writ! .)
4:13.! v
.5» .1.
£3. 1
\ =71... .
. 13......
I

5‘.
. mafia...
in? 5
r“... .y z; . 1!. .I I...

A‘CUJ'I . . .‘13‘ 1
v.94 a an. firfihh; .0441).
- “swaruufifia
)5» x...

5..» i
1.1 a
.

. 3.519?
2 I.

. 2.:
. 4...

$9.“? .3. «$41!.
1 , wick?

in .. 74..
2 .. E- 5
in}; afvwflg
.,313$3~.‘43p£ .73:‘.’\-." n...
iii-‘59.. i.3’\.u.y§‘ 5 El‘
.3 $QV 2471‘“: h‘i‘I “ga’J.
.I A . hriviuuu‘
51.9.1}.

4.. .‘. M
WW. MfidfiWJ. ..
uni... army...

x

. fin
. 1
a, l lt§ufi ‘73
22:53,: .. .5;
.....i a I:
i

 

. a! .u .u 33“?!
i... it... s ‘1‘.

y . , -1...) I
n ‘snsilll - L‘P‘ly .1 .t.
”HUN—B. 1!. I. . . .
t u 4 . .
c, . .

 

 

LIBRARY
Michigan State
University

This is to certify that the
dissertation entitled

Analysis of rigidity, stability, and activity in Thennotoga
neapolitana adenylate kinase

presented by

Harini Krishnamurthy

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Cell and Molecular Biology

 

l/fl flax/W

ajor Professor’s Signature

/ fi/ 7/56

Date

 

MSU is an Affirmative Action/Equal Opportunity Institution

 

 

o---~-----n-—o-o---o-a--o--o—o-c-o----o-n—--.--._.-.-.-.-.-.--1-.-.. —.-._.-

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DateDue.indd-p.1

 

ANALYSIS OF RIGIDITY, STABILITY, AND ACTIVITY IN
Thermotoga neapolitana ADENYLATE KINASE

By

Harini Krishnamurthy

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Program in Cell and Molecular Biology

2006

ABSTRACT

ANALYSIS OF RIGIDIT Y, STABILITY, AND ACTIVITY IN Thermotoga neapolitana

ADENYLATE KINASE
By

Harini Krishnamurthy

Proteins from hypertherrnophiles are adapted to be stable and active at temperatures 2
80°C, conditions in which mesophilic proteins denature. Because the sequence and
structural similarity between hyperthermophilic proteins and their mesophilic
homologues is very high, it is not readily apparent as to what causes the enhanced
stability of hyperthermophilic proteins. Understanding the mechanisms of protein
thermostability has been one of the most intensely studied problems in the last three
decades. The activity—temperature profiles of hyperthermophilic and mesophilic proteins
show that the profile is right-shifted for hyperthermophiles. In other words, they attain
similar activity as their mesophilic homologues at a much higher temperature and they
are inactive at low temperatures. These facts have lead to the proposal of the ‘rigidity’
hypothesis regarding hyperthermophilic proteins. Hyperthermophilic proteins achieve
high thermostability through increased structural rigidity. These proteins are highly rigid
in their native structure that allows them to maintain their structural integrity at high
temperatures. The increased rigidity of hyperthermophilic proteins freezes out
fluctuations required for activity at room temperature making them inactive at low

temperatures.

 

f

The studies described here use adenylate kinase from the hyperthermophilic
bacteria Thermotoga neapolitana (TNAK) as a model system in which to test the rigidity
hypothesis. What makes TNAK a really interesting model system is the fact that it is
highly active at 30°C, an unusual property for a hyperthermophilic protein. Special
attention was paid to the techniques used to investigate motions in TNAK. NMR and MD
simulations, techniques that can access a wide range of timescales in atomic detail, were
used to compare dynamics in TNAK with its mesophilic homologue from Escherichia
coli (ECAK).

Results from 15 N NMR relaxation data shows that TNAK is uniformly more rigid
than ECAK in the ps-ns timescales as well as us timescale. Although, overall, TNAK has
higher rigidity than ECAK, several residues in the AMP-binding and lid domains exhibit
high flexibility in the ps-ns timescale. Residues in the hinge regions between the lid and
core domains of TNAK exhibit flexibility in the ps-ns timescales as well us timescales.
H-D exchange data, which provide information on timescales greater than seconds, show
that TNAK’s lid and AMP-binding domains are more stabilized compared to ECAK.
Again, in spite of this increased rigidity, several residues in these domains of TNAK
show considerable local fluctuations.

The increased rigidity coupled with localized flexibility may explain the
simultaneously high activity at low temperatures and stability at high temperatures of
TNAK. The results also show that TNAK is not uniformly rigid across the entire
structure at all timescales. Instead, TNAK domains are differentially stabilized at
different timescales. Finally, the results show that the rigidity hypothesis may not hold

true for all proteins.

iv

To my Mother, Father, and Suba

ACKNOWLEDGEMENTS

I would like to start by sincerely thanking everyone who has helped me carry out
this research. I thank Dr. Greg Zeikus, my thesis advisor, for believing in me and for
giving me the independence to explore new scientific perspectives. It was a great fortune
for me to have Dr. Claire Vieille in the lab who was a mentor and friend to me
throughout these years. Without her support and guidance this project would not have
seen the light of day. Both of them have been the driving force behind this
interdisciplinary thesis and I look forward to using the skills and enthusiasm imparted to
me in this lab for an exciting scientific career.

I would like thank my thesis guidance committee: Dr. Leslie Kuhn, Dr. Honggao
Yan, Dr. Sheilagh Ferguson-Miller, Dr. Rawle Hollingsworth and Dr. Tom Schmidt for
their many substantive suggestions and thoughtful discussions. I am especially grateful to
Dr. Kuhn and Dr. Yan for always being available to answer my questions. I am grateful
to Dr. Aizhou Liu for his invaluable support and guidance in NMR experiments. My
thanks to Lishan Yao, who was instrumental in completing this research. We have both
worked on the same techniques in our respective labs and our long discussions have
helped me understand the theoretical aspects of NMR. I also extend my appreciation to
the members of the Cell and Molecular Biology Program and the Department of
Biochemistry and Molecular Biology for their help and support.

I thank all the present and past members of the Zeikus laboratory. Special thanks
to Jake McKinlay for all the interesting discussions, both scientific and non-scientific,

and for his friendship over the years. I extend an enthusiastic thank you to Maris

 

Laivenieks and Karla Fjeld for creating an enjoyable working environment and providing
useful feedback. My thanks also go to Dr. Pil Kim, Dr. Suil Kang, and Dr. Yoshiharu
Sasaki, all former members of the lab, for sharing their expertise and friendship.

