This aspartate deprotonates above about pH 5.0 and thus makes

the cavity ideal for the binding of arginyl and lysyl sidechains. The

12

fact that thrombin is 'trypsin-like’ in its ability to cleave Arg-X,
Lys-X bonds and that it contains an aspartate at position 189 suggests
that thrombin and trypsin share a similar specificity pocket [1].

The diverse functions and regulation thrombin provides in
hemostasis, however, clearly indicate that it is a unique trypsin-like
serine proteinase which requires more than a specificity pocket to
govern its selectivity. Vhile trypsin is omnivorous in its choice of
protein substrates and in the number of peptide bonds it will attack
[1], thrombin is highly selective in its choice of macromolecular
substrates and interactions. Thrombin achieves such high selectivity by
using active site region subsites, and exosites which are distinct from
the active site [17,40-46]. Moreover, it is these exosites in addition
to the catalytic site region which allow thrombin to function in
hemostasis by means of enzymatic (proteolytic), nonenzymatic (hormonal),
and coupled mechanisms [47]. Coupled mechanisms require the involvement
of the thrombin active site and one or more exosites.

The high degree of specificity of a—thrombin interactions with
macromolecular substrates is believed to result from binding at three
distinct regions on the thrombin surface: the primary binding or
specificity pocket [1], an apolar binding site adjacent to the catalytic
site [40,41], and an anion-binding exosite which is responsible for
fibrinogen recognition [42,44]. The anion—binding exosite appears to
contain or be in the vicinity of the B—cleavage loop (residues
Arg67-Arg77A [48]) of dethrombin [44]. The integrity of this loop is
important to the interaction of thrombin with fibrinogen,
thrombomodulin, and protein C [49-52]. Additional secondary binding
Sites have been identified for heparin-binding [45] and growth factor

activity [46].

a
. .0!
r_ ‘

.soU~‘

o-AII

 

,I. .II‘v J

'C‘,"
Is»... .

T"""

~..4;.

'Vfiht‘ "
\

-.vV—v

9. '“vn
....,

 

 

 

 

13

In an effort to truly understand the structure-function
relationships underlying thrombin specificity, this enzyme has been the
focus of x—ray crystallographic studies for some time [6,52-55].
Although crystals of bovine [52] and human thrombin [53] had been
reported for well over ten years, no protein structure has resulted from
these crystals. In the absense of definitive crystallographic
structures, three-dimensional models of the thrombin 8 chain were
proposed based on the sequence homology with bovine chymotrypsin and
trypsin [56,57]. Recently however, nearly identical crystals of human
a-thrombin inhibited by D-Phe-Pro-Arg-chloromethylketone (hereafter
referred to as PPACK) were obtained by us [55] and independently at the
Max-Planck-Institut fur Biochemie in Martinsried, Germany [6]. The
German research group led by Dr. Volfram Bode and Professor Robert
Huber eventually solved the structure of PPACK-inhibited thrombin at 1.9

A resolution [6].
B. Hirudin

Hirudin is a small protein which occurs in the salivary glands of
the the European medicinal leech, Hirudo medicinalis and is the most
potent thrombin inhibitor known [28]. The complex of hirudin with
arthrombin is high-affinity, noncovalent, stoichiometric (1:1) and has a
dissociation constant in the picomolar to femtomolar range [59,60].

Crude hirudin from the leech is a mixture of polypeptides
containing 65 or 66 amino acids which possess a molecular weight of
approximately 6900 daltons [61]. The first three hirudin sequences to
be determined were classified as hirudin variant 1 [62], hirudin variant
2 [63], and hirudin variant 3 [64] based on their chronological

appearance in the literature [65]; these sequences are often referred to

14

.A_oo_ souu :oxmuv enamouau

moonwasm m moumofivca .» .mo cowuwmoa um osvwmeu Hmcowuwvvm

cm mcwmucoo m>m .mesvwmeu acmwum> 0» Home» AIV madam egg was

smoHoeo: uo mnemmou oueocfiaoo moxon one .m>= use .~>= .~>=
”mucmwmm> sauna“: amusumz ooush may «0 moucosaom ofioc ocma< .o ousmfim

 

 

 

 

 

 

 

 

 

 

 

a 6 6C
zIzImIo xImIeIoImIoIaI>IUIOIz.qu no:
zImIqu z-AIRIoImIoIaISIUIOIzioIa «on
znzImIo aImIanImIoIaI>IUI0Iz-aIa up:

on me am mm

a a 4‘.
IOImIoIsIHIoIaIzIo onIorsuzImIquIoIqIoIqIzIOIoImImIaIUIoIaIsIaIH no:
IzImIoIoIHIOIzIzIoIzIoIoI>IzImIquIoIqIoroIzIOIoIquIaIUIoIaIquIu a»:
IoImIoIQIHIUIzIquIo oIUI>IzImIoImIUIuIUIaIzIoroImImIaIUIoIaIsI>I> up:

 

 

 

 

 

 

mm 0m .1w c~ mu on I m A

rhy Fl.
I CL r a
9 .. . D.
.N . «Au EU.

I l . h...“ a:

 

\_

'I

15

as HV1, HV2, and HV3, respectively, and are given in Figure 6 [66]. The
hirudin sequences are approximately 80% homologous and contain only 13
variable positions [67]. The amino acid compositions also reveal a high
content of asparagine (N), glutamine (Q), aspartate (D) and glutamate
(E), as well as the absence of arginine (R), methionine (M), and
tryptophan (V) [67]. The COOH-terminal region of these hirudin
sequences is rich in acidic residues; nearly half of the last 11
homologous amino acid positions are occupied by an aspartate or
glutamate residue. A unique acidic residue found in all three hirudin
variants is the sulfated tyrosine (Y’) in the antepenultimate position
[62,68,69]. This tyrosine acyl sulfate enhances the ability of hirudin
to inhibit thrombin; desulfato-hirudins display three-to ten-fold
decreased affinity for thrombin with respect to the sulfated forms
[59,60,70,71]. Presently, twenty naturally—occurring hirudin sequences
have been determined, and they display an overall sequence homology of
approximately 66% [61]. .

Hirudin variants also share a common disulfide connectivity
pattern. All hirudin molecules contain six conserved cysteine (C)
residues in the NHZ-terminal region which are involved in 3
characteristic cystine linkages: Cys6-Cysl4, Cys16-Cys28, and
CysZZ-Cys39 [72]. Interestingly, all known serine proteinase inhibitors
can be grouped into one of ten structural families based on sequence
homology, and the topological relationship among disulfide bonds and the
inhibitor reactive site [73]. However, hirudin displays no topological
or sequence homology to any known serine proteinase inhibitor and thus
could represent a heretofore unknown family of inhibitors [72]. Hirudin
does share some global features with epidermal growth factor (EGF),

which is a structural motif found in numerous proteins [74]. Alignment

16

..a@. Eczema gauche Haauoeaam
van cacao“: mo mcueuumm zua>uuoooooo econ okuasmua one

.5 ousmflm

 

 

sauna“:
I m 3. we mm mm mm
mm UH Mr M. .— —

 

Louuem guaouu Heeuevfiqm

NV mm PM ON 3.

I

 

m
—I Ia mm. I— a. H I

F

 

17

of the hirudin and EGF disulfide linkage patterns in opposite directions
reveals a similar connectivity, as is seen in Figure 7 [67]. It is of
note, however, that hirudin and EGF share no functional or sequence
homology.

Because leeches contain mixtures of naturally occurring hirudin
variants which are difficult to purify and isolate in large quantities,
recombinant DNA technology has been employed to obtain homogeneous
preparations in ample amounts. To date, five different microbial
expression systems have been made to produce recombinant hirudin variant
1 (rHVl) [67,75—77] and two systems have been developed to produce
recombinant hirudin variant 2 (rHV2) [63,65]. Recombinant hirudins
differ from their natural counterparts in that they lack the sulfated
tyrosine. The source of hirudin used in this study is the recombinant
hirudin variant 2-Lys47 mutant (rHV2-K47) resulting from a yeast
expression system [65]. The sequence and covalent structure of rHV2-K47
are given in Figure 8.

Secondary and tertiary structural information on hirudin in
solution has been obtained independently by two research groups
employing 20 NMR techniques to study recombinant hirudin variant 1
[78-80]. Qualitatively, the results of both groups are in general
agreement. These studies revealed that hirudin is a two domain protein
characterized by a compact, well-defined NHZ—terminal core and a
flexible, disordered COOH-terminal tail. The core domain (residues 3-30
and 37-48) contains the cystine linkages, several turns and two
antiparallel B-strands. Residues 31-36 within the NHZ—terminal domain
and residues 49—65 of the COOH-terminal tail exhibited no preferred
conformation in solution. A stereo drawing displaying nine optimally

superimposed DISMAN distance geometry solutions resulting from the 20

 

Figure 8. The Sequence of Recombinant Hirudin Variant 2-Lysine 47
(rHV2-K47).

.v.
‘6.

Figure 9.

19

 

Stereo Drawing of Nine Optimally Superimposed DISMAN Distance
Geometry N, CA, C Structures from the 20 NMR Constraint Data
of Haruyama and Vuthrich [80]. Disorder exists in the
residue ranges 1-3, 31-36 and 49—65 (taken from [80]).

‘.--AF“

\\
..-~.

new-91‘

5.”!

l

v.5

3‘1
- .
M...

Or.

a-

[H‘I

.I...

fish-0‘!

f‘

"-04

»A'_-

..~

 

a:

A
w

"‘e.

:5.
u...

«\—

20

NMR constraint data of Haruyama and Vuthrich is shown in Figure 9 [80].
Hirudin reacts rapidly with human a-thrombin (kon > 1 x 109 M‘ls'l)
[59] and forms an amazingly high-affinity, noncovalent complex with a
dissociation constant estimated at 80 pM [28] and 20 fM [59]. The apolar
binding site and the anion-binding exosite or fibrinogen recognition
site of thrombin appear to be important thrombin determinants in complex
formation. Hirudin binding to human a~thrombin displaces proflavin from
the apolar binding site [41]. In addition, the dissociation constant of
hirudin and PPACK-thrombin, in which PPACK alkylates the active site
histidine and appears to bind in the apolar binding site [40], is 1 x
106 higher than with a-thrombin [81]. Both of these observations
implicate the apolar binding site in the hirudin-thrombin interaction.
Evidence supporting the importance of the anion-binding exosite of
thrombin to its interaction with hirudin has come from binding studies
involving the proteolytic derivatives of a-thrombin: B—, y—, and
s-thrombin. Schematic drawings of human a-thrombin and these
proteolytic derivatives, with the cleavages indicated, are given in
Figure 10. B—Thrombin and y-thrombin are produced by autolysis of
native a—thrombin or by limited proteolysis with trypsin [82].
B—Thrombin generated by autolytic degredation is characterized by a
cleavage at Arg77A followed by Arg67, and by the concomitant loss of the
Ile68-Arg77A undecapeptide [48] (Figure 10 and Table 1). The B—thrombin
derivative produced by trypsinolysis, designated BT-thrombin, contains a
single cleavage at the Arg77A-Asn78 peptide bond [52]. The exact
covalent structure of y-thrombin appears to depend on the method of
preparation. Boissel et al. [48] report y-thrombin to possess
cleavages at Arg67 and Arg77A, as in B—thrombin, and additional breaks

at Arg126 and Lysl49E, followed by the loss of the resulting 11- and

 

21

 

 

 

 

 

 

 

 

 

 

1H 15
(:1 l
T H57CHO A01: 5195 “I?
BI rises 6122 l D 0L
Al I
cuo l j
I . I
Bl RS7 thA 7 lg
Al |
one I l
8' :---4 4 ----- .2 I
V R67 RNA R120 RI49E ' 2}

 

.. Al I

 

 

Figure 10.

l
l
A i4” I Q

Human a-Thrombin and its Proteolytic Derivatives.

The numbering of residues is taken from Table 1. The
A-chain and B-chain of thrombin are designated by "A" and
"B": respectively. CHO represents the polysaccharide linked
to nitrogen N02 of Asn(N)6OG. The location of the catalytic
triad residues is noted (H57, 0102, and $195). Excised
proteolytic fragments are denoted by broken lines.

22

31-amino acid peptides (Figure 10 and Table 1). Fenton et al. [82], on
the other hand, have reported a y-thrombin with only two major cleavage
sites - Arg77A and Lysl49E. s-Thrombin results from limited proteolysis
with elastase and contains a single cleavage at the Ala149A-Asn149B bond
(Figure 10 and Table 1) [52]. Hirudin complexes with B-thrombin and
v-thrombin, proteolytic derivatives of a—thrombin which possess little
or no fibrinogen activity (0.05%) due to a disrupted anion binding
exosite [52], have significantly higher dissociation constants with
respect to the a-thrombin-hirudin complex [81]. e-thrombin, which
displays only about 40% diminished fibrinogen activity with respect to
native thrombin [52], in complex with hirudin has only a slightly
increased dissociation constant [81]. Hirudin complexes with native
human a-thrombin as well as with some of its chemical and proteolytic
derivatives are not retained on nonpolymerized fibrin-agarose resin in
contrast to free thrombin [44]. The importance of the hirudin
interaction with thrombin at the anion-binding exosite is further
emphasized by the observation that an antibody raised against a peptide
corresponding to the B—cleavage site undecapeptide, Ile68-Arg77A, has
been shown to inhibit hirudin binding to a—thrombin [83].

Hirudin is unique among macromolecular substrates for thrombin, and
serine proteinase inhibitor proteins in general [84], in that
interaction with the primary specificity pocket of thrombin does not
appear to be crucial for inhibition. A hirudin lysyl residue was often
thought to interact with the thrombin specificity pocket, especially
Lys47; residues (39-48) of hirudin display 50% homology with residues
(148-157) of PT, which contains an a-thrombin cleavage site utilized in
human prothrombin activation [85] (cleavage site "a" in Figure 2). When

the basic residues on hirudin were altered by site-directed mutagenesis

533

51

j .

.1.

,
7130.

I

upDA
,-
.n..h

;..n,.

I‘V'ab

 

 

 

 

3“
F'
In“?-

as

C

~.U

8‘!

A

 

23

and the binding of the resulting mutants was tested with bovine and
human thrombin, only mutation of Lys47 was found to be sensitive
[86,87]. However, the increase in the dissociation constant was only
minor: four- to ten-fold . Moreover, the fact that a Lys47 is not
conserved in natural hirudin sequences further suggests that this
interaction is not necessary [61].

The NH2- and COOH-terminal regions of hirudin both contain residues
important to the hirudin-thrombin interaction. The electrostatic and
acidic nature of the carboxy-terminal tail appear to be vital. Removal
of the last 7 COOH-terminal amino acids of hirudin - which includes two
Glu residues and a sulfated Tyr in natural hirudin - results in a loss
of about 90% of inhibitory activity, and removal of the last 22 amino
acids nearly abolishes inhibition [70]. Single and multiple mutations
of the carboxy-terminal glutamates (57,58,60 and 61) to glutamine
caused increases in the dissociation constant of the complex which
increased with the number of mutations [87], and each negatively charged
Glu appears to contribute equally-favorably to the binding energy [88].
Desulfato Tyr63 hirudin-thrombin complexes possess a higher dissociation
constant compared to those from sulfated forms, albeit only three- to
ten-fold higher [59,60,71]. The hydrophobic residues in the anionic
tail also appear to be involved in the interaction. In inhibition
studies with a hirudin COOH-terminal peptide mimetic corresponding to
residues 55-64 , which is the minimum size required to abolish
fibrinogen recognition, it was shown that Phe56, I1e59, Pro60, and Leu64
as well as Glu57 were all sensitive to modification [89]. In the
NHZ-terminal region of hirudin, it appears that the hydrophobic nature
of the first two amino acids, a positively charged amino terminus, and

the integrity of the three disulfides are crucial [68,90].

x
I“
(h

.
\n h
.'

In.
[F
..A.U

uh.
\ “

“I\fle

~.I..‘

‘ROVA

‘0‘...

‘0»...

‘nn."

, L
'flno.
~.u.

‘4

2,. P

 

 

 

24

In its ability to interact with thrombin at both the apolar binding
site and the anion-binding exosite, hirudin appears to be a bivalent or
'double-headed' inhibitor. Kinetic analysis of complex formation has
revealed that the first step is ionic strength-dependent and unaffected
by binding at the active site while the second step is influenced by
binding at the active site [59]. These results in light of further
studies suggest that the first step involves the interaction of the
negatively-charged carboxy-terminal tail of hirudin with the anionic
binding exosite, while the second step involves amino-terminal domain
interaction in the active site region [81,88]. In support of this
notion, it has been found in thrombin binding studies that the
COOH-terminal hirudin fragments (49-65,53-65) protect against tryptic
cleavage, while the NH2-terminal fragment (1-52) is a competitive
inhibitor of the thrombin substrate D-Phe-pipecolyl-Arg-p-nitroanilide
[91,92]. Peptide synthesis of the hirudin segment (45-65) first
confirmed that the COOH-terminal region of hirudin alone in complex with
thrombin was sufficient to inhibit fibrin formation and the release of
FPA without affecting amidase activity [93]. Since then it has been
reported that amino acid residues (56-64) and (53-64) are minimally
required for these results [89,94]. In addition, the finding that a
(56-65) hirudin peptide produces a conformational change in thrombin as
judged by circular dichroism which does not occur upon PPACK binding
suggests that the COOH-terminal tail of hirudin may trigger a
conformational change prior to binding of the NHZ-terminal region
[95,96]. It is important to note that these carboxy-terminal hirudin
lleptide mimetics in complex with thrombin have dissociation constants
Which are about 6 orders of magnitude higher and anticoagulant

aetivities 3-4 orders of magnitude lower than that of intact hirudin

en, 6
an
E?

,.

Ah
‘,
AL
V
5““‘9
add".

RR

I...

.

'5:
~1§V
‘A'
5

 

2‘ 6 “NI. 5 s H H a.

 

25

[93].

The ability of hirudin in complex with thrombin to completely
abolish the conversion of fibrinogen to fibrin has lead to a great deal
of study and interest in the use of hirudin as an antithrombotic medical
agent. Many factors make hirudin an attractive alternative to
conventional anticoagulant medical therapies. Currently,
pharmacological control of thrombin activity in vivo is attained by
means of heparins and bezamidine-type thrombin inhibitors, and both of
these treatments have drawbacks [97]. Heparins, which are sulfated
glycosaminoglycans, inhibit thrombin indirectly by enhancing the
inhibitory power of antithrombin III and heparin cofactor II [26,27,97].
Thus, heparin efficacy is dependent on the plasma levels of these
proteins. In addition, heparins can be neutralized by natural
antiheparins in the blood such as platelet factor 4, can interact with
other blood components such as platelets, fibrinolytic constituents, and
lipoprotein lipases, and can lead to hemorrhagic side effects [97].
Oral benzamidine-based anticoagulants interfere with the biosynthesis of
clotting factor proteins [97]. Hirudin, in contrast, is a selective,
direct thrombin inhibitor which is not dependent on or affected by other
plasma proteins, and which does not interfere with the biosynthesis or
action of these components [28,98]. In addition, in animal studies
modeling five pathogenic thrombosis conditions, hirudin exhibited
antithrombotic efficacy at the lowest concentrations in all five
examples compared to heparin and the best synthetic inhibitor [97].
Hirudin also displays low antigenicity, is rapidly cleared from the
bloodstream, and can be produced in ample amounts to accomodate high
medical demand through recombinant DNA methods [68,99]. Two main

problems with using hirudin as a therapeutic anticoagulant are the

CECESE

oi
owhyvv
.oUbAI

,
'3rtnl
e'u».'v’¢

 

 

26

unparalleled affinity of hirudin for thrombin (Rd in the picomolar to
femto molar range), which would be a challenge to counteract if
necessary, and the fact that hirudin blocks amidolytic as well as
fibrinolytic activities in thrombin [59]. Such concerns are not a
factor in the promotion and development of hirudin COOH-terminal peptide
mimetics for anticoagulant use. These molecules can effectively inhibit
the thrombin-catalyzed fibrinogen to fibrin conversion without
disrupting thrombin amidolytic acitivity, have significantly higher
dissociation constants in complex with thrombin compared to hirudin, and

can easily be synthesized by non-recombinant means [89,93-95].

 

w"
r. '1

""4'-.

“[IYIII‘

~dheg

130311.

“a

 

 

II EXPERIMENTAL

A. Sample Preparation

Human a—thrombin was generously supplied to us by Dr. John V.
Fenton II of the New York State Department of Health in Albany, New
York. The thrombin samples were prepared by the method of Fenton et al.
[100], and came as frozen vials which were 3.1 mg/mL in 0.75 M NaCl.
The recombinant hirudin sample, rHV2-K47, was kindly supplied to us by
Dr. Carolyn Roitsch of Transgene,S.A. in Strasbourg, France. The
rHVZ-K47 was produced by the method of Loison et a1. [65], and was
purified by the method of Bischoff et al. [101]; the samples were a
lyophilized powder of better than 95% purity.

A frozen solution of thrombin was allowed to thaw on ice. A 10-202
molar excess of rHV2-K47 was dissolved in 50 mM sodium phosphate buffer,
1 mM NaNs, pH 7.3, and was added to the thrombin solution. The
inability of several 5 uL hirudin-thrombin aliquots to produce clot
formation in a 0.5 mL human plasma sample verified that thrombin
inactivation had occurred.

The hirudin-thrombin sample was then transferred to a dialysis bag
made of Spectra/For 4, 12-14 k0 cutoff dialysis membrane, and dialyzed
against 25 mM sodium phosphate buffer, 0.375 M NaCl, 1mM NaNz, pH 7.3.
Dialysis was carried out at 4 °C for several days with several buffer
changes to affect a 10‘9 to 10'11 dilution of the excess hirudin. The
protein solution was then concentrated in a refrigerated centrifuge
using an Amicon Centricon 10 microconcentrator to approximately 6.0

mg/mL.

27

r».

.43

I:

nay

 
 

 

A..I‘

28
8. Sample Molecular Veight Estimation

Discontinuous polyacrylamide gel electrophoresis [102, 103] was
used to estimate the molecular weight of the hirudin-thrombin complex.
Slab gels were prepared using the BIORAD Mini-PROTEAN II dual slab cell
unit and the power source for electrophoresis was a BIORAD Model 200
constant voltage power supply. The gels were prepared, stained, and
destained using Hoefer Scientific Instruments protocols [104]. The
procedure used to estimate the molecular weight of electrophoresed
proteins is a Sigma Chemical Company protocol [105], and is based on the
methods of Bryan [106] and Davis [103]. Four different slab gels of
varying separation gel pore size - 7, 8, 9, and 10 %T - were prepared,

where ZT is defined as :

%T = [(mass acrylamide monomer + mass crosslinker)/ total volume} X 100
.(1)
The pore size and the ZT of a gel are inversely correlated; a large
value of ZT indicates a small pore size. Three microgram samples of the
hirudin-thrombin complex, pre- and post-dialysis, as well as of the
following calibration standard proteins were applied to each gel :
bovine milk a-lactalbumin (MV 14,200), bovine erythrocyte carbonic
anhydrase (MV 29,000), chicken egg albumin (MV 45,000), and bovine serum
albumin (MV—monomer 66,000; MV-dimer 132,000). Bromophenol blue
tracking dye was added to all of the samples. The gels were run at a
constant voltage of 200 V. It took approximately one hour for the dye
front to move the length of the gel, which concludes the electrophoretic
separation. The Rf, or electrophoretic mobility, of each protein
oligomer was calculated for each of the four gels, and 100Log(fo100)

versus percent gel concentration (Figure 11) was plotted for each

29

100Log(fo100)

QCMZ,

“90.,

 

Figure 11.

 

v

é 4 é e to
ZGel Concentration (2T)

100Log(fo100) vs. 2T Gel Concentration. The

slope for each line gives the retardation coefficient (Kr)
for the corresponding protein oligomer. Abbreviations:
L(1ysozyme), CA(carbonic anhydrase), CEA(chicken egg
albumin), BSA-M(bovine serum albumin-monomer), BSA-0(bovine
serum albumin), HT(hirudin-thrombin)

I4

:u

.b»
\ I s

 

‘v~~

“A

 

protein, where Rf is defined as :

Rf = distance of protein migration ...(2)
distance of tracking dye migration

 

The slope for each protein oligomer in such a plot is known as the
Retardation Coefficient (Kr). The negative logarithm of Kr was then
plotted versus the logarithm of the molecular weight for the standard
proteins, producing a linear plot (Figure 12). Using the
experimentally-determined Kr value for the complex (-5.4), extrapolation
from this plot gave a hirudin-thrombin complex molecular weight of
38,500. Although this molecular weight estimate for the complex is only
1900 daltons larger than that of native human a-thrombin and is about
5000 daltons less than that expected, the discrepancy could be
attributed to the difficulty to accurately estimate protein molecular

weights from gel electrophoresis.
C. Crystallization

Crystals of the rHV2-K47-human a-thrombin complex were initially
obtained by the hanging drop, vapor diffusion method [107] in an
incomplete crystallization parameter factorial search [108] using 30
(w/v)% polyethylene glycol (PEG) 4000, 0.2 M MgClz, 0.1 M sodium acetate
buffer, pH 4.5, 1 mM NaN3, and a protein concentration of about 6 mg/mL.
Crystals appeared in 5-7 days. Refinement of these conditions found
that the largest crystals resulted from precipitant solutions which
contained 27-28Z PEG 4000, 0.2 M MgClz, 1 mM NaNa, and any of the
following buffers at a concentration of 0.1 M: sodium acetate buffer,
pH 4.7 or 5.0; sodium citrate buffer, pH 5.5; and ADA
(N-l2-Acetamido]-Iminodiacetic Acid) buffer, pH 6.0. Larger crystals

(maximum dimensions 1.0 x 0.4 x 0.4 mm3) suitable for x-ray diffraction

31

“K r
/00..

CEA
CA
BSA-M
BSA-D
HT

Ofi‘DOO

l0]

 

 

I 1 $4 I I I

/0 20 50 /00 200
M. W.

Figure 12. Log(-Kr) vs. Log(Molecular Veight). Linear regression best
fit line to the standard protein retardation coefficient
data. Molecular weight (M.V.) in kilodaltons. Abbreviations
used have the same meaning as in Figure 11. The extrapolated
molecular weight estimate for the hirudin-thrombin complex
is 38.5 k0.

..:\_A

mvl
I

u,
(lv
1"
, \

I‘V'H
~4au

.7”,-
\

’Irc

v.‘

7".

I
f] I

'r!

'V

 

32

studies were obtained by macroseeding protein-precipitant solutions
[109]. Precipitant solutions used for macroseeding experiments were
identical to those mentioned above, but were 26% in PEG 4000. Crystals
selected to be used as seeds for macro-seeding in protein-precipitant
droplets were rinsed in a solution prior to transfer which was identical
to the precipitant solution of the crystallization, but which was
23.5-24.0% in PEG 4000. A photograph of a data collection-quality
hirudin-thrombin crystal is shown in Figure 13; the trace of the seed

crystal is visible in the center of the large crystal.
0. Crystal Characterization

Crystals of the complex were characterized using Ni-filtered Cu K“
x—rays. Still and precession photographs were taken using a Rigaku
RU-ZOOB Standard Rotating Anode Source with a 0.5 x 3.0 mm filament
operating at 8.8 kV power (55 kV, 160 mA), and a Charles Supper
precession camera. Ve thank Drs. Cele Abad-Zapatero, John Erickson,
and Chang Park of the Protein Crystallography Lab of Abbott Laboratories
for the use of their facility to carry out these experiments. Frame
data were also collected using a Xentronics Area Detector on a Nicolet
P3/F diffractometer operating at 1.75 kV (50 kV, 35 mA) with a fine
focus x-ray tube. The area detector data were processed using XENGEN
data reduction software. Ve would like to thank Mr. Steven Muchmore of
the Protein Crystallography Lab at The Upjohn Company for carrying out
the area detector work for us.

The unit cell parameters and relevant crystal data are summarized
in Table 2. Still x-ray pictures displayed diffraction to better than
2.5 A resolution. From (Okl) and (hkO) precession photographs (Figure

14) and/or the reduced area detector data, the rHV2-K47-human a-thrombin

33

 

 

Figure 13. Photograph of an rHVZ-K47-Human a-Thrombin Crystal.
Crystal size is approximately 1.0 x 0.4 x 0.4 mm. Note the
trace of the seed in the center of the crystal.

34

Table 2. Summary of Crystal Data of rHV2-K47-Human a-Thrombin.

Crystal system
Space group

Number of molecules
per asymmetric unit

a (A)
c (A)
Solvent fraction

Matthews number
(Vm; A3/dalton)

Tetragonal

P41212 or P43212

1
90.39(2)
l31.97(8)
61%

3.10

35

 

 

Figure 14. hkO Precession Photograph of a Hirudin-Thrombin Crystal.
The resolution limit at the edge of the photo is 3.7 A. (mu
angle = 12.0°.)

36

crystals were shown to be tetragonal, space group P41212 or P43212. The
unit cell dimensions measured with a Nicolet P3/F diffractometer using
10 high order reflections are: a = b = 90.39(2), c = l31.97(8) A.
Assuming eight hirudin-thrombin complexes (Mr = 43,500 0) per unit cell,
and a protein specific volume of 0.74 cm3/g gives a solvent fraction
(fs) of 61% in the crystals; protein crystals generally have a solvent
fraction ranging from about 27 to 65%, with 43% being the most common
value [110]. Given the cell parameters and assuming one complex
molecule per asymmetric unit, the Matthews number or Vm [110], which is
the crystal volume per unit of protein molecular weight, is calculated
to be 3,10 A3/dalton; the most commonly observed Vm value is 2.15
A3/dalton [110]. The high solvent fraction and high Matthews number of
the hirudin-thrombin crystals are, however, nearly identical to the
values obtained from crystals of the asymmetric molecules human Ig
immunoglobulin (fs = 0.61, Vm = 3.14 A3/dalton) [111] and yeast
phenylalanine tRNA (fs = 0.60, Vm = 3.20 A3/da1ton) [112], WhiCh
suggests that the hirudin-thrombin complex could be asymmetrical in
shape.

A storing solution of 30% PEG 4000, 0.375 M NaCl, 0.2 M MgClz,
0.1 M sodium acetate buffer, pH 4.6, 1 mM NaN3 was found to
stabilize the crystals and maintain diffraction quality. The
rHV2-K47-thrombin crystals used for x-ray diffraction experiments were
gradually equilibrated in this storage solution prior to use by slowly

exchanging the crystallizing solution with the storage solution.

37

E. 2.8 A Resolution Diffractometer Intensity Data Collection

A 2.8 A resolution set of diffractometer intensity data was
collected on the hirudin-thrombin crystals using monochromatic Cu K“
radiation (1.5418 A) and a Nicolet P3/F four-circle diffractometer
controlled by a Data General Nova/4 computer. X-rays were generated
using an AEG sealed x-ray tube operating at 2 kV (50 kV, 40 mA), and
were monochromated by means of a graphite monochromator crystal. A
helium evacuated collimator and beam tunnel were used to minimize the
air absorption of the incoming beam and the diffracted x-rays,
respectively. The complete data set was collected in two parts, each
part using a different crystal. The first data collection contained
reflection measurements from 3.63 to 2.80 A resolution (reflections had
26 values ranging from 24.5° to 32.0°), and will be referred to as HTN2.
The second set contained intensity data to 3.56 A resolution (reflection
measurements from 2.0° to 25.0° in 20), and will be referred to as HTN3.
(The first actual 3.56 A resolution intensity data set collected, HTNl,
was used to generate a file of observed reflections.)

The crystal used for the HTN2 and HTN3 data collections had
dimensions of approximately 0.95 x 0.40 x 0.35 mm. These crystals were
mounted in siliconized glass capillaries of diameter 1.0 mm. The
morphology of an idealized hirudin-thrombin crystal is shown in Figure
15. A crystal of the complex was mounted such that the c* axis (along
the longest dimension of the crystal) was coincident with the capillary
axis. A drawing of a hirudin-thrombin crystal mounted in a capillary is
shown in Figure 16. The crystal was positioned in the capillary bathed
in the storing solution and, just prior to sealing the capillary, the
crystal was partially ’dried’ with a filter paper strip so that only a

small amount of liquid remained under the crystal to affix it to the

38

 

 

It»

 

 

 

 

 

Figure 15. Morphology of an Idealized Hirudin-Thrombin Crystal.

39

 

P ”7“ it I“

Crystal Mother Liquor

_Figure 16. A Hirudin-Thrombin Crystal Mounted in a Capillary
for X-ray Crystallographic Analysis.

4O

capillary wall. A small band of storage solution was left at one end of
the capillary before sealing it off with wax; as protein crystals
contain a significant solvent fraction, the presence of this band of
storage solution in the sealed capillary ensured, by vapor
equilibration, that the crystal would not dry out. The capillary
containing the crystal was then inserted into plasticene mount on a
goniometer head.