But for all the great friends I have made as a graduate student, my life at MSU
would not have been as enjoyable. I especially thank Ajith Vengellur, Clarisa Bejar,
Anastasia Gridasova, Sandhya Payankulam, and Radhika Iyer for their close friendship
and support. There are many more friends who I would like to thank for the great
moments that I have had during these years but I am unable to due to lack of space. My
most cherished time in graduate school has been the many good friends I have made from
diverse cultures. All these people have infinitely enriched my life.

Finally, I would like to thank my family: mom, dad, and my sister, Subashini
without whose support I would not be where I am today. They have provided me with the
courage and inspiration to keep going when things seemed impossible. I also thank my

uncles and aunts, especially Krishnan mama and Radha mami for their support.

vi

TABLE OF CONTENTS

List of Tables ......................................................................................... x
List of Figures ....................................................................................... xi
Key to Abbreviations ............................................................................. xiii
CHAPTER I .......................................................................................... 1
1.1 Hyperthermophiles: Definition and evolution ........................................... 1
1.2 Hyperthermophilic enzymes ............................................................... 2
1.3 Mechanisms for protein thermostability .................................................. 4
1.4 Protein stability .............................................................................. 6
1.5 Rigidity, thermostability, and activity .................................................... 9
1.6 Protein dynamics from NMR ............................................................. 15
1.6.1 '5 N NMR relaxation ............................................................... 16
1.6.2 H-D exchange, protein stability, and flexibility ............................... 22
1.7 Work presented in this thesis ............................................................. 28
1.8 References ................................................................................... 3 1

CHAPTER II: Thermotoga neapolitana adenylate kinase is highly active at 30°C...38

2.1 Abstract ............................................................................. 39
2.2 Introduction ........................................................................ 40
2.3 Materials and Methods ............................................................ 44
2.3.1 Bacterial strains and plasmids .......................................... 44
2.3.2 Library construction and screening .................................... 44
2.3.3 Manipulation of DNA ................................................... 45
2.3.4 Subcloning of AK gene from T. neapolitana into an expression
vector ...................................................................... 45
2.3.5 AK purification ........................................................... 46
2.3.6 Gel—permeation chromatography ....................................... 47
2.3.7 AK assays .................................................................. 47
2.3.8 Kinetic stability assays ................................................... 49
2.3.9 Microcalorimetry ......................................................... 49
2.3.10 Spectroscopic methods .................................................. 50
2.3.11 Other analytical procedures ............................................. 51
2.4 Results .............................................................................. 51
2.4.1 Cloning of Adk gene from T. neapolitana in E. coli ................. 51
2.4.2 Nucleotide sequence of the adk gene from T. neapolitana ......... 52
2.4.3 Sequence comparisons ................................................... 52
2.4.4 TNAK overexpression in E. coli and purification ................... 55

2.4.5 Effects of temperature and pH on TNAK activity and stability. . .56
2.4.6 TNAK stability studied by calorimetry and CD spectroscopy. . ...59
2.4.7 Kinetic properties of TNAK ............................................. 62
2.4.8 Determination of Zn2+ content of TNAK .............................. 62

vii

2.5
2.6

2.4.9 Thermostability of apo-TNAK ......................................... 64
Discussion .......................................................................... 66
References .......................................................................... 72

CHAPTER III: Backbone dynamics in a hyperthermophilic adenylate kinase

that is highly active at 30°C ................................................. 76
3. 1 Introduction ........................................................................ 77
3.2 Materials and Methods ............................................................ 84
3.2.1 Sample preparation ....................................................... 84
3.2.2 NMR spectroscopy ....................................................... 85
3.3 Results .............................................................................. 92
3.3.1 Resonance assignments of TNAK ...................................... 92
3.3.2 Relaxation data ............................................................ 99
3.3.3 Molecular rotational diffusion from hydrodynamics calculate-

ions and NMR data ...................................................... 102
3.3.4 Backbone internal dynamics ........................................... 103
3.4 Discussion ........................................................................ 1 17
3.5 Conclusion ........................................................................ 122
3.6 References .......................................................................... 123
CHAPTER IV: Stability and flexibility in a mesophilic and hyperthermophilic ........
adenylate kinase .............................................................. 128
4. 1 Introduction ....................................................................... 129
4.2 Materials and Methods .......................................................... 136
4.2.1 Protein samples ......................................................... 136
4.2.2 NMR sample preparation and data collection ....................... 137
4.2.3 H-D exchange rate analysis ........................................... 139
4.3 Results and Discussion ......................................................... 140
4.3.1 The exchange mechanism at the experimental pH* values ....... 140
4.3.2 Determination of P values ............................................. 141

4.3.3 The slowest exchanging amide hydrogens are in the ECAK
and TNAK cores ........................................................ 143

4.3.4 Inhibitor binding stabilizes the AMPbd and lid domains of
ECAK*Ap5A ............................................................ 148
4.3.5 Hinge regions ............................................................ 153
4.3.6 H-D exchange and 15N NMR relaxation ............................. 154
4.4 Conclusion ........................................................................ 157
4.5 References ........................................................................ 1 59

CHAPTER V: Associative mechanism for phosphoryl transfer: A molecular

5.1
5.2
5.3

dynamics simulation of Escherichia coli adenylate kinase com-

plexed with its substrates ................................................... 163
Abstract ........................................................................... 164
Introduction ....................................................................... 165
Methods ........................................................................... 168
5.3.1 Starting structure ........................................................ 168

viii

 

5.3.2 Force field ............................................................... 170

5.3.3 Molecular dynamics .................................................... 171
5.3.4 Equilibration ............................................................ 172
5.3.5 H-bond-salt bridge analysis ........................................... 174
5.4 Results and Discussion ......................................................... 174
5.4.1 Validation of simulation ............................................... 174
5.4.2 Substrate specificity .................................................... 182
5.4.2.1 AMP binding site .............................................. 182
5.4.2.2 ATP binding site ............................................... 188

5.4.2.3 Role and conformation of active site residues around
Mg2+ ...................................................................... 194
5.4.3 Catalytic mechanism ................................................... 199
5.5 Conclusions ....................................................................... 201
5.6 References ........................................................................ 204
CHAPTER VI : Conclusions and future directions ....................................... 209
6.1 Dynamics in ECAK and TNAK on the ps-ns timescales .................. 211
6.2 H-D exchange rate analyses of ECAK and TNAK .......................... 215
6.3 MD Simulations of ECAK and TNAK ....................................... 219
6.4 References ........................................................................ 223

ix

 

LIST OF TABLES

11-] Percent identity (percent similarity) of TNAK with selected prokaryotic

AKs .......................................................................................... 53
11-2 Tm values of ECAK, B. subtilis AK, BSAK, and TNAK ............................. 61
II-3 Comparison of ECAK, BSAK, and TNAK kinetic parameters ...................... 61
11-4 Zn2+ content in TNAK as determined by ICP-AES .................................... 63
III-1 Average 82 values of the secondary structures in the various domains of

ECAK, TNAK, ECAK*Ap5A, and TNAK*Ap5A .................................. 109
111-2 Residues with Rex values > 1.5 s'] in the various domains of ECAK, TN -

AK, ECAK*Ap5A, and TNAK*Ap5A ................................................ 115
IV-l Distribution of protected amides in the free and inhibitor forms of ECAK
.......... and TNAK146
V-l Water—Mg2+ Van der Waals parameters for the Mg2+ coordination water

during the MD equilibration and simulation periods ................................. 173

x

LIST OF FIGURES

[-1 Comparison of theoretical protein stability curves for hyperthermophilic and

mesophilic proteins .......................................................................... 8
I-2 H-D exchange data for E. coli and Thermus thermophilus IPMDHs ............... 13
1-3 Sensitivity of NMR experiments to motions in different time regimes ............. l7
[-4 The unfolding model for hydrogen exchange in a hypothetical protein... .. . .......25
II-l Alignment of T. neapolitana, A. fulgidus, B. stearothermophilus, A. aeolicus,

E.coli, and P. abyssi AK sequences ...................................................... 54
II-2 Effect of temperature on TNAK activity and stability ................................. 57
11-3 Effect of pH on TNAK activity and stability .......................................... 58
II-4 TNAK unfolding in the presence of GdnHCl .......................................... 60
11-5 Titration of Zn2+ in TNAK with PMPS ................................................. 65
111-1 Alignment of TNAK and ECAK sequences using ClustalW ........................ 81
III-2 Activity profiles of ECAK and AAAK and the kca, values of TNAK at 30°C

and 80°C .................................................................................... 82
III-3 Ribbon diagrams of the homology-modeled structures of TNAK (open form)

and TNAK*Ap5A (closed form) ........................................................ 91
III-4A 600 MHz 'H-‘SN HSQC spectra of TNAK at 30°C. . . . .. ....................................... 94
111—41; 600 MHz ‘H-‘SN HSQC spectra of TNAK*ApSAat 30C97
III-4C HNCA strip plots of TNAK and TNAK*Ap5A .......................................... 98
III-5 Relaxation parameters - R1, R2, and NOE of TNAK and TNAK*Ap5A ......... 101
III-6 Order parameters, 82, of TNAK and ECAK; and TNAK*Ap5A and ECAK-

*ApSA ......................................................................................... 106
III-7 Exchange contributions, Rex (3") in TNAK and ECAK; and TNAK*Ap5A

and ECAK*ApSA ........................................................................ 1 l3
III-8 Backbone ps-ns motions (S2 values) summarized for TNAK and ECAK open

forms ....................................................................................... 1 19

xi

III-9

IV-1

IV-2

IV-3

V-l

V-2

V-5

V-6

V-7
V-8

us timescale dynamics in ECAK and TNAK backbones open forms ............. 1 l9
Ribbon diagrams of the homology modeled structures of TNAK (open form)
and TN AK*Ap5A (closed form). ...................................................... 134
Sequence dependence of log of protection factors, P, in ECAK, TN AK, EC-
AK*ApSA and TNAK*Ap5A. ......................................................... 145
Protection factors summarized for ECAK, TNAK, ECAK*Ap5A, and TN A-
K* ApSA backbones ...................................................................... 150
Structures of IAKE and 1AKE* ........................................................ 169

Comparison of lAKE crystallographic B-factors and simulation RMSF
values ....................................................................................... 177

The ATP-AMP-Mg2+-4(H20) complex and several of the (charged) residues
that are within 7A radius of Mg2+ from a snapshot around 2.5 ns. ................ 181

Hydrogen bonds and salt bridges between the protein and ATP and between
the protein and AMP during the course of the simulation. ......................... 184

Stable H-bonds and salt bridges between ECAK residues and AMP and bet-
ween ECAK residues and ATP during the 3-ns simulation ......................... 185

Mg2+ coordination sphere ............................................................... 197
Distance between ATP y-P and AMP a-P atoms during the 3-ns simulation. ...200

Geomtery of Mg2+ association with its coordinating waters, ATP and AMP
phosphates, and Lysl3 .................................................................. 202

xii

<82)

2D

3D

AAAK

AK

AMPbd domain
ApSA

BGAK
BStAK
BSubAK

CD

DSC

ECAK
ECAK*ApSA
EcRNase
Gdanl

H-D exchange
HSQC
ICP-AES
IPMDH

MD

NMR

NOE

ns

LIST OF ABBREVIATIONS
mean S2
2-dimensional
3-dimensional
Aquifex aeolicus AK
adenylate kinase
AMP-binding domain
diadenosine pentaphosphate
Bacillus globisporus AK
Bacillus stearothermophilus AK
Bacillus subtilis AK
circular dichroism
differential scanning calorimetry
Escherichia coli AK
ECAK complexed with ApSA
E. coli RNase H
guanidine hydrochloride
hydrogen-deuterium exchange
heteronuclear single quantum coherence
inductively coupled plasma-atomic emission spectroscopy
isopropyl malate dehydrogenase
molecular dynamics
nuclear magnetic resonance
nuclear Overhauser effect

nanoseconds

xiii

 

 

P values

PAR

pdb

PMPS

ps

RMSF

RNase

TNAK
TNAK*Ap5A

TtRNase

us

protection factor values
4-(2-pyridylazo) resorcinol

protein data bank
p-hydroxy-mercuriphenyl sulphonate
picoseconds

root mean square fluctuations
ribonulease

Thermotoga neapolitana AK

TNAK complexed with ApSA
Thermus thermophilus Rnase H

microseconds

xiv

Chapter 1

1.1 Hyperthermophiles: Definition and evolution

Bacteria that have an optimal growth temperature above 80 °C are termed
hyperthermophiles. All but two of the hyperthermophilic genera are archaea,
Thermotogales and Aquificales being the only bacteria (66, 67). These two genera are the
most ancient of bacteria (6). They are the first to have diverged from the
Archaea/Eukarya lineage, and are they are also the most slowly evolving of all extant life
forms (67). These facts together led many scientists to conclude that life has first

appeared on Earth under hyperthermophilic conditions.