The protein crystal on a goniometer head was positioned onto the
goniostat of a four-circle diffractometer. A schematic drawing of this
goniostat is shown in Figure 17 [113]. The goniostat orients the
crystal and detector with respect to a stationary x-ray beam, and brings
Bragg planes into reflecting position. V.L. Bragg developed the
analogy of the the diffraction of x-rays by crystals to ordinary
reflection, and he deduced a simple equation - nk = 2dsin9 - by treating
diffraction as "reflection" from planes of atoms in a lattice [114]. In
this equation, d represents the interplanar spacing of the Miller planes
and A is the wavelength of the x-ray radiation. Thus, diffracted
x-rays are often referred to as reflections. A reflection is identified
by the Miller indices h,k, and l, where the reflection plane (hkl) makes
intercepts a/h, b/k, and c/l with the unit-cell axes a, b, and c of the
crystal. The four circles of the goniostat refer to the three circles
swept out by rotation of the crystal about the w, 6, and X axes (Figure
17), and the circle swept out by the arm of the detector through
rotation about the 0 axis (which is coincident with the w axis). A
goniometer head with the mounted protein crystal is placed on a base in
the circle which is coincident with the axis. Each diffracted x-ray
beam collected with a diffractometer is therefore defined by a 20, w, ¢.

and X angle as well as by an (hkl) index.

41

 

Figure 17. The Goniostat of a Four-Circle Diffractometer [111].

42

The 20, w, o, and X angles which the computer-controlled
diffractometer drives to in the course of making an intensity measurent
result from the unit cell parameters and the orientation of the crystal.
The unit cell constants and crystallographic orientation of the crystal
are stored in the computer in the form of an orientation matrix A, which
relates the reciprocal lattice vectors a*, b*, and c* in terms of their
components in the coordinate system of the instrument.
{/ a*x b*x c*x

a*y b*y c*y
\\ a*z b*z c*z

.(3)
In the HTN2 and HTN3 data collections, the orientation matrix was
initially determined by manually locating three strong and approximately
mutually orthogonal reflections, (0,0,32), (1,16,0), and (16,1,0).
These reflections are stored in the reflection array, and were used to
calculate the unit cell constants and the initial orientation matrix.
As the hirudin-thrombin crystals were mounted with the c* or (001)
direction of the crystal approximately coincident with the capillary
axis, it was possible to align the c* axis to coincide with the 6 axis
of the diffractometer. This alignment facilitated the location of
(1,16,0) and (16,1,0), which could be found at X = 0, since the the
crystal lattice is an orthogonal system. Axial intensity distributions
of the (001) as well as the (h00) and (hh0) directions were then
recorded. These measurements, which are displayed in Figure 18, allow
one to assess the quality and integrity of the diffraction pattern from
crystal to crystal. After this initial orientation matrix was obtained,
a more accurate matrix was found by including nine additional high order
reflections into the reflection array and by performing a least squares

calculation between the calculated and observed angles of the

43

.mHmumauo caneouceIa cmesmIcficzumm
“confineooom mg» no mcoflusnfimamwn zuwmcoucH Hmfix< .wH ousmwm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mm mm mm
9. on om 9 a 2. on 8 Q G 9. on cm 9 c
_ a 5 1 5 w _ _l _ .1 _
m
a
low I am a 16m
9.
R a
H 2 1% .2 I on a Lon
. .4 ONE N 2 = Q n.
8 _. 9
m
193. I EN . loam
look 1 sea 102
v
..
18.: w! 8.: look

3qu $qu A :qu

Be 93 2E

44

reflections in the array. These additional reflections were all of
relatively high 20 (15° < 20 < 22°), possessed sufficiently strong
intensities so that the reflection centering routine would be able to
center them after typical decay (5-15%), and collectively were
distributed uniformly in the region of reciprocal space used for
intensity measurements - most notably in x. The cell dimensions of the
crystal were updated each time a least squares calculation was performed
on the reflections in the reflection array. The average cell parameters
obtained by least squares in the HTN2 and in the HTN3 data collections
are shown in Table 3.

The method used to measure the intensity data was first proposed by
VYckoff et al. [115], and is known as the Vyckoff w—step procedure.
The Vyckoff procedure is a stationary crystal-stationary detector peak
sampling method where the top of the reflection peak is sampled by
making a limited number of intensity measurements at small intervals in
00, as opposed to the full profile scan from background to background.
It is often used in cases where reflections are closely spaced or where
QUiCk! accurate measurements are sought to minimize x-ray-induced
CFVStal decay - situations frequently encountered in macromolecular data
collections. Using this method in the HTN2 AND HTN3 data collections,
the Peak was sampled 0.09° on either side of the calculated value for
a given reflection by taking intensity measurements at seven 0.03° step
increments, and it took 18 seconds to complete the scan. The four
largest intensities measured were then summed, and this sum was taken to
be Preportional to the integrated intensity. Background measurements
were made on either side of the peak by offsetting the I» value 0.40",
and it took approximately 7 seconds to complete the left and right

background measurements. Thus, it took approximately 25 seconds to

Table 3.

45

The Average Cell Parameters of the HTN2 (3.6-2.8 A)
and of the HTN3 (2.8 A) Crystals. Cell lengths in
in A units, angles in degrees.

A B C a 6 Y
HTN2 91 .07 90.99 132.60 90.05 89.99 90.03
a 0.02 0.01 0.04 0.01 0.01 0.01
HTN3 90.95 90.95 132.56 89.99 90.04 90.03

0 0.02 0.02 0.03 0.04 0.03 0.02

46

measure each reflection intensity. This procedure also can compensate
for minor crystal misalignments by allowing the diffractometer to take
additional steps if the peak maximum is too close to an edge of the scan
range; six additional steps were allowed in the HTN2 and HTN3 data
collections.

Althought the Vyckoff to step scan was used to collect both sets of
hirudin-thrombin intensity data, the manner of data collection was
different in the two sets. In the HTN2 (3.63-2.80M data collection,
the intensity data were collected in three 26 shells of (24.5-27.5)°,
(27.5—30.0)°, and (30.0-32.0)°, in the order of lowest to highest
resolution. As only 1/16th of the sphere of reflection needs to be
measured to collect a unique set of intensity data, only reflections
with the k index greater than or equal to the h index were measured. In
each shell, data were collected by ascending h,k,l - with 1 varying
fastest , h slowest. A total of 7914 reflections were measured in the
HTN2 data collection, and it took approximately 55 hours to obtain these
data. In the HTN3 3.56 A data collection, a reflection file was used
based on the observed reflections of the HTNl 3.56 A data collection;
81% 0f the possible HTNl reflections were judged to be observed and were
included in this file. The file was created by dividing the unique,
observed data into 19 29 ranges of approximately 305 reflections per
range, Each range was sorted in order of descending 29, with h varying
810West, 1 varying fastest, and these individual files were merged in
the Order of descending average 29, so that the data collection file
wmild allow reflections to be measured from high to low resolution.
Since high order data are generally weaker, and are more strongly
affected by x-ray-induced decay than low order data, this manner of data

collection is preferred to that used in the HTN2 data collection. In

47

the HTN3 data collection, 5983 reflections were measured over the course
of about 42 hours.

During the course of each data collection, the alignment of the
crystal was monitored by measuring the intensity of three nearly
orthogonal reflections every 100 reflections; reflections (1,16,0),
(16,1,0), and (0,0,32) were chosen for this purpose. If the intensity
of these monitor reflections diminished by more than 15%, the crystal
was considered misaligned, and this triggered the re-measurement of the
12 reflections in the reflection array, followed by the calculation of a
new orientation matrix and cell parameters by a least squares
calculation. Slopes resulting from a least squares fit line of
intensity versus exposure hour data for each monitor reflection
were used in data processing to obtain decay corrections to the data of
29 > 15.0°.

A number of different measurements preceded and followed the actual
intensity data collection; these measurements included the 0) profile of
the reflection (6,16,17), the intensity absorption of (0,0,8) reflection
versus 4» angle, and the (hkO) zonal reflections to 5.9 A resolution.
The 0° profile of a reflection helps determine the parameters in the
Vyckoff to step scan, and provides an indication of the suitability of
the Crystal for intensity data collection. A split profile, for
example, may imply that the crystal is cracked, and that two closely
oriel‘tted but distinct crystals are contributing to the profile. In the
HTN2 and HTN3 data collections, the a) profile of the reflection
(6,16,17) was determined by making 10 second intensity measurements at 00
Values of +/- 0.50° from the calculated value in 0.05° increments of w;
the a) profile obtained for reflection (6,16,17) prior to beginning the

HTN2 data collection is shown in Figure 19. The width of the profile

400-

200-

FigUre 19.

IZOD-

I000_

600.

48

TCOL/N TS//O SEC.

I4OOJ

1
800.

 

 

 

 

90° ' 92° V 95° /0.2° 10.5" no“

The w—Profile of (6,16,17) Taken Before the 3-D
HTN2 Data Collection. Peak width at half-height
approximately O.18°.

49

peak at half the intensity maximum defines the region of the peak which
should be scanned in the Vyckoff w step scan; since this width is
approximately O.18° in the profile (Figure 19), a scan range of O.18°
was chosen. From this profile one can also determine a suitable w
offset value at which left and right background measurements can be
made; an offset value of 0.40° was chosen from the Figure 19 profile.
The intensity absorption versus ¢ angle for the reflection (0,0,8)
was measured to determine the optimum region of ¢ to carry out the data
collection. Ideally, one would like to collect data in the region with
the lowest overall x-ray absorption or the highest overall intensity.
The (0,0,8) reflection was chosen because it has a large intensity
(Figure 18) and, more importantly, because each crystal was aligned on
the diffractometer so that the c* axis was coincident with the o axis.
This alignment ensured that the (001) Bragg planes were always in
reflecting position for all values of d at X = 90°. Intensity
measurements were made every 20° in ¢ over the entire 360° range, and
then every 10° over a 150° range containing the 45° data collection ¢
region. A plot of I(¢) versus ¢ taken before the HTN3 3.56 A data
collection is shown in Figure 20; the 45° range indicated in the plot
was the ¢ range employed for the data collection. This region was
chosen for the data collection because it was of high intensity or low
x-ray absorption. The curve obtained from averaging the absorption
curves determined before and after each data collection was used to
correct the intensity data for absorption in the data processing.
Intensity data for (hkO) zonal reflections of 29 (2-15)° were
measured before and after the three-dimensional data collections. These
data were used in the three—dimensional data processing to assess decay

due to x-ray exposure for reflections of the corresponding 29 range.

/(¢>) 1*

50

 

5000.. hOO h/10 OkO

4000..

3000..

2000..

IOOO-

 

Figure 20.

/00° 60° 20° —20° -60° -/00°
91>

Intensity Absorption Plot of (0,0,8) Taken Before the

HTN3 3.56 A Resolution Data Collection. I(¢) in total
counts/measurement. The 45° ¢ range employed in the
data collection indicated by "D.C.”.

 

A.

u

 

51

Moreover, the 5.9 A resolution zonal intensity data measured before data
collection for the HTN2 and HTN3 data sets were used to obtain an

initial scale factor for scaling the two data sets.
F. 2.3 A Resolution Area Detector Data Collection

A 2.3 A resolution set of area detector intensity data was
collected from a recombinant hirudin-thrombin crystal in the laboratory
of Professor Robert Huber at the Hax—Planck-Institut fur Biochemie in
Martinsried, Germany. The Enraf—Nonius data collection system consisted
of a CAD-4 four-axis goniostat and a FAST television area detector. Cu
K“ x-rays were produced by a Rigaku RU-ZOOB rotating anode x-ray
generator fitted with a 0.3 x 3.0 mm fine focus filament and operating
at 5.2 kw (45 kV, 120 mA). The x-rays were monochromated by means of a
Ni-filter which removed 99.95% of the Cu K“ radiation. A
hirudin-thrombin crystal of approximate dimensions 0.85 x 0.50 x 0.50
mm, mounted in the previously described manner with the c* axis
coincident with the capillary axis, was used to collect the intensity
data; all measurements were made at 6 °C.

The Enraf-Nonius CAD-4 goniostat is a 4-axis diffractometer which
makes use of the same 29, m, and ¢ angles as the Siemens P3/F system,
but which uses a K mechanism, as opposed to the X circle, to position
the goniometer head. A drawing of a 'K-geometry' goniostat is shown in
Figure 21 [116]. The goniometer head is held on an arm which can rotate
about an axis 50° from the vertical. The base of the arm rotates about
the 9 axis, and the motion is identical to that of w on a conventional
goniostat. The upper portion of the arm is held with the ¢ axis inclined
50° to the K swivel axis. Combined motions of the w and K can produce

Crystal orientations analagous to —100° < X < 100° on a conventional

52

._oHH_ acuowuon Hucofiucw>coo a nu“:
uwuoEOHomuumfin msfizoznumucm an no mafiaoun .aN ousufim

 

 

 

 

 

 

diffractometer [116].

The Munich Area Detector (NE) System or MADNES software package is
an inclusive area detector x-ray diffraction data collection system of
programs which was used to carry out all steps involved the FAST
experiment. MADNES is an area detector independent system which allows
one to perform the following diverse tasks [117]: approximately align
crystals; find reflections; autoindex found reflections and use them to
determine unit cell parameters and an orientation matrix; refine crystal
and detector orientation; predict reflection positions; collect
reflections; evaluate reflections as they are collected; monitor and
update the crystal orientation matrix and detector alignment throughout
data collection; and plot predicted and/or observed reflections.

The data collection geometry used in a MADNES-controlled area
detector x-ray diffraction system is similar to that used in rotation or
oscillation x-ray photography: the crystal is rotated about an axis
perpendicular to the incident x-ray beam. In order to measure a nearly
complete set of data from a hirudin—thrombin crystal, two crystal
orientations and two data collection ¢ ranges were used. (The ¢ angle
in-MADNES terminology is identical to the w angle of the CAD-4
goniostat.) Initially, data were collected by rotating nearly 90° in ¢
(-60° to 24°) about the c* or capillary axis of the mounted crystal. In
addition, data were collected by rotating 100° in ¢ (-10° to 90°) about
an axis 30° oblique to the c* axis.

Unlike conventional scintillation counter diffractometers which
measure reflection intensities individually, area detectors have the
Capability of measuring several hundred reflection intensities at a
time. The FAST area detector is a phosphor screen coupled to a

television camera, and it is composed of picture elements called pixels.

54

The pixels are arranged on the screen in a 512 x 512 grid. FAST area
detector x—ray intensity data are collected by rotating a crystal
through a small angle while x—ray quanta are collected on the phosphor
screen. The x-ray quanta collected at each pixel location are
integrated, and a two—dimensional array is generated whose elements
correspond to the amount of x—rays detected at each detector location;
this array is called an image or frame, and it is the basic
informational entity used by MADNES to perform its numerous
aforementioned tasks. In the hirudin-thrombin FAST experiment, frames
of data were generated by rotating O.1° in ¢ at a rate of 90 seconds per
frame. The crystal-to-detector distance was 49 mm, and it took
approximately 56 hours to complete the frame data measurements at the
two previously mentioned orientations.

The flow which transpires during a MADNES intensity data collection
is indicated in Figure 22 [118]. The initial step in the data
collection process is a pre-alignment, and it is very similar to the
preliminary alignment which must be conducted prior to x-ray precession
photography. The purpose of the pre-alignment is to find two major
zones of the crystal, and to reposition the crystal so that the center
of the zones approximately coincides with the center of the x—ray beam
[117,118]. In this experiment, the (Okl) and (hOl) zones were aligned.
A least squares fit to the spots which defined the diffraction cones
gave the circle center, and the new goniostat angles were calculated to
bring the zone and beam centers into alignment. After this step, the
missetting angles were approximately zero and rough unit cell parameters
could be obtained.

The second step in the FAST experiment utilizes the FIND command

and involves determining the centroids of a few hundred reflections at

55

[1. PBS-ALIGNMENT]

 

 

 

' 2. FIND
Yms, st, Phi for several hundred
reflections at two phi positions

 

 

 

 

3. ALIGNMENT 1

A. AUTO-INDEX
found reflections and obtain 3
approximate orientation matrix i
3. REFINE !
cell parameters, orientation matrix, & ;
detector-related experimental parameters_J

W

4. PREDICT
a batch of reflections

v

S. COLLECT & EVALUATE
, intensity data
get and process next image

 

 

 

 

 

 

 

 

 

 

 

 

 

yes !<:fiis-alignment detect3§2:>

 

 

 

 

 

<:Reed more predictionsz>>———————-yes

no <<flfii駧§2>

yes

 

 

END

Figure 22. A Flow Chart of Operations in a MADNES
Area Detector Data Collection.

56

the pre-aligned ¢ positions. Three coordinates define a FAST reflection
position: Yms, st, and ¢. The coordinates Yms and st (’ms’ stands
for ’mass store') correspond to the vertical and horizontal directions
of the detector, and the w coordinate refers to the crystal rotation
angle (the ¢ angle is identical to the w goniostat angle, as was
mentioned earlier). Finding a hundred or so reflections at each of the
(hOl) and (Okl) zonal positions was a 4-stage process [118]. First, an
initial image was produced by rotating 0.4° in ¢ about each zonal ¢
position. Second, each resulting image was searched for reflections and
the Yms, st values were extracted. Third, ten images spanning 1.0° in
v (O.1° rotation per frame), and centered on the zonal o value of step
1, were measured. Finally, a list of Yms, st, and ¢ coordinates for
the reflections contained in these frames was produced. The purpose of
the third step was to ensure that any partial reflections measured in
step 1 would be completely defined.

The third step of the area detector data collection is an alignment
stage which includes an auto-indexing of the found reflections, an
orientation matrix calculation, and a refinement of various crystal and
experimental parameters. In the rHVZ-human a-thrombin data collection,
the AUTJ (auto-indexing of reflections by the James Pflugrath method)
command was used to auto—index the found reflections and obtain an
approximate orientation matrix. The REFINE command was then implemented
to refine the crystal orientation matrix, unit cell parameters, the
missetting angles, the crystal-to-detector distance, the detector tilt
and twist angles, and the center of the primary beam [117]. The unit
cell parameters obtained as a result of this alignment step for
hirudin-thrombin were: a = b = 90.53 A; c = 132.05 A; a = B = v =

90.00° and the error estimates were approximately 0.05 A and 0.05°,

57

respectively.

Having completed the pre—alignment, reflection-finding, and
alignment steps, one is now ready to actually collect and evaluate the
needed frame data. Obtaining a three-dimensional set of reflection data
using MADNES basically involves the iterative use of two commands:
PREDICT and COLLECT. PREDICT uses the output from the previous aligment
step to calculate for a reflection the (hkl) index, central ¢ value, ¢
range, starting ¢ value, Lorentz—polarization correction factor, oblique
incidence correction factor, and the mass store coordinates Yms, st
[117]. Reflections are predicted in small ¢ increments called batches.
The major restriction on the batch size is that it contain less than
3000 reflections; in the hirudin-thrombin data collection a batch size
of 2.5° was used. Once a batch of reflections has been predicted it is
sorted on starting o value and checked for overlaps.

By means of the COLLECT command, frame data collection and data
processing are carried out, and controlled by the appropriate parameter
input. The data collection proceeds by the exposure of a series of
images as the crystal rotates, and involves the extraction of
information around each predicted reflection occurring on the images.
MADNES stores all the reflection information in a double-linked list
[117]. Each time a reflection begins to approach the Ewald sphere, the
storage size of data for the reflection is calculated and a contiguous
amount of space in this list is allocated. Information for this
reflection from a series of adjacent images is transferred to this
allocated location as it is obtained. Once the reflection has finished
passing through the Ewald sphere, the information for that reflection in
the double-linked list is complete and that reflection is processed; the

data for the reflection can be written to disk if desired, but most

58

importantly the data space for this reflection is cleared for use by a
new reflection. As an image is being generated by the area detector,
MADNES can process the previous image.

Collected (hkl) reflection intensity data were processed to
observed structure amplitudes (uncorrected for scaling, absorption, or
decay), denoted lFobsl, during the course of data collection by

application of the following relationship:
|Fobs|2 = LORPOL x I(hkl) ...(4)

where I(hkl) is the background-corrected intensity of a reflection (hkl)
and LORPOL is the Lorentz-polarization correction factor, which is a
geometric-polarization correction term. The size and shape of
reflections was determined by means of the dynamical masking method of
Sjolin and Wlodawer [119]. This method generates a mask which is used
to evaluate the pixel data. Mask generation proceeds via five stages
[118]. In stage 1, the 3-D data are treated with a smoothing algorithm,
and the average and standard deviation of the smoothed array are
determined. In stage 2, a mask is produced from the smoothed data using

the formula:

MASK = (SMOOTHv — Average)/ Std. Dev.

v.2.6 2.:

..(5)

In stage 3, a new average background and standard deviation are
calculated from the smoothed array by leaving out pixels which deviate
too much from the previous average and step 2 is repeated. In step 4, a
contiguity criterion is applied to sufficiently positive mask values:
the contiguous region is considered to be the peak and non—contiguous

regions due to noise or interfering reflections. In step 5, the mask is

59

used to assign the original data into peak, background, and neither peak
nor background pixels so that peak integration and intensity calculation
can be achieved. Masks are not produced for low-intensity reflections;
these reflections are processed using a mask made from the masks of
medium and strong reflections from the same region of the detector.
(The detector surface is divided into 16 regions.)

As a result of the COLLECT command, an observed structure factor
amplitude (lFobsi) file is generated and continually added to during the
course of active data collection. The file contains the following
information for an individual reflection record: (hkl), lFobsl,
o(Fobs), the average background, o(background), error codes, crystal
identifier codes, calculated and observed: Yms, st, ¢ coordinates and
the width of the spot center (observed values are only listed if
found), and exposure time in hours at the spot center [117]. Each
reflection is given a simplified character as well as a detailed numeric
error code. Reflections can be given one of four simplified error
codes: good, weak, edge, and bad. 'Good' reflections are measurements
which were obtained under no obvious error conditions and will be used
in further processing (to be described in a later chapter) to generate
|F(hkl)| values. Reflections which are labelled with the other three
simplified error codes are ignored in later processing, and have
detailed numerical error codes which reveal the exact source of the
error. 'Veak’ reflections were not observed in the data collection.
’Edge’ reflections were measured on the edge of a data array in one or
more of the Yms, st, or ¢ coordinates. A 'bad’ reflection was labelled
as such if one or more of the following problems occurred with regard to
a peak: appeared too far from one or more of the predicted detector

coordinate values. peak too wide in a particular coordinate value. bad

60

background or o(background), bad pixel non-uniformity in the data array
or a problem with pixel overflow — i.e. too dark [117]. In the
hirudin-thrombin data collection, 183,791 reflection measurements were
recorded in the output IFobsl file, of which 164,250 (89.4% of the
recorded reflection data) were labelled as 'good’ and 19,541 were
labelled with the bad, edge, or weak error codes.

MADNES also has a means to continually monitor whether a
mis-alignment has occurred and a re-centering is needed [118]. During
the course of data collection, the average difference between the
predicted and observed Yms, st, and ¢ coordinates is calculated for
each of the 16 regions of the detector. Moreover, the evaluation status
of the 500 most recently processed reflections is stored. If average
differences in the predicted and observed coordinates get too high, or
if the fraction of 'good’ reflections in the 500 most currently
evaluated reflection list gets too low, the data collection
automatically returns to the REFINE command for a re-alignment. Up to
250 reflections from the current list with the highest |Fobs| and
suitable error codes are selected for refinement of missetting angles
and detector position. (The unit cell parameters are not refined in a
realignment as the ¢ range and the sampling of reciprocal space are too
small to yield an accurate orientation matrix calculation.) After the
refinement has been conducted, two quantities are calculated to assess
if the re-centering has been successful [117]. The root mean square
(r.m.s.) difference between the actual and the predicted spot positions
on the detector in millimeters (rmscfd), and the r.m.s. difference in
angles between the reciprocal lattice vector and the predicted one
(rmsdeg) for the reflections selected to perform the refinement are

calculated and checked against maximum acceptable values stored in

61

RmsChk. In the hirudin-thrombin data collection, the acceptable values
for for these quantitites were 0.2500 and 0.2000, respectively. If the
calculated values were below the tolerance limits, the data collection

would proceed; otherwise, the data collection would halt and prompt for

user intervention.

III DATA REDUCTION

A. Converting the 2.8 A Resolution Diffractometer Intensity Data to

Structure Amplitudes

The conversion of each reflection intensity, I(hkl), to its
corresponding structure factor modulus, |F(hkl)l, is the initial step in
the data processing of x-ray intensity data. The generation of a 2.8 A
resolution set of structure factor amplitudes from the diffractometer
x-ray data of the hirudin-thrombin crystals proceeded in two stages.
Initially, the 3.56 A resolution HTN3 and the (3.63-2.80)A resolution
HTN2 data sets were reduced to structure factor amplitudes. Afterward,
the data sets were scaled and averaged to produce a unique set of
amplitudes to 2.8 A resolution.

The intensity data were reduced using the program P-DATA, written
by C.D. Buck, formerly of this laboratory. The program is able to
accomplish this task, using input data from either a Nicolet P3/F or a
Picker FACS-I four circle diffractometer, by means of the following

equation:
|F(hkl)|2 = consr x ABS x DEC x LORPOL x I(hkl) ...(6)

where CONST is a scaling factor, ABS is an absorption correction factor,
DEC is an x-ray induced decay correction factor, LORPOL is the
Lorentz-polarization correction factor, and I(hkl) is the
background-corrected intensity of the reflection (hkl).
Background-corrected reflection intensities were obtained by
averaging the background data associated with the reflections in shells
of 29. The integrated intensity output for each reflection during the

course of data collection is computed from the following equation [120]:

62

63

I = [total scan count — sum of background counts ] x scan rate
background to scan time ratio

 

-(7)

where measurements of the left and right background as well as of the
total scan count for the peak are used in the calculation.
Background-averaging of intensity data results in more accurate
measurements for two reasons. First, each left and right background was
measured only one—fifth as long as the peak in these intensity data
collections. Second, since these background count values are generally
low, statistical fluctuations can significantly affect the intensities
calculated for weak reflections. Averaging the background of the
intensity data for the HTN2 and the HTN3 data collections revealed that
there was a definite 29 dependence to the background measurements. The
background data were also averaged in shells of 6 as well as 29;
however, since no 6 dependence could be discerned, the intensity data
resulting from the 29 shell background averaging were used in the P-DATA
processing.

The formulation used to correct intensity measurements for
absorption in the P-DATA program was based on the method proposed by
North, Phillips, and Mathews [121]. The program uses absorption tables
of Imax/I(¢) versus 6 to apply the absorption correction, and such a
table was obtained from the absorption curve depicted in Figure 20.

The decay correction factor, DEC, corrects for the intensity
deterioration of a reflection due to x—ray exposure to the crystal, and
it is generally a function of both exposure time and 29 angle. The DEC

factor has the form 1/(1-St) in the following equation:
I'(t) = I(t)[1/(1-St)] ...(8)

In equation (8), I(t) is an intensity measurement recorded at time t in

64

the data collection, I'(t) is an intensity measurement corrected for
decay as if it were recorded at time zero. Furthermore, S = -s/I'(0),
where s is either the slope from an intensity versus exposure time plot
for a single, periodically-measured monitor reflection, or the slope
from an average intensity versus exposure time plot for a
periodically-measured group of reflections, and I'(O) is the
corresponding extrapolated intensity at zero time.

The decay corrections for the HTN2 (3.6—2.8) A data set were
obtained from the linear plot of 8 versus average 29 (Figure 23). Four
points were used to generate the line; three points came from intensity
decay plots of the three monitor reflections [(1,16,0), (16,1,0), and
(0,0,32)], and the fourth point was the slope calculated by comparing
peak heights from 5.9 A resolution (hkO) Patterson projection maps
computed using intensity data collected before and after the 3—D data
collection. The three decay— and alignment-monitoring reflections
decayed approximately (12—16)Z over the course of the 57 hour data
collection, while the 5.9 A resolution (hkO) zonal reflections decayed
8% on an average. The HTN2 intensity data were collected from low to
high resolution in shells of (24-27.5)°, (27.5-30.0)°, and (30.0-32.0)°
in 29, respectively. From the S versus average 29 plot, S values were

extrapolated for the 29 values 27.5° (S = 0.00373 hours-1) and 30-0° (5

= 0.00404 hours—1)- The (24.0—27.5)° intensity data were corrected for
decay using S = 0.00373 hours'l. the (30.0-32.0)° intensity data were
corrected for decay using S = 0.00404 hours-1. and the (27-5-30-0)° data
were corrected for decay by interpolating between the two S values.

In the 3.56 A resolution HTN3 data set, which was collected from
high to low resolution using an observed reflection file, the DEC

correction factors had a slightly different form than in equation (8),

S

65

x /O.3 hours"

 

 

4h
5..
4..
3..
..
2fl_ I.
/.L
4° 8° /2° 16° 20° 24° 28° 32°
<29>
Figure 23. Dependence of X-ray Decay Correction with Scattering

Angle for the HTN2 Hirudin-Thrombin (3.6-2.8) A Resolution
Data Collection.

66

and are contained in the expressions listed below:

I'(t)

I(t)[1/(1-29kt)] 29 > 12.0° ...(9)

I'(t)

I(t)[1/(1-12kt)] 29 g 12.0° ...(10)

To obtain DEC corrections factors for this data set an S versus average
29 plot was made, as the one shown in Figure 23, but now the slope of
the 5 versus average 29 plot, k, as well as the actual 29 value of the
reflection were used to obtain the appropriate decay correction. In the
HTN3 data collection the three alignment— and decay-monitoring
reflections decayed (5-11)% over the course of the 44 hour data
collection, and the 5.9 A resolution (hkO) zonal reflections decayed on
an average of 4%. The k value determined from the S versus average 2A
curve for the HTN3 data collection was 9.04 x 10'5 hours-1.

The factor CONST was used to approximately scale the HTN2 intensity
data to the HTN3 intensity data. The CONST value was obtained by
finding the mean ratio of corresponding Patterson peak heights from the
HTN2 and HTN3 5.9 A resolution (hkO) Patterson projection maps
calculated using intensity data obtained before the start of each 3-D
data collection. The value for CONST determined to put the HTN2
intensity data roughly on the same scale as the HTN3 intensity data was
1.04. Moreover, the fact that the CONST value is so close to unity
indicates that the x-ray scattering of the two crystals was comparable.

Once the P-DATA reflection processing was completed, it was
possible to determine the percentage of collected data which were
observed in the HTN2 and HTN3 intensity data collections. The HTN2
(3.6-2.8)A resolution data were processed to 3483 unique structure
factors out of 7597 possible observations, thus this data set had 46% of

the unique data observed. In the HTN3 3.56 A resolution data

67

collection, 5624 unique measurements were observable out of the 5769
reflections present in the reflection file. This number was unusually
high because the data file contained only reflections observed to 3.56 A
resolution from a previous data collection, HTNI. However, if one
considers that this observed reflection file represented 81% of the
unique data to 3.56 A resolution, the HTN3 data set had 79% of the
unique data observed.

The two sets of reduced data contained 190 common structure
factors, and these were used to scale the HTN2 structure amplitudes more
accurately to those of HTN3. Before carrying out any further scaling,

an R-factor was calculated for the overlapping reflections using the

equation:

R-factor -

I
M
A
"'1
O
U—
U)

- lFobSI I)
HTN3 HTN2

...(11)

and the value was found to be 0.098. In an effort to reduce the above
Refactor and scale the overlapping structure factors (and ultimately the
two sets of structure factors more closely), the following ratio was
calculated:
£(IFobsl )/£ (lFobs| ) ...(12)
HTN3 HTN2

which was found to be 0.96. Applying a scale factor of 0.96 to the HTN2
|F(hkl)| values and recalculating the R-factor for the overlapping
amplitudes yielded a value of 0.091, indicating that the re—scaling
brought about improvement. To create a unique set of structure

amplitudes to 2.8 A resolution, all the HTN2 structure factors were

68

multiplied by the scale factor of 0.96, combinded with the HTN3 reduced
data, and the overlap amplitudes were averaged; the result was a set of
8917 unique |F(hkl)| values. The <|F(hkl)|2> versus <29) distribution
for these hirudin—thrombin structure factors is shown in Figure 24. The
completeness of this 2.8 A resolution set of |F(hkl)| values was
determined by loading these data into the PROTEIN [122] package of
macromolecular crystallographic programs, and inserting a CHECK
COMPLETENESS subcommand into the LOAD command file; the set of structure
factors was found to correspond to 62% (8917/14362) of the possible

unique reduced data.

B. Post—Data Collection Processing of the 2.3 A Resolution

Area Detector Stucture Factors

Although the MADNES-controlled FAST area detector diffractometer
generated lFobs(hkl)| values for each reflection measured in the
hirudin-thrombin data set, these structure factors still had to be
corrected for absorption, decay, and scaling before they could be used
for x-ray crystal structure analysis. However, the previously described
means of carrying out absorption and decay corrections cannot generally
be conducted with an area detector system. Azimuthal angle w scan data
cannot be collected for an empirical absorption correction, nor can
specific reflections be measured periodically to carry out a
conventional decay correction. Nonetheless, one can make use of the
real power of the area detector - its ability to measure large numbers
of symmetry—equivalent refletions in a data set — to correct for
absorption and decay, as well as for scaling.

By means of the program ABSCOR [123] the hirudin-thrombin FAST area

detector diffractometer data were corrected for scaling, absorption, and

69

</F/2>
IZCH;

ICMDl
8C7»
SCI.
4CL.

20..

 

<:£?62>

Figure 24. The <|Fobs|2> versus <29) Distribution for the 2.8 A
Resolution Hirudin-Thrombin Diffractometer Data Set.