This thesis focuses on enzymes isolated from hyperthermophilic bacteria.
Pioneering work on thermophilic and hyperthermophilic bacteria was done by
Dr Thomas D. Brock at Indiana University in the late 19605, who studied life in the
boiling hot springs of Yellowstone National Park, Wyoming. The first organism to be
identified was Sulfolobus acidocaldarius, both a hyperthermophile and an acidophile (5).
Since then, more than 70 hyperthermophilic species have been isolated (53) from
terrestrial and marine hot environments, with the record of high temperature for life held

since 2003, by an as-yet-unnamed microbe closely related Pyrodictium occultum (30),

Strain 121, which can grow at temperatures up to 121 °C. Previously, the most heat
tolerant organism known was Pyrolobus fumarii discovered in 1997 from a thermal pool
in Italy, which was reported to grow at temperatures up to 113 °C (4). All these findings
raise the question: What is the upper limit of temperature for life? When T. D. Brock first
discovered microorganisms in hot springs he found that they were not just surviving but
growing and flourishing at these temperatures. Subsequent biochemical studies of
thermophiles led to the discovery of thermostable protein-synthesis enzymes in Thermus
aquaticus and the conclusion that thermophiles evolved enzymes that function optimally
at their optimal growth temperature (74). Since high temperatures are required rather than
tolerated by hyperthermophiles, proteins from these organisms (or hyperthermophilic
proteins) are inherently adapted to be stable at high temperatures. Hyperthermophilic
enzymes can therefore be used as model systems for studying molecular evolution and
molecular mechanisms for protein thermostability and enzyme function. Such studies can
lead to the development of new and efficient strategies for protein engineering and to a

wide range of biotechnological applications.

1.2 Hyperthermophilic enzymes

Enzymes can be classified into three broad categories based on their thermostability

properties:

1. Mesophilic: These are enzymes isolated from organisms that grow below 50 °C.

They are typically active between 25 °C and 50 °C.

2. Thermophilic: These enzymes are isolated from organisms that grow optimally

between 50 °C and 80 °C. They are typically stable between 60 °C and 80 °C.

3. Hyperthermophilic: These enzymes are isolated from hyperthermophiles,
organisms that grow optimally above 80 °C. They are typically active and stable

above 80 °C.

With the discovery of hyperthermophilic enzymes, studies were initiated to understand
the unusual thermal properties of these enzymes. Initial biochemical and structural

studies comparing homologous mesophilic and hyperthermophilic enzymes showed that:

1. Under their respective physiological conditions the molecular properties of
hyperthermophilic and mesophilic enzymes are comparable (i.e., they maintain
“corresponding states” (52) with respect to overall topology, solvation, activity,

etc).

2. Their sequences are typically 40% to 85% similar (13, 60) and their 3D structures
are super-imposable. Particularly, the core structures of hyperthermophilic
proteins and their mesophilic homologues are similar. The high level of similarity
in the core region suggests that even mesophilic proteins are packed very
efficiently, not leaving much room for further improvements in stabilization.
Sequence comparisons between hyperthermophilic proteins and their mesophilic
homologues show that stabilizing interactions are often found in the less

conserved regions of the structure (61).

3. When expressed in Escherichia coli, the majority of the hyperthermophilic
enzymes retain all of their native biochemical properties including proper folding
(19), thermostability, and optimal activity at high temperatures (57, 60). This fact
has greatly facilitated their study, since hyperthermophilic organisms are hard to

grow.

The discovery that hyperthermophilic enzymes can be expressed in mesophilic
bacteria like E. coli without loss of their thermostability or activity highlighted the
inherent nature of their thermostability. Most hyperthermophilic enzymes are intrinsically
thermostable with the exception of a few enzymes that are stabilized by intracellular
factors such as salts, cofactors, substrates, activators, and general stabilizers such as
thermamine. Remarkably, the main differentiating feature between hyperthermophilic
and mesophilic proteins is their temperature ranges for stability and activity. Subsequent
to this discovery, several studies were conducted to identify the molecular mechanisms

by which thermostability is achieved.

1.3 Mechanisms for protein thermostability

A powerful method for identifying molecular mechanisms that confer thermostability is
comparing a number of families of homologous hyperthermophilic, thermophilic, and
mesophilic proteins. As of 2005, there are 241 completed genomes listed in The Institute
for Genomic Research website (www.tigr.org/tigr—scripts/CMR2/CMRGcnomicsspl) out
of which 18 belong to thermophilic and hyperthermophilic organisms. The structures of

hundreds of hyperthermophilic proteins have been deposited in the Protein Data Bank

(pdb). The temperature range of stability for these proteins spans 70 °C to 200 °C, and
their sizes vary from 53 residues (rubredoxin) to 419 residues (a monomeric unit of the
hexameric glutamate dehydrogenase). Numerous studies have therefore focused on
identifying the mechanisms of adaptation of proteins to high temperatures by- first
comparing the structures of hyperthermophilic-mesophilic homologous proteins, and

second by testing stability (3, 10, 28, 29, 34-36, 55, 56, 59).

Based on these studies, several mechanisms have been proposed to contribute to
the greater stability of hyperthermophilic proteins, the most commonly cited being:
(1) Optimal hydrophobic packing, including deletion or shortening of loops (48) to reduce
conformational entropy; (ii) increase in the number of hydrophobic residues, which may
increase the change in enthalpy at melting temperature (10, 47); (iii) increased occurrence
of proline residues in turns or loops, increase in the number of charged residues, increase
in helix-stabilizing residues such as Arg and decrease in helix-destabilizing residues such
as Cys and Pro (10, 21, 48); (iv) increase in the number of hydrogen bonds (62) and salt
bridges (34, 48, 69, 72). While there are several strategies to increase thermostability,
nature seems to have employed electrostatic interactions (H-bonds and saltbridges) most
frequently. The role of salt bridges as a stabilizing mechanism in proteins has long been
controversial (22). However, given the high temperatures to which hyperthermophilic
proteins are exposed, and given the fact that the desolvation penalty to form a salt bridge
decreases with increasing temperature (the dielectric constant of water decreases with
increasing temperature), salt-bridges could be stabilizing factors in these enzymes (34,

69). Sequence comparisons and interaction analyses of mesophilic-hyperthermophilic

enzyme pairs agree that no single molecular mechanism dominates the interaction pattern

in hyperthermophilic proteins.

1.4 Protein stability

Protein stability can be described in terms of two properties: thermodynamic stability and
kinetic stability. The thermodynamic stability of a simple two-state protein is described

as:

K
Folded 4; Unfolded ( 1 )

where Ku is the equilibrium constant for unfolding. The stability of a protein is simply the
difference in the Gibbs free energy between the folded and unfolded states (AGsmb). The
two most common ways of studying thermodynamic stability are via thermal
denaturation using differential scanning calorimetry (DSC) and chemical denaturation
followed by circular dichroism (CD) spectroscopy. Thermal denaturation experiments
using DSC yield the enzyme’s melting temperature (Tm, the temperature at which 50% of
the enzyme is unfolded) and AGsmb. AGstab can be determined only if the enzyme unfolds
reversibly. Studies have shown that AGsmb is typically of the order of 5-15 kcal/mol for a
globular protein, a number that corresponds to only a small number of weak interactions.
Since the stabilizing enthalpic energy created by the noncovalent bonds in the secondary
structures and the hydrophobic core competes with the loss in configurational entrOpy

upon folding, AGstab, in effect represents a small difference between large numbers. The

increase in the free energy of stabilization for a hyperthermophilic protein
(AAGsmb= AGstab, hypenhemophmc - AGsmb, mesophilic) is of the same order of magnitude as
AGsmb indicating that minute structural alterations within the protein may be sufficient to
cope with high temperatures. It has been shown that differences in AGsmb values as small

as 3 to 6.5 kcal/mol can lead to increases in Tm of about 12 °C (31, 43).

Figure I—1 indicates the three theoretical ways in which a higher Tm can be
achieved for a hyperthermophilic protein: (1) AGsmb of a hyperthermophilic protein may
be up—shifted to higher AGsmb values (ii) the curvature of the stability curve (specified by
the difference in the heat capacity between the folded and unfolded states, ACP) may be
broader, leading to a higher Tm, or (iii) the stability curve of the hyperthermophilic
protein may be right-shifted with respect to the curve of the mesophilic protein. The most
common stabilization mechanism among hyperthermophilic proteins is through a
combination of up-shifting and broadening the AGsmb vs. temperature curve (33). Several
examples have shown that the up-shifted stability curve in hyperthermophilic proteins is
achieved by a greater enthalpy change at Tm. The greater enthalpy change is derived from
a larger number of interactions (e.g., electrostatic, hydrophobic, cation-1t) (24, 25, 33,

37).

Kinetic stability, which is expressed as the half-life (tr/2) of the enzyme at defined
temperatures, is a measure of how rapidly an enzyme unfolds. Kinetic stability of a

protein depends on the energy barrier to unfolding. It is a particularly important

AGsmb (kcal/mol)
4L

/ Temperature (K)

/ / / \ \

(a) (b) M (C)

 

 

Figure I-l Comparison of theoretical protein stability curves for hyperthermophilic and
mesophilic proteins. Curve M represents a typical AGstab vs. Temperature curve for a
mesophilic protein. Curves (a), (b), and (c) represent the theoretical AGsmb vs.
Temperature curves for hyperthermophilic proteins. (a) The AGsmb vs. Temperature curve
of the hyperthermophilic protein is broadened. The hyperthermophilic and mesophilic
proteins have the same temperature of maximum stability (Ts) and the same AGsmb at the
Ts. (b) The stability curve of the hyperthermophilic protein is up-shifted and broadened
(this appears to be the most common mechanism of stabilization in hyperthermophilic
proteins). The T8 is the same for both proteins but the AGstab value for the
hyperthermophilic protein is higher. (c) the stability curve is right-shifted to higher
temperatures but the AGsmb at the TS are the same for the mesophilic and
hyperthermophilic proteins.

consideration for proteins that denature irreversibly or slowly, as is often observed for
hyperthermophilic proteins. Irreversible loss of protein folded structure is described by:
"1 "3

F 2:; U —-) [(nactive ) (2).
It has been shown that several hyperthermophilic proteins are kinetically trapped (they
unfold more slowly than their mesophilic homologues) and their higher stability is due to
their slower unfolding rate constants (8, 9, 49). In the case of the T. maritima cold shock
protein, entropic (and not enthalpic) factors contribute to its high thermostability (49). In
the past two decades, several studies have presented evidence that residual entropy in
folded proteins can significantly contribute to the free energy balance of protein stability

(54). The only methods currently available to address these issues are molecular dynamic

simulations and NMR spin relaxation rate measurements (38, 54, 71).

1.5 Rigidity, thermostability, and activity

A current working hypothesis is that, to be stable at high temperatures, hyperthermophilic
enzymes are more rigid than their mesophilic homologues at mesophilic temperatures.
This hypothesis stems from the observation that most hyperthermophilic enzymes are
inactive at low temperatures (20 °C — 40 °C), the temperature range of optimum activity
for their mesophilic homologues. Proteins from hyperthermophilic organisms are adapted
to function at temperatures greater than 80°C, at which their mesophilic homologues are

denatured. The 3D structures of many homologous hyperthermophilie-mesophilic

 

enzyme pairs are highly similar and their active site residues are conserved suggesting
that the reduced catalytic rate of the hyperthermophilic enzyme at low temperatures does
not arise from a dissimilar energy profile of the catalyzed reaction. Instead, the most
popular theory invoked to explain the difference in activity at low temperatures involves
enzyme motions required for activity. The surprising constancy in the catalytic rates of
homologous mesophilic, thermophilic, and hyperthermophilic enzymes at the optimal
growth temperatures of their source organism led G. N Somero to propose the
“corresponding states” hypothesis (52). It is thought that homologous hyperthermophilic-
mesophilic enzyme pairs have similar ‘catalytically relevant’ flexibility (corresponding
states) at their respective optimum temperatures. The stability of hyperthermophilic
enzymes is optimized to maintain corresponding states under a given set of
environmental growth conditions. Enhanced thermal stability of hyperthermophilic
enzymes would then be the result of enhanced conformational rigidity in their folded

native state.

Two questions arise from this hypothesis: (1) Do hyperthermophilic enzymes have
increased conformational rigidity as a consequence of their stability? (ii) Is the low
activity of hyperthermophilic enzymes at ambient temperatures a consequence of their
increased rigidity? Several studies have been conducted on mesophilic-hyperthermophilic
enzyme pairs in the past two decades to answer these questions. I will present here a few

representative examples that either support or do not support this hypothesis.

10

Supporting evidence:

a)

b)

C)

Figure I-2 illustrates the slow hydrogen-deuterium (H-D) exchange in
Thermus thermophilus 3-isopropyl malate dehydrogenase (IPMDH) compared to
the faster exchange in its mesophilic E. coli homologue at 25°C, each at pD 7.15
and 8.15. At temperatures close to their respective activity Optima (48 °C for
E. coli IPMDH and 70 °C for T. thermophilus IPMDH), the exchange rates
become super-imposable indicating similar degree of flexibility for the two

enzymes and lending support to Somero’s “corresponding states” concept. (73).