‘A'

70

decay as well as for the non-uniformity of response of the detector.
The intensity data output file from the MADNES program (the so-called D
13 file) serves as the input file for ABSCOR.

In a generalized macromolecular crystallographic data collection,
it is possible that the crystal will not be fully bathed in the primary
x—ray beam, and that variable scattering volumes of the crystal will be
irradiated during the course of the data collection. (In the
hirudin-thrombin data collections discussed in this dissertation, the
crystals should have been bathed uniformly in the x-ray beam throughout
the course of the experiments.) In this situation, it is reasonable to
assume that reflections measured in a batch or in a block, which
consists of one or more contiguous batches, should have similar scale
and temperature factors which correct for changes in irradiated volume
and for crystal decay due to x-ray exposure. One can represent a
structure factor amplitude corrected for absorption, decay, scaling, and

non-uniformity of response of the detector by the following formula

[123]:

lFlcorr-h(i) = IFIobs-h(i)*Si*exp[_Bi(sin(e)2/x2)l
X [A'O'h(i)*(Tll,...,T33)*A's_h(i)*(T11,...,T33)]-1/2

O 1 2
x (Knonunlform) /

...(13)

where |F|obs-h(i) is the observed structure factor amplitude of
reflection h of the ith batch; h is the triple index (hkl); i is the
index of the reflection batch; Si is the individual scale factor of the
ith reflection batch; Bi is the temperature factor of the ith reflection
batch; A'O-h(i) is the transmission factor of the primary beam of

reflection h(i); A's-h(i) is the transmission factor of the secondary

71

beam of reflection h(i); T11,...,T33 are the components of the symmetric
tensor T which describes the transmission properties of the crystal;
and, Knonuniform is the post-correction factor for the nonuniformity of
response of the FAST area detector.

The transmission factors A'O—h(i) and A’s-h(i) are obtained from

the tensor ellipsoids:

A'o_h(i)‘”2 = (Tllmo2 + T22*v02 + T33*wo2 +
2*T12*u0*v0 + 2*T13*u0*w0 + 2*T23*v0*w0)

...(14)
A's—h(i)-1/2 = (T11*us2 + T22*vs2 + T33*ws2 +

2*T12*us*vs + 2*T13*us*ws + 2*T23*vs*ws)

...(15)

where these factors are generated by the following operations [123]:

A'O—h(i)'1/2 = [vO-h(i)]T*T*v0-h(i) ...(16)
A's-h(i)-l/2 -.- [vs-h(i)]T*T*vs—h(i) ...(17)
with
v0-h(i) = u0 and vs-h(i) = us ...(18)
v0 vs
w0 ws

being the primary and secondary direction vectors of the normalized beam
in the goniometer head coordinate system for reflection h(i).

The scale (Si) and temperature (Bi) factors for each block of
reflections, and the symmetric transmission tensor matrix elements (T11,
T22, T33, T12, T13, T23) for the crystal are determined by sorting the
reflection structure factors into symmetry-equivalent sets (including

the Friedel mates, where |F(hkl)| and |F(-h—k-l)| constitute a Friedel

72

pair) and minimizing by least sqares methods the following sum of
squared differences of pairs of symmetry—equivalent reduced data:
k[h(i)l who): ,
Z Z Z (lanlcorr - lanlcorr )‘ = min
h(i) j=1 l=j+1 h(i),j h(i),l

...(19)
where k[h(i)] is the number of reflections in a set of symmetry
equivalent reflections of index h of the ith reflection block. Natural
logarithms are used so that the difference is linear with respect to Si
and Bi, and the scale and temperature factors of the first block are
fixed at 1.0 and 0.0, respectively.

It is quite common for area detectors to display a non-uniformity
in detector response; this problem is generally corrected by means of a
pixel—to—pixel correction table obtained by an appropriate calibration
method prior to the initiation of the data collection. The
aforementioned least squares minimization allows one to introduce a
non-uniformity of response post-correction. After the mimization of the
scaling and absorption correction parameters has been completed,
corrected structure factor amplitudes of the symmetry-equivalent
reflections are compared. A deviation from the average value is
expressed as a correction factor for particular pixel values. The table
which results from this is composed of 64 8x8 pixel boxes and it can be
smoothed by various methods. The values from this table are the basis
of the correction factor Knonuniform.

In the ABSCOR processing of the hirudin-thrombin FAST output data,
183,791 reflections were input, of which 164,250 were flagged as 'good'
and used for further processing and 19,541 with error flags were
ignored. The block size used to restrict scale and temperature factors

was 2 sequentially measured reflection batches, and since the batch size

73

used in the data collection was 2.5° in 6, the block size was 5°.
Previous work has shown that block sizes of 3° to 6° in 6 give reliable
and reproducible results [123]. The end result of this was that the
164,250 FAST lFobsl were sorted into 31 blocks, and that after a
complete least squares minimization of equation (19), the six matrix
elements of the symmetric transmission tensor, as well as the 31 scale
factors, and 31 temperature factors were determined.

The complete ABSCOR processing of the hirudin-thrombin FAST
reflection output data proceeded via 3 stages, and the results are
summarized in Table 4. The progress at each stage was assessed by the

calculation of the R-factor presented below:

Rmerge = Z X |I(h)j -<I(h)>|/ Z Z I(h)j ...(20)
h J' h J'
where h represents the triple index (hkl) and j represents the number of
equivalent reflections for a given I(h). Prior to any minimization of
equation (19), the ABSCOR output revealed that the 164,250 reflection
intensities corresponded to 23,521 unique reflections, and that the
initial R-merge between individual reflections was 0.168. In the first
stage of the ABSCOR processing, scale and temperature factors were
refined and applied to the input lFobsl, and the result was that the
Rmerge after scaling was 0.120 between individual reflections. In the
second stage of the ABSCOR processing, the scale factors, temperature
factors, and the six components of the symmetric transmission tensor T
were refined and applied to the lFobsl, which resulted in an Rmerge of
0.111 between individual reflections. Finally, in the third stage of
the ABSCOR processing, the Knonuniform detector non-uniformity of

response correction factors were applied and the resultant Rmerge was

74

Table 4. Reflection Statistics on the Hirudin—Thrombin
FAST Data Set Evaluated by MADNES and Corrected
using ABSCOR

Total number of reflections: 164,250
Number of unique reflections: 23,521
Rmerge*-uncorrected: 0.168
Rmerge*-scaling applied: 0.120

Rmerge*-scaling & absorption
corrections applied: 0.111

Rmerge*—scaling, absorption,
& post-correction of
detector non—uniformity

 

of response applied 0.108
Rmerge** 0.042
Completeness of data
to 2.3 A resolution 0.91
Rmerge = 2 (I - <I>)/ Z I

* = between individual reflections

** = between averaged Friedel pairs

75

0.108.

The resultant output file from the ABSCOR processing contained data
for 41,062 reflections, in which there were 23,521 unique data and
17.541 Friedel mate data. The unique and Friedel mate data were
averaged by means of the PROTEIN program package to generate a set of
23,521 unique structure factors. The Rmerge calculated in PROTEIN
between averaged Friedel pairs was 0.042 and the data set was found to

be 91% complete to 2.3 A resolution.

IV MOLECULAR REPLACEMENT CALCULATIONS

A. Introduction

In any macromolecular crystallographic investigation, once a
requisite set of structure factor amplitudes have been obtained, the
next major task is to work toward generating an electron density map for
structure interpretation and model-building. The electron density at a
point (x,y,z) in the crystal can be determined by computing the
following Fourier summation:
p(x,y,z) = l/V Z Z ZIP(hkl)I*exp{ia(hkl)}*exp[-2n*i(hx + ky +lz)}

all h, k, l
...(21)
where |F(hkl)| is the amplitude of the reflection (hkl), «(hkl) is the
phase of the reflection, V is the volume of the unit cell, and the
summations extend over all of the observable reflections. However,
calculating such an electron density map is non-trivial, as it requires
a knowledge of the phase angles which cannot be measured experimentally.
In this investigation, the molecular replacement method [124] or
Patterson search method [125] was used to overcome this "phase problem".

The molecular replacement method has two main types of applications
[126]. In crystals which possess more than one molecule per asymmetric
unit, the technique can be employed to determine the orientation of
non-crystallographic symmetry elements within an asymmetric unit of the
unit cell. The method can also be used to generate a set of trial
phases for an unknown structure by orienting and positioning a similar
molecule of known tertiary structure in the unit cell of the unknown
structure. This latter application was employed to solve the structure

of the rHVZ-K47-human a—thrombin complex using PPACK-human a—thrombin as

76

77

a model [127]. The structure solution of the hirudin-thrombin complex
was a problem ideally suited for the molecular replacement method.
First of all, both the model and the crystal contained the same protein,
human a-thrombin. Second, thrombin accounts for roughly 84% of the
total molecular weight of the complex. Thus, it seemed reasonable to
assume that one could generate a good initial Fourier map based on a
Patterson search using the PPACK-thrombin structure as a model.

The molecular replacement method is often referred to as a
Patterson search method since calculations are performed using the
Patterson functions of the known model and the unknown structure of
interest. A Patterson function is a Fourier synthesis which uses the

square of the structure factor amplitudes, has the form:

+h +k +1
P(uvw) e 2/v t z 2 |F(hkl)|2*cos{2n*(hu + kv + lw)}, ...(22)
o o 0

and produces peaks corresponding to all of the interatomic vectors
[128]. A peak at the point (u,v,w) implies that there exist atoms at
(x1,y1,zl) and (x2,y2,22) such that u = x1 - x2, v = yl - y2, and w =
21; 22. The height of a peak in the Patterson function is approximately
proportional to the product of the atomic numbers of the atoms which
define the ends of the vector, (21 x 22). Moreover, if the unit cell of
a molecule contains N atoms there will be N2 peaks in the Patterson map,
of which there will be N origin or self-vector peaks and (NZ-N)
non-origin peaks (half of which are centrosymmetric). Thus, the
Patterson function is a unique Fourier synthesis which contains
structural information (relative structural information), but which
requires no knowledge of the phases and can be calculated using

experimentally—obtainable quantities, the structure factor amplitudes.

78

The Patterson search conducted with PPACK—human a-thrombin as a
model to generate a trial structure of the hirudin-thrombin complex
proceeded in three stages common to molecular replacement calculations
[124]: (1.) The relative orientation or rotational relationship of the
thrombin model with respect to the unknown hirudin-thrombin structure
was determined by "rotation function" [126] calculations involving the
intramolecular Patterson vectors of the model and the crystal Patterson
synthesis. (2.) The correct translation of the properly oriented
thrombin model in the unit cell of the hirudin-thrombin crystal was
determined by "translation function" [129,130] calculations involving a
modified crystal Patterson map and intermolecular model Patterson
vectors. Once the first two stages were successfully achieved, the
equivalence between a point x’ in the unknown hirudin-thrombin structure
and the corresponding point x in the model thrombin structure could be

expressed by the following relationship:
X, = [Clx + d ...(23)

where [C] represents the rotation matrix determined from the first stage
of the Patterson search, and d represents the translation vector
determined from the second stage. (3.) The properly oriented and
translated thrombin model in the hirudin-thrombin unit cell was used to
calculate a set of phase angles which were combined with 8.0 to 3.0 A
resolution observed structure factors to produce a trial electron
density map for model-building.

The three stages in the Patterson search calculations will now be

discussed in more detail.

79

B. The Rotation Function Calculations

A model Patterson vector set was produced from the coordinates of
PPACK—thrombin [6]. The thrombin model (284 residues) contained all but
11 of the protein residues in the PPACK-thrombin crystal structure; the
first five residues (lH-lD) and the last four residues (14K-15) of the
thrombin A chain, as well as the last 2 residues of the thrombin B chain
(246-247) were deleted as they had poor or no density in the
PPACK-thrombin electron density maps (Table 1). The search model also
contained 24 internal water oxygen atoms, but it lacked the
covalently-bound peptidyl inhibitor PPACK. Triclinic Pl structure
factors were obtained by Fourier inversion of a model electron density
map using the program FFTSF, which contains two main sub-programs,
RHOGEN and PISF. RHOGEN calculated an electron density map of the model
positioned in a triclinic P1 orthogonal cell of 120 x 120 x 120 A3 using
8.0 to 3.0 A resolution data and an overall temperature factor for the
atoms of 20 A2. PISF then calculated structure factors by fast Fourier
inversion of the electron density. Finally, these calculated structure
factors were used to generate an 8.0 to 3.0 A resolution Patterson
function of the thrombin model in a 120 x 120 x 120 A3 orthogonal cell-

However, not all of the peaks in this model Patterson map were used
for the rotational search with the hirudin-thrombin crystal Patterson
function. In the rotation function calculation one is only interested
in intramolecular vectors, thus only Patterson vectors of length greater
than 3.0 A and less than 20.0 A were selected. The lower limit of 3.0 A
eliminates the huge origin peak from the calculation. The upper limit
of 20.0 A is meant to prevent intermolecular vectors from entering the
rotation function calculation. Moreover, as PPACK-thrombin is a nearly

spherical molecule of dimensions 45 x 45 x 50 A3 [5], an upper limit 0f

 

 

80

20 A for intramolecular vectors appeared to be a good estimate. 'Weak'
or small magnitude Patterson peaks were excluded from the calulation by
using an appropriate cutoff limit. Finally, as only half of the
complete triclinic model Patterson map is unique, the peak search was
restricted to 60 A along the w axis (although any of the three axes
could have been chosen). The end result was that 7565 model Patterson
peaks were selected for a rotation function calculation. The
hirudin-thrombin crystal Patterson function was calculated using
observed diffractometer structure factor amplitudes over the same
resolution range.

The rotational search was conducted with the SEARCH routine in the
PROTEIN program package. The model thrombin Patterson vector set was
rotated and interpolated to the nearest grid points in the
hirudin-thrombin crystal Patterson function, and a product function was
calculated. This process was repeated until all angles over a specified
angular range and increment size were examined. The set of coordinates
yielding the highest product function corresponded to the transformation
necessary to bring the model and crystal Patterson vectors into the same
orientation. The Huber ("ROH") angles 6, 9, and 6 [122] were used to
conduct the rotational Patterson search. A rotational operation 6, 9, 6
corresponds to a rotation of the the Cartesian axes by 6 about the z
axis, a rotation 9 about the new x axis, and a rotation 6 about the new
y axis. The Huber angle rotation range for the calculation was: 6,
0-100°; 9, 0-180°; and 6, 0-180°. All angles were incremented by 5° in
the calculations. The highest peak in the search occurred at the Huber
angles (5.0°, 175.0°, 80.0°), and it was 7.70 above the mean. To refine
this solution, a finer rotational search was carried out over the range:

6, 0-10°; 9, 170-180°; and 6, 75-85° using the angular increment of

81

1.0°. The refinement improved the peak height to 8.10 above the mean,

and the refined coordinates were (3.5°, l76.0°, 81.5°).
C. The Translation Function Calculations

Once the orientation of the model molecule is found in the unit
cell of the target crystal, the approximate position is found by
conducting a translation search. However, in achieving this translation
result one has actually positioned not just one molecule, but the entire
set of symmetry-related molecules in the crystal unit cell. This is due
to the fact that the relative position between symmetry-related
molecules is fixed. For example, the hirudin—thrombin complex
crystallizes in space group P43212 with one molecule per asymmetric unit
and 8 molecules in the unit cell. The relative positions of the 8
symmetry-equivalent molecules in the unit cell are given in the

following relations [131]:

Molecule 1: x y 2 Molecule 5: y x -z
Molecule 2: —x -y 1/2+z Molecule 6: -y -x 1/2-z
Molecule 3: 1/2-y 1/2+x 3/4+z Molecule 7: 1/2-x 1/2+y 3/4-z
Molecule 4: 1/2+y 1/2-x 1/4+z Molecule 8: 1/2+x 1/2-y 1/4-z

..(24)

Moreover, the relative positions of the 7 intermolecular vectors are

fixed as well; the complete intermolecular vector set is listed below:

1—2: 2x 2y -1/2 1-6: (x+y) (y+x) -1/2+22
1-3: -1/2+(x+y) —1/2-(x-y) -3/4 1-7: -1/2+2x -1/2 —3/4+22
1-4: —1/2+(x-y) -1/2+(x+y) -1/4 1-8: —1/2 -l/2+2y -1/4+22
1-5: (x-y) (y-x) 22

...(25)

82

The positioning of the thrombin model in the hirudin complex unit
cell was achieved by means of a set of programs written by E.E. Lattman
(and modified by J. Deisenhofer, R. Huber, and M. Schneider) which
calculates a translation function as described by Crowther and Blow

[130]. The Crowther and Blow translation function has the form:

n-l
Ts(t) = x (lFobs(h)|2 - 2 |Fm(h*Ai)|2)
h i=0
x Fm(h)*Fm’(h*A)*exp(-2n*i*h*t) ...(26)

where h represents the triple index (hkl); Fobs(h) and Fm(h) are the
observed and model structure factors at triple index h and are on an
absolute scale; Fm'(h*A) is the complex conjugate of the model structure
factor of a symmetry-mate molecule related by the symmetry operator A; t
is the translation vector; and the term |Fm(h*Ai)|2 subtracts out the
intramolecular vectors from the observed Patterson function at each of
the n—l symmetry related positions using the matrix Ai, A0 being the
identity matrix of the model. The above translation function Ts(t) is
specific for the intermolecular vector at the translation t between a
model molecule and its symmetry-mate related by the symmetry operator A;
there will be n-l such functions for the n-l unique intermolecular
vectors relating a given molecule to its n-1 symmetry-mate molecules in
the unit cell of a crystal.

Thus, the Crowther and Blow translation function makes use of the
intermolecular vector equations defined by the space group of the
crystal (relation 25 above) to determine the correct position of the
properly-oriented model in the crystal cell. Basically, for each unique
intermolecular vector translation function, all the intermolecular

vectors were calculated for the model and its appropriate symmetry mate.

83

Each resulting set of intermolecular vectors is then translated
throughout the crystal Patterson synthesis modified to remove the
intramolecular vectors. A product function is calculated between the
modified crystal Patterson map and the model intermolecular vector set
at each point in the translation search. At the correct position of the
model in the crystal, the product function will have a maximum value.
In many cases, as in the hirudin-thrombin Search, if all the translation
functions for the set of unique intermolecular vector equations are
calculated, the correct position of model in the crystal cell will be
overdetermined. In other words, the correct position of the model
should account for a high product function (if not the highest) peak in
each of the intermolecular translation function searches.

The translational searches were conducted using 8.0 to 3.0 A
resolution crystal and thrombin model structure factors, and a
translation function calculation was conducted for each unique
intermolecular vector in the P43212 space group (relation 25 above).
The results of the search are listed in Table 5. The highest peak in
all of the calculations occurred in Harker sections at consistent
locations which positioned the molecule at x = 67.42 A, y = 3.00 A. and
z = 52.91 A, and verifed the correct enantiomorphic space group to be
P43212 as opposed to P41212. The correct position of the molecule also
gave an outstanding peak in each Harker section; the highest peak in
each section was on the average 13.30 above the mean, and 7.50 higher
than the next highest peak. The R-value for this solution, where the

R-value is defined as

R-value = Z IIFobs(h)|-|Fcal(h)ll/ Z |Fobs(h)l,
h h

...(27)

84

was found to 0.39.

The rotational and translational parameters were further refined
with the rigid body refinement-Fourier transform fitting program TRAREF
[132]. The program TRAREF refined the orientational and translational
parameters of a thrombin model by fitting its molecular Fourier
transform or search model structure factors to the observed
hirudin-thrombin structure factor amplitudes. Model structure factors
were calculated based on the position of the model in the crystal cell
determined from the translation search (Table 5). The fitting was done
by calculating derivatives of the molecular Fourier transform with
respect to the three orientational angles, the three positional
parameters, a scaling factor, and an overall temperature factor. The
system of equations was then reduced to normal equations and improved
parameters were determined. The TRAREF refinement shifted the center of
gravity position for the thrombin model in the crystal to x = 67.44 A, y
= 2.87 A, and z = 52.87 A, and the R-value for this new postion was

0.34.
D., The Initial Hirudin—Thrombin Electron Density Hap Calculation

An initial hirudin-thrombin Fourier synthesis at 8.0 to 3.0 A

resolution was calculated using coefficients of the form

V*(2|Fobs(h)|-|Fcal(h)])*exp(i*ac(h)),

...(28)

where h represents the triple index (hkl); [Fobs(h)] is the observed
structure factor amplitude at triple index h; |Fcal(h)| and uc(h)
represent the model structure factor amplitude and phase at triple index

h calculated from the TRAREF-refined position of the thrombin model in

85

Table 5. Results of the Translation Search

Second
Harker Highest peak highest peak Position of
Vector section Position (A) Height Height the molecule
U V V (a) (a) (A)
1—2 0:1/2 43.94 6.00 66.30 13.8 6.7
1-3 V=3/4 24.87 72.01 33.21 13.9 4.7
1-4 V=1/4 18.98 24.87 99.38 13.9 4.8 x=67.42
1-5 - 64.52 26.46 105.75 12.5 6.7 y: 3.00
1-6 - 70.28 70.41 39.36 13.0 6.6 2:52.91
1-7 V=1/2 89.26 45.50 6.36 13.6 6.2
1-8 U=1/2 45 50 51.72 72 81 12 3 5.2

Peak height is given in standard deviations above
of the function.

The vectors are as follows:

the mean value

1-2: 2x, 2y, -1/2 1-6: (x+y), (y+x), -1/2+22
1—3: -1/2+(x+y). -1/2-(x—y), -3/4 1-7: -1/2+2x, -1/2, -3/4+22
1-4: -1/2+(x~y), -1/2+(x+y), -1/4 1-8: -1/2, -1/2+2y, -1/4+22

1-5: (X-y). (y-x). 22

86

the crystal; and V represents the Sim weight [133,134] for the model
phase angle ac(h). The map size was 100 x 100 x 140 grids, which
corresponded to 0.91 A/grid along the a and b axes, and 0.94 A/grid
along the c axis.

Such so-called "2Fo-Fc" maps are commonly used in macromolecular
crystallography when one is working with an incomplete structure, as in
this situation where the phasing model lacked hirudin and solvent
structure. The map is basically a point-by-point summation of an F0 and
an (Fo-Fc) or delta F synthesis, and is useful for several reasons.
First of all, like an F0 synthesis, it displays the input model. In
addition, the map also indicates how additional structure can be
accomodated, or whether existing structure can be reoriented or removed
from the present model, like a delta F synthesis [135].

The Sim weighting scheme for the electron density map calculation
was designed to weight each ac(h) by the probability of it being
correct, and on the notion that the best weights for structure
amplitudes minimize the mean square error in the electron density map
due to phase angle error. Based on previous work by Blow and Crick

[136], Sim showed that such weights are defined as

2n
V = I cosEp(E)dE, ...(29)
0
where
E = a(h)- ac(h) ...(30)

and the probability that the phase difference lies between E and E + dB
is given by the probability function p(E). This probability function

has the form

87
p(E) = exp(X*cosE)/2n*IO(X), ...(31)
where 10 is a zero-order modified Bessel function, and X is given by
X = 2|Fobs(h)|*|Fcal(h)|/Iu ...(32)

in which Iu represents the contribution from the "unknown structure" or
the atoms in the structure not included in the model. In the Sim
weighting scheme used to produce the hirudin-thrombin trial electron

density map, Iu was approximated by the quantity

~ 2
Iu = <||Fobs|‘ _ |Fcal|2|> = fillFobs(h)2 - Fcal(h) Il/N.
...(33)

where N is the number of observed structure factors included in the
calculation, and Iu was calculated in 10 shells of equal thickness in
reciprocal space. The Sim weight definition given above in equation

(29) can be further simplified to the form
v e Il(X)/IO(X), ...(34)

where 11 is a first-order modifed Bessel function. The Sim weight is
zero when X=0 and 1.0 when X is infinity; a plot of the Sim weighting
factor V against X is given in Figure 25 [137].

The first Fourier map was displayed on an Evans and Sutherland
P5390 interactive graphics display system and it was encouraging, as it
clearly had density to accomodate hirudin as well as thrombin. The
initial model-building which was conducted using this density map, as
well as further graphics interventions and EREF energy refinement which
lead to a mildly-refined 2.3 A resolution hirudin—thrombin structure

[127], will be summarized next.

88

 

0.9 '-

08 r-

06 r-

oa~

0.2 r—

O.| r-

 

 

 

Figure 25.

Plot of Sim weighting factor V versus X where

H-Il(X)/IO(X) and X=2|Fobs(h)|chal(h)|/Iu (taken from
[137]). Il(X) and IO(X) are first- and zero-order modified
Bessel functions, respectively. Iu is an estimate of the
contribution from the unknown structure, or that which is
not accounted for in the model.

V NOBEL-BUILDING AND EREF REFINENENT

The hirudin-thrombin complex was initially refined to an R—value of
0.193 with 7.0 to 2.3 A resolution data by an iterative process of
model-building and EREF refinement [127]. The model-building work was
conducted interactively using the PSFRODO version [138] of the molecular
graphics program FRODO [139] by fitting protein and solvent atoms in
Fourier maps viewed on a Evans and Sutherland P5390 graphics display
system. The model—building was initiated on the 3.0 A (2Fo-Fc)
Sim-weighted Fourier map resulting from the Patterson search
calculations with the thrombin model. The map clearly displayed density
for thrombin, which most of the model thrombin structure could easily
accomodate, and had a significant fraction of additional density for the
recombinant hirudin. At this point, the coordinates of the hirudin
structure determined using two—dimensional NMR techniques by Folkers et
al. [79] proved to be very useful in tracing residues 3’ to 48’ ( where
a prime after a number designates a hirudin residue) of the NHZ-terminal
domain of hirudin. The most exciting aspect of these initial
model-building studies was to see continuous density for the last ten
COOH-terminal residues of the hirudin molecule, especially for the
sidechains of Phe 56’ and Tyr 63'. This initial hirudin-thrombin model
was refined using the energy-restraint crystallographic refinement
program EREF [140], and the resulting EREF structure was inspected and
modified at the interactive graphics display. Sim-weighted density maps
- positive (2Fo-Fc) and negative as well as positive (Fo-Fc) Fourier
maps - were used in the course of model-building. The cyclic
model—building and refinement procedure was repeated six times with

gradual increases in resolution to 2.3 A; the pertinent details of this

89

90

work are summarized in Table 6. In the second and later stages of this
process, the FAST area detector structure factors were used solely.
Water molecule oxygen atoms were inserted in the third stage, and in
later stages. If a strong peak (generally > 3.50(Ap)) occurred in both
the 7 to 2.3 A resolution (2Fo-Fc) and (Fo-Fc) Fourier maps, and was in
a stereochemically reasonable position for a water molecule, a water
oxygen atom was inserted. Individual temperature factors were also
refined. After the six rounds of model-building and refinement, the
structure of the complex had an R-value of 0.193 for 21,960 reflections
between 7.0 and 2.3 A resolution. The structure included 207 water
molecules, had a mean individual temperature factor of 35.7 A2, and the
r.m.s. deviations for bond lengths and bond angles were 0.013 A and
2.4°, respectively. The final model parameters of the EREF-refined
hirudin-thrombin complex are summarized in Table 7.

The EREF macromolecular refinement program simultaneously minimizes
a realistic potential energy function and a crystallographic residual.
The complete potential energy function includes terms for bond
stretching, bond angle bending, torsion potentials, and non- bonded and
electrostatic interactions, and has the following form [141]:

E e z 1/2Kb(bi — b0)2 + r 1/2Kt(ti — :0)2
bonds bond angles

2 Ko{1 + cos(m9i + d)} + r (A/ri12 + B/ri°) + z qi*qj/r
torsion angles non-bonded
interactions

...(35)
where Kb is the bond stretching force constant; b0 the equilibrium bond
length; Kt is the bond angle bending force constant; to is the

equilibrium bond angle; K0 is the torsion barrier; m is the periodicity

of the torsion barrier; d is the phase of the barrier; and A and B are

91

Table 6. Course of Model-Building and EREF Refinement

Stage R-value Resolution

No.
of
atoms

(kcal/mole)

Energy

(A)
0 0.349 3.0
1 0.245 2.3
2 0.291 2.5

3 0.261 2.3

4 0.221 2.3

5 0.205 2.3

6 0.193 2.3

2495

2495

2677

2787

2848

2915

-1458

-1721

-1763

-2175

-2128

-2221

Rotation, translation
structure; Nicolet P3/F
data set used.

Model rebuilt on

display (1); 167 protein
atoms added; 8 cycles

of EREF including
resolution extension.

Model rebuilt on

display (2); 9 cycles

of EREF including
resolution extension and
2 cycles of B-factor
refinement; FAST data set
used.

Model rebuilt on

display (3); 182 protein
atoms added; 14 cycles
of EREF including
resolution extension and
4 cycles of B-factor
refinement.

Model rebuilt on display
(4); 17 protein atoms
and 93 water molecules
added; 15 cycles of EREF
and 4 cycles of B-factor
refinement.

Model rebuilt on display
(5); 10 protein atoms
deleted and 71 water
molecules added; 19
cycles of EREF
refinement and 4 cycles
of B-factor refinement.

Model rebuilt on display
(6); 24 protein atoms
and 43 water molecules
added; 13 cycles of EREF
and 4 cycles of B-factor
refinement.

92

Table 7. Final Model Parameters of the EREF-refined

rHV2-K47 Human aeThrombin Structure.

Number of protein atoms
a) Thrombin
b) Hirudin
Number of solvent atoms
r.m.s. deviation from target values

Bond lengths (A)
Bond angles (deg.)

Total internal energy (kcal/mole)
Resolution range (A)

Number of unique reflections
used for refinement

R-value

Mean B value (A2)

2296
411
207

0.013
2.36
-2221

7.0 — 2.3

21960
0.193
35.7

93

the coefficients of the Lennard—Jones potential for non-bonded
interactions. The Lennard-Jones potential parameters used were those
specified by Levitt [141], and the values of the force constants were
derived from the vibrational spectra of small molecules. Moreover, the
bond angle and torsion angle force constants were corrected for the
missing hydrogen atoms in the structure [142]. The crystallographic

residual has the form

x e r (|Fobs(h)| — |Fcal(h)|)2.
h

...(36)

A cycle of EREF refinement proceeds in three steps [142]. In the
first step, the system of normal equations for the diagonal-matrix
least-squares refinement of atomic parameters is developed from an
(Fo-Fc) difference Fourier map. In step two, the potential energy of
the model is minimized along with a crystallographic residual to refine
atomic positions. If atomic temperature factors are refined, this
minimization is bypassed. In the third step, new structure factors are
calculated from the model resulting from step two.

The system of normal equations generated in step 1 of a refinement
cycle uses the (Fo-Fc) difference Fourier map to minimize the
crystallographic residual in equation (36). These equations can be

compactly represented by the expression
A*dp = b ...(37)

where dp represents the vector of desired parameter shifts (positions,
occupancies, B-values); the matrix A is approximated by a diagonal
matrix in which the diagonal elements are estimated for each atom type.

The components of the vector b, which represent the shifted parameters

94

of vector dp, are obtained by convoluting the Fourier transform of the
atomic form factor (i.e. the density about a given atom) with the
gradient of the (Fo-Fc) map around the atomic site in the case of a
positional refinement, or with the difference map itself in the case of
occupancy or B-factor refinement. The convolution is generally
calculated in a box of 3 x 3 x 3 grid points around the atomic
positions, and the grid spacing used is approximately one-third of the
maximum resolution in angstrom units.

The function which is minimized in step 2 of an EREF cycle is
R = E + kX ...(38)

where E represents the conformational energy of equation (35), X
represents the crystallographic residual of equation (37), and k is a
weighting factor which can be used to emphasize or de-emphasize the
diffraction pattern in the minimization process. For example, when k a
0 a pure energy refinement is conducted, and when k is large, a pure
crystallographic refinement is performed. Generally, during the initial
cycles of a refinement run, the k used was 5 x 10", and it was
gradually increased in the following cycles from 3 x 10‘5 to up to 7 x
10'5 in the last cycles.

The final step of the EREF cycle run involved generating a new set
of calculated structure factors based on the energy minimization results
or B-factor refinement conducted in the cycle. The FFT (Fast Fourier
Transform) program of Ten Eyck [143,144] was used for this purpose. The
program initially produces a model electron density map from which it

calculates the structure factors by a Fourier inversion.

VI MODEL—BUILDING AND PROLSO REPINENENT

Using the final coordinates of the EREF-refined hirudin—thrombin
structure as a basis (Table 7), the iterative model-building and
refinement procedure was continued using the restrained least squares
refinement program of Hendrickson and Konnert [145]. This program,
PROLSO (an acronym for PROtein Least SOuares), incorporates
stereochemical information into the refinement process by treating this
knowledge as additional observations. Basically, PROLSO minimizes a
grand function 6 a E 6i, where contains 61 functions for: a)
structure factors, b) distances, c) planar groups, d) chiral centers, e)
non-bonded contacts, f) torsion angles, g) temperature factors, and h)
non-crystallographic symmetry. Each piece of information is treated as

an observational equation,
g-obs = g-calc(x) + e ...(39)

where g-obs is the observation, g-calc(x) is the theoretical value
computed from the parameters x of the model, and c is the discrepancy
between the two [146]. Moreover, as the refinement is based on the
principle of least squares, the ’best' set of parameters for equation
(39) are those which minimize the weighted sum of squared residuals

taken over all the observations,

6(x) = 2 whlgh—obs - gh-calc(x)]2 ...(40),
h

in which the appropriate weights wh are the inverses of the variances of

the observations.

The decision was made to further refine the hirudin-thrombin

coordinates, and to use the PROLSO program in particular, for several

95

96

reasons. First of all, PROLSO has been shown over the years to be
effective in producing stereochemically-sound structures of biological
macromolecules. Moreover, in particular, PROLSO refinement circumvents
two serious shortcomings in the EREF refinement. Temperature factors
are not restrained during B-factor refinement in EREF, and in some
instances the difference in B-factor between connected atoms in the
complex is unreasonably large, as much as 30 A2 or more. Also, the
final coordinates of the complex contain 40 w angles which differ by
more than 10° from either side of 180° - one angle as
small as 148° - since only moderate energy penalties are placed
on such deviations from planarity. As the function minimized by PROLSO
has terms for thermal parameters and planar groups, the aforementioned
problems can be averted by making an appropriate choice of target a
values, and therefore weighting factors for equation (40). In addition,
upper and lower temperature factor limits were set in the program so
that B-factors were not allowed to drop below 15 A2 or rise above 60 A2.

Another important reason for choosing to refine the complex further
by means of PROLSO is the inherent flexibility of this program compared
to to EREF. By virtue of the different (a-h) functions which comprise
the grand function minimized by PROLSO, as well as the ability to choose
the weighting factors for these functions, a refinement protocol can be
tailor-made. Through a judicious choice of weighting factors, one can
tightly, moderately, or loosely restrain any or all of the (a-h)
observational functions; one can also refine atomic occupancies
(generally only performed on solvent molecules) in addition to positions
and temperature factors, as well as introduce and adjust the extent of
shift damping to be applied to parameters during the refinement. The

EREF program, on the other hand, affords very little refinement

97

flexibility; one can simply adjust the crystallographic residual
weighting factor k in equation (38), choose positional or B-factor
refinement, and control the extent of damping applied to the recommended
parameter shifts.

Two new steps were also taken in this second phase of
model-building and refinement. The first unit of the carbohydrate
moiety in thrombin (Figure 3), N-acetylglucosamine (NAGl), was modeled
into a sufficiently large region of 2Fo-Fc electron density near the
side chain atom NDZ of Asn6OG and included into the refinement (Figure
26). The manner in which water molecules were located and refined in
the crystal structure was different. In addition to the criteria used
in the EREF refinement and model-building for choosing possible solvent
atom positions {good 2Fo-Fc density and significant (generally > 3.50)
Fo-Fc density over the resolution range being studied (7 A to the
Inaximum resolution, 2.5 or 2.3 A), as well as proximity to a
Daydrogen-bonding donor or acceptor atom], it was also required that the
[mosition have significant (generally > 3.50) Fo-Fc density over the
resolution range 8 A to the maximum resolution. Refinement of the

tscalvent positions also included alternating block cycles of B—factor and
Occupancy refinement along with positional refinement.

The interactive model-building, as in the previous work, was
cairried out on an Evans and Sutherland P5390 graphics display using the
PSF'RODO version of the graphics program FRODO; however, the 2Fo-Fc and
I:'<>~Fc maps used were not generated using Sim-weighted phases. In
addition, the restrained least squares refinement was almost exclusively
Q0“ducted using a modified version of PROLSO known as PROFFT, which
incorporates fast Fourier transform-based algorithms to speed—up the

Q0'Ilputation of structure factors and least squares matrix elements, and

98

  
   

    
  

I D
01A)

:0,
'6‘

   
 
 

“ - mung“
I .‘

 
  

       

V4}

1

Stereoview of the First Unit of Thrombin Carbohydrate
Modeled into Map Density. Displayed in bold is
N-acetylglucosamine (NAGl) linked to the side chain atom ND2
of AsnGOG. Map density is from the final 2Fo—Fc map used
for model-building; contours at 0.80. Atomic positions are
from the final coordinates.

Figure 26.

99

which can restrain intermolecular contacts [147].

A summary of the model-building and PROLSO refinement is presented
in Table 8. Refinement was initiated using the final coordinates of the
EREF-refined hirudin-thrombin structure (including protein atoms and
water molecules) in which all the atoms were given an individual
B-factor of 35.7 A2, the average individual B-factor of the final EREF
structure (Stage -3, Table 8). After two initial rounds of refinement
(Stages —2 to -1, Table 8) involving 27 cycles {19 cycles employing
tight (T) restraints, 8 employing loose (L) restraints; target 0 values
used to produce the T and L restraint weights in the refinement are
given in Table 9} and a model-building session, however, a new direction
had to be taken as the refinement was ill-conditioned. Large coordinate
and solvent occupancy shifts were occurring which got progressively
larger in size and number, leading to a severe degradation in the
stereochemistry and divergence in the refinement. In an effort to avoid
these refinement problems, two steps were taken. After difficulties
arose in the first round of refinement (Stage -2, Table 8), 857 weak
reflections which gave rise to extremely poor R-factors were removed
from the reflection file; however, problems persisted in the next stage
of refinement (Stage -1, Table 8). As a second attempt to resolve these
problems, the refinement was re-started without any of the solvent atoms
of the final EREF structure; only the protein and NAGI carbohydrate atom
coordinates from the first graphics rebuilding session (Stage 0, Table
8) were used.

Removal of all the previous EREF-based water molecule oxygen atoms
terminated the occurrence of large positional shifts in the protein
atomic positions. In the first stage of refinement after this re—start,

following only 10 cycles of refinement (8 T and 2 L restraint cycles,

100

Table 8. Summary of Model-Building and PROLSO Refinement.

Abbreviations: T, M, L a tight, medium, and loose refine-
ment restraints, respectively (see Table 9); 0 = occupancy;
NAGl a N-acetylglucosamine; Bi = individual B (temperature)
factor; <> s average; d = d-spacing. Final model parameters
given in Table 10.

Resolution
Stage R—value Maximum Total Comments
(A) Atoms/Waters

-3 0.282 2.5 3056/228 EREF structure, all Bi a 35.7 A2.

-2 0.237 2.5 3002/174 17 refinement cycles conducted (12
T, 5 L), 4 used (all T). (Bi)-
36.5 A . Refinement ill-
conditioned: cannot recover from L
cycles; large positional shifts on
protein & solvent atoms; low occu-
pancy (Q < 0.4) solvents deleted.

-1 0.224 2.5 3022/228 Model rebuilt on display (1): 14
atom NAGl sugar unit positioned
near Asn60G; 54 new waters added;
48 protein atoms removed. 857
reflections (weak or yielding poor
R-factors) removed. 10 refinement
cycles conducted (7 T, 3 L). <Bi>=
36.0 A2. Refinement ill-condi-
tioned: large positional and occu-
pancy shifts. REMOVE ALL SOLVENTS
AND RESTART.

O 0.274 2.5 2795/0 Protein & NAGl coordinates from
model rebuild (1). (Bi): 35.9 A2.

1 0.232 2.5 2795/0 10 cycles of refinement (8 T, 2 L)
conducted; (81): 35.5 A2.

2 0.202 2.5 2855/76 Model rebuilt on display (2): 76
new waters located; 16 protein
atoms removed. 30 refinement cycles
(22 T, 8 L) conducted; 16 (14 T,
2 L) used; variable structure factor
weighting applied hereafter. <Bi>=
34.5 A2.

3 0.199 2.3 2924/132 Model rebuilt on display (3): 56 new
waters located; 13 protein atoms
added. 35 refinement cycles (29 T.
6 L) conducted; 32 (27 T, 5 L) used.
Resolution extension to 2.3
<Bi)= 34.7 A2.

Table 8.

Stage R-value

4

5

(cont’d.)
Resoluti
Maximum
(A)
0.185 2.3
0.173 2.3

101

on
Total
Atoms/Waters

2987/202

3048/265

Comments

Model rebuilt on display (4): 70 new
waters located; 7 protein atoms re-
moved. 47 weak reflections removed,
2.56 A > d > 2.30 A. 12 refinement
cycles (12 T) conducted; 9 used.
<Bi>= 35.6 A .

Model rebuilt on display (5): 63
new waters located; 2 protein atoms
removed. 32 cycles of refinement
(11 T, 21 M) conducted; 17 used

(a T, 13 M). <Bi>= 35.3 A2.

102

Table 9. Summary of Hirudin-Thrombin PROLSO

Refinement Restraint Parameters.

Target 0 values listed for loose, medium, and tight
refinement cycles. Veights used in refinement correspond
to 1/02. Two values separated by a ’/' indicate

that the first listed 0 value was used in early refinement

situations while the second listed value was used later.

Loose Medium Tight
Distances (A):
bond lengths 0.030/0.025 0.020 0.015
bond angles 0.050/0.040 0.035 0.030
planar 1-4 0.060/0.050 0.045 0.040
Planes (A):
peptides 0.040 0.020 0.015
aromatic groups 0.040 0.020 0.015
Chiral volumes (A3): 0.200 0.150 0.150
Non—bonded contacts (A):
single torsion 0.55 0.55 0.50
multiple torsion 0.55 0.55 0.50
possible (X..Y)
hydrogen bond 0.55 0.55 0.50
Isotropic thermal
parameters (A2):
main chain bond 0.5 0.5 0.5
main chain angle 1.0 1.0 1.0
side chain bond 0.5/1.0 1.0 0.5/1.
side chain angle 1.0/1.5 1.5 1.0/1.

Target 0 Values

UIO

103
Table 10. Final Model Parameters of the PROLSO—refined
Hirudin-Thrombin Complex.
The directional arrow and value enclosed in brackets indicate
the change with respect to the final hirudin-thrombin EREF

structure value.

Protein atoms:

Thrombin: 2322 (T 26)
Hirudin: 447 (T 36)
Carbohydrate (NAGl): 14 (T 14)
Solvent atoms: 265 (T 58)
Resolution range (A): 7.0-2.3
Unique reflections: 21056 (1 904)
R-value: 0.173 (6 0.020)
Mean B-value (A2): 35.3 (I 0.4)

Geometrical Conformity

r.m.s. deviation in
Bond angles (deg.): 2.8° (t 0.4)

r.m.s. deviations from target values:

target model
Distances (A):

bond lengths 0.020 0.021

bond angles 0.035 0.051

planar 1-4 0.045 0.052
Planes (A):

peptides 0.020 0.017

aromatic groups 0.020 0.015
Chiral volumes (A3): 0.150 0.198
Non-bonded contacts (A):

single torsion 0.55 0.22

multiple torsion 0.55 0.27

possible (X..Y)

hydrogen bond 0.55 0.27

Isotropic thermal
parameters (A2):

main chain bond 0.5 0.6
main chain angle 1.0 1.1
side chain bond 1.0 1.2
side chain angle 1.5 1.9

 

104

Stage 1, Table 8) the R—factor dropped 0.042 from 0.274 to 0.232, and
there was an overall improvement in the stereochemical conformity of the
structure. These results clearly imply that at least some of the
solvent molecules in the final EREF structure coordinates were the
source of the previous ill-conditioned refinement behavior. Moreover, a
possible reason why some of the EREF structure water molecules could be
causing problems in the PROLSO refinement is likely linked to the
dramatic difference in the functions being minimized by the two
programs. The EREF program minimizes a potential energy function along
with a crystallographic residual, and likely the energetics of salvation
play a strong role in refined water positions. The program PROLSO, on
the other hand, has no direct energy restraints, with refinement being
guided primarily by stereochemical restraints. In addition, when one
also considers that these crystallographic water molecules have a
scattering power of about ten electrons, are generally not fully
occupied, and have on an average higher thermal parameters than the
protein, it seems reasonable that the EREF and PROLSO refinement
programs would treat such minor contributors to the overall scattering
power of the protein crystal differently.

The model-building and refinement progressed very well over the
next four stages (Stages 2-5, Table 8). Water molecule oxygen atoms
were added and refined in each stage until a total of 265 were included.
Also of note is the fact that the resolution of the refinement was
extended to 2.3 A for the last 3 stages of the work, and 47 weak
reflections (many yielding poor R-factors) of resolution 2.56 A > d >
2.30 A were removed from the reflection file. Finally, a set of
refinement restraints intermediate to loose (L) or tight (T), so called

'medium' (M) restraints (Stage 5, Table 8; Table 9), were used in the

105

last stage of refinement to arrive at the final hirudin-thrombin
coordinate set. Overall, in the course of these last four rounds of
model—building and refinement ( 54 T, 13 M, and 7 L restraint cycles
used), the R—factor decreased by 0.059 from 0.232 to 0.173, 12 protein
atoms were removed, 265 water molecules were added, and the average
individual B—factor decreased slightly from 35.5 to 35.3 A2 (Table 8).

An examination of the final model parameters of the PROLSO-refined
hirudin-thrombin complex, along with changes in these values with
respect to the final EREF-based complex coordinates (Table 10), reveals
that the second phase of model-building and refinement work was clearly
worthwhile. The final PROLSO-based complex coordinates contain 3048
total atoms: 2769 protein atoms (2322 in thrombin, 447 in hirudin), 14
carbohydrate atoms, and 265 water molecules. This atom total represents
an increase of 134 over the number of atoms present in the final
EREF-based coordinates; moreover, the increase is distributed as
follows: 62 more protein atoms (26 in thrombin, 36 in hirudin), 58 more
water molecules, and 14 N-acetylglucosamine atoms. Of particular
importance is the fact that the R-factor was lowered significantly and
the mean B—value decreased as well; the R-factor improved 0.020 from
0.193 to 0.173 and the mean B—factor dropped from 35.7 to 35.3 A2.
Also, the final PROLSO refinement reflection file had 904 fewer
reflections than that used in the EREF refinement work; these
reflections were removed because they were of low intensity and
generally yielded high R-factors.

The geometrical conformity of the PROLSO-refined hirudin-thrombin
structure is also quite good (Table 10). The r.m.s. deviation of bond
angles is 2.8°, 0.4° larger than the value for the final

EREF complex structure. The r.m.s. deviations on many restrained

106

paramaters are also generally in good agreement with the moderate
restraint target values used in the final stages of the refinement
(Tables 9, 10). The deviations on bond, angle, and planar 1-4 distances
are close to, though generally larger than, the target values, with the
deviation on bond angle distances displaying the greatest discrepancy
(Table 10). The peptide and aromatic group planarity is excellent, and
the r.m.s. deviations on these planar units are below the target values
(Table 10). The histogram of w angles (twist along the C-N peptide
bond) in Figure 27 attests to the high degree of peptide bond planarity
as better than 91% of the bonds have m angles of 180° +/- 5°. In
addition, the Ramachandran plot of the 6 (twist about CA-N axis) and 6
(twist about CA-C axis) angles for non-glycine residues in these
coordinates, shown in Figure 28, indicates that the vast majority of the
residues have ¢.W angles within conformationally-allowed regions. Only
three residues in the complex, which all reside in the thrombin A-chain
(SerlE, Phe7, and Asp14L), deviate significantly from the Ramachandran
allowed regions.

The stereoconfiguration at chiral centers is satisfactory; the
r.m.s. deviation on chiral volumes is 0.198 A3, which is reasonably
close to, but noticeably larger than, the target value of 0.150 A3
(Table 10). On the other hand, the r.m.s. deviations on all non-bonded
contacts are outstanding, as all three contact types have deviations
which are less than half the target value of 0.55 A (Table 10). The
low r.m.s. deviations on non-bonded contacts could in part be due to
the fact that an EREF structure was used to inititate the PROLSO
refinement, and that the EREF program, which minimizes a potential
energy function containing terms for non—bonded contacts, did well in

idealizing these non-bonded interactions. Finally, the deviations on

107

.xoaqeoo :Beougetfiuafiz 05 5 mafia: 3 we 33:3“: .3 6.28:
mmozq comzo

 

 

 

 

 

 

 

 

 

 

& .mmd D .Nh— S fix..— 0 du— O .muul a 68.7 0 .nhul 5 .Dhnl
_ —I

all!

0“ GM ILII

.Mn I?!

if.

11

III—dull. l4].

.mm .l:l.

LI-

Lr.

.1"

11

1T

III

#1 Lu

 

.0
.n
d«
flu
6N
.nN
dn
.mn
6v
.nv
in
.nn
do
60
6h
6h
.60
.no
in
.mm
.8 n
.3"
.0. n
.n— n
.8 n
.8 a
.GM u
.3 a
.3"
.0. n
.8"

SUOBHO JO HBBHON

 

107

.5338 533351533: 6.: 5 moan: 3 mo 30.333:

a .mm.- 0 Nb" a .05—
_ _

msozq cowzo

ads—

0 .m~ — I a 6%.:

0.9::
pl

.x~ assess

a .27.

 

13!!

.M n

 

 

.mm

 

 

 

 

.Nfifi

111414
117

ITIIIIUTI

lllllll

[Fill

[IT]

 

JllJllllllglllllJllllll

.9
.n
.0 n
.n «
6N
.nN
dn
.nn
.0?
.0?
6n
.mn
dc
6m
.Dk
fit
60
do
6m
.nm
.8 n
.8"
.0. n
.m— n
.8 n
.nNn
.8 n
.3 a
.00 u
.9 n
.8"

SUSBHO 30 838NON

108

 

 

 

 

 

 

 

 

a. an. lea.
I L 1 1 1 1 m.
I
1
I 1'-
I
1
I _
l
1 ,1®I4L Asp
I I 90.
1 d '
1 1 1
| 1 1 _
1 "I 4
I
K g ..
\\ 1'
\II
o.
1"”.
-3. .......... ,
"”- l 1' 1 1 1 r 1 1 1 f 1 -1ao.
-10a 40. 0. so. 180.
PHI

Figure 28. The Ramachandran Plot of Hirudin—Thrombin. Only non-glycine
residue angles are displayed.

109

isotropic thermal parameters among main chain and side chain atoms
involved in bonds and angles are all fairly close to the target values,
though they are slightly larger (Table 10).

Data pertaining to the R-factor versus resolution shell (seven
shells were used, containing approximately 3000 reflections per shell)
for the final model parameters of the complex are presented in Table 11
and plotted in Figure 29. The nearly linear, direct correlation between
the shell-R and resolution is typical of well-refined structures.
Moreover, the fact that the R-factor versus resolution data point curve
of the complex falls between the lines corresponding to the Luzzati
coordinate error estimates [148] of 0.20 A and 0.25 A, suggests that the
average error in the atomic positions for the final PROLSO-refined
coordinates is likely close to these estimates (Figure 29). Luzzati
plots are based on the assumption that the discrepancies between
observed and calculated structure factor amplitudes are due solely to
the coordinate error, Ar; other possible sources of error, such as
inaccuracies in the measurement of Fobs, are neglected [148]. Thus, the
actual error in the coordinates of the complex is probably somewhat
larger than 0.25 A.

While the average individual isotropic temperature factor, <Bi>,
for all the atoms in the hirudin-thrombin complex is 35.3 A2 (Table 10),
it is much more informative to consider the <Bi> of each component of
the crystal and how the (Hi) varies within each of these components.
Values of <81) for the various constituents of the hirudin-thrombin
crystals are listed in Table 12. Not surprisingly, the <Bi> for
thrombin (32.1 A2) is significantly lower than that for hirudin (43.3
A2), and the thrombin B-chain atoms have a lower <Bi> than those of the

A-chain (31.5 A2 for the B-chain versus 37.3 A2 for the A-chain). Since

110

Table 11. The R-factor vs. Resolution for the
Final Hirudin-Thrombin Model Parameters.

The dmin vs. shell R-factor is plotted in Figure 29. The
directional arrow and number enclosed in brackets next to
shell R-factor values for the the unhydrated structure
indicate the change with respect to the corresponding
values for the hydrated structure.

dmin reflection <IFo-Fc|> sphere shell shell
(A) number R-factor R—factor R-factor

no waters
4.13 3043 89.19 0.150 0.150 0.239 (T 0.089)
3.39 3380 63.62 0.143 0.136 0.198 (T 0.062)
3.00 3335 55.61 0.151 0.176 0.243 (T 0.067)
2.75 3160 46.67 0.157 0.190 0.261 (T 0.071)
2.56 3112 42.60 0.163 0.212 0.267 (T 0.055)
2.42 2826 39.19 0.168 0.235 0.285 (T 0.050)
2.30 2200 38.25 0.173 0.265 0.304 (T 0.039)
overall:

21056 54.47 0.173 0.242 (T 0.069)

111

._qu_ “Lawns; cu mcfivuouuw Auqv muouuo uumcwvmoou

wmmuw>m cowmwowam om one name aucmawmumwc ovsufiaaem Louomu

wmsuusuum ucwmouamu moHouwo voHwa Hausa one .xwageoo

may no mucfioq name may unomwuaou moaomfiu vwaafiu omens
.cwneomzeuzwczmfim Lou :owusflowwm .m> monumWIm mo Loam .mu ousmfim

 

 

 

QQZGJ\IHZ~w

DON .o nNN .0 SN .0 n: .0 00‘ .6 DN— .6 00— .0

_ p e _ _ _ .on
_ a q _ _ a

v
11.n_
11.0“
<0N.OL< IIAhN

<mw.o.._<
Ila.”
11.nn
[1.0+
I 1. t. t. n u u I a.
and on.« “dad 8.8 mu." 8” 2.0

nauonPDJOwwm

Z NOIOUJ-d

112

Table 12. Average Individual B-factor and Average Occupancy
Factor Tabulations for the Hirudin-Thrombin Model Parameters.

Stage ’x' waters refer to water molecules located during
stage x of the hirudin-thrombin model-building and
refinement as outlined in Table 8. (The average occupancy
is not given for all the atoms as this value is of no
significance.)

Atom Type Atom <Bi> <0)
Number (AZ)

All atoms 3048 35.3 ----
All protein atoms 2769 34.7 1.00
Thrombin 2322 32.1 1.00
Thrombin A-Chain 253 37.3 1.00
Thrombin B-Chain 2069 31.5 1.00
Hirudin 447 48.3 1.00
All water molecules 265 39.8 0.78
Stage 2 waters 76 31.1 0.91
Stage 3 waters 56 37.8 0.81
Stage 4 waters 70 44.1 0.76
Stage 5 waters 63 47.2 0.63

113

hirudin is a small inhibitor protein and thrombin is a large serine
proteinase with a molecular architecture like chymotrypsin, one would
expect atomic vibrations of the atoms in the serine proteinase to be
lower than those of the inhibitor. Moreover, since the thrombin A-chain
is much smaller than the thrombin B-chain one might expect the B-factors
of the covalently liked peptide to be higher, on the average. However,
what is startling is the fact that the <Bi> for the hirudin atoms is
even higher than the average <Bi> for all the
crystallographically—located water molecules (39.8 A2) by 8.5 A2.

An examination of the <Bi> versus residue number for all of the
protein components in the crystal is even more informative. As can be
seen in Figure 30, the <Bi> varies significantly among the main chain
and side chain atoms within each protein unit. The smooth variation in
<Bi> for main and side chain atoms is likely a result of the isotropic
thermal parameter restraints used in the refinement (Table 9). However,
the magnitude and variation of <Bi> are likely reflective of the
conformational flexibility, crystal disorder, and stabilizing
interactions in the protein. The <Bi> for the side chain atoms is in
general larger than for main chain atoms because side chain atoms, by
and large, have more conformation flexibility than the main chain atoms.
Moreover, in several of the residues where the side chain <Bi> is lower
than that of the main chain, this occurs because the sidechain atom is a
cysteine sulfur involved in a disulfide linkage, which could certainly
explain the diminished thermal motion compared to that of the main chain
atoms (Figure 30).

Many of the ’high peaks’ in the Figure 30 plot correspond to
residues (most of which are in loop segments) which are poorly defined

in electron density, or are near disordered regions. In the thrombin

as
a.
gal.
.—
8.3.
Li.
C.
no
0:
Dis.
.—
(I
(Zn.
[LI
0.
z: z:
LLI
.—
fl.
1:
II.
a
I.
Figure

114

m1“ TWIN I-CMIN

 

 

 

HIRUDIN
A-CHAIN
X
1 x t
I x .
1 1 a: x v. “F. .7
I ‘ X
. xx xx x x ‘8 I I
-- i ‘ X ‘ &‘ 3 x 1, :
.> E n a . g X] :7 x ‘ x
- v ‘ ! In. Ax 5; I b h & ‘
1x - ' " A x I .
.. i ‘ § ' ‘ 3
__*x . 6 '. ‘1
~ .' k .
’ - ' a l‘ ' X
"" x x‘
0 x
l l I l l l J l l [I L l LLJ l l l I l l I l l I [I I ll
rII'I—IIFIIIFrIIIITIIIIIIIIIIII'I'II VII—T—
I“ I. I9 . on " . I. II, I“ I” I” I” III ””I' .' ‘0' “'0'
RESIDUE NUNBER
30.

Plot of the Average Individual B-factor of Main and
Side Chain Atoms for the Protein Atoms in the Complex. The
<Bi> for main and side chain atoms is represented by ’+’ and

'X' symbols, respectively. Main chain atom <Bi> values
connected by a line. The break in <Bi> in hirudin
corresponds to the completely undefined segment

Ser32'—Lys35’. Bold arrows point to main and side chain
<Bi> values of Cys residues engaged in disulfide linkages.

115

molecule, the N- and C—termini of the A-chain and the C-terminus of the
B—chain are all missing residues due to disorder in the crystals, and
the peak at Glu97A is indicative of the poor density for this residue
(weak main chain density and essentially no side chain density). In the
hirudin molecule, the residues possessing the largest main and side
chain <Bi> either adjoin the completely undefined loop residues
Ser32'—Lys35’ or are near the poorly defined tripeptide segment
Asn52'-Asn53’-Gly54'.

Finally, stabilizing interactions in the protein molecules of the
complex are probably responsible for some of regions of diminished
thermal motion in the Figure 30 plot. The stabilizing or damping effect
which the disulfide linkages have on the thermal motion of residues near
them is clearly visible, as the cystine linkages generally occur at or
near the minima. Also, the low main chain and side chain <Bi> values
for residues Ile1'-Tyr3’ and the trough for residues Asp55’-Gln65’ in
Figure 30 is probably due to the numerous hirudin—thrombin interactions
involving these regions of the hirudin molecule, as will be elaborated
on in the following chapter.

All the water molecules in the final PROLSO-refined coordinates of
the complex were located and refined over the course of four stages
(Table 8), as was mentioned earlier. Upon examination of the average
individual B-factor and average occupancy (<0)) for the water molecules
added in each of these four steps, an interesting trend was discerned
(Table 12). The magnitude of <Bi> increases and the magnitude of <0)
decreases with each successive round of water molecules added, with
values <Bi> = 31.1 A2 and <0) = 0.91 for the water molecules added in
the first stage of hydration (Stage 2, Table 8) and <Bi> = 47.2 A2 and

<0) = 0.63 for water molecules added in the last stage (Stage 5, Table

116

8). This trend seems reasonable if one considers that the difference
density peaks selected to be water molecules in the first round of water
addition were large in magnitude and significance, thus to
satisfactorily account for these peaks a water molecule would generally
refine to a low B-factor and a high occupancy factor. However, in the
last stage of hydration, the majority of bound solvent atoms have
already been located, and frequently many of the significant difference
density peaks are likely due to errors. In addition, the largest
difference densities are lower in magnitude than they were in the first
stage of hydration, and many of the suitable difference density peaks
will correspond to second hydration sphere water molecules (more
interaction with other water molecules than with protein atoms); thus,
the latter water molecules frequently refine to significantly higher
B-factors and lower occupancies. The histogram of water individual
B-factors in the complex (Figure 31) indicates that although the range
in individual temperature factors for water molecules is quite broad,
with 15 A2 < Bi < 55 A2, nearly half of the crystallographic waters have
40 A2 < Bi < 55 A2. The histogram of water occupancies in the complex,
Figure 32, shows that majority of the located water molecules are of
fairly high occupancy, with nearly 40 water molecules per occupancy
shell in each of the five shells between 0.5 and 1.0, and the largest
number of waters (55) having the highest possible occupancy, 1.0.

An examination of the R-factor versus resolution data in Table 11
calculated without the water molecules included clearly indicates that
the water structure makes a significant contribution to the overall
R-factor, and to the the R-factor of each resolution shell. The overall
R-factor for the unhydrated complex is 0.242, 0.069 larger than that of

the hydrated structure. In addition, while the effect of the water

117

.ousuosuum caneouze1cfivauum may :« moaaumuoz
Loans mnu we mucuomuum Hmacfi>ficcm may we enumODmfim .Hm ousmfim

maoeoqu mzwh

0.0» 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00. 0.0N 0.0N 0.0. 0.0" 0.0 0.0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_ r e _ p

_ _ _ _ a
.o. 6" 11

a." ,
.0“ 1.1
on
11.
dc
do

an . if

 

SNBIUH JO 338NOH

118

.o.H mo moccazooo

cm o>mn .H.H 1 o.H .Hawzm mucmasuoo ummH on“ a“ measuwaoe

kumxa

HH<

.musuusuum caneoune1eficsuwm one a“ measumaoz
Loam: mzu mo muouomm mocmqsuuo may no EMLNODmfim

>oz¢anooo

0.0 0.0 5.0 0.0 0.0 0.0 0.0 «.0 «.0
_
l

.Na ..eaae

 

 

 

on .—

 

 

On“

 

ann

 

 

 

 

 

 

.0» .00 Fl

out

 

 

 

on?

 

SHBIUH JO 838NON

119

structure is most significant in the lowest resolution shell (0.089 for
the 7.00 to 4.13 A resolution data), where ordered solvent typically
makes its largest contribution to x-ray scattering, the water
contribution is surprisingly large in all the resolution shells (Table
11). Even in the highest resolution shell (2.42 to 2.30 A), removal of
the water molecules causes the R-factor to increase by 0.039 for the
shell. The considerable impact of the water structure at all resolution
levels is probably a result of the protein structure re-adjusting in the

refinement to accomodate the water molecules optimally.

VII RESULTS AND DISCUSSION

A. The Hirudin Structure

Hirudin appears to possess two distinct domains in the complex
(Figure 33). (The hirudin sequence is given in Figure 34.) The
NH2—terminal domain is compact, and contains residues Ilel' to Pro48'.
The COOH-terminal domain adopts an extended conformation involving two
stretches of polypeptide, residues Glu49’—Gly54' and Asp55’-Pro60’, and
terminates with a near 310 type III helical reverse turn (Glu6l’-Gln65')
[149]. The secondary structural elements present in the hirudin
molecule are summarized in Table 13.

The globular conformation of the NHZ-terminal domain of hirudin
(Figure 33) is intimately related to the presence of the three-disulfide
core (Figure 35). The disulfides Cy56'-Cysl4' and Cysl6'-Cy528' orient
nearly perpendicular to one another with the distance between the
midpoints of the disulfides being 4.97 A. In addition, Cysl6'-Cy528'
and Cy322'-Cys39' are nearly parallel, and the distance between the
disulfide bond midpoints is 5.35 A. Similar close disulfide core
interactions are also observed in the kringle structures of prothrombin
fragment 1 [150] and plasminogen K4 [151], C3a anaphylatoxin [152], and
squash seed trypsin inhibitor [153]. A complete list of the
sulfur-sulfur distances in the hirudin molecule is presented in Table
14. It is of particular note that, excluding the actual disulfide bonds
themselves, the smallest inter-sulfur distance is 3.74 A (Cys6’
SG-Cy528' SG), the largest distance is only 8.32 A (Cysl4' SG—Cys39’
SG), and the mean distance is 5.91 A +/-1.30 A(0). The result of these
close cystine interactions is that the loop segments B, C, and D, which

comprise a double loop in hirudin (Figure 34), fold into three unique

120

121

. 32-
35!]

 

Figure 33. Stereoview of the Folding of Hirudin in the Complex.
N, CA, C atoms only; hirudin disulfides in bold; disordered
or poorly defined regions indicated by dashed lines.

122

 

Figure 34. The Sequence of Recombinant Hirudin Variant 2-Lysine
47 (rHV2-K47). Isolated loop designated 'A’ and loop
segments comprising the two interconnected loops designated
I r ICI, and ID].

123

Table 13. Secondary Structural Elements of Hirudin.

Element
B—Structure
1
2
Helix
H1

Reverse Turns
T1
T2
T3

T4

Type

Antiparallel

Antiparallel

Polyproline

Type II'
Type II

?

Type III

Residues

Cysl4’-Cysl6'
Asn20’-Cy522'

LysZ7'-Gly31'
Gly36'-Val40'

II Pro46'-Hi551'

Glu17'—Gly18'-Ser19'—Asn20’
Gly23’-LysZ4'-Gly25’-Asn26'
Ser32'-Asn33'—Gly34’-Ly535'

Glu61'-Glu62'-Tyr63’-Leu64'

124

28' 28'

6.

   

Figure 35. Stereoview of the Cystine Disulfide Core of Hirudin.
Disulfide linkages in bold.

Table 14.

Sulfur
Atom

6!

14'

16’

22'

28'

39'

125

Sulfur-Sulfur Distances (A) in Hirudin.
A '*' indicates a disulfide bond.