Low concentrations of denaturants such as guanidium hydrochloride increase the

activity of hyperthermophilic enzymes at suboptimal temperatures (32, 64).

15N relaxation dispersion experiments on the thermophilic Aquifex aeolicus
adenylate kinase (AAAK) and the mesophilic E. coli adenylate kinase (ECAK)
show that the lid-opening rate is the rate-limiting step in catalysis in the two
enzymes. AAAK has a much slower lid-opening rate at 20 °C than ECAK. This
slow opening rate accounts fully for difference in the kcm values between the two
enzymes at 20 °C. This difference in the lid-opening rates between AAAK and
ECAK suggests the possibility that the closed conformation in much more
stabilized relative to the open conformation in AAAK than in ECAK due to more

intramolecular interactions in AAAK (68).

ll

Figure [-2 H-D exchange data for E. coli and Thermus thermophilus IPMDHs at
25 °C is shown in the upper panel. E. coli IPMDH, pD 7.15 (A) and 8.15 (Cl); Th.
thermophilus IPMDH, pD 7.15 (V) and 8.15 (O). X represents the percentage of
unexchanged peptide hydrogens, t is the time, and k0 is the chemical exchange
rate constant. X at different tko values is higher for the hyperthermophilic enzyme
compared to that of the mesophilic IPMDH indicating greater structural rigidity
for the hyperthermophilic enzyme. The lower panel shows the exchange data for
the E. coli and Th. Thermophilus IPMDHs’ respective temperature optima. E. coli
IPMDH, 48°C, pD 7.15 (A) and 8.15 (D); Th. thermophilus IPMDH, 70°C, pD
7.15 (V) and 8.15 (O). The exchange curves are similar indicating similar levels of
flexibility for the mesophilic and hyperthermophilic enzymes. (Image reproduced
from (73), copyright (1998) National Academy of Sciences, USA)

12

0.8

0.6

0.4

0.2

0.8

0.6

0.4

0.2

 

 

   

 
    
 

 

 

 

 

    
 

 

75W:-
.. ' J Th. the ~ ophilus
E. coli IPMDH ‘
_ p=10’3 10‘4 104 104 ..
I l 1 l 1 l 1 l l
2 3 4 5 6
I I I I
" p=1o‘4 10" 'J
3 m thenno ilus IP DH, 70 c -
E. COII IPMDH, 48°C
K 1 I l I l I l 1 l
4 5 6 7 8
I09(kot)

13

 

Evidence against the hypothesis

a)

b)

Using H-D exchange rate measurements in Pyrococcus furiousus rubredoxin,
Hernandez et al showed that (i) at 28°C, conformational opening for solvent
access occurred in the millisecond timescale for the entire protein. This timescale
corresponds to the turnover rate of many enzymes; (ii) distribution of amide
protection factors were typical of those observed for a large number of
mesophilic proteins; and (iii) all native hydrogen bonds involving amide
hydrogens were disrupted within 1 3 close to its temperature of maximal

thermodynamic stability (23).

ISN relaxation data on E. coli and T. thermophilus RNase H (EcRNase and
TtRNase, respectively) show that the difference in mobility across multiple
timescales is complex: backbone order parameters show that certain regions of the
mesophilic EcRNase are more rigid on sub-nanosecond timescales, while other
regions are more flexible. In contrast, the thermophilic TtRNase has a higher
activation barrier for structural fluctuations in the substrate-binding handle region

in the micro-seconds timescale (7).

Conflicting reports such as the above mentioned examples reflect the complex nature

of the energy landscape of enzymes. A correlation between rigidity required for protein

stability and flexibility required for activity does not have to exist per se. Moreover, there

is no single measure of flexibility, since proteins exhibit motions ranging from

picosecond (ps) to nanosecond (ns) timescales (36). Flexibility implies increased

conformational entropy in the folded state, and should therefore lead to thermodynamic

l4

stability of the folded state. Besides, to satisfy requirements for activity and stability
imposed by their high physiological temperatures, hyperthermophilic enzymes may
partition conformational flexibility between the active site and the rest of the protein
differently from what occurs in mesophilic enzymes. Proteins show individual features of

adaptation and more examples need to be studied to draw general conclusions.

The motional properties of a number of hypertherrnophilie-mesophilic enzyme
pairs have been compared using a variety of techniques, including neutron scattering (16,
17), tryptophan fluorescence (18), H-D exchange (24, 27, 73), and more recently NMR
(7, 23) and MD simulations (11, 20, 36). Many of these techniques measure only the
average global dynamics in a protein. Few proteins have been characterized on multiple
time scales using site-specific, rather than global, measures of flexibility. NMR and
molecular dynamics (MD) simulations are two techniques that are uniquely suited to
report site-specific protein motions on multiple timescales. Both of these techniques are
ideal tools to obtain a deeper understanding of the relationship between stability, activity,

and flexibility in hyperthermophilic proteins.

1.6 Protein dynamics from NMR

The principal import of the corresponding states hypothesis is that stabilizing factors in
hyperthermophilic proteins are associated with a decrease in structural flexibility.
Conformational flexibility not only impacts protein stability, it also plays a significant

role in catalytic activity. Therefore, a reasonable balance between rigidity and flexibility

15

of the protein structure is a key element for thermal adaptation. While a number of
experimental techniques are available to probe motional properties of proteins, NMR
spectroscopy has become popular because of the range of motional timescales that are
accessible and the spatial resolution offered by isotopic labeling of proteins. NMR
techniques for studying protein dynamics span more than 12 orders of magnitude in
timescale, as seen in Figure I-3. Of relevance to this thesis are 15N NMR relaxation and
H-D exchange measurements to probe protein dynamics. An overview of these two

techniques is presented in the following sections.