6' 14' 16'
0.0 2.01* 4.72
0.0 6.43

0.0

22'

6.11

6.56

4.70

0.0

28'

3.74

5.43

2.06*

5.48

0.0

39'

7.65

8.32

5.22

2.05*

6.58

0.0

126

loops of protein folding. This, combined with the ordinary Cys6'-Cysl4’
loop, produces four three-dimensional loops (Figure 36).

The hirudin NHZ-terminal domain appears to be stabilized by
antiparallel B—structure. A short segment (91, Table 13) results from
the interaction of strands Cysl4'-Cysl6' with Asn20’-Cy522’ (Figure 37).
The 81 structure is maintained by the hydrogen bonds Cys22' N-Cysl4' 0
(2.83 A) and Cysl6’ N-Asn20' O (3.17 A). A short hydrogen bond
interaction (2.60 A) between residues Leu13’ N and CysZZ' 0 immediately
precedes these two strands. The polypeptide strands of this B—structure
are connected by a type II' reverse turn (T1, Table 13) [149]. A type
II reverse turn (T2, Table 13) [149] is also present in loop segment C
involving residues Gly23'- Asn26'. A longer antiparallel B-finger (92,
Table 13) occurs in loop segment D, and involves residues Ly527'-Gly31'
and Gly36’-Va140' (Figure 38). Four hydrogen bonds are apparent in this
structural element: Lys27' N-Val40’ O (2.39 A); Val40' N-LysZ7' O (2.42
A); Ile29’ N-Gln38' o (2.53 A); and Gln38’ N-Ile29' o (2.72 A). As with
91, a reverse turn (T3, Table 13) joins these strands; however, as in
solution [78-80], it is completely disordered in the crystal structure
of the complex (Figure 33). The structure of the 62 strand does give
some indication that its associated turn will likely be disordered. The
strands of peptide appear to diverge just prior to the turn (Figure 38),
and this is further evidenced by the fact that what should be a hydrogen
bond preceding the turn, Gly31’ N-Gly36' O, is a 3.25 A long contact -
too long to be a viable hydrogen bond. The flexibility of this turn
could be due to the fact that it is hinged to glycines 31' and 36'.

In addition to the disulfide core, many intramolecular hydrogen
bonds also appear to stabilize the N—terminal domain of hirudin in the

complex (Table 15). In all, 25 intramolecular hydrogen bonds are found

Figure 36.

127

 

Stereoview of the Folding of NHZ—terminal Domain

of Hirudin in the Complex. N, CA, C atoms only;
designations ’A', ’B', ’C’, and 'D’ correspond to those of
Figure 34; disorder indicated by dashed lines.

128

 

 

 

Figure 37. Stereoview of the 81 Antiparallel B-Strand in
Hirudin. Hydrogen bonds indicated by dashed lines.

129

 

Figure 38. Stereoview of the 92 Antiparallel B—Strand in
Hirudin. Hydrogen bonds indicated by dashed lines.

Table 15.

A.

8.

Element

130

Intramolecular Hydrogen Bonds in Hirudin.
Hydrogen atoms are assigned geometrically idealized
Donor atoms are denoted ’D’ and acceptor

positions.
atoms ’A'.

Donor

Cys6’

Gly10'
Asn12'
Leul3'
Leu15’
Cysl6'
Asn20'
Cy522'
Asn26'
LysZ7'
Cy528'
Ile29'
Gln38'
Val40’
Gly42’
Gly44'
Thr45’
Lys47'
Tyr63’
Leu64'

Donor

Thr7’

Glu8'

Glnll’
Leu30'
Cys39'
Lys47'
Lys47'

ZZZZZZZZZZZZZZZZZZZZ

N
N
NE2
N
N
NZ
NZ

Acceptor

Leu15'
Cy528’
Thr45'
Cys22’
Thr4’

Asn20'
Glu17’
Cysl4'
Gly23'
Val40’
Glnll'
Gln38'
Ile29'
LysZ7’
Gly25'
Asn26’
Glle’
Asn12’
Pro60'
Glu6l’

OOOOOOOOOOOOOOOOOOOO

Mean:
Standard Deviation:

Hydrogen Bonds involving Side

Acceptor

Glnll'
Glnll'
Glu8’
Ser9'
Glu17'
Thr4’
Asp5'

0E1
0E1
081
OG
081
001
0

Mean:
Standard Deviation:

1.
0.

46

.69
.34
.52
.83
.53
.34
.34
.91
.89
.95
.03

93
35

70

.71
.62
.69
.94

1.75
0.

18

NNNNNWNU ON

ON

NNNNNWNNNWNWNWWNNNNNU

Hydrogen Bonds involving Main Chain Atoms.

Distances (A)
H..A
1.79
1.44
1.38
1.72
1.85
2.17
2.63
1.89
2.
1
2
1
1
1
2
2
1
1
1
2

..A
.75
.38
.24
.60

Chain Atoms.

Distances (A)
H..A
1.56
2.06
1.
1
1
1
1

..A
.48
.01
.63
.60
.57
.64
.90

.69
.19

DHA
164
159
141
148
161
162
130
149
137
128
130
167
158
147
134
123
145
144
159
134

146
14

DHA
144
160
158
150
169
149
152

155

Angles(deg)
CAH CAD
145 142
137 145
117 130
166 171
119 125
157 154
109 122
152 155
130 130
128 146
125 138
156 160
140 148
168 167
128 120
138 132
137 148
174 170
124 128

92 106

Angles(deg)
CAH CAD
133 137

83 87
116 114
120 130
133 129
130 121
145 150

131

in the domain, of which 18 involve main chain atoms only (eight reside
in turns or B—structure), and seven involve side chain atoms. In
particular, while an examination of Figure 36 might lead one to wonder
what maintains the conformation of the Val40'-Pro48’ polypeptide,
impressively nine hydrogen bonds along this nine residue stretch link
the chain to the rest of the N-terminal domain.

Hydrogen bonds involving Lys47' also appear to be important in
extending the domain to residue Pro48' (Figure 39). The e-amino group
of Lys47’ is involved in two hydrogen bonds with residues of the amino
terminal pentapeptide of hirudin: Thr4’OG1 (2.64 A) and Asp5'0 (2.90 A)
(Figure 39). In addition, Lys47' N hydrogen bonds to Asn12’ O (2.70 A)
(Figure 39). Such interactions bring the N- and C-terminals of the
domain in relatively close proximity, increasing the size but
maintaining the compactness of the domain. The rigidity of the lysine
side chain might well be related to the two prolines which flank it,
Pro46’ and Pro48’. These three residues also initiate the polyproline
II helix (81, Table 13) [154] which extends from Pro46' to HisSl’
(Figure 40); the six residues in this stretch exhibit Ramachandran ¢, w
angles similar to those in a collagen helix [155].

A hydrogen bond is a two bond system, D-H..A, consisting of a donor
(D-H) and an acceptor (H..A) bond, where D and A represent the
electronegative donor and acceptor atoms, respectively. As such, the
hydrogen bond contains a large positive (H atom) and a large negative
(acceptor) partial charge at close distance. Moreover, since a positive
charge assumes its lowest potential energy between two negative charges
when all three are aligned, hydrogen bonds are ideally linear.
Experimentally, however, it has been shown that the angle decreases from

180° for strong hydrogen bonds to around 130° for weak

Figure 39.

132

 

Stereoview of the Intramolecular Lys47' Hydrogen

Bonds in the Hirudin N-terminal Domain. CA structure of
residues 1’-48' including disulfides (disulfide linkages in
bold), and hydrogen bonding residues (bold). The disorder
in residues 32’—35' as well as the hydrogen bonds are
indicated by dashed lines.

Figure 40.

133

46' 46'

5V

Stereoview of the Polyproline II Helix in Hirudin.
N, CA, C atoms indicated in bold.

51'

134

ones [156,157]. Hydrogen bond interactions are commonly described in
terms of two distances, r(D..A) and r(H..A), and the DHA angle.
(Additional angles can also be used to more fully describe the
interaction; the CAB and CAD angles can be reported, where ’C'
represents the carbon atom to which the acceptor atom A is bonded, but
these are not as critical as the DHA angle.)

The mean values of the hydrogen bond descriptors r(H..A), r(D..A),
and of the angle DHA were determined for the intramolecular hydrogen
bonds in hirudin (25 of the 27 bonds are in the hirudin NHZ-terminal
domain), and in general they agree fairly well with mean values
determined from organic crystal structures (Table 15). The mean values
of the distances H..0 and N..O, and of the angle NHO for the 20
intramolecular main chain atom (N, CA, C, 0) hydrogen bonds in hirudin
were 1.93 A, 2.79 A, and 146°, respectively (Table 15). These
values are in good agreement with those determined for intramolecular
main chain atom hydrogen bonds {r(H..O) a 1.988 A; r(N..O)=2.755 A; NHO
angle - 132.5°} and all (intra- and intermolecular) main chain
atom hydrogen bonds in general {r(H..O) = 1.921 A; r(N..O) . 2.878 A:
NHO angle . 158.3°} from the literature [153]. Similarly, the
corresponding mean values for the intramolecular hirudin hydrogen bonds
involving amino acid side chains {r(H..A) = 1.75 A; r(D..A) = 2.69 A;
DHA angle a 155°} (Table 15) are in good agreement with those
determined from organic crystal structures involving nitrogen donor and
oxygen acceptor atoms {r(H..A) a 1.93 A: r(D..A) . 2.858 A: DHA angle -
160.4°} [159]. Finally, while the ideal bond length for general
hydrogen bonds involving oxygen and nitrogen-oxygen atoms is considered
to be 2.80 A and 2.85 A [145], respectively, the 1-4 hydrogen bonds in

turns are often accepted to be as long as 3.2 A [149]. The fact that so

 

hFu
11..

 

 

 

 

 

 

135

many of the intramolecular hirudin hydrogen bonds in Table 15 deviate
significantly from the ideal values suggests that the diffraction data
are not of high enough resolution or quality to resolve them accurately.

The folding of the NHZ-terminal domain of hirudin in the
crystalline complex is very similar to that which has been observed for
hirudin in solution by ZD NMR studies [78-80]. An optimal superposition
of 114 hirudin N, CA, C atoms from the complex (residues 4'-30' and
38'—48') with the average coordinates of 32 NMR structures [79] gave an
r.m.s. difference of 0.86 A, with an r.m.s. difference of 1.95 A for
side chains. The r.m.s. difference of 0.86 A agrees well with the
average atomic r.m.s. differences between the 32 individual NMR
structures and with the mean NMR structure, which was approximately 0.7
A for backbone atoms [79]. The NMR hirudin structures also had disorder
in the vicinity of the 32'—35' reverse turn [78-80]. Based on these
comparisons, the close correspondence between the NMR and the crystal
structures indicates that the hirudin N-terminal domain possesses a
similar structure in the free and thrombin-bound states.

Surprisingly, while the main chain atom agreement can be considered
to be quite good between the NMR and crystalline complex hirudin
structures, the corresponding r.m.s. difference between cystine sulfurs
is quite large, 2.43 A. An optimal superposition of the two disulfide
cores based on the N, CA, C fitting mentioned previously is displayed in
Figure 41; not surprisingly, the orientation of the disulfides is fairly
different between the two structures. The dihedral angles and dihedral
energies of these disulfide bridges are presented in Table 16. The
Cys6-Cysl4 and the CysZZ-Cys39 disulfides are displaced approximately a
half-bond length and a full bond length in the NMR structure from where

they are positioned in the crystal. Moreover, the orientation of the

136

 

Figure 41. Stereoview of the Disulfides in the Crystal Complex
and in the NMR Hirudin Structure [79]. NMR hirudin
disulfide side chains indicated by dashed lines; disulfide

linkages in bold.

137

Table 16. Dihedral Angles and Dihedral Energies of the Disulfide
Bridges in Hirudin. A ’*' indicates the energy value
was excluded from the calculation of the mean.

CA-—CB--SGl--SG2--CB--CA
X1 X2 X3 X1' X2'

Disulfide Source x1 x2 X3 Xl' x2' CA—CA(A) E(kcal/mole)
6-14 NMR -161 -62 87 108 65

16—28 NMR —60 —169 75 -168 -40
22-39 NMR -128 110 165 -83 -121

97 3.32
21 1.92
49 15.05*

89 2 62

Mean Values: .
81) (+/-0.99)

4.

6.

6.

5.
(+l-Oo
6'-14’ Crystal -51 -66 97 86 54 4.99 2.07
5.

5.

16'-28’ Crystal ~165 81 85 76 -175 94 1.70
22'-39' Crystal -175 70 80 65 174 55 2.23
Mean Values: 5.49 2.00

(+/-0.48) (+/—0.27)

138

the Cysl6-Cy528 disulfide is different in the two structures: the x1,
x1' angles and the x2, X2’ angles differ by approximately 120° and 247°,
respectively (Table 16). Available protein structure
coordinates indicate that two general disulfide families can be defined
based on the chirality of the dihedral x3 angle about the sulfur-sulfur
bond: right-handed (x3 approximately 90°) and left-handed (x3
approximately -90°) [160]. Five of the six disulfides presented
in Table 16 are right-handed - all of the disulfides from the crystal
structure and two of the disulfides from the NMR structure.

The CysZZ-Cys39 disulfide of the NMR structure has an unusually
large x3 value (X3 a 165°) which has not been observed in protein
structures [160] most likely due to its high dihedral energy. The
dihedral energies for the disulfides presented in Table 16 were

calculated using a formula from the program AMBER [161]:

E(kcal/mole) 2.0*(1 + cos(3*X1)) + 2.0*(1 + cos(3*X1'))

+

1.0*(1 + cos(3*x2)) + 1.0*(1 + cos(3*x2’))

+ 3.5*(1 + cos(2*x3)) + O.6*(1 + cos(3*x3))

...(41)

The dihedral energy for the CysZZ-Cys39 disulfide in the NMR structure
is 15.05 kcal/mole, drastically larger than the mean value of 3.19 (+/—
1.43) calculated for right-handed disulfides using available protein
crystal structure data [160]. Excluding this outlier value, the mean
dihedral energy of 2.62(+/- 0.99) kcal/mole for the remaining two
disulfides in the NMR structure, and of 2.00(+/-0.27) kcal/mole for the
disulfides in the crystal structure agree well with mean value
calculated by Katz and Kossiakoff [160]. The mean CA-CA(A) distance of
5.49(+/-0.48) A (Table 16) observed in the complex also agrees well with

the value of 5.07(+/—0.73) A for right-handed disulfides [160]. It is

139

of note that the mean CA-CA(A) value of 5.89(+/-0.81) A for the
disulfides in the NMR structure is on the high side due to the fact that
two of these distances were long — 6.29 A (Cysl6 CA-CysZB CA) and 6.49 A
(CysZZ CA-Cys39 CA).

The COOH-terminal domain of hirudin exists as two extended
stretches of polypeptide with a bend at Asp55' (Figure 33). The first
segment of this domain is approximately 18 A long and contains residues
Glu49’ to Gly54'. The first three residues of this segment,
G1u49'-Ser50'—HisSl’, complete the polyproline helix displayed in Figure
40. The last three residues of this stretch, Asn52'-Asn53’-Gly54', have
very poorly defined electron density in the 2Fo-Fc maps. The second
segment of the COOH—terminal domain is approximately 19 A long and
consists of 15 A in extended conformation (Asp55'-Pro60') followed by a
type III 310 helical reverse turn [149]. This reverse turn (T4, Table
13) involves residues Glu61'-Leu64', contains a hydrogen bond between
Leu64' N-Glu61' O (2.82 A) (Table 15), and is shown in Figure 42. A
hydrogen bond interaction between Tyr63' N—Pro60’ 0 (2.90 A) (Table 15)
precedes this turn. A general helical structure was also deduced for
residues Glu61'-Leu64’ by NMR with a synthetic peptide fragment of the
hirudin tail [162]. The elongated nature of the C-terminal tail with no
stabilizing interactions back to the N-terminal domain probably explains

why this portion of the hirudin molecule is disordered in solution.
B. The Hirudin-Thrombin Interaction

The CA structure of the rHV2-K47-thrombin complex is shown in
Figure 43, from which it can be seen that the NH2-terminal tail of
hirudin binds in the active site region, while the 39 A long

COOK—terminal tail extends across the surface and finishes almost

140

 

Figure 42. Stereoview of the T4 Type III Helical Reverse Turn
in Hirudin. Hydrogen bonds indicated by dashed lines; main
chain atoms in bold; G1n65’ also included.

141

 

Figure 43. Stereoview of the CA Structure of the Hirudin-Thrombin
Complex. Hirudin CA and disulfides of hirudin and thrombin
in bold. N— and C—terminals of thrombin designated.

142

diametrically opposed to the hirudin N-terminus.

A particularly unique and unexpected interaction occurs between the
NHZ-terminal tripeptide of hirudin and the the active site region of
thrombin from the standpoint of classical serine proteinase inhibition
by macromolecular inhibitors. All SERPIN (SERine Proteinase
INhibitor)-serine proteinase complexes characterized to date [84,163]
display a common mode of interaction in the active site which is also
observed in the PPACK-thrombin structure [6] (Figure 44). The
polypeptide chain of PPACK aligns antiparallel to the 214-217 strand in
thrombin, and the arginyl side chain, which corresponds to the P1
residue of thrombin, sits in the specificity pocket in a substrate-like
manner - making an ion pair with residue Asp189 at the base of the
pocket. The hirudin NH2-terminal tripeptide, however, forms a parallel
B—strand interaction with the 214-217 thrombin segment and makes no use
of the specificity pocket (Figure 45). The hirudin tripeptide forms
three hydrogen bonds with the thrombin 214-217 strand (Table 17).
Residue Ile 1’ engages in two hydrogen bonds (Ilel' N-Ser214 0, 3.02 A;
Ile 1' 0-Gly216 N, 3.21 A) and Tyr3’ participates in one (Tyr3' N-Gly216
0, 3.01 A) (Table 17). In addition, the carbonyl oxygen of Tyr3' makes
a hydrogen bond with the amide nitrogen of Gly219 (2.84 A) (Table 17).
Most notable, however, is the fact that the amide nitrogen of Ilel' also
interacts through a hydrogen bond with Ser195 06 (3.06 A) (Table 17).

A consideration of the relative spatial positioning of the hirudin
N-terminal tripeptide and PPACK in the thrombin active site reveals both
similarities as well as differences. The side chain atoms of Ilel’ and
Tyr3’ occupy basically the same spatial position in the active site
region as do Pro and D-Phe of PPACK, respectively. The isoleucyl side

chain sits in the $2 subsite of thrombin and makes numerous hydrophobic

Figure 44.

 

143

217

Stereoview of PPACK in the Active Site Region of Thrombin.
PPACK, from the PPACK-thrombin structure [6], optimally
superimposed into thrombin active site based on N, CA, C
fitting of PPACK—thrombin and hirudin-thrombin complex
thrombin structures; PPACK in bold.

144

Figure 45. Stereoview of the Interaction of the Hirudin N-terminal
Tripeptide in the Active Site of Thrombin. Hirudin
N-terminal tripeptide in bold: hydrogen bonds of the
N-terminal nitrogen indicated by dashed lines.

145

Intermolecular Hydrogen Bonds in the Hirudin—Thrombin

Complex. Hydrogen atoms are assigned geometrically
Donor atoms are denoted

for an unprotonated Ilel' amino terminus.

ID!

and

A protonated amino terminus is
designated 'N+' and an unprotonated amino terminus is
A '*' indicates the mean
and standard deviation were calculated using values

Table 17.
idealized positions.
acceptor atoms 'A’.
designated 'N’ for Ilel'.
A.

8.

Hydrogen Bonds involving Main Chain Atoms.

Donor Acceptor
Ilel' N+ Ser214 0
Ilel' N Ser214 O
Tyr3’ N Gly216 O
Glu57' N Thr74 O
Gly216 N Ilel’ O
Gly219 N Tyr3' O
Ly3224 NZ Ser19' 0

Mean*:
Standard Deviation*:

Hydrogen Bonds involving Side Chain Atoms.

Ilel’ N+ Ser195 OG
Ilel' N Serl95 OG
Va121' N Glu217 0E1
Lys60F NZ Glu49’ OEl
Arg73 NHZ Asp55’ ODl
Arg73 NHl Asp55' OD2
Tyr76 N Glu57’ 0E2

Mean*:
Standard Deviation*:

Distances
H..A
2.05
2.04
2.04
1.84
2.32
2.17
2.17
2.10 3
0.16 O

2
2
2
2
1
2
1

.45
.26
.10
.16
.99
.32
.89

.12
.16

CW

WNWNUOLIJUJU

NWNUJUJWUJ

(A)
A

.01
.01
.01
.82
.21
.84
.10

.00
.15

Angles(deg)
DHA CAH CAD
154 144 152
155 149 152
165 161 159
173 165 164
149 153 158
123 138 155
149 150 150
152

17

117 126 132
132 121 132
163 133 128
176 94 94
142 101 110
153 123 116
169 130 129
156

17

146

contacts with the thrombin residues HisS7, Tyr60A, Trp60D, Leu99, and
Trp215. The aromatic moeity of Tyr3' is in the hydrophobic cleft
defined by Leu99, Ile174 and Trp215. This hydrophobic region is likely
the apolar binding site for indole reported by Berliner and Shen [40],
elaborated upon by others [41,47,58], and delineated by Bode et al., [6]
in the PFACK-thrombin structure. The nonpolar contacts of the Tyr3'
side chain with thrombin are also complemented by those involving nearby
residues in the hirudin molecule: in all, this tyrosyl group is involved
in eight van der Vaals contacts of less than 4.0 A with the side chains
of Ilel’, Leu15’, and Val 21'. The side chain of Thr2' only penetrates
the edge of the specificity pocket and well below the point of entry of
the arginine of PPACK. The fact that hirudin interacts in such a novel
fashion in the thrombin active site coupled with the fact that it
displays no close sequence or topological similarity to the ten existing
classes of serine proteinase inhibitors [73] suggests that hirudin truly
represents a heretofore unknown family of inhibitors [72].

It is likely that all naturally occurring hirudins interact in the
active site as does rHV2-K47, since they all possess Ile-Thr-Tyr or
Val-Val-Tyr NHZ-terminal sequences [61]. Furthermore, the importance of
maintaining nonpolar character in the first two positions has been
demonstrated experimentally in binding studies with specific 1-2
position site-directed mutants [90]; these experiments showed that
replacing N-terminal Val-Val residues with polar amino acids resulted in
an increase in the inhibition constant (Ki), while replacement of these
residues with aromatic or hydrophobic amino acids had little effect on
the value of Ki. This is consistent with the nonpolar character of the
$2 subsite.

More difficult to reconcile with the crystal structure is

147

experimental evidence obtained by Stone, Braun, and Hofsteenge which
supports the importance of having a positive charge on the hirudin
amino-terminus [81]. In Table 17, appropriate hydrogen bond distances
and angles are presented for the two amino-terminal Ilel' N hydrogen
bonds to Ser214 0 and Serl95 OG assuming both a protonated and an
unprotonated hirudin N-terminus. The significantly improved H..A
distance (2.26 A vs. 2.46 A) and DHA angle (132° vs. 117° )
for the Ile1' N-Ser195 OG hydrogen bond with an unprotontated
amino terminal nitrogen supports the the notion that Ilel' N is
uncharged. Since 81557 of the catalytic site is also protonated at pH
4.7, a neutral hirudin amino-terminus could alleviate a positive charge
build-up in the active site region. Moreover, a protonated Ilel’
nitrogen would be only 3.23 A from a protonated H1557 NEZ in the crystal
structure.

Although most of the NHZ-terminal domain of hirudin is not in
contact with the thrombin surface (Figure 43), many hydrophobic and
polar interactions exist at the interface of the two. For instance,
three hirudin residues participate in nonpolar interactions at this
interface (Figure 46). Residues Leu13’ and Pro46’ form hydrophobic
contacts of 3.34 A and 3.73 A, respectively, with Pro6OC of the nine
residue 60A-6OI insertion loop which narrows the active site cleft [6].
In addition, Va121' has two van der Vaals contacts of less than 4.0 A
with Ile174. Two ion pair and two hydrogen bond interactions are also
found at the interface (Figure 47). The salt bridges are:
Asp5'-Arg221A (3.6 A), and Glu17'-Arg173 (5.0 A). Hydrogen bonds occur
between residues Ly5224 NZ-Ser19' OG (3.10 A) and Va121’ N-Glu217 031
(3.05 A) (Table 17). In all, of the 48 amino acids in hirudin

N—terminal domain, 15 residues are directly involved in contacts of less

148

15./174 )T174
\

‘37, fi?‘ @600
Q Q.-

Figure 46. Stereoview of Hydrophobic Contacts of the N-terminal
Hirudin Domain with Thrombin. Hirudin residues in bold.
Residues in figure: I1e174, Pro6OC, Leu13', Va121', and

Pro46’.

149

173

 

Figure 47. Stereoview of Electrostatic Interactions of the N-terminal
Hirudin Domain with Thrombin. Hirudin residues in bold.
Hydrogen bonds indicated by dashed lines, and ion pairs by

+/-o

150

than 4.0 A with thrombin.

A number of water-mediated hirudin—thrombin and hirudin-hirudin
hydrogen bond interactions involve residues in the N-terminal hirudin
domain (Tables 18, 19, and 20). The carbonyl oxygen of Gly18', side
chain oxygens of Asp5’and Asn20', and the ring hydroxyl group of Tyr3’
all interact with thrombin residues via water molecules (Tables 18 and
19). In particular, while Asp5’ and Arg221A appear to form a strong
salt-bridge directly, a nearly pyramidal-coordinate water molecule in
the vicinity, V622, appears to bridge this contact as well as bond to
the Gly18' 0 (Figure 48). The result is that the water molecule makes
three hydrogen bonds with the three residues involving Asp5’ OD2 (2.56
A), Gly18' O (2.87 A), and Arg221A NHZ (2.65 A). The remaining
hydrogen-bonding interactions with thrombin through water molecules are
Tyr3'-Tyr60A and Asn20'-Arg173. There are also several long range (d >
3.2 A) water-mediated hirudin N-terminal domain-thrombin interactions
which involve the following residues: Leu13'—Trp60D; Asn20' with
Thr172-Arg173, Glu217, Ly5224; and Va121'-Arg173. Lastly, there are
four solvent-bridged hydrogen bond interactions between residues within
the domain: Thr4'-Leu13'; Asp5'-Gly18'; Leu13'-Lys47': and
Glu17'-Asn20' (Table 20).

The tripeptide segment Pro46'-Lys47'-Pro48' appears to be important
in facilitating the binding of the hirudin N-terminus in the active site
region of thrombin (Figure 49). However, it is not because this stretch
of residues resembles a thrombin substrate-like sequence [164] or
because the Cys39'-Pro48' segment bears 502 sequence homology to
residues Arg148-Ser157 of the thrombin cleavage site on prothrombin
[86]. The hydrogen bonds made by the s-amino group of Lys47' to Thr4'

(MS and Asp5’ O, as well as the solvent-bridged interaction with Arg221A,

Table 18.

Donor

Thr2'
Thr2’
Tyr3'
Thr4'
Thr4'
Asp5’
Glu17'
LysZ4'
Gly25'
Glu58'
Glu62'
Tyr63’
V401
V467
V469
V472
V483
V502
V508
V541
V543
V573
V622
V622
V623
V672
V672
V676
V726

151

Hirudin-Vater Molecule Hydrogen Bonds in
the Complex. Hydrogen atoms are assigned
geometrically idealized positions. Donor
atoms are denoted 'D', acceptor atoms ’A'.

Acceptor Distances (A) Angles(deg)
H..0 D..A DHA COH

V498 2.13 3.05 173

V410 1.94 2.85 163

V606 2.15 3.00 152

V469 1.82 2.72 147

V612 2.13 3.10 173

V534 2.07 3.05 168

V502 2.04 2.99 168

V673 2.14 3.08 162

V493 1.79 2.70 163

V722 1.95 2.99 160

V472 2.00 2.86 146

V657 1.83 2.71 154

Glu57' 0 2.36 133

Glu58' 082 2.75 121

Leu13’ 0 2.78 133

Glu62' 082 2.86 101

Thr2’ 0G1 2.40 121

Asn20’ 0 2.85 126

Glu49’ 081 3.17 104

Leu15' 0 3.08 135

Asp55' o 2.56 150

Asn20' 001 2.33 130

Asp5’ 002 2.56 121

Gly18' 0 2.88 111

81851’ 0 3.03 89

Leu13' 0 3.07 119

Lys47' o 2.56 139

01058' 081 2.30 144

Cys6' 0 2.75 157

Table 19. Solvent-Bridged Hirudin-Thrombin Hydrogen
Bond Interactions.
involved in the interaction are represented
as ’A' and ’B', and the mediating water 'V'.
A Vater B
Molecule
Tyr3' 0H V606 Tyr60A
Asp5’ 0D V622 Arg221A
Gly18’ 0 V622 Arg221A
Asn20' 0D1 V573 Arg173
Glu49' OEl V508 Arg35
Glu57' 0 V401 Arg67
Glu57’ 0 V401 Arg67
Glu58' 082 V467 Arg77A
Table 20. Solvent-Bridged Intramolecular Hydrogen
Bond Interactions in Hirudin.
atoms involved in the interaction are
represented as 'A' and 'B', and the mediating
water 'V'.
A Vater B
Molecule
Thr4' N V469 Leu13' 0
Asp5' 0D2 V622 Gly18' 0
Leu13' 0 V672 Lys47' 0
Glu17' N V502 Asn20' 0
Glu62' N V472 G1u62' 0E2

152

The protein atoms

0H
N82
NHZ

NHl
NHl
NH2
NB

Distances (A)

NNNwNNNw>
w
(a)

The protein

Distances (A)

A..V

V..B

2.78
2.88
2.56
2.85
2.86

V..B
2.38
2.66
2.66
2.84
2.93
3.00
2.94
2.94

153

I.ws22
221A ’ 15' Ml

W622

I‘\
fife 9

Figure 48. Stereoview of a Hirudin-Thrombin Vater Mediated Interaction.
Hirudin residues in bold. Hydrogen bonds indicated by
dashed lines, ion pairs by +/-, the water molecule by *.

 

Figure 49.

154

 

Stereoview of Hirudin Pro46'—Lys47'-Pro48' and the

Hirudin N-Terminus-Thrombin Active Site Interaction.
Hirudin residues in bold. Hydrogen bonds indicated by
dashed lines, ion pairs by +/-, and water molecules by *.

155

could help stabilize the active site binding by "clamping down" on the
4'-5’ dipeptide. Moreover, the van der Vaals contact between Pro46’ and
Pro60C, and the polar contacts between Lys47' and Trp6OD could help
anchor the N-terminal domain in the proper position for the
amino-terminal tripeptide to associate intimately with active site
region residues. It is of note that Lys47' has often been viewed as the
P1 residue of substrate in the hirudin—thrombin complex [67,70,87].
Though the lysyl side chain of 47' does not occupy the specificity
pocket of thrombin and is located approximately 11 A away (Figure 49),
it appears to be a factor in the hirudin-active site region binding.

The COOH-terminal tail in hirudin, in contrast to the NHZ-terminal
head, adopts an unusual, extended conformation in the complex (Figure
43). In addition, by means of this extended conformation, residues of
the tail can interact intimately with a wealth of residues on the
thrombin surface; in all, the side chains of 12 of 17 hirudin C-terminal
domain residues are involved in electrostatic or nonpolar contacts with
thrombin. The region of thrombin where these residues bind has become
known as the anion-binding exosite [47]. In addition, this region,
which is defined by the loop segments Phe34-Ser41 and Lys70-Glu80 [6],
is particularly rich in charged side chains - especially
positively—charged arginyl and lysyl side chains [165,166] (Figure 50).
Another extensive and highly electropositive exosite appears to exist on
the opposite side of the thrombin molecule and consists of Ly387, Arg93,
‘Arg97, Lysl69, Arg175, HisZ30, Arg233, Lys236, and Ly5240 (Figure 51);
this might be the site of heparin binding since the latter binds at an
exosite different from that of fibrinogen [45].

In the first segment of the COOH-terminal hirudin tail, the initial

three residues, Glu49'-Ser50’-Hi351', are involved primarily in

156

 

Figure 50. Charged Residues of the Anion-Binding Exosite Region of
Thrombin. Thrombin CA structure only with thrombin
disulfides and charged anion-binding exosite residues in
bold.

Figure 51.

157

 

Basic Residues of the Postulated Heparin Binding Site of
Thrombin. Thrombin CA structure only with thrombin
disulfides and basic residues of heparin binding site in
bold.

158

electrostatic contacts with thrombin (Figure 52). Residue Glu49' forms
a hydrogen-bonding ion pair with Lys60F (3.20 A) (Table 17) and
interacts with Arg35 via the water molecule V508 (Tables 18, 19). The
OG of Ser50’ appears to be involved in close, bifurcated polar contacts
of 3.36 A and 3.68 A with the side chain oxygen atoms of Glu192. Also,
the imidazolium ion of HisSl' is involved in an ion pair with Glu39 of
thrombin (HisSl' ND1-Glu39 0E2, 3.0 A). The final three residues of
this C-terminal tail segment, Asn52'-Asn53’-Gly54', are poorly defined
in the electron density maps.

The second segment of the hirudin COOH-terminal tail,
Asp55'-Gln65’, can be viewed as being composed of a 15 A extended
stretch of polypeptide from Asp55'-Pro60' followed by a type III 310
helical reverse turn from Glu61'-Leu64'. Impressively, nine of the 11
residues in this hirudin tail region interact with residues on the
thrombin surface.

Four of the last eleven residues in the hirudin C-terminal tail
participate in electrostatic contacts with thrombin (Figure 53). The
carboxylate of Asp55' is involved in a double hydrogen-bonding salt
bridge with Arg73 (Arg73 NHZ-Asp55' OD1, 2.93 A; Arg73 NHl—Asp55' 002,
3.22 A) (Table 17) and forms an ion pair with Lysl49E of 4.9 A. The
Asp55'-Lysl49E salt bridge is in agreement with a successful lysine
cross-linking experiment carried out using a synthetic
dinitrofluorobenzyl-Gly54-Leu64 peptide [167]. In addition, Lysl49E is
the position of the y-cleavage site in thrombin [48,82] (Figure 10).
The residue Glu57' is engaged in hydrogen bonds with main chain atoms of
residues Thr74( Glu57' N-Thr74 o, 2.93 A) and Tyr76 (Tyr76 N-Glu57’ 032,
2.89 A) (Table 17), and has 10 contacts of less than 4.0 A with main

chain and side chain carbon atoms of Arg75. The carbonyl oxygen of

159

 

Figure 52. Hirudin(Glu49'-Ser50'-Hi551')-Thrombin Interactions.
Hirudin residues in bold. Hydrogen bonds indicated by
dashed lines, ion pairs by +/-, and water molecules by *.

160

flew we “5'
as?

73 I ‘Iw 76

26;: 77A 77A
74 5 73%:
M14967 7 Russ W737;

x 5

Figure 53. Electrostatic Hirudin(Asp55'—Gln65')-Thrombin

Interactions. Hirudin residues in bold.
indicated by dashed lines,
molecules by *.

Hydrogen bonds
ion pairs by +/-, and water

161

GluS7' also interacts with the Arg67 side chain atoms NHl and NH2
through the water molecule V401 (Table 19). However, the most
interesting aspect of this residue is that it participates in a double
hydrogen-bonding ion pair with Arg#75 of a symmetry-related molecule in
the crystal (Arg#75 NE-Glu57’ 031, 2.81 A; Arg#75 NH2-GluS7' 032, 2.93
A); the proximity of Arg75 to Glu57’ suggests that this salt—bridge
could likely occur within the complex in solution. The Glu57'—Glu58’
residues are also involved in an electrostatic interaction with Arg77A,
which is the position of the B—cleavage site in thrombin [48,52].
The side chain of Glu58' forms an ion pair with Arg77A directly while
GluS7’ can interact with the arginyl group through the mediating water
V467, although the distance is long for a hydrogen bond (Glu57’0E2—V467,
3.72 A). Finally, the C-terminal carboxylate of Gln65’ forms an ion
pair of 3.33 A with the e-amino nitrogen of Lys36.

Glutamate residues 61' and 62’ have no interactions with thrombin;
this was somewhat surprising as site-directed mutagenesis studies have
shown that single and multiple mutations of the carboxy-terminal
glutamate residues 57',58', 61', 62'to glutamine caused increases in the
hirudin-thrombin dissociation constant [87]. The water molecule V472
appears to bridge an intramolecular and intraresidue hirudin hydrogen
bond interaction between G1u62' 0E2 and G1u62’ N (Table 20).

An unexpected number of van der Vaals contacts also occur between
the latter half of the extended C-terminal tail and the anion-binding
exosite. The residue Phe56’ penetrates into a depression on the
thrombin surface and makes 11 close contacts with thrombin residues:
one with the sulfur of Met32, three with Phe34, one with Leu40, and
seven with Thr74 (Figure 54). Most notably, as the planes of Phe56’ and

Phe34 are nearly perpendicular to one another, edge-on aromatic stacking

162

56'

 

74 74

Figure 54. Hirudin Phe56' in a Hydrophobic Thrombin Cavity.
Phe56’ in bold.

163

occurs between the two phenyl groups (Figure 54). There is also a
substantial concentration of nonpolar hirudin—thrombin interactions in
the vicinity of the type III helical reverse turn (Figure 55). Hirudin
residue Ile59’ has three close intramolecular contacts with the side
chain of Leu64', as well as one contact each with the side chain atoms
of Leu65 and Ile82. Residue Pro60' is involved in five intermolecular
contacts of less than 4.0 A with Tyr76 and has one close contact with
Ile82. The residue Tyr63’, which possesses an acyl-sulfate in natural
hirudin, participates in a sum total of five close hydrophobic
approaches with Leu65 and Ile82. As a result of the type III 310
helical reverse turn, Leu 64' makes two close contacts each with Ile59’
and main chain atoms of Glu61’, as well as one close contact with the
side chain of Leu65 of thrombin. The interactions in the vicinity of

the turn terminate with the Gln65'-Lys36 ion pair.

As the lack of a sulfated tyrosine at position 63' in hirudin has
been shown to lead to a six-fold increase in the dissociation constant
of the hirudin-thrombin complex with respect to natural Tyr63’-sulfated
hirudin [86], it is of some importance to consider residues in the
vicinity of a sulfated Tyr63' which could conceivably be involved in a
stabilizing interaction. Initially, it seemed reasonable that the
nearby thrombin residues Lys81 and Lyle9-LysllO could form ion pairs
with a sulfated residue via free bond rotation. In the crystal complex,
the e-amino nitrogen of LysBl forms a hydrogen bond with the carbonyl
oxygen of Ala113. The terminal amino groups on the side chains of
Lyle9 and LysllO, on the other hand, are not engaged in stabilizing
interactions — making these two lysyl side chains more likely

participants in an ion pair with a sulfated-Tyr63’. However, the

164

34 34

 

Figure 55.

Hirudin—Thrombin Interactions near the Type III
Helical Reverse Turn of Hirudin. Hirudin residues in bold,
with N, CA, C of hirudin bolder. Ion pairs indicated by

+/-.

165

structure of the hirugen-thrombin complex at 2.2 A resolution [168]
(hirugen is Asn54'-Leu64' with a sulfated Tyr63’) reveals that the
oxygen atoms of the Tyr63' sulfate make hydrogen bonds to Tyr76 0H and
Ile82 N, and that no definitive ion pairs occur. Moreover, the spatial
similarity of Tyr63’ and Tyr76 in the hirudin-thrombin structure to the
corresponding tyrosine rings in the hirugen-thrombin structure suggests
that this interaction is likely responsible for the additional binding
affinity observed in a sulfated hirudin-thrombin complex.

The aforementioned electrostatic interactions certainly support
experimental evidence that acidic residues of the hirudin C-terminal
tail [70] and basic residues of thrombin [18,165—166] are important to
hirudin binding. Recently, Chang has identified inaccessible lysyl
residues in hirudin—blocked thrombin ( lysines 36, 60F, 70, 109, 110,
149E) and additionally implied the participation of the arginine
residues 67, 73, 75, and 77A in the interaction [166]. The involvement
of the lysine residues 36, 60F, 149E, and the arginine residues 67, 73,
75, and 77A in electrostatic interactions with hirudin has already been
mentioned. However, the side chains of Lyle9 and LysllO are completely
exposed to solvent and have no interactions with hirudin. Moreover,
while Lys70 has no contacts with hirudin in the crystal structure of the
complex, the Lys 70 c—amino nitrogen (N2) is involved in two internal
ion pair interactions with Glu77 (3.5 A) and with Glu80 (3.4 A).
Similar lysyl inaccessibility studies have been carried out by reductive
methylation which imply hirudin protection for the lysines 36, 60F, 70,
81, 145, and 224. [169]. Of these newly implied lysine residues, LysBl
forms an internal hydrogen bond with Ala113 (LysBl NZ—Ala 113 0, 2.73
A), Lysl45 interacts with Glu18 through a solvent-bridged hydrogen bond
interaction (Lysl45 NZ-V705, 2.87 A; V705-Glu18 0E2, 2.80 A), and Ly5224

166

N2 forms a hydrogen bond with Ser19' 0 (3.10 A) of hirudin (Table 17),
as well as internal hydrogen bonds with Ser171 0 (2.76 A) and with
Glu217 032 (2.60 A).

The COOH-terminal tail of hirudin appears to be firmly anchored to
thrombin at the end of the anion—binding exosite through hydrophobic
interactions of the helical turn and the terminal Gln65’-Lys36 salt
bridge. Although removal of Gln65' has little affect on the Ki
inhibition or dissocation constant (increases by a factor of 1.3),
deletion of Tyr63'-Leu64’ raises the Ki by a factor of 40 [170]. The
bulky 31° helical turn may very well be important for hirudin-thrombin
recognition, but whether this region of thrombin is part of the
fibrinogen binding site remains uncertain.

An aspect of the hirudin COOH terminal tail-thrombin interaction in
the anion-binding exosite region to emerge in importance as a result of
the structure solution of the complex is the extent of the hydrophobic
contribution. In the latter half of the C—terminal hirudin tail
(Asp55'-Gln65’), nearly half of the residues (five of 11) are
hydrophobic or aromatic in nature, and all five are engaged in nonpolar
hirudin—thrombin interactions. The importance of these interactions to
the anticoagulant activity of hirudin C-terminal tail analogues (like
hirugen) has already become evident through thrombin binding studies
which indicate that the minimal peptide necessary for detection of
anticoagulant activity is Phe56'-Gln65' and, furthermore, that the
activity is sensitive to modification of residues Phe56', Ile59',
Pro60’,Leu64', as well as Glu57’ [89,95]. Based on the information in
the crystal structure of the rHVZ-K47-thrombin complex, it is clear that
the hydrophobic contacts are at least as important as the polar

interactions in the latter half on the hirudin COOH-terminal tail.

167

It is likely that hirudin binds with such high affinity to thrombin
- forming tight, noncovalent complexes with dissociation constants in
the pico-femtomolar range [59,71] - due to the fact that as it meanders
across the the thrombin surface it is able to interact with a wealth of
residues in a very intimate manner (Figure 56). All in all, 26 of the
65 residues of hirudin and 33 of the 259 residues in the thrombin
B-chain are involved in 216 contacts of < 4.0 A, of which eight are ion
pairs and 12 are hydrogen bonds.

A summary of the hirudin-thrombin interactions is presented in
Table 21. The hirudin N-terminal tripeptide is engaged in a remarkable
41 close contacts which are largely hydrophobic or neutral (30 out of 41
or 73%), as well as five hydrogen bonds. The remainder of the
N-terminal hirudin domain, residues Thr4'-Pro48', have 62 close contacts
with thrombin in which there are two hydrogen bonds, two ion pairs, and
four nonpolar or neutral contacts. Thus, the N—terminal domain is
responsible for 103 or 47% of the close contacts with thrombin; it is,
however, important to also recognize that 41 or 402 of these 103
contacts are concentrated in the first three residues of hirudin.

As the C-terminal tail begins, there are 32 close contacts within
the first three residues, Glu49’-His51', including two ion pair
interactions, and one hydrogen bond as well as 19 close contacts along
the length of the Glu49' and Trp6OD side chains. The last three
residues of this first segment of the C-terminal tail, Asn52’-Gly54’,
have poorly defined density. As the latter half of the C-terminal tail
begins, there is again a large number of interactions within a small
number of residues; 60 close contacts occur between the Asp55’—Glu58’
stretch and thrombin. Among these contacts, the residues Asp55',

Glu57’, and Glu58' are largely involved in electrostatic interactions -

168

 

Figure 56. Space-filling Drawing of the rHV2—K47-Human a—Thrombin
Complex. Hirudin in sky blue.

169

H H
I
H

HHN
ll

MNHHWNNWHHMQPWM

H

7 v? 7
M‘DWFVIDMVIDMGQOIDQN
N HF‘H

.5.

 

mmeIlHflllém
m m m c

Amy M

mu no om an
O N m H

no canonaoucm: ”veep cowouuzg m
wumochH

nmmv m

 

«muwma
umHsomHoeuwucH mum madnesz

wméllcmlltmm
N N
MH mH NH mH

AHv

.m.
m Him

7. N a
w «L:

a

 

mellelltmv
H
m m MN

HIAv.
mH

Aw.

Hm on av

 

 

#0 N»
H H

N H w o mH NH
Him

v

Him N m

VH

H
H

H H

AH.

5' @v VN HN ON aH QH bH mH MH
K m I > z m U a H A

.vHOD Cw vUC«Hu0@H—D mud MHUQHGOU Hflhuaflc

co“

m“ “mass:

m an vwaoHH0u
mononucmumq “4 o.« v muomucoo
.mcoHuomawucH :Hnsounsla :HvsuHm

v N
.vv

0 v
a B

m OH NN

u
HiN H

N
Tu N TM
a
Ta
TH
N
N

#411

 

.HN mHan

wellHVIiOH mA‘HOB
N n mnzomim

29m
vNNx
<HNN¢
ONNU
mHNQ
bHNm
mHNU
mHNS
vHNm
mmHm
NmHm
HmHU
vhHH
Mme
mmvHx
mVHm
mm
mm
Nm
dbb

In
\D
ahxuzxmsxammamumnsa

170

accounting for 48 or BOX of the contacts, of which four are hydrogen
bonds and three are ion pairs. Moreover, Phe56' interrupts these
electrostatic interactions by engaging in 12 close contacts, including
10 neutral or hydrophobic interactions. Lastly, in the region of the
type III helical reverse turn, residues Ile59'—Gln65’ produce 21 close
contacts with thrombin in which 12 are neutral or nonpolar. The last
hirudin-thrombin interaction, Lys36 NZ—Gln65 OXT (the terminal
carboxylate oxygen), is an ion pair. In all, the hirudin C-terminal
residues participate in 113 close contacts with thrombin. Moreover, as
in the case of the hirudin N-terminal tripeptide, these contacts are
concentrated in small stretches - 32 among Glu49'-Hi551’, 60 among
Asp55'-Glu58', and 21 among Ile59'-Pro60' and Tyr63'-Gln65’.

It is also worth noting that 212 (seven of 33) of the thrombin
residues involved in contacts with hirudin of less than 4.0 A are
'insertion residues', and therefore were judged to be inserted amino
acids with respect to the sequence of chymotrypsinogen [6]. The major
insertions in the thrombin sequence with respect to chymotrypsinogen are
indicated in Figure 57, as well as the insertion residues which are
directly involved in hirudin-thrombin interactions. Moreover, these
residues account for 272 (58 of 216 contacts) of the total contacts.
All of the interactions have been reviewed previously, and many of them
appear to be important. The fact that insertion residues are
responsible for a significant fraction of the hirudin-thrombin contacts,
and considering that hirudin does not bind to other serine proteinases
or blood proteins, suggests that the insertion residues are likely
important to the diverse functionality of thrombin and to its
specificity toward macromolecular substrates.

As another means to appreciate the extensive and intimate

171

.cmHouHu
mum waqeoo cHnaounuacfivsuH: may a“ :Hvsan nu“: 4 o.<
v «o muumucoo mchmmmmoa mmsvwmmu :oHuummcH new .cmumuHccH
mum mcoHuumccou mwmxcwH mvHuHsmHo .cwmzo mucwsawm
umocHH 0:» Scan mconsuaoua mm vmumouccH NHHmonnmum
mum .0. comocfimnzuuoezzu mo mucoscmm may cu uowqmou an“:

mosvamu :oHuuwmcH .cgneousaaa sees: no mucoscwm any

ION IAN INN

.3 33:"

filllda

gIIIIOIHDIONIIIIIIFIIIIIIIII®OIIII-

 

 

 

 

 

a 1:1 a: .2 a:
apneaoueauanuxeuuoo-eo- «nouns-4 nanauunu-uoaa an:
as non
.-
nnn ace and and au—
urnsnapped»-.oflu-u-aaensure-eauenoaeu-na-ao>~unua-
I II
I I
I
..A on no an
IIDIIIIIIIINIIIIIIIIIIIInINIIIIIMIIIJQIIIIIIIIINI.
I
3 3 a, a 3 3 3
IdIIId®I II‘IIFIIIINIIII IIIOIIIIII’OIIIIIuI‘IIII’IIN‘
I I
a a
. a.
av
a
31?-
a
I I
I I
u a o
u a u n
a a 4 o
o n a a
88-. and-unnuauuaeo T:-
2 on u u
31?.

 

 

172

hirudin-thrombin interaction, one can examine results obtained from
solvent accessible surface calculations; these calculations were
performed using a probe radius of 1.4 A [171] on hirudin, thrombin, and
the hirudin-thrombin complex, and the information gathered from this
work is summarized in Tables 22-24. In its complex with thrombin,
hirudin blocks 11.9% of the thrombin solvent acessible surface - masking
9.9% of the main chain atom accessibility and 12.4% of the side chain
atoms (Table 22). This result is not surprising, for, as Figure 56
clearly inidicates, hirudin covers only a small fraction of the large,
nearly spherical thrombin surface. More impressive are the results
presented in Tables 23 and 24. The information in Table 23 indicates
that in the complex, thrombin masks 37.7% of the total solvent
accessibility of hirudin, in which there is a 23.6% loss in main chain
atom accessibility and a 42.4% loss of side chain atom accessibility.
In addition, Table 24 reveals that the bulk of this loss is concentrated
in the same small stretches of hirudin polypeptide which had a large
number of close thrombin contacts in Table 21.

The hirudin N-terminal tripeptide, which displays the highest
concentration of close thrombin contacts, also shows the greatest loss
in solvent accessibility due to complexation - with a 90.7% total
accessibility loss, of which 85.4% and 92.6% are due to a loss of main
chain atom and side chain atom accessibility, respectively. As most of
the remainder of the hirudin N-terminal domain, Thr4'—Pro48', is located
away from the thrombin surface, this region shows the lowest loss of
accessibility due to complex formation; among these residues there is a
loss of only 20.2% in total atom accessibility, in which 7.2% is due to
main chain atoms and 25.0% is due to side chain atoms.

The two short C-terminal hirudin polypeptide stretches which

173

Table 22. Solvent Accessible Surface Area of Thrombin Residues
Alone and in Complex with Hirudin. Abbreviations
used: TASA: total atom surface accessibility; MASA:
main chain atom surface accessibility; SASA: side chain
atom surface accessibility.

Thrombin Alone Thrombin in the Loss of Thrombin Residue

(A2) Hirudin-Thrombin Accessibility Due to
Com lex Complexation with Hirudin
( 2> (A2) (7.)
TASA 13,669.3 12,044.8 1,624.5 (11.9%)
MASA 2,650.0 2388.9 261.1 (9.9%)
SASA 11,019.3 9655.9 1,363.4 (12.4%)

Table 23. Solvent Accessible Surface Area of Hirudin Residues
Alone and in Complex with Thrombin. Abbreviations
used are as defined in Table 22.

Hirudin Alone Hirudin in the Loss of Hirudin Residue
(A2) Hirudin-Thrombin Accessibility Due to
Complex Complexation with Thrombin
(A ) (AZ) (Z)
TASA 4,957.3 3,090.6 1,866.7 (37.7%)
MASA 1,243.0 949.7 293.3 (23.6%)

SASA 3,714.3 2,140.9 1,573.4 (42.4%)

174

Table 24. Percent Loss of Hirudin Solvent Surface Accessibility
Due to Complexation with Thrombin. Results presented
with regard to hirudin residue ranges. Abbreviations
used are as defined in Table 22.

%Loss TASA %Loss MASA %Loss SASA
Hirudin Residue
Range:
1-3 90.7 85.4 92.6
4-48 20.2 7.2 25.0
49-51 46.8 30.0 51.0
55—58 56.4 31.9 62.3

59-65 35.7 34.5 36.0

175

display a high concentration of thrombin interactions ( Glu49'-HisSl'
and Asp55'-Glu58') in Table 21 also showed large losses in accessibility
due to complexation. In the Glu49’-Hi551' and Asp55’-Glu58' segments
there are 46.8% and 56.4% losses in total surface accessibility,
respectively. Most notable, however, is the high loss in side chain
atom accessibility in these regions which is 51.0% (Glu49'-HisSl’) and
62.3% (Asp55’-Glu58'), and is consistent with the wealth of interactions
involving the side chains of these residues. Finally, in the vicinity
of the type III helical reverse turn of hirudin, one sees a modest 35.7%
total loss in surface accessibility (34.5% main chain; 36.0% side chain)
resulting from complexation. This result is likey due to the fact that
intramolecular van der Vaals contacts between side chains atoms both in
and near the turn restrict access to residues in this region.

A listing of the residue surface accessibility of free hirudin is
presented in Table 25. The table includes the total, main chain, and
side chain atom surface accessibilities for each residue present in the
crystal structure, as well as two ratios: the total atom accessibility
(TASA) divided by the theoretical side chain atom accessibility (SASAt)
and the side chain atom accessibility (SASA) divided the theoretical
side chain accessibility. A TASA/SASAt ratio greater than one implies
that there exists at least some main chain atom accessibility, while the
SASA/SASAt ratio indicates the fraction of actual side chain atom
accessibility with regard to the maximum possible. In general, residues
with high TASA/SASAt and SASA/SASAt ratios are solvent accessible, and
many of these participate in contacts in the complex. It is of note
that while the mean values of the TASA/SASAt and SASA/SASAt ratios for
most of the hirudin residues in the complex were 0.57 (+/- 0.33) and

0.47 (+/— 0.27), respectively, (residues poorly defined in electron

176

Table 25. Residue Surface Accessibility of Free Hirudin.
Conformation of hirudin as it appears in the complex.
Abbreviations used: SASAt= theoretical side chain
atom surface accessibility. All other abbreviations
used are as defined in Table 22. All surface
accessibility results reported in A2. SASA/SASAt
ratios indicate how the actual side chain
accessibility compares to the maximum possible.
TASA/SASAt ratios greater than one indicate that

that there is at least some main chain accessibility.
These ratios are not given for glycines as glycine has
no non-hydrogen side chain atoms. A '*’ preceding a
line indicates that the values given are susceptible
to error due to disorder in or near these residues.

A ’+’ preceding a line indicates a residue which has
at least one direct contact of less than 4.0 A with

 

thrombin.

Res. MASA SASA TASA TASA SASA
SASAt SASAt

+ Ile 1’ 72.7 125.2 197.9 1.14 0.72
+ Thr 2' 29.0 94.9 123.9 0.98 0.75
+ Tyr 3' 8.9 84.2 93.1 0.43 0.39
+ Thr 4' 3.1 61.6 64.7 0.51 0.49
+ Asp 5' 1.0 90.5 91.5 0.73 0.72
Cys 6' 11.3 0.2 11.5 0.10 0.00
Thr 7' 23.7 66.9 90.6 0.72 0.53
Glu 8' 3.8 95.4 99.2 0.63 0.60
Ser 9' 25.2 42.5 67.7 0.71 0.45

Gly10’ 19.6 0.0 19.6 *** ***
Gln11' 0.0 2.4 2.4 0.01 0.01
Asn12' 0.0 4.6 4.6 0.03 0.03

+ Leu13’ 7.6 39.3 46.9 0.28 0.24
Cysl4' 0.0 0.0 0.0 0.00 0.00

+ Leu15' 3.8 34.2 38.0 0.23 0.20
Cysl6' 10.1 0.0 10.1 0.09 0.00

+ Glu17' 7.6 70.5 78.1 0.49 0.45

+ Gly18' 62.9 0.0 62.9 *** ***
+ Ser19' 23.8 77.3 101.1 1.06 0.81
+ Asn20' 2.9 98.8 101.7 0.72 0.70
+ Va121' 14.8 47.5 62.3 0.44 0.33
CysZZ' 0.2 0.0 0.2 0.00 0.00

Gly23' 30.2 0.0 30.2 *** ***

+ LysZ4' 21.5 141.2 162.7 0.83 0.72

Gly25' 44.9 0.0 44.9 *** ***
Asn26' 0.0 44.1 44.1 0.31 0.31
LysZ7' 0.0 81.1 81.1 0.41 0.41
Cy828' 0.0 0.0 0.0 0.00 0.00
Ile29’ 0.8 45.6 46.4 0.27 0.26
Leu30' 19.2 42.4 61.6 0.37 0.25

* Gly31' 77.3 0.0 77.3 *** ***

* Gly36' 102.4 0.0 102.4 *** ***
Asn37' 15.7 37.0 52.7 0.38 0.26
Gln38' 0.0 104.7 104.7 0.64 0.64

Table 25.

+ + + + + +

(cont'd.)

Cys39'
Val40’
Thr41’
Gly42’
Glu43’
Gly44’
Thr45'
Pro46'
Lys47’
Pro48'
Glu49'
Ser50'
HisSl'
Asn52'
Asn53'
Gly54'
Asp55’
Phe56’
Glu57'
G1058’
Ile59’
Pro60’
Glu61'
G1u62'
Tyr63’
Leu64'
Gln65'

l6

4.

7.
24.
19.
13.

0.

2.
11.
17.
27.
15.
33.
28.
44.
62.
38.
30.
18.
28.

12.

.9 20.4 37.
4 28.8 33.
3 116.6 123.
7 0.0 24.
4 70.3 89.
1 0.0 13.
1 94.7 94.
9 53.4 56.
1 50.8 61.
7 85.0 102.
8 110.8 138.
0 72.2 87.
5 121.2 154.
8 93.5 122.
0 59.8 103.
2 0.0 62.
0 95.3 133.
1 127.8 157.
4 127.1 145.
3 125.2 153.
5 82.6 92.
1 64.6 76.
2 122.2 126.
8 103.7 130.
7 122.0 147.
2 80.3 110.
0 122.4 202.

All residues:

mean:

standard deviation:

All residues which have

contacts of < 4.0 A
with thrombin:

mean:

standard deviation:

177

bUVU‘J-‘NHUTMOLONCDWVNO‘VVOWQH\JNONW

0.33
0.23
0.98

***

0.57

.75
.41
.32

.88
.92

.87
.74
**

.07
.79
.92
.97
.53
.56
.80

.68
.66
.24

id<D<D<D<D<D<3<D<D<Dtd *<D<D<D<D<D<D<D<D<D

0.72
0.29

C>C>

<D<3<D<D<D<D<D<D<D<D<D *<D<D<D<D<D<D<D<D<D

.58
.20

178

density and glycines were excluded), those same ratios for residues
directly involved in the hirudin-thrombin interaction were significantly

higher - 0.72 (+/-0.29) and 0.58 (+/-0.20).
C. Crystal Packing

Intermolecular forces and interactions between molecules in the
unit cell are an important part of any crystalline molecular structure.
Such forces not only stabilize the three-dimensional crystalline array
but are also likely important in influencing the symmetry within the
unit cell and the structure of the component molecules. The
rHVZ-K47-thrombin complex crystallizes in the tetragonal system, space
group P43212, with eight molecules in the unit cell and one molecule in
the asymmetric unit. As such, a 43 screw axis along c, a 21 screw axes
along a and b, and a 2-fold symmetry axis along the a/b diagonal relate
the crystallographically-independent molecules in the unit cell. The
eight symmetry-related equivalent positions, along with that of the
reference molecule (x,y,z), are given below [131]:

x, y, z
-x, -y, 1/2+z

1/2—y, 1/2+x, 3/4+z
1/2+y, 1/2-x, 1/4+z

y, X: -2

-y, —x, 1/2-z
1/2-x, 1/2+y, 3/4-z
1/2+x, 1/2-y, 1/4-z

#LAJNH
muom

...(42)

In the crystals of the complex, both electrostatic and van der
Vaals forces appear to be important in the crystal packing interactions
in the unit cell. A particularly striking hydrophobic cluster of
residues occurs at a two-fold symmetry axis between the reference
molecule (symmetry position 1, relation 42) and a two-fold
symmetry-related molecule (symmetry position 5, relation 42) (Figure

58). Seven hydrophobic/aromatic groups from the two participating

179

 

Figure 58.

Stereoview of a Hydrophobic Crystal Packing Interaction

at a Two-Fold Symmetry Axis. Top: Molecules involved are
at symmetry positions 1 and 5 (relation 42, text).
Disulfides and hirudin in bold. Two—fold axis indicated.
The hydrophobic interactions occur in the circled region and
are detailed below. Bottom: Nonpolar/aromatic residues
which comprise the hydrophobic cluster. Groups involved:
Leu14G, Tyr14J, I1e14K, Leu129C, Tyr134, Pr0204, and Phe204A
(all bold) and the corresponding symmetry-equivalent groups.

180

thrombin molecules comprise this cluster - Leu14G, Tyr14J, Ilel4K,
Leu129C, Tyr134, Pr0204, and Phe204A (Figure 58). It is of particular
note that three residues at this interface are from the thrombin A
chain, a portion of thrombin surface which does not adjoin hirudin in
the complex. Residue Leu14G has two nonpolar contacts < 4.0 A with
Ile#14K (#14K indicating residue 14K of a thrombin molecule at symmetry
position 5), and Ilel4K has two close van der Vaals contacts with
Ile#14K in addition to the two with Leu#14G. Moreover, residue Tyr14J
stacks against Tyr134 (five close contacts), and has two close
approaches each with Pr0204 and Leu14G. In the thrombin B chain, the
side chain atoms of Leu129C have three van der Vaals contacts of < 4.0 A
with ring carbon atoms of Phe 204A, and residue Pr0204 has three
nonpolar contacts with Phe 204A as well as those with Tyr134. Finally,
Phe204A has 13 close hydrophobic approaches with residues of the
neighboring thrombin molecule - three with Leu#129C, four with Tyr#134,
and two each with residues Tyr#14J, Pro#204, and Phe#204A. It is also
of interest that this nonpolar cluster occurs in a region on the
thrombin surface on the opposite side from the extended site of hirudin
binding (Figure 58).

Another packing interaction related by two-fold symmetry is
electrostatic, and is also a hirudin-thrombin contact. It involves
residues Glu57'-Arg#75 and Glu#57'-Arg75 of the complex (positions 1 and
6, relation 42) (Figure 59). These ion pairs are of special interest
for two reasons. First, the fact that they are double hydrogen-bonding
ion pairs with good donor-acceptor distances (Arg#75 NHZ-Glu57' OEZ,
2.93 A; Arg#75 NE-Glu57' OBI, 2.81 A) and good donor-proton-acceptor DHA
angles (176° and 173°) (Table 26) suggests that the forces are

strong and stabilizing packing interactions. Second, the

Figure 59.

181

 

Steroview of an Electrostatic Crystal Packing Interaction

at a Two-Fold Symmetry Axis. Top: Molecules involved are
at symmetry positions 1 and 6 (relation 42, text). Hirudin
and all disulfides are in bold. Two-fold axis indicated.
The electrostatic crystal packing interactions are within
the circled region and are detailed below. Bottom: The
double hydrogen-bonding ion pairs involved in the crystal
contacts. Dashed lines indicate hydrogen bonds; hirudin
glutamate residues are in bold. Two—fold symmetry axis
offset approximately 30 degrees from the perpendicular to
the plane of the paper.

182

Table 26. Hydrogen Bonds Between Symmetry-Related Molecules.
Hydrogen atoms are assigned geometrically idealized
positions. Donor atoms are denoted ’D' and acceptor
atoms 'A'. The symmetry molecule number refers to
the symmetry equivalent position of the symmetry-related
molecule in the interaction, as given below. The
symmetry-related molecule residue number is preceded
by a ’#’.

Symmetry-equivalent Positions in P43212:

1. x, y, z 5. y, x, -z
2. -x, -y, 1/2+z 6. -y, —x, 1/2-2
3. 1/2-y, 1/2+x, 3/4+z 7. 1/2-x, 1/2+y, 3/4-z
4. 1/2+y, 1/2-x, 1/4+z 8. 1/2+x, 1/2-y, 1/4-z
Symmetry Donor Acceptor Distances (A) Angles(deg)
Molecule H..A D..A DHA CAH CAD
Number
7 Lys145 NZ Gln#244 0 2.46 3.12 123 134 147
7 Trp148 N Gly#10 0 1.69 2.58 147 138 140
7 Thr149 N Gln#244 0E1 1.99 2.96 161 120 114
7 Asn149B ND2 Leu#123 0 1.69 2.73 172 145 148
7 AlafilB N Trp148 0 2.04 2.94 155 160 160
6 Arg#75 NH2 Glu57 0E2 1.89 2.93 176 115 114
6 Arg#75 NE Glu57 0E1 1.80 2.81 173 117 119
7 Lys#235 NZ Asn149B 0 2.04 2.92 141 148 159
7 Gln#244 NE2 Thr149 0 1.94 2.87 155 122 129
7 Gln#244 NE2 Va1149C 0 1.73 2.45 127 160 149
Mean: 1.93 2.83 153
Standard Deviation: 0 23 0.20 19

183

hydrogen—bonding ion pairs form between side chains of adjoining complex
molecules, while the spatial proximity of these residues to one another
within the complex suggests that in solution the ion pairs could form
intramolecularly by free bond rotation.