1.6.1 15N NMR relaxation

If a protein sample containing magnetically active nuclei (i.e., lH, l5N, and/or 13C) is
placed in a static magnetic field, a net magnetization will be induced within the sample.
The induction is not instantaneous; instead it accumulates at an exponential rate. If we
allow such a magnetization to establish itself, and then apply a radio frequency pulse to
disturb the equilibrium magnetization, we find that the system will relax back to the
original state at the same exponential rate as the induction. The NMR relaxation process
is therefore the propensity of the magnetic moment to re-establish its equilibrium
polarization subsequent to its perturbation by an external pulse. There are two types of
relaxation processes in NMR. The return of the z magnetization component to
equilibrium (Boltzmann distribution) after perturbation is described by the time constant
T, (i.e., longitudinal relaxation). The return of the xy magnetization component (due to

loss of coherence or dephasing) is described by the time constant T2 (i.e., transverse

16

.moEEB o8: «sesame E 3288 9 03:38 can 35 mucoEtoqxo M22 2:
macaw Bead Seton BE. .3288 .5329: «o 2388: 339$ of £505 edema 2: 5 3:3 QB 25 m; ensur—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N
W
a
9
X
.0
m
8:298 of 598% ceaxsam a» 5563 m02 ah a» w
l T i r k 1’ U
l ‘ 4 ‘ d ‘ my?
00m 00 _‘ GO —. n-O—. 90 r ab _. N70 —. W.
_ p . . . _ a
. . J _ _ _ m
a.
495592: 9.5609331 1 95:99. ccom 1 I.
1 8:338. 05.8.2 t 3552 .596 w
l V O
l 9.59 t m.
Al V I 7 10+

 

 

 

l7

relaxation). The relaxation rates are defined by the fluctuating magnetic fields at the site

of nuclear spins, the fluctuation being caused by internal motions in the protein.

A majority of studies of protein dynamics through NMR have been performed
using the amide 15N nucleus to probe backbone dynamics. The 15N probe has a number of
advantages: (1) ISN-enriched protein samples are easy to produce; (ii) resolution offered
by 2D heteronuclear correlation spectrosc0py is high; (iii) procedures for reliable
measurement of relaxation rates are well established; and (iv) interpretation of the
relaxation data of a 15N-enriched protein is made easy because relaxation of the
15N-nucleus is primarily determined by dipolar interactions with the only other spin-V2
nuclei, the directly bonded protons. All other 1H nuclei will be too far away to make any
significant contribution. Contrast this to 13C uniform enrichment wherein most l3C nuclei
will be directly bonded to two other magnetically active C nuclei. The presence of
additional spin-V2 nuclei adds substantial complexity to the relaxation kinetics.
Nevertheless, recent studies have made use of other backbone probes such as 13CO, (15,
70) and carbonyl spins (12, 14, 75) to obtain a more complete picture of protein backbone

dynamics.

The extraction of motional parameters from 15N NMR relaxation data involves
two steps. First, the appropriate pulse sequences are chosen to accurately measure
R1 (1/T.), R2 (1/T2), and the nuclear Overhauser effect (NOEs), which are related to the
heteronuclear cross-relaxation rates. Secondly, an analytical method is employed to relate
the experimental relaxation rates to the dynamics of the protein. The measured 15N NMR
relaxation rates are directly related to the rotational fluctuations of the ”NH bond

vector with respect to an external magnetic field, B0. The fluctuations of each NH vector

18

—-

   

are described by the autocorrelation function, C(t), which describes how quickly the
orientation of the NH dipoles change as a function of time, t. As t increases, the NH
vector undergoes motions such that C(t) decreases. The manner in which C(t) decays is
described by the spectral density function, J((t)), of the NH fluctuations. J(o)), the Fourier
cosine transform of C(t) (65), gives the relative contribution that frequencies in the range
of (a), u) + dm) make towards C(t) for each NH vector. Essentially, it is J ((0) that gives us
information about the NH vector dynamics. The orientational fluctuations of the NH
vectors, due to the overall tumbling of the protein and internal motions, create time-
varying fluctuating magnetic fields at each NH site, which lead to the relaxation process.
Since these fields are driven by internal motions, the frequency distribution in these fields
is what is described by J(u)). Since the form of J ((1)) is dictated by the NH bond dynamics,
the central aim of heteronuclear relaxation studies is to characterize the shapes of the
individual spectral density functions. The relaxation rates are linear combinations of the
spectral densities at 0, (0H, CON, (0” + CON, and a)” - 0)N(1). They are given by:

2
R1 =dT[3J(a)N)+J(a),, —a)~)+6J(a)H +a),,)+c21(a)~ )] (3)

2 2

R2 =%[4J(O)+3J(a)N)+J(a),, —a)N)+6J(a),,)+6J(a),, —w~)]+%[41(0)+3J(w~)l
+RM (4)

d2
N015 = 1+ (Hxfixwtw. + w. > — 1(a). — w. )1 (5)

l 7N

l9

_ (UNO'A

‘73

the Planck’s constant; y“ and 7N are the gyromagnetic ratios of 1H and '5 N , respectively;

where d = d = (fig—allier ’3), c
75

 

, 110 is the permeability of free space; h is

rNH = 1.02 A is the mean NH bond length; and (0H and cm are the Larmor frequencies of
lH and 15N, respectively. The phenomenological Rex term in Eq. 4 represents
conformational exchange and pseudo-first-order processes occurring on the microsecond

(us) to millisecond (ms) time scale.

Relaxation data analysis proceeds by directly analyzing the measured relaxation
rate constants or by analyzing 1(0)) values determined from the relaxation rate constants
by spectral density mapping. In either case, determination of J(0)) proceeds by using
specific physical models or by using “model-free” functional forms containing a limited
number of parameters. There are many reasonable physical models for the motion of
bond vectors in proteins, most of them decomposing the motions into global and local.
The global motion can be isotropic or anisotropic, anisotropic motion being further
categorized as axially symmetric or fully anisotropic. The models invoked for local
motions are either jumping type of motions between specific sites or restricted wobbling
in a cone type of motion. The latter model is the most widely used. Relaxation data are
most commonly analyzed by using the model-free formalism of Lipari and Szabo (40,
41). In the model-free formalism, J ((1)) for a single timescale is modeled as:

521,"
ZTmZ

(l-Sz)1'

J<w>=3u 2 21} (6)
T

5 1+0)

l+[

1+a)

with t" = 1m" + 12,"; where 1m is the effective correlation time for global motion and I, is

20

the effective correlation time for the internal motion. For internal motions on two

timescales, J((t)) is given by:

-Sz)r'5
2 l} (7)

2 . 2
2. 2 2.
If 1+a) 2's

2
2 S 2'
J(w)=-{[——2’" 21+1
5 Ha) rm 1+0)

where t’f = tam /(Tf + rm); 1’, = 151:", /('ES + Tm); If is the effective correlation time on the
fast timescale (If < 100 ps); I, is the effective correlation time on the slow timescale (If <
I, < rm); 82 = SfZSsz, is the square of the order parameter characterizing the amplitude of
internal motions (0 S S2 S 1); and S,2 and sz are squares of the order parameters for
internal motions on the slow and fast timescales, respectively . Motions characterized by
S2 are referred to as ps-ns timescale motions, those associated with 8,3 are referred to as
sub-ns timescale, while motions characterized by 8,2 are in the ns time scale. As stated in
the paper by Lipari and Szabo (40, 41), 32 is a model-independent measure of the degree
of the NH bond vector spatial restriction, in that the exact nature of the motion
experienced by each NH bond vector does not have to be identified. On the other hand, re
depends on both the rate and amplitude of internal motions and it requires a physical
model, such as diffusion in a cone model, for its interpretation. In summary, the model-
free formalism supplies a set of motional parameters that are optimized in a non-linear

least squares fitting procedure to reproduce the experimental relaxation data.