Finally, there is an extended crystal packing interaction involving
residues of the thrombin Lysl45-Gly150 undecapeptide loop of one
molecule (symmetry position 1, relation 42 ) and the NHZ- and
COOH-terminal chain residues of another molecule (symmetry position 7,
relation 42) (Figure 60). At this interface, a remarkable eight
hydrogen bonds are formed with residues in the insertion loop (Table
26). The amide nitrogen of Ala#1B and the carbonyl oxygen of Gly#1D
form hydrogen bonds with the carbonyl oxygen and and amide nitrogen of
Trp148, respectively. The C—terminal residue Gln#244 engages in four
hydrogen bonds with residues of this thrombin loop: the carbonyl oxygen
forms a hydrogen bond with Lysl45 NZ, atom 081 of the gluatamyl side
chain interacts with N of Thr149, and both protons of the side chain
atom NE2 participate in hydrogen bonds with the carbonyl oxygens of
Thr149 and Va1149C. An aromatic stacking interaction also takes place
between the phenyl ring of the C-terminal residue Phe#245 and Trp148.
The loop conformation is further stabilized by two hydrogen bonds
involving Asn149B and Leu#123, Lys#235. The wealth of interactions
involving residues Lysl45-Gly150, which is the largest protruding loop
structure in the nearly spherical thrombin molecule (Figures 43, 56),
could explain in part why it is so well-defined in the crystal structure
of the hirudin-thrombin complex. In addition, it is also plausible that
in the absence of this extensive network of stabilizing interactions, as

in solution, the loop could be disordered.

    

Figure 60.

184

I 1493 7 #235 I 1493 7 #235
Jr #123 #123

Stereoview of a Crystal Packing Interaction Involving

the Thrombin Lysl45-Gly150 Undecapeptide Loop. The crystal
contact involves molecules at symmetry positions 1 and 7
(relation 42, text). Residues involved: Lysl45, Trp148.
Thr149, Asn149B, Va1149C, Gly#1D, Ala#1B, Leu#123, Lys#235,
Gln#244, and Phe#245. N, CA, C atoms and residues of the
Lysl45-Gly150 loop engaged in hydrogen bond interactions in
bold; hydrogen bonds indicated by dashed lines.

VIII A COMPARISON OF THE THROMBIN STRUCTURE IN THE

HIRUDIN-THROHBIN COMPLEX AND IN PPACK-THROHBIN

Having examined the structure of hirudin in the complex and the
hirudin-thrombin interaction in detail, the thrombin structure of the
complex and and the manner in which it differs from that in
PPACK-thrombin [6] will now be considered. An optimal superposition of
the two thrombin structures on the basis of the main chain atoms N, CA,
C was obtained by means of the program MOLFIT. This program calculates
the rotation matrix and translation vector which 'best’ superimpose one
orthogonal coordinate set upon another by minimizing the root mean
square (r.m.s.) deviation between the basis and transformed coordinate
sets. The MOLFIT results for the two thrombin structures are shown in
Table 27. The agreement on CA and on the main chain atoms N, CA, C is
quite good (Table 27). The r.m.s. difference in N, CA, C reduces to
0.27 A (609 atom pairs) when deviations larger than 10 are excluded from
the comparison. An examination of Figure 61 verifies that in general
the overall folding is very similar between the two molecules. Not
surprisingly, the r.m.s. differences of non-main chain atoms are larger
(Table 27). Though the r.m.s. difference on sulfur atoms is 0.68 A,
both cysteine and methionine sulfurs are included in this value, and it
is more meaningful to consider these different sulfur atoms separately.
A calculation of the average distance between corresponding sulfurs
reveals that for cysteine sulfur atoms the value is only 0.24 A, while
for methionine sulfurs the value is 0.60 A. The small deviation with
Cys sulfur atoms is likely due to the fact that they are all engaged in
disulfide linkages, which are both conformationally restricted and

important in maintaining the folding of the thrombin molecule. The

185

186

Table 27. Results of an Optimal Superposition of the Thrombin
Structures of the Complex and PPACK-Thrombin [6] using
the Program MOLFIT. Fitting based on N, CA, C Atoms.
Results expressed in A units as root mean square
deviations (r.m.s.d.). Numbers in the '{]' brackets
represent the number of atom pairs contributing to
the r.m.s.d.

r.m.s.d. (A)
Main Chain (N,CA,C) Atoms 0.45 [818]
Alpha Carbon Atoms 0.46 [272]
Carbonyl Oxygens 0.70 {274]
Sulfurs (Cys, Met) 0.68 {15]
Side Chain Atoms 1.43 {1103}

All Protein Atoms 1.08 [2195]

187

 

Figure 61.

Stereoview of the Optimally Superimposed Thrombin

CA Structures of the Complex and PPACK—Thrombin.
Superpositioning corresponds to the MOLFIT N, CA, C fitting
results of Table 27. Residues of the complex in bold.

188

larger deviation with Met sulfur atoms is consistent with the fact that
these sulfurs are located in side chains which have significantly more
flexibility than a cystine linkage. To be more exact, the average
deviation in methionine sulfur positions is large mainly because of a
single difference between the SD sulfurs of Met84 (2.42 A), a surface
residue.

To determine the location of the significant differences, and in an
effort to assess the possible causes for the differences, the mean
r.m.s. deviations of main chain and of side chain atoms versus residue
number were examined (Figures 62 and 63). All residues with r.m.s.
differences greater than 10 (> 0.45 A for all main chain atoms, > 1.43
A for all side chain atoms) were visually examined using computer
graphics in the optimally superimposed thrombin strutures with the aid
of an Evans and Sutherland P5390 stereographics display and the PSFRODO
version of the molecular graphics program FRODO. Following this
extensive comparison, most of the significant structural differences can
be accounted for in terms of the following factors or influences: (a)
tentatively placed residues, poorly defined residues in the electron
density of either structure, or alternate interpretations of density
maps; (b) affect of crystal packing interactions; (c) interaction with
hirudin; and (d) variability of surface residue conformation. Listings,
with pertinent comments for the residues possessing significant main
chain and/or side chain r.m.s. deviations ( > 10 ) believed to result
from each of the four categories of factors or influences, are presented
in Tables 28 to 31.

Inspection of Figures 62 and 63, as well as a more careful
examination of Figure 61, reveals that the most outstanding main chain

and side chain differences in the two structures occur at the N- and

 

189

 

 

 

 

 

 

 

 

 

 

 

 

 

A J Autumn!!!
0|‘lf- Berlol-Olvldfl .
5‘. _1_Aop|A
‘15 _"" q-
H.242-
Dana‘s
«a _F
2
5 l5 —~
i
- 10 -~
1 0
i‘
2.5 --i-
o'
I Prowl-
‘ °'"‘a c 191- a :03-
‘- 2.. "'0 Luuo out“ 4: s
i _ s
‘ f,..°.. AI||W \ . 0‘
Y 1.3 --.""_[ A.” [ Lulu 1‘ “:33“;
L001?
e .3.- hr 79- '01
I t3.“ } “:70 i "I‘79 '\ one" 1A
‘4. "i- I i . \v .I. "
r ‘ ‘ A AA- 1 "
es-E ' . +
._. lll [JillllILllillllllll
'llllllIleIlllllllllrlTllIll
I» la 19 as a. to so nor :21 104 :99 171 194 2:2 233 249
Itlflll I'll!
Figure 62. Average R.M.S. Deviation of Main Chain Atoms versus

Residue Number for the Superimposed Thrombin Molecules of
the Complex and PPACK—Thrombin. Superpositioning as in
Figure 61. The horizontal lines represent la and 20
deviations. Most of residues with deviations > la are

identified.

5..

4.5

4..

3.5

3..

190

 

2.5

KAN I H In roe CINCMIN A7015

0.5

I.‘

IN

Figure 63.

A

Luello J
._.] "H!-
«v.77. Q—ThrtGT-Valaoc,
LuelQn.
Borlfi. Olulc > OIHISI
A Anala-fheals
lie!“-
""" clnaavzlgcato
l Thrgol— ’ \1'
A. 7"‘73 Lqeit 'IOZO‘A
\3 r”
FA A b
‘ “W'so 2' '7 Thri77
I?” \‘f‘ i .'.12. 1
Lawn ”I"? 1
H0004 Oluxqa L002“
Q...“ LIU|I°

 

 

Ola:-

 

T. .T T

 

 

 

 

 

 

 

 

 

.4.

13 I. :0 fl .9 so 107 137 I“ I” 1” 1°‘ :12 232 245

mm: W

Average R.M.S. Deviation of Side Chain Atoms versus

Residue Number for the Superimposed Thrombin Molecules of
the Complex and PPACK-Thrombin. Superpositioning as in
Figure 61. The horizontal lines represent 1a and 20
deviations. Many of the residues with deviations > la are
identified.

Table 28.

Residues(s)

GlylF
SerlE
Glle
Ilel4K
Asp14L
Gly14M
Asp60E
Glu97A

Leu99

LysllO

191

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin

[6] due to Residues Poorly Defined in Electron
Density or Tentatively Placed in Either Structure,
or due to Alternate Interpretations of the Density
Maps. Optimal superpositioning of thrombin
molecules as in Table 27. Abbreviations: HT =
hirudin-thrombin complex; PT = PPACK-Thrombin;

MCA r.m.s.d. s main chain atom average root mean
square deviation; SCA r.m.s.d. a side chain atom
average r.m.s.d.; S a surface; AS a active site;
conf. . conformation; rot. a rotation. Deviations
> la on main chain (0.45 A) and/or side chain (1.43 A)
atoms are listed.

Location MCA SCA Comments
r.m.s.d. r.m.s.d.
(A) (A)

S 12.34 ---- ——Gly1F-Gly1D undefined

8 13.16 19.94 in PT & tentatively

S 9.71 --—- placed along B-chain [6].

S 1.62 2.66 --Ile14K-Gly14M undefined

S 5.28 3.63 in PT 8 arranged with

5 11.13 ---- regard to crystal packing
[6].

S 1.01 1.67 --no side chain density
beyond atom CB in HT

5 1.29 —--— --no side chain density in
HT new conf. for side chain

AS 0.24 1.67 --approximate 180° rot.

CB-CG bond; could be due to
density map interpretation.
5 1.50 5.00 --different main chain &
side chain conf. due to
cis-Prolll in PT, trans-
Pr0111 in HT; density map
interpretation difference.

192

Table 29. Structural Differences Between the Thrombin Structures
of the Complex and PPACK-Thrombin [6] due to Crystal
Packing Interactions in the Complex. Optimal
superpositioning of thrombin molecules as in Table 27.
Abbreviations: as = aromatic stacking; hb =
hydrogen bond; hip = hydrogen-bonding ion pair; hc s
hydrophobic contact; pc . polar contact; sympos a
symmetry molecule positions (see Table 26), wmi- water
mediated interaction. All other abbreviations are
defined in Table 28. Residues in brackets [] are not
involved in crystal contacts but have significant
conformational differences which could be due to packing
interactions of nearby residues. A '*' denotes the
average MCA and SCA r.m.s.d. for residues Asn143-G1n151.
Table 32 gives a complete listing of the MCA and SCA
r.m.s.d.'s for the residues. Deviations > la on main
chain (0.45 A) and/or side chain (1.43 A) atoms listed.
Residues(s) Location MCA SCA Packing Interactions/
r.m.s.d. r.m.s.d. Comments
(A) (A)
GlylF S 12.34 ---- --sympos-l,7; wmi 0—V534-
Glu#146 0E2.
Serlfi 5 13.16 19.94 --sympos-1,7; OG-Lys#47’ NZ
(pc) & OG-Thr#4'OG(pc)
Glle S 9.71 ---- --sympos-1,7: Trp#148(hb)
[Glu1C] S 6.79 7.92
AlalB S 1.91 0.55 --sympos-1,7: Trp#148(hb)
[AsplA] S 0.55 0.53
[Ser14I] S 0.57 0.20
Tyr14J S 0.74 0.97 --sympos-1,5: Tyr#134(as),
Phe#204A(as).
Ilel4K S 1.62 2.66 --sympos-1,5: Leu#14G(hc),
Ile#14K(hc).
Asp14L S 5.28 3.63 --sympos-1,5: 0-Arg#14D NH1
(PC)
[Gly14M] S 11.13 ----
Arg75 S 0.61 1.96 --sympos-1,6: Glu#57'(hip)
Asn143-Gln151 S 3.99* 3.71* --sympos-1,7: 8 hydrogen
bonds involving these
residues (see Table 26).
[Ser203] S 0.65 1.03
Pr0204 S 1.65 0.62 --sympos-1,5: Ile#14K(hc),
Pro#204A(hc)
Phe204A S 1.15 3.16 —-sympos-1,5: Tyr#14J(hc),
Leu#129C(hc), Tyr#134(hc),
Pro#204(hc), Phe#204A(hc)
[Asn204B] S 1.23 0.56
[Asn205] S 0.54 1.04
[Ile242] S 0.67 0.72
[Asp243] S 1.19 5.35
Asn244 S 5.83 11.31 --sympos-1,7: Lys#145(hb),
Thr#149(hb), Val#149C(hb).
Phe245 S 6.57 13.52 --sympos-1,7: Trp#148(as)

Table 30.

Residues(s)

Phe34
Arg35
Lys36
Pro6OC

Trp6OD
Lys60F
Arg75
[Tyr76]
Arg77A

[Asn78]

Asn143-Gln151

Arg173
Ile174
[Arg175]
Cysl91
Glu192

[Gly193]
[Asp194]
Ser195
[Gly196]
Gly219

193

Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6]
due to the Interaction with Hirudin. Optimal
superpositioning of the thrombin molecules as in
Table 27. All abbreviations are defined in
Tables 28 and 29. Residues in brackets [] do not
interact with hirudin, but have significant
conformational differences which could be due to
hirudin interactions of nearby residues.
Deviations > 10 on main chain (0.45 A) and/or
side chain (1.43 A) atoms listed.

Location MCA SCA Interacting Hirudin
r.m.s.d. r.m.s.d. Residues/Comments
(A) (A)
S 0.46 0.87 Phe56'(hc)
S 0.68 1.05 Glu49’ via V508
S 0.68 0.82 Gln65’(ip)
S 0.85 0.85 Leu13'(hc), Ly824'(pc),
Pro46’(hc)
S 1.15 0.87 Ile1’(hc), Lys47'(pc),

Glu49’(mostly he)

5 0.71 1.28 Glu49’(hb)

S 0.61 1.96 Glu#57'(hip—sympos-1,6)

S 0.64 1.08

S 0.78 5.79 Glu58’(ip), Glu57’ via
V467

S 0.61 1.01

S 3.99* 3.71* Lg. shift possibly
explains CD change
occurring upon hirudin
C-terminal peptide bind-
ing but not with PPACK.

S 0.60 0.96 Glu17'(ip), Asn20'(pc)

S 0.82 0.42 Va121'(pc)

S 0.64 0.71

AS 0.84 0.34 Thr2'(pc)

A8 0.88 1.70 Thr2'(pc), Thr4'(pc),
Ser50'(pc)

A5 0.70 ---

AS 0.60 0.75

AS 0.50 0.77 Ile1'(hb)

AS 0.49 ---

AS 0.53 --- Thr2'(pc), Tyr3'(hb),

Leu15'(pc)

Table 31. Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin
[6] due to Surface Residue Conformational
Variability. Optimal superpositioning of the
thrombin molecules as in Table 27.
Abbreviations: fs = free side chain; ident. =
identical. All other abbreviations defined

in Tables 28 and 29.

Deviations > 10 on

main chain (0.45 A) and/or side chain (1.43 A)

atoms listed.

Residues(s) Location MCA
r.m.s.

(A)

ProS S 0.54
Serll S 0.69
Leu12 S 0.67
Lysl4A S 0.34
Arg14D S 0.59
Glu14H S 0.30
Glu14E S 0.61
Gly19 S 0.64
Ser20 S 0.53
Asp21 S 0.48
Gln38 S 0.31
Leu41 S 0.20
Arg50 S 0.23
Asn6OG S 0.52
Phe60H S 0.58
Thr60I S 0.74
Glu61 S 1.69
Asn62 S 1.61
His71 S 0.51
Ly581 S 0.45
Met84 S 0.23
Ly587 S 0.40
Trp96 S 0.67

SCA

.m.s.d.

(A)

0.75
1.34

1.11

1.38

2.43
1.40
0.78

0.50
0.55
1.98

1.45
2.63
1.03
0.99
1.59
1.53
4.62
0.51
3.30
1.89
1.67

1.39

Comments

--CA,C,O offset; fs, side
chain nearly ident.
--peptide atoms offset; fs,
rot. approx. 90° CA-CB

bond

-- new side chain conf.;

new pc: NZ-Asp 21 O, 3.53 A.
--new ip w/ Glu14H, 3.8 A

—-peptide atoms offset; side
chain conf. nearly ident.
new ip w/ Glu14E, 3.8 A.
--main chain atoms of
surface

tripeptide Gly19-Asp21
offset

--approx. 90° rot. CA-CB
bond

--approx. 90° rot. CA-CB
bond; new hc's w/ Phe60H
-—new conf. for side chain
—-Thr6OI-Asp63 turn shifted;
new ¢, 0 angles for Asn62
classify turn as type 1; en-
hanced hc’s between Phe6OH,
Leu64, Leu41 in new conf.
--main chain atoms offset;
approx. 20° twist to
histidine ring

--new hb: NZ-Ala 113 0,

2.73 A

--approx. 30° rot CA-CB

bond

--fs; approx. 30° rot.

CB-CG bond

--main chain atoms offset;
side chain conf. nearly
ident.

195

Table 31 (cont’d.)

Arg97 S 1.14 2.32 --main chain atoms offset;
new conf. for side chain;
new pc: NH1-Asn95 0D1,

3.35 A
Asn98 S 0.56 1.35 --main chain atoms offset;
new conf. for side chain
Arg126 S 0.26 2.08 --new side chain conf.
Glu127 S 0.34 1.42 --new side chain conf.
Leu160 S 0.21 1.59 --approx. 45° rot.
CA—CB bond
Va1167 S 0.51 0.66 -—peptide atoms offset; side
conf. nearly ident.
Lysl69 S 0.32 1.45 --fs; new side chain conf.
Thr177 S 0.41 2.27 --approx. 180° rot.
CA—CB bond
Asp186A S 0.46 1.46 -—fs; new side chain conf.
Glu186B S 0.86 0.57 --peptide atom conformation
significantly different
Gly186C S 0.92 ---- for G1u186B-Ly5186D
Lysl86D S 0.60 1.22 --new ip w/ Glu14E, 3.8 A
Leu234 S 0.37 1.56 --approx. rot. 180°
CB-CG bond.
Trp237 S 0.47 0.44 --shifting of type III turn
Ile238 S 0.56 0.52 Lys235—Ile238
Gln239 S 0.62 3.69 --new side chain conf.
LysZ40 S 0.68 3.56 --fs; new side chain conf.

196

Table 32. Structural Differences Between the Thrombin
Structures of the Complex and PPACK-Thrombin [6]
in the Vicinity of the 149A-149E Insertion Loop.
Optimal superpositioning of the thrombin molecules
as in Table 27. Abbreviations: r.m.s.d. = root
mean square deviation; MCA r.m.s.d. = average
main chain atom r.m.s.d.; SCA r.m.s.d. =
average side chain atom r.m.s.d. All residues
listed have deviations > la on main chain (0.45 A)
and/or side chain atoms (1.43 A).

Residue MCA SCA
r.m.s.d.(A) r.m.s.d.(A)
Asn143 0.65 0.98
Leu144 0.63 0.62
Lysl45 0.85 3.24
Glu146 1.52 1.11
Thr147 4.48 5.25
Trp148 7.22 8.07
Thr149 7.54 9.52
Ala149A 7.59 1.43
Asn149B 8.84 5.88
Va1149C 6.62 2.33
Gly149D 5.19 ---—
Lysl49E 2.82 4.56
Gly150 0.97 ----
Gln151 0.38 1.51

197

C-termini of the thrombin A-chain, the C-terminus of the thrombin
B—chain, and in the loop containing the l49A—149E insertion. Moreover,
it appears that the the aforementioned factors (a), (b) and (c) are at
the core of these large deviations. The drastic differences in
conformation of the tripeptide segments GlylF—SerlE-Glle and
Ilel4K-Asp14L-Gly14M at the N- and C-termini of the thrombin A-chain in
the two structures (Figure 61), and the large average r.m.s. deviations
for main chain (Figure 62, Table 27) and side chain atoms (Figure 63,
Table 27) for these polypeptide segments results from the fact that
these residues were totally undefined in the PPACK-thrombin structure,
and were arranged in reasonable orientations — along the thrombin
B-chain for residues lF-lD, and in a conformation allowed by crystal
packing for residues 14K—14M [6]. The probability that the conformation
of these positioned residues would be similar to that observed in the
hirudin-thrombin crystals is low.

Packing interactions in the crystals of the complex appear to play
a role in the conformation of residues at the N- and C- termini of the
thrombin A-chain, the C-terminus of the thrombin B-chain and in the
Ala149A-Lysl49E insertion loop (Tables 26 and 29). As was seen in
Figure 60 and mentioned in the previous chapter, at a crystal contact
between molecules at symmetry positions 1 and 7 (Table 26), A-chain
residues Gly#1D and AlafilB engage in two hydrogen bonds and B-chain
residues Gln#244 and Phe#245 engage in four hydrogen bonds and an
aromatic stacking interaction with residues in this insertion loop
(Lysl45, Trp148, Thr149, Asn149B, and Va1149C). Such strong
intermolecular forces will clearly have an impact on the conformation of
the short polypeptide stretches Gly#1D-Ala#lB and Gln#244-Phe#245, as

well as on nearby residues in these regions. In particular, Figure 64

 

198

 

Figure 64.

Stereoview of the Superimposed Residues LysZ36—Phe245

of the Complex and PPACK-Thrombin. Superpositioning as in
Figure 61. Residues of the complex in bold. Symmetry
molecule interactions of the complex displayed; symmetry
molecules indicated by a '#' preceding the residue number.
Hydrogen bonds indicated by dashed lines.

 

199

reveals how interactions at an analagous crystal contact, involving
Asn244 and Phe245 with residues in the 140-150 thrombin loop of a
symmetry-related molecule, are most likely responsible for the
conformation of the C-terminal residues Asp243-Phe245 in complex; the
C-terminal B-chain folding of residues LysZ36—Ile242 agrees quite well
in the two structures but at Asp243 the termini diverge. In addition,
the conformation of the N-terminus of the thrombin A-chain is likely
influenced by the hydrophobic and aromatic stacking interactions
involving Tyr14J and Ilel4K (Figure 58) and by a close polar contact
between the carbonyl of Asp14L and Arg#14D NH1 (3.57 A) at a crystal
contact involving symmetry-related molecules at crystal positions 1 and
5 (Table 26).

The wealth of hydrogen bonds (8) along the protruding thrombin
insertion loop from residues Lysl45 to Va1149C (Table 26, Figure 60) and
the aromatic stacking interaction between Trp148 and Phe#245 (Figure 60)
at the crystal contact between molecules at symmetry positions 1 and 7
(Table 26) suggests that the conformation of this loop is
well-stabilized by, if not dependent on, these interactions. The
drastically different conformation this loop adopts in the two
structures is shown in Figure 65. The most striking feature of the two
loop conformations is the relative position of the indole ring of
Trp148. In the complex, the indole is at the center of the loop and
makes many contacts, while in PPACK-thrombin it is directed outside of
the loop and toward the solvent. In addition, 14 residues in this loop
have the r.m.s. differences > lo on main and/or side chain atoms
between the two structures; these values are presented in Table 32.

Crystal contacts alone may not be the only factor involved in

influencing or provoking the dramatic shift in the conformation of this

 

200

146 150

 

Figure 65. Stereoview of the Superimposed Residues Glu146-Gly150
of the Complex and PPACK-Thrombin. Superpositioning as in
Figure 61. Residues of the complex in bold.

201

thrombin loop in the complex. The orientation of this loop in the
complex cannot be the same as in the PPACK-thrombin structure, as it
would result in steric problems and would prevent binding of the hirudin
N-terminus in the active site region of thrombin; this point is
graphically made in Figure 66. If the insertion loop adopted the same
conformation in the complex as in PPACK-thrombin, the indole of Trp148
would practically collide with the Thr4'-Asp5’ dipeptide of hirudin, and
would obstruct the hirudin N-terminal tripeptide from penetrating into
the thrombin active site. Thus, a change in the loop conformation
between the two structures is necessary. Moreover, since the presence
of the hirudin N-terminus in the thrombin active site region is
important to the potency with which hirudin binds and inhibits thrombin,
it is not surprising that the position of the loop and the indole of
Trp148 in the complex are compatible with this important
hirudin-thrombin interaction; the thrombin 140-150 loop residues are
well away from the five hirudin N-terminal residues and the region where
they interact with thrombin (Figure 66).

As hirudin is a 'double-headed' thrombin inhibitor, it is worth
considering whether the binding of the N-terminal head or the C-terminal
tail is responsible for triggering this necessary loop shift. On the
basis of steric restraints and on the fact that the hirudin NH2-terminus
must enter the thrombin active site region for optimal binding, one
could argue that the hirudin N-terminal head is responsible for the
shift in the 149A-149E insertion loop. However, two different types of
experimental evidence point to the likelihood of hirudin C-terminal tail
binding to thrombin as the cause of the loop shift. Circular dichroism
(CD) studies indicate that the binding of hirudin or a dodecapeptide

analogue of the hirudin COOH—terminal tail to thrombin produces a change

Figure 66.

202

Q51?

1' “(795

   

150

Stereoview of the Superimposed Residues G1u146-Gly150

of PPACK-Thrombin and the Complex, including the Hirudin
N-Terminal Pentapeptide. Superpositioning as in Figure 61.
Residues of PPACK-thrombin in bold. Thrombin catalytic
triad of the complex displayed.

203

in CD which does not occur upon PPACK inhibition [95,96]. Vhile the
exact structural transitions repsonsible for these CD changes are not
known, it is possible that the CD spectal results reflect the altered
conformation for the indole of Trp148 and its associated 100p.
Moreover, the lack of change in CD upon PPACK binding to thrombin and
the presence of one upon whole hirudin or C-terminal analogue binding
suggest that hirudin binding in the vicinity of the anion-binding
exosite, and not the thrombin active site, is responsible for the
spectral change. In addition, studies of the kinetics of hirudin
inhibition of thrombin indicate that C-terminal tail binding precedes
N-terminal head binding, as the rate of the first step in complex
formation is decreased by increasing the ionic strength and is
unaffected by binding at the active site of thrombin [59]. Taken
together, the experimental data suggest that initial C-terminal tail
binding of hirudin to the anion-binding exosite region of thrombin is
responsible for inducing a CD change, which could be due to a
conformational change involving Trp148 and its associated loop. Thus,
all in all, it appears as though the hirudin-thrombin interaction as
well as crystal contacts are influential in determining the structure of
the 149A-149E thrombin insertion loop.

Vhile the largest differences in the thrombin structures have been
examined, the majority of the significant differences have yet to be
considered in terms of the four categories of influences likely involved
in these changes. In category (a), two differences appear to result
from poorly defined residues in the complex and two appear to be due to
alternate interpretations of electron density maps (Table 28). The
residues Asp60E and Glu97A of the complex show large side chain and/or

main chain deviations with respect to the PPACK-thrombin structure, most

 

204

likely because there is little or no side chain density for these
residues in hirudin—thrombin (Table 28). The structural deviations
associated with Leu99 and LysllO appear to stem from map interpretation
(Table 28). Residue Leu99, in the active site region, has nearly the
same main chain conformation in the two structures but differs by 1.67 A
in its side chain. Though this deviation is large in magnitude, in
actuality the side chains in the two structures occupy the same space
and simply differ by a near 180° rotation about the CB-CG bond;
thus, either could be suitable for modeling in the density map. A much
more dramatic difference, resulting from alternate density map
interpretations, involves residues LysllO-Prolll (Figure 67). The
peptide bond preceding residue Prolll was interpreted to be cis-, making
this a cis-proline (CP), in PPACK-thrombin while in the complex a more
typical trans-peptide bond was modeled. Moreover, while a cis- versus a
trans-peptide bond does affect the configuration of Prolll, it places
much more severe restraints on the conformation available to the
preceding lysine residue, and is why the main and side chain deviations
are so large for LysllO (Figure 67).

Though many of the significant structural differences linked to
crystal contact interactions have already been mentioned, there are a
few more worth pointing out. The double-hydrogen bonding ion pair
between Arg#75 and GluS7' which takes place at the two-fold axis between
molecules at symmetry positions 1 and 6 (Table 26, Figure 59) is likely
the source of the large main and side chain deviations between the two
thrombin structures at Arg75 (Table 29). Also, the hydrophobic cluster
created at another two-fold axis by residues from the thrombin A— and
B-chains of molecules at symmetry positions 1 and 5 (Table 26, Figure

58) is probably the cause of the significant main and side chain

Figure 67.

205

111

110

 

Stereoview of the Superimposed Residues LysllO-Prolll

of the Complex and PPACK-Thrombin. Superpositioning as in
Figure 61. Residues of the complex in bold. Peptide bond
preceding Prolll is cis- in PPACK-thrombin, trans- in the
complex.

 

206

deviations with respect to PPACK-thrombin for residues Pr0204 and
Phe204A, and those which adjoin them (Ser203, Asn204B, and Asn205). The
heptapeptide segments LysZOZ-Arg206 taken from the main chain fitted
thrombin structures are displayed in Figure 68, which reveals the
noticeably different positioning of Phe204A in each molecule.

An examination of the entries in Tables 28 to 31 reveals that the
majority of the conformational differences in the thrombin structures
appear to arise from the hirudin-thrombin interaction and the
variability of surface residues. The notion that the hirudin-thrombin
interaction is responsible for producing numerous significant structural
perturbations in the thrombin molecule with respect to the
PPACK-thrombin structure is certainly reasonable when one considers the
potency of this interaction, and the fact that only 26 of the 65 hirudin
residues are responsible for producing a remarkable 216 contacts < 4.0 A
with thrombin. Also not surprising is the fact that the bulk of the
affected thrombin residues are associated with or adjoin important
hirudin (and in general macromolecular) binding sites on thrombin -
namely, the active site cavity and the anion-binding exosite region
(Table 30). Moreover, these hirudin residues are largely found in short
peptide stetches displaying high interaction per residue densities -
Ile1'-Tyr3’ (41 contacts/3 residues or 13.7 contacts/residue),
Glu49'-HisSl' (32 contacts/3 residues or 10.7 contacts/residue), and
Asp55'-Glu58' (60 contacts/4 residues or 15.0 contacts/residue) (Table
21).

In an effort to more carefully examine the conformational
differences between the thrombin structures in the active site binding
cavity, consider Figure 69 which displays the superimposed active sites

from the hirudin-thrombin (bold) and PPACK-thrombin structures, and

 

207

 

Figure 68. Stereoview of the Superimposed Residues Lys202-Arg206 of the
Complex and PPACK-Thrombin. Superpositioning as in Figure
61. Residues of the complex in bold.

 

208

10

.15 21

 

Figure 69. Stereoview of Superimposed Active Site Residues of
PPACK-Thrombin and the Complex, including the Hirudin
N-Terminal Tripeptide. Residues of the complex in bold.
Superpositioning as in Figure 61. Hydrogen bonds indicated
by dashed lines.

209

includes the hirudin N-terminal tripeptide (bold). It is interesting to
note that while the residues involved in the B—strand interaction with
the hirudin N-terminal tripeptide, Ser214—Gly219, are in nearly the same
conformation in the two structures, residues Cysl91-Ser195 and in the
Tyr60A-Trp60D loop have quite different conformations between the two.
The basis for the structural similarity of the Ser214-Cy5220 polypeptide
segments is likely due to the fact both hirudin and PPACK are involved
in strand interactions with this rigid thrombin stretch and, even though
the the strand interaction is antiparallel in PPACK and parallel in the
complex, most likely the inhibitor tripeptides conform to the thrombin
strand rather than vice versa. However, the significant structural
perturbations at positions Cysl91—Ser195 and Tyr60A-Trp60D most likely
arise from hirudin-thrombin interactions which have no analogous match
in the interaction of PPACK with thrombin. The large deviation in
Cysl91-Ser195 strand probably results from the need for Glu192 to adopt
a different conformation in the complex, as well as from a unique
interaction between Ser195 and the hirudin NHZ-terminus. The Glu192
side chain cannot have the same position in the complex as in
PPACK—thrombin for its carboxylate would come too close to the Thr2'
side chain of hirudin. This problem coupled with the fact that the side
chain can form a stabilizing close polar contact with Ser50' is
sufficient to induce the Glu192 main chain and side chain atoms to
shift. Similarly, the structural perturbation at Ser195 is a result of
the unique hydrogen bond interaction between Ilel' N and Ser195 00. The
residues which adjoin Ser195 and Glu192 - Cysl91, Gly193, and Asp194 —
undergo significant main chain structural changes due to the Glu192,
Ser195 hirudin-thrombin interactions. Moreover, while the proline

residue of PPACK and the Ilel' residue of hirudin both have hydrophobic

210

contacts with Tyr6OA and Trp6OD of the thrombin 60A-60I insertion loop,
the significant structural differences in the conformations of Pro60C,
Trp60D, and Lys6OF also result from unique hirudin-thrombin interactions
which are unmatched in PPACK-thrombin; these include the van der Vaals
interaction between Pro46' and Pro60C, the close polar contacts between
Lys47' and Glu49' with Trp6OD, and the ion pair between Glu49' and
Lys60F (Table 30).

As PPACK has no interactions with the thrombin anion-binding
exosite while hirudin has numerous hydrophobic and electrostatic
contacts here, there are expectedly many significant deviations between
the two thrombin structures in main and side chain conformation at this
binding region (Table 30). Among the Phe34-Ser41 loop of residues which
comprise the anion-binding exosite, Phe34, Arg35, and Lys36 display
significantly altered main chain and somewhat altered side chain
configurations in the complex. These deviations are probably the result
of the following contacts: the hydrophobic stacking interaction between
Phe34 and Phe56', the water-mediated contact between Arg35 and Glu49',
and the ion pair involving the C-terminal carboxylate of Gln65' and
Lys36. A number of such structural changes can be found in the
Lys70-Ly381 loop of the anion-binding exosite; these differences can be
appreciated from Figure 70, which displays Arg75—Ly581 of both thrombin
structures superimposed, as well as some pertinent hirudin residues.
The difference in the conformation of Arg75 is probably due the double
hydrogen-bonding ion pair between this residue and Glu 57'of a symmetry
molecule. The major main and side chain deviations at Arg77A stem from
the direct ion pair formed between this arginyl side chain and Glu58' of
hirudin, as well as from the longer range interaction with Glu57' via

the water molecule, V467. Lastly, the differences at Tyr76 and Asn78

Figure 70.

211

 

Stereoview of the Superimposed Anion-Binding Exosite
Residues Arg75-Lys81 of the Complex and PPACK-Thrombin.
Additional residues of the complex (thrombin Ala113, hirudin
Glu57'-Glu58', and the water molecule V467) included.
Residues of the complex in bold. Superpositioning as in
Figure 61. Hydrogen bonds indicated by dashed lines.

212

are likely a by-product of the hirudin-thrombin interactions of the
nearby residues.

Also present in Figure 70 is another type of conformational
difference; the side chain NZ of LysBl participates in a hydrogen bond
with the carbonyl oxygen of Ile82 in PPACK-thrombin (2.84 A), while in
the complex it forms a hydrogen bond with the carbonyl oxygen of Ala113
(2.73 A). This difference is a clear example of the conformational
variability possible with surface residues [172-174], and represents the
most common type of structural alteration found in the two thrombin
molecules; overall 40 thrombin residues have significant conformational
differences in the two structures which appear to be due to this (Table
31). Another interesting example this of variability involves the
residues Leu41, Phe60G-Leu64 (Table 31), and is displayed in Figure 71.
The Thr6OI-Asn63 surface turn shifts in the complex with respect to that
in PPACK-thrombin. Moreover, while the conformational 0, w angles of
Asn62 in PPACK-thrombin do not conform with any of the three major turn
classifications [149], these angles are quite different in the complex,
and identify this turn as type I. The shifting of the turn is the main
cause of the large side chain deviations for residues Thr6OI-Glu61,
while the drastic change in the 0, w conformational angles at Asn62 as
well as the turn shift are responsible for the large main and side chain
differences for this residue. In addition, several of the new side
chain conformations in this loop of the hirudin-thrombin complex appear
to stabilize the shift in turn. In particular, the large side chain
conformational changes at Leu41 and Leu64 in addition to the significant
main chain shift for Phe60H in the thrombin structure of the complex
result in there being six close nonpolar contacts of < 4.0 A between

Phe6OH and Leu41, Leu64 (three with each residue); in the PPACK-thrombin

213

I! -'- i 63
f /
\ 64
3% 60H
$4 41
i!
/

 

Figure 71. Stereoview of Residues Leu41 and Phe60H-Leu64 of the Complex
and PPACK-Thrombin. Residues of the complex in bold.
Superpostioning as in Figure 61. Hydrogen bonds indicated
by dashed lines.

214

structure there is only one close contact < 4.0 A between Phe60H and
Leu64, and none involving Leu41. The shift in the side chain
orientation at Thr6OI in the complex also results in a new close polar
contact with Asp63 (Thr6OI 001-Asp63 ODl, 3.27 A) which is not present
in the PPACK-thrombin structure.

A careful examination of Table 31 reveals that, all in all, not
only is the number of residues with main/side chain structural changes
likely due to surface residue conformational variabilty large (40), but
so is the extent to which this variability manifests itself. Ten
residues are engaged in new contacts: four are involved in ion pairs
(Arg14D, Glu14H, Glu14E, and Ly5186D), three form close polar contacts
(Lysl4A, Arg97, and Thr6OI), two engage in close van der Vaals contacts
(Leu41, Phe6OH), and one is involved in a hydrogen bond interaction
(Ly581). The conformation for eight residues is significantly affected
by a shifting in the position of two surface turns: five (Phe6OG-Asn62)
are affected by the shifting of a type I turn involving residues
Thr6OI-Asn62, and three (Trp237-Gln239) are affected by the shifting of
a type III turn involving residues LysZ35-Ile238. In addition, all but
11 residues in Table 31 have significantly altered side chain atomic
positions, with six residues displaying dramatically new side chain
conformations because they are unstabilized and directed into the
solvent (Serll, Leu12, Ly587, Lysl69, Asp189,and Ly5240) and six more
displaying new side chain conformations simply 'by reason of free
rotation about a single bond (Gln38, Leu41, Lys87, Leu160, Thr177, and
Leu234).

It has been possible to associate most of the major main and side

 

chain structural differences between the thrombin molecules in the .

hirudin-thrombin complex and in PPACK—thrombin with one or another of

215

the (a)—(d) influences. However, all of these differences have involved
solvent-accessible residues located on the thrombin surface - and have
been classified in Tables 28 to 31 as belonging to the general thrombin
surface (S) or the specialized surface of the active site binding cavity
(AS). If one excludes the problematic differences associated with
category (a), which probably all appear on the surface only
fortuitously, some clear implications can be drawn from this comparison.
First of all, only surface residues display significantly altered
main/side chain structure in the two thrombin molecules. Second, the
deviations associated with these surface residues appear to result not
only from binding or crystal contact interactions but also from an
inherent conformational flexibility which is likely due to being on a
solvent-accessible surface [172-174]. Finally, the main chain and side
chain folding of 'internal’ or non-surface residues agreed very well -
displaying no major (> la ) deviations.

As thrombin is a large, globular, and well—defined serine
proteinase, the conclusions drawn cannot be expected to be of general
applicability to all structural comparisons involving the same protein
or» enzyme in free or complexed states. However, it is of particular
note that the observations made in this thrombin comparison, in general,
agree well with conclusions drawn from a very careful, comprehensive,
and detailed structural comparison made on a close relative of thrombin,
a—chymotrypsin [174]. In this study, Blevins and Tulinsky examined the
structural differences in the two independent molecules of the
aschymotrypsin dimer crystal structure at 1.67 A resolution, and further
compared these independent molecules with monomeric y—chymotrypsin.
Among the conclusions they drew from this study were that the main chain

folding was basically identical between the two independent molecules in

216

the dimer, and that there was marked asymmetry in the side chain
structure of the surface residues - some which could be explained by
interfacial interactions but much of which had to be attributed to
general flexibility in surface side chain structure. Moreover, when the
independent molecules of the chymotrypsin dimer were compared to
y—chymotrypsin, although the two molecules of the dimer were more like
one another than y-chymotrypsin, the same conclusions could be drawn
from the independent molecules in the dimer comparison. The
chymotrypsin comparison, in particular, elegantly displays that even in
a situation where the molecules being compared are exposed to the same
macroscopic chemical and pH environment, there is still significant
surface side chain variability which cannot be explained by interface
interactions - emphasizing that protein surface structure is variable as
well as adaptable.

One important difference in the conclusions drawn from the thrombin
and chymotrypsin structural comparisons is that there were no
significant main chain deviations involving surface residues among the
chymotrypsin molecules while there were between the two thrombin
structures. This discrepancy is probably due to two factors. In the
chymotrypsin (CHT) comparison, basically the same free enzyme molecule
is examined in all the situations. In the comparison of unique
molecules within the dimer of a—CHT, differences simply due dimerization
were examined, and in the aHCHT versus y-CHT comparisons, differences
due to pH (aHCHT, pH 3.5; y-CHT, pH 5.5) and aggregation state (aHCHT,
dimer; y-CHT, monomer) were assessed. In the thrombin comparison, on
the other hand, thrombin structures were contrasted which not only
differed in pH and chemical environment but, more importantly, were

differently inhibited molecules: the thrombin inhibitors differed

 

217

significantly in their extent of interaction with the thrombin surface.
In particular, it appears that unique hirudin-thrombin interactions are
likely the source of many of the significant main (as well as side)
chain deviations between the two structures. Also, as the thrombin
molecule can be viewed as possessing 12 insertions with respect to
chymotrypsinogen (Figure 57) [6], the significant number of main chain
differences between the thrombin structures could be due to the fact
that the thrombin surface is inherently more adaptable than that of

chymotrypsin as a result of these inserted sequences.

LI ST 01’ REFERENCES

10.

11.

12.
13.
14.

15.

16.

17.

LIST 0? REFERENCES

Magnusson, S. (1971) in The Enzymes (Boyer, P.D., Ed.) pp 277-321,
Academic Press, New York.

Nesheim, M.E., Hibbard, L.S., Tracy, P.B., Bloom, J.V., Myrmel,
K.H., & Mann, K.G. (1980) in The Regulation of Coagulation
(Mann, K.G., & Taylor, P.B., Eds.) pp 145-149, Elsevier-North
Holland, New York.

Mann, K.G. (1976) Methods Enzymol. £2, 123-156.

Butkowski, R.J., Elion, J., Downing, M.R., & Mann, K.G. (1977) J.
Biol. Chem. 222, 4942-4957.

Thompson, A.R., Enfield, D.L., Ericsson, L.H., Legaz, M.E., &
Fenton II, J.V. (1977) Arch. Biochem. Biophys. 222, 356-367.

Bode, V., Mayr, I., Baumann, U., Huber, R., Stone, S.R., &
Hofsteenge, J. (1989) EMBO J. 8, 3467-3475.

Jackson, C.M. & Nemerson, Y. (1980) Annu. Rev. Biochem. 32, 765-
811.

Nilsson, B., Horne, M.K. III, & Gralnick, H.R. (1983) Arch.
Biochem. Biophys. 222, 127-133.

Doolittle, R.F. (1984) Annu. Rev. Biochem. 22, 195-229.

Hantgan, R.R., & Hermans, J. (1979) J. Biol. Chem. 222, 11272-
11281.

Blomback, B., Hessel, B., Hogg, D., & Therkildsen, L. (1978)
Nature 222, 501-505.

Blomback, B. (1986) Ann. N.Y. Acad. Sci. £22, 120-123.
Casassa, E.F. (1955) J. Chem. Phys. 22, 596-597.
Carr, M.E., Shen, L.L., & Hermans, J. (1977) Biopolymers 2g, 1-15.

Ferry, J.D., & Morrison, P.R. (1947) J. Am. Chem. Soc. g2, 388-
400.

Lorand, L., Downey, J., Gotoh, T., Jacobsen, A., & Tokura, S.
(1968) Biochem. Biophys. Res. Commun. 22, 222-230.

Fenton II, J.V. (1981) Ann. N.Y. Acad. Sci. 220, 468-495.

218

18.
19.

20.
21.
22.

23.

24.

25.

26.
27.

28.
29.

30.

31.

32.

33.

34.

35.

36.

37.
38.

219

Fenton II, J.V. (1986) Ann. N.Y. Acad. Sci. 222, 5-15.

Esmon, C.T., Esmon, N.L., Kurosawa, S., & Johnson, A.E. (1986)
Ann. N.Y. Acad. Sci. 522, 215-220.

Rick, M.E. & Hoyer, L.V. (1978) Brit. J. Haematol. 22, 107-119.
Nesheim, M.E. 8 Mann, K.G. (1979) J. Biol. Chem. 222, 1326-1334.

Seegers, V.H., Novoa, E., Henry, R.L., & Hassouna, H.I. (1976)
Thromb. Res. 2, 543-552.

Kisiel, V., Canfield, V.H., Ericcson, L.H., & Davie, E.V. (1977)
Biochemistry 22,5824-5831.

Lorand, L. (1975) in Proteases and Biological Control (Reich, E.,
Rigkin, D.B., & Shaw, E., Eds.) pp 79-84, Cold Spring Harbor
Laboratory Press, Cold Spring Harbor.

Travis, J. & Salvesen, 0.5. (1983) Annu. Rev. Biochem. 22, 655-
709.

Bjork,I. 6 Lindahl, U. (1982) Mol. Cell. Biochem. 22, 161-182.

Pratt, C.V., Vhinna, H.C., Meade, J.B., Treanor, R.E., & Church,
F.C. (1989) Ann. N.Y. Acad. Sci. 222, 104-115.

Markwardt, F. (1970) Methods. Enzymol. 22, 924-932.
Shuman, M.A. (1986) Ann. N.Y. Acad. Sci. 222, 228-239.

Veksler, B.B., Ley, C.V., & Jaffe, E.A. (1978) J. Clin. Invest.
22, 923-930.

Prescott, S.M., Zimmerman, G.A., & McIntyre, T.M. (1984) Proc.
Natl. Acad. Sci. USA 22, 3534-3538.

De Groot, P.G., Gonsalves, M.D., Loesberg, C., van Buul-
Vortelboer, M.F., van Aken, V.G., & van Mourik, J.A. (1984)
J. Biol. Chem. 222, 13329-13333.

Gelehrter, T.D., & Sznycer-Laszyk, R. (1986) J. Clin. Invest. 22,
165-169.

Bender, M.L., & Kezdy, F.J. (1964) J. Am. Chem. Soc. 22, 3704-
3714.

Inward, P.V., & Jencks, V.P. (1965) J. Biol. Chem. 222, 1986-
1996.

Stroud, R.M., Kay, L.M., & Dickerson, R.E. (1974) J. Mol. Biol.
22, 185-208.

Birktoft, J.J., & Blow, D.M. (1972) J. Mol. Biol. 22, 187-240.

Shotton, D.M., & Vatson, H.C. (1970) Nature 222, 811-816.

 

39.

40.
41.

42.

43.

44.

45.

46.

47.

48.

49.

50.
51.

52.

53.

54.

55.

56.

220

Sawyer, L., Shotton, D.M., Campbell, J.V., Wendell, P.L.,
Muirhead, B., Vatson, H.C., Diamond, R., & Ladner, R.C. (1978)
J. "01. 8101. E, 137-2080

Berliner, L.J. & Shen, Y.Y.L. (1977) Biochemistry 22, 4622-4626.
Sonder, S.A., & Fenton II, J.V. (1984) Biochemistry 22, 1818-1823.

Berliner, L.J., Sugawara, Y., & Fenton II, J.V. (1985)
Biochemistry 22, 7005-7009.

Olson, T.A., Sonder, S.A., Vilner, C.D., & Fenton II, J.V. (1986)
Ann. N.Y. Acad. Sci. 222, 96-103.

Fenton II, J.V., Olson, T.A., Zabinski, M.P., & Vilner, C.D.
(1988) Biochemistry 22, 7106-7112.

Church, F.C., Pratt, C.V., Noyes, C.M., Kalayanamit, T., Sherrill,
G.B., Tobin, R.B., & Meade, J.B. (1989) J. Biol. Chem., 222,
18419-18425.

Bar-Shavit, R., Kahn, A.J., Mann, K.G., & Wilner, C.D. (1986)

Fenton II, J.V. (1988) Semin. Thromb. Hemost. 22, 234-240.

Boissel, J-P., Le Bonniec, B., Rabiet, M-J., Labie, D., &
Elion, J. (1984) J. Biol. Chem. 222, 5691-5697.

Levis, S.D., Lorand, L., Fenton II, J.V., & Shafer, J.A. (1987)
Biochemistry 22, 7597-7603.

Bezeaud, A. & Guillin, M-C. (1988) J. Biol. Chem. 222, 3576-3581.

Braun, P.J., Hofsteenge, J., Chang, J-Y. & Stone, S.R. (1988)
Thromb. Res. 22, 273-283.

Hofsteenge, J., Braun, P.J. & Stone, S.R. (1988) Biochemistry 22,
2144-2151.

Tsernoglou, D., Valz, D.A., McCoy, L.E., & Seegers, V.H. (1974) J.
Biol. Chem. 222, 999 (1974).

McKay, D.B., Kay, L.M., & Stroud, R.M. (1977) in Chemistry and
Biology of Thrombin ( Fenton II, J.V., Lundblad, R.L. & Mann,
K.G. Eds.) pp 113-121, Ann Arbor Science, Ann Arbor.

Edwards, B.F.P., Kumar, V., Bedford, B.A., Martin, P.D., &
Kunjummen, R.D. (1986) Ann. H.I. Acad. Sci. 222, 411-413.

Skrzypczak-Jankun, E., Rydel, T.J., Tulinsky, A., Fenton II, J.V.,
& Mann, K.G. (1989) J. Mol. Biol. 222, 755-757.

 

57.

58.

59.
60.
61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

221

Magnusson, S., Peterson, T.E., Sottrup-Jensen, L., 8 Claeys, H.
(1975) in Proteases and Biological Control (Reich, E., Rifkin,
D.B., 8 Shaw, E., Eds.) pp 123-150, Cold Spring Harbor Laboratory
Press, Cold Spring Harbor.

Furie, B., Bing, D.B., Feldmann, R.J., Robison, D.J., Burnier,
J.P., 8 Furie, B.C. (1982) J. Biol. Chem. 222, 3875-3882.

Stone, S.R. 8 Hofsteenge, J. (1986) Biochemistry 22, 4622-4628.
Valsman, P., 8 Markwardt, F. (1981) Pharmazie 22, 653-660.

Scharf, M., Engels, J., 8 Tripier, D. (1989) FEBS Lett. 222,
105-110.

Dodt, J., Muller, H.P., Seemuller, U., 8 Chang, J.Y. (1984) FEBS
Lett. 222, 180-183.

Harvey, R.F., Degryse, E., Stefani, L., Schamber, F., Cazenave,
J.P., Courtney, M., Tolstoshev, P., 8 Lecocq, J.P. (1986) Proc.
Natl. Acad. Sci. USA 22, 1084-1088.

Dodt, J., Machleidt, V., Seemuller, U., Maschler, R., 8
Fritz, H. (1986) Biol. Chem. Hoppe-Seyler 222, 803-811.

Loison, C., Findeli, A., Bernard, 5., Nguyen-Juilleret, M.,
Marquet, M., Reihl-Bellon, N., Carvallo, D., Guerra-Santos, L.,
Brown, S.H., Courtney, M., Roitsch, C., 8 Lemoine, Y. (1988)
Bio/Technology 2, 72-77.

Courtney, M., Loisson, C., Lemoine, Y., Riehl-Bellon, N.,
DeGryse, E., Brown, S.U., Cazanave, J.P., Defreyn, C.,
Delebassee, D., Bernat, A., Maffrand, J.P., 8 Roitsch, C.
(1989) Semin. Thromb. Hemost. 22, 288-292.

Johnson, P.H., Sze, P., Vinant, R., P.V. Payne, and J.B. Lazar,
Semin. Thromb. Hemost. 22, 302 (1989).

Bagdy, D., Barabas, E., Graf, L., Peterson, T.E., 8 Magnusson, S.
(1976) Methods Enzymol. 22, 669-678.

Fink, E. (1989) Semin. Thromb. Hemost. 22, 283-287.
Chang, J.Y. (1983) FEBS Lett. 222, 307-313.

Seemuller, U., Dodt, J., Fink, E., 8 Fritz, H. (1986) in
Proteinase Inhibitors (Barrett, A., 8 Salvesen, C., Eds.)
pp 337-359, Elsevier, New York.

Dodt, J., Seemuller, U., Maschler, R., 8 Fritz, H. (1985) Biol.
Chem. Boppe- Seyler 222, 379-385.

Laskowski, M. Jr., 8 Kato, I. (1980) Annu. Rev. Biochem. 22,
593-626.

 

74.

75.

76.

77.

78.

79.

80.
81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

92.

222

Apella, E., Weber, I.T., Blasi, F. (1988) FEBS Lett. 222, 1-4.

Dodt, J., Schmitz, T., Schafer, T., 8 Bergmann, C. (1986) FEBS
Lett. 222, 373-377.

Fortkamp, E., Rieger, M., Heisterberg-Moutses, G., Schweitzer, S.,
8 Sommer, R. (1986) DNA 2, 511-517.

Bergman, C., Dodt, J., Kohler, 8., Pink, E., 8 Gassen, H.G.
(1986) Biol. Chem. Hoppe-Seyler 222, 731-740.

Sukumaran, D.K., Clore, C.M., Preuss, A., Zarbock, J., 8
Gronenborn, A.M. (1987) Biochemistry 22, 333-338.

Folkers, P.J.M., Clore, C.M., Driscoll, D.C., Dodt, J.,
Kohler, S., 8 Gronenborn, A.M. (1989) Biochemistry 22, 2601-2617.

Haruyama, H., 8 Uuthrich, K. (1989) Biochemistry 22, 4301-4312.

Stone, S.R., Braun, P.J., 8 Hofsteenge, J. (1987) Biochemistry 22,
4617-4624.

Fenton II, J.V., Landis, B.H., Valz, D.A., 8 Finlayson, J.S.
(1977) in Chemistry and Biology of Thrombin (Fenton II, J.V.,
Lunblad, R.L., 8 Mann, K.G., Eds.) pp 43-70, Ann Arbor Science,
Ann Arbor.

Noe, G., Hofsteenge, J., Rovelli, G., 8 Stone, S.R. (1988) J. Biol.
Chem. 222, 11729-11735.

Read, R.J., 8 James, M.N.G. (1986) in Proteinase Inhibitors
(Barrett, A. 8 Salvesen, C., Eds.) pp 301-333, Elsevier, New York.

Peterson, T.E., Roberts, H.R., Sottrup-Jensen, L. 8 Magnusson, S.
(1976) in Protides of the Biological Fluids (Peeters, H., Ed.)

pp 145-149, Pergamon Press, Oxford.

Dodt, J., Kohler, S., 8 Baici, A. (1988) FEDS Lett. 222, 87-90.

Braun, P.J., Dennis, S., Hofsteenge, J., 8 Stone, S.R. (1988)
Biochemistry 22, 6517-6522.

Stone, S.R., Dennis, S., 8 Hofsteenge, J. (1988) Biochemistry 22,
6857-6863.

Krstenansky, J.L., Owen, T.J., Yates, M.T. 8 Mao, S.J.T. (1987)
J. Med. Chem. 22, 1688-1691.

Wallace, A., Dennis, S., Hofsteenge, J., 8 Stone, S.R. (1989)
Biochemistry 22, 10079-10084.

Dodt, J., Kohler, 8., Schmitz, T. 8 Wilhelm, B. (1990) J. Biol.
Chem. 222, 713-718.

Dennis, 5., Wallace, A., Hofsteenge, J. 8 Stone, S.R. (1990)
Eur. J. Biochem. 222, 61-66.

 

93.

94.

95.

96.

97.
98.
99.

100.

101.

102.
103.
104.

105.

106.
107.

108.
109.
110.

111.

112.

223

Krstenansky, J.L. 8 Mao, S.J.T. (1987) FEBS Lett. 222, 10-16.

Maraganore, J.M., Chao, 8., Joseph, M.L., Jablonski, J. 8
Ramachandran, K.L. (1989) J. Biol. Chem. 222, 8692-8698.

Mao, S.J.T., Yates, M.T., Owen, T.J. 8 Krstenansky, J.L. (1988)
Biochemistry 22, 8170-8173.

Konno, S., Fenton II, J.W., 8 Villanueva, G.B. (1988) Arch.
Biochem. Biophys. 222, 158-166.

Markwardt, F. (1986) Ann. N. Y. Acad. Sci. 222, 204-214.
Markwardt, F. (1985) Biomed. Biochim. Acta 22, 1007-1013.

Markwardt, P., Hauptmann, J., Nowak, C., Klessen, C., 8
Walsmann, P. (1982) Thromb. Baemostasis 22, 226-229.

 

Fenton II, J.W., Fasco, M.J., Stackrow, A.B., Aronson, D.L.,
Young, A.H., 8 Finlayson, J.S. (1977) J. Biol. Chem. 222,
3587-3598.

Bischoff, R., Clesse, D., Whitechurch, 0., Lepage, P., 8 Roitsch,
Co (1989) Jo Chi-0.0 fl, 245‘255.

Ornstein, L. (1964) Ann. N.Y. Acad. Sci. 222, 321-349.
Davis, B.J. (1964) Ann. N.Y. Acad. Sci. 222, 404-427.

Hoefer Scientific Instruments catalogue (1986), pp 110-116, Hoefer
Scientific Instruments, San Francisco.

Technical Bulletin No. MKR-137 (1986), pp 1-8, Sigma Chemical
Company, St. Louis.

Bryan, J.K. (1977) Anal. Biochem. 22, 513-519.

Reid, B.R., Koch, G.L.E., Boulanger, Y., Hartley, B.S., 8 Blow,
Do". (1973) J. "01. 31°10 8_0, 199-2010

Carter, C.W. Jr. 8 Carter, C.W. (1979) J. Biol. Chem. 222,
12219-12223.

Thaller, C., Weaver, L.H., Eichele, C., Wilson, E., Karlsson, R.,
8 Jansonius, J.N. (1981) J. Mol. Biol. 222, 465-469.

Matthews, B.V. (1968) J. Mol. Biol. 22, 491-497.

Edumndson, A.B., Wood, M.K., Schiffer, M., Hardman, K.D.,
Ainsworth, C.F., Ely, K.R., 8 Deutsch, H.F. (1970) J. Biol.
Chem. 222, 2763-2764.

Ichikawa, T. 8 Sundaralingam, M. (1972) Nature New Biol. 222,
174-176.

113.

114.

115.

116.

117.

118.

119.

120.

121.

122.

123.

124.

125.
126.
127.

128.
129.

130.
131.

224
McPherson, A. (1982) in Preparation and Analysis of Protein
Crystals, p. 241, John Wiley 8 Sons, New York.
Stout, G.H. 8 Jensen, L.H. (1989) in X-ray Structure

Determination - A Practical Guide, pp. 23-25, John Wiley
8 Sons, New York.

Vyckoff, H.W., Doscher, M., Tsernoglou, D., Inagami, T.,
Johnson, L.N., Hardman, K.D., Allewell, N.M., Kelly, D.M., 8
Wyckoff, H. (1985) Methods Enzymol. 222, 330-386.

Pflugrath, J.W. 8 Messerschmidt, A. (1989) Munich Area Detector
NB System User’s Guide, Enraf-Nonius, The Netherlands.

Pflugrath, J.W. 8 Messerschmidt, A. "Structured Software for Data
Collection with Area Detectors", from Crystallography in Molecular
Biology Conference, Bischenberg, France, September 1985.

Sjolin, L. 8 Wlodawer, A. (1981) Acta Crystallogr. 222, 594-604.

Data Collection Operation Manual (1982), Section D.9 pg 5, Nicolet
XRD Corp., Madison.

North, A.C.T., Phillips, D.C., 8 Mathews, F.S. (1968) Acta
Crystallogr. 222, 351-359.

Steigemann, W. (1974) Ph. D. Thesis, Technische Universitat
Munchen.

Messerschmidt, A., Schneider, M., 8 Huber, R. (1990) J. Appl.
Cryst. 22, 436-439.

Rossmann, M.G. (1972) in The Molecular Replacement Method
(Rossmann, M.G., Ed.) pp 1-42, Gordon and Breach, New York.

Huber, R. (1965) Acta Crystallogr. 22, 353-356.
Rossmann, M.G. 8 Blow, D.M. (1962) Acta Crystallogr. 22, 24-31.

Rydel, T.J, Ravichandran, K.G., Tulinsky, A., Bode, W., Huber, R.,
Roitsch, C., 8 Fenton II, J.W. (1990) Science 222, 277-280.

Patterson, A.L. (1935) Z. Krist. 229, 517-542.

Rossmann, M.G., Blow, D.M., Harding, M.M., 8 Coller, E. (1964)
Acta Crystallogr. 22, 338-342.

Crowther, R.A. 8 Blow, D.M. (1967) Acta Crystallogr. 22, 544-548.

International Tables for X-ray Crystallography (1952) (Henry,
N.F. 8 Lonsdale, K., Eds.) Vol. I, p 186, The Kynoch Press,
Birmingham.

 

132.
133.
134.

135.

136.

137.

138.

139.
140.
141.

142.

143.
144.

145.

146.
147.
148.
149.

150.

151.

152.

225

Huber, R. 8 Schneider, M. (1985) J. Appl. Cryst. 22, 165-169.
Sim, G.A. (1959) Acta Crystallogr. 22, 813-815.

Sim, G.A. (1960) Acta Crystallogr. 22, 511-512.

Stout, G.H. 8 Jensen, L.H. (1989) in X—ray Structure
Determination - A Practical Guide, p 337, John Wiley 8

Sons, New York.

Blow, D.M. 8 Crick, F.H.C. (1959) Acta Crystallogr. 22, 794-802.

Blundell, T.L. 8 Johnson, L.N. (1976) in Protein Crystallography,
p 419, Academic Press, New York.

Pflugrath, J.W., Saper, M.A., 8 Quiocho, F.A. (1984) in
Methods and Applications in Crystallographic Computing

(Hall, 5. 8 Ashida, T., Eds.) pp 404-407, Oxford Univ. Press,
Oxford.

Jones, T.A. (1978) J. Appl. Cryst. 22, 268-272.

Jack, A. 8 Levitt, M. (1978) Acta Crystallogr. 222, 931-935.
Levitt, M. (1974) J. Mol. Biol. 22, 393-420.

Deisenhofer, J., Remington, S.J., 8 Steigemann, W. (1985) Methods
Enzymol. 222, 303-323.

Ten Eyck, L.F. (1973) Acta Crystallogr. 222, 183-191.

Ten Eyck, L.F. (1977) Acta Crystallogr. 222, 486-492.
Hendrickson, W.A. 8 Konnert, J.H. (1980) in Computing in
Crystallography (Diamond, R., Ramaseshan, S. 8 Venkatesan, K.,
Eds.) pp 13.01-13.26, The Indian Academy of Sciences, Bangalore,
India.

Hendrickson, W.A. (1985) Methods Enzymol. 222, 252-270.

Finzel, B.C. (1987) J. Appl. Cryst. 22, 53-55.

Luzzati, P.V. (1952) Acta Crystallogr. 2, 802-810.

Crawford, J.L., Lipscomb, W.N., 8 Schellman, C.G. (1973) Proc.
Natl. Acad. Sci. USA 22, 538-542.

Tulinsky, A., Park, C.H., 8 Skrzypczak-Jankun, E. (1988) J. Mol.
Biol. 222, 885-901.

Mulichak, A.M. 8 Tulinsky, A. Blood Coagulation 8 Fibrinolysis,
manuscript in press.

Huber, R., Scholze, H., Paques, E., 8 Deisenhofer, J. (1980)
Boppe-Seyler's z. Physiol. Chem. 222, 1389-1399.

 

153.

154.
155.
156.
157.
158.

159.

160.

161.

162.

163.
164.
165.
166.
167.

168.

169.
170.
171.
172.

173.

174.

226
Bode, W., Greyling, H.J., Huber, R., Otlewski, J., 8 Wilusz, T.
(1989) FEBS Lett. 222, 285-292.
Arnott, S. 8 Dover, S.D. (1968) Acta Crystallogr. 222, 599-601.
Yonath, A. 8 Traub, W. (1969) J. Mol. Biol. 22, 461-477.
Pedersen, B. (1974) Acta Crystallogr. 222, 289-291.
Brown, I.D. (1976) Acta Crystallogr. 222, 24-31.

Taylor, R., Kennard, 0., 8 Versichel, W. (1984) Acta Crystallogr. 222,
280-288.

Gorbitz, G.H. (1989) Acta Crystallogr. 222, 390-395.

Katz, B.A. 8 Kossiakoff, A. (1986) J. Biol. Chem. 222, 15480-
15485.

 

Weiner, S.J., Kollman, F.A., Case, D., Singh, U.C., Ghio, C., i
Alagona, C., Profeta, S. Jr., 8 Weiner, P. (1984) J. Am. Chem.
Soc. 222, 765-784.

Ni, F., Konishi, Y., 8 Scheraga, H.A. (1990) Biochemistry 22,
4479-4489.

Huber, R. 8 Bode, W. (1978) Acc. Chem. Res. 22, 114-122.
Chang, J.-Y. (1985) Eur. J. Biochem. 222, 217-224.
Chang, J.-Y. (1986) Biochem. J. 222, 797-802.

Chang, J.-Y. (1989) J. Biol. Chem. 222, 7141-7146.

Bourdon, P., Fenton II, J.W., 8 Maraganore, J.M. (1990)
Biochemistry 22, 6379-6384.

Skrzypczak-Jankun, E. 8 Tulinsky, A., unpublished results of
this laboratory.

Zuck, V. 8 Owen, W., unpublished observations.
Degryse, E., unpublished observations.
Lee, B. 8 Richards, F.M. (1971) J. Mol. Biol. 22, 379-400.

Tulinsky, A., Vandlen, R.L., Morimoto, C.N., Mani, N.K., 8
Wright, L.H. (1973) Biochemistry 22, 4185-4192.

Tulinsky, A. (1980) in Biomolecular Structure, Conformation,
Function, and Evolution (Srinivasan, R., Ed.) Vol. 1, pp. 183-199,
Pergamon Press, Oxford.

Blevins, R.A. 8 Tulinsky, A. (1985) J. Biol. Chem. 222, 4264-4275.

MICHIGAN STATE UNIV. LIBRARIES
11111111111111W111111111111111111111111
3129300901 1549