In the model-free formalism described above, the spectral density function at five
frequencies (0, (on, (ON, (on - (ON, (0“ + cm) is being evaluated using only three

experimental observables: T1, T2, and NOE, at a single magnetic field. Hence, Lipari and

21

Szabo made some simplifying assumptions about NH motions without the use of an
explicit motional model. They assumed that the auto-correlation function, C(t), can be
described as a sum of exponential decays, and that global and local motions are totally
uncoupled. It is argued, with merit, that, since the information supplied by three
relaxation parameters is not sufficient to accurately specify J(u)), the errors in the
motional interpretations stem not only from experimental errors but also from potentially
incorrect assumptions. In an attempt to do away with the assumptions in characterizing
J((t)), an alternate method called spectral density mapping has been proposed (45, 46).
This method increases the experimental sampling of 1(a)) by measuring the relaxation
data at multiple fields, thereby providing a clearer view of the shape of J((t)). Although
spectral density mapping has been used in a number of studies, the model-free approach
continues to dominate the current thinking in NMR relaxation data analysis. Work
reported in this thesis uses the model-free analysis of relaxation data acquired at a single
magnetic field. The reason for this is two-fold: (i) during the course of this work, an
NMR spectrometer at 600 MHz was the only one available. (ii) Relaxation data reported
in chapter 3 of this thesis require comparison with previously published work that used
the same method. The use of model-free analysis is justified here due to the qualitative

and comparative nature of the account.

1.6.2 H-D exchange, protein stability, and flexibility

While 15N NMR relaxation experiments provide information on the ps-ns and us

timescales, measurements of amide proton exchange can provide information on global

22

stability as well as on local conformational fluctuations that occur at timescales greater
than milliseconds. Since amide NH groups are distributed uniformly throughout a
protein, access to their exchange behavior can, in principle, provide detailed residue-level
information on structure change, dynamics, and energetics of proteins. For many years,
however, this level of resolution was not realized since the rate measurements relied on
techniques that measure H-D rates in a summed, structurally unresolved way.
Measurement of H-D exchange rates through NMR has made accessible the full power of
site-resolved H-D exchange. In a typical H-D exchange experiment, 15N-labeled protein
is expressed and purified under conditions favoring the native state. A 2D hetero-nuclear
single quantum correlation (HSQC) spectrum of the protein in undeuterated ('H2O)
buffer is recorded, and the observed chemical shifts are assigned to specific backbone
amides. An HSQC spectrum is then recorded for the protein in deuterated (D20) buffer.
Several HSQC spectra are recorded at definite time intervals during which the protein is
allowed to exchange with D2O in the buffer. Because deuterated amide hydrogens will
not produce a signal in HSQC, exchange is observed as decay in signal intensity for each
amide proton as time progresses. An exponential fit is made to the signal intensity decay

of the HSQC spectra, and the exchange rate for each amide proton is computed.

Under native conditions a protein will experience dynamic fluctuations that range
from local breathing motions to global unfolding events. To exchange, protected amides
must be exposed to the solvent by either solvent penetration into the protein interior or by
transient opening that exposes the amide to the exterior. Potential opening events include
fluctuations that expose single amides, cooperative partial unfolding, as well as complete

unfolding. Figure I-4 illustrates the unfolding model for protein H-D exchange proposed

23

Figure L4 The unfolding model for hydrogen exchange in a hypothetical protein.
Three types of structural changes that cause transient separation of H-bond donor
and acceptor groups and solvent accessibility can lead to hydrogen exchange:
local fluctuations (bottom), partial unfolding (center), and complete unfolding
(top). kc; and k0p are the closing and opening rates, respectively, and kch represents
the chemical exchange rate.

24

k

9
‘1

ch

H

   

global
subglobal

‘60?
%
k0p
kcl

it

 

local

25

by Linderstrom-Lang (who first conceived the H-D exchange approach) and his
colleagues (26). In this view, hydrogens involved in hydrogen bonds exchange during a
small fraction of time when the H-bond donor and acceptor are transiently separated in a
dynamic opening-closing reaction or breathing movements. These movements have the
effect of exposing the amide hydrogens to the solvent during which the peptide group is
attacked by either an OH' or H30+ ion. Local breathing motions can arise as a result of
H-bonds breaking individually, as in the local pathway, in the absence of strong bond
forces or as a result of allosteric or functional dynamics in the protein. Figure I-4 also
considers large, more cooperative opening modes for global unfolding which can be
measured by H-D exchange even in conditions favoring the native state. Proteins
molecules in solution are required by thermodynamic principles to occupy all possible
higher energy (relative to the native state) states according to the Boltzmann distribution
and, over time, to cycle through these states. Thus, the protein molecule unfolds and
refolds even under native conditions. This behavior is invisible to most techniques one
can think of since they are swamped by signals from the majority population of the native
state. H-D exchange is unique because the native state does not contribute significantly to
the measured exchange rates. Instead, measurable exchange rates are produced by the

cycling of protein molecules through high energy states.

The general scheme for amide hydrogen exchange with solvent as proposed by

Linderstrdm-Lang (39) is:

k0p kch
Closed 4:» Open __> Exchanged (8)
cl

26

In the above scheme, closed represents the exchange-incompetent state, and open
represents the open exchange-competent state with kOp and kc; representing opening and
closing rates, respectively. Once in the exchange-competent state, exchange of the amide
proton with the solvent occurs with an exchange rate, kch. Kch values, also called
Bai factors, are typically measured using short model peptides, and they can be easily
predicted from the protein sequence (2). In folded proteins, observed exchange rates are
often much smaller than the predicted Bai factors. The extent of retardation of the
exchange rate is expressed as the protection factor, F, the ratio of the random coil
exchange rate constant kch to the observed rate constant kex. From Eq. (8), the observed

rate constant, kex, can be expressed as:

__ kOP kch

ex - (9)
kcl + kch

Depending on the relative rates of closure (kd) and chemical exchange (kch), Eq. (9) can
be further simplified. EXl type exchange occurs when kch >> kc] (the rate-determining
step is the opening), and kex = kop. EX] is rarely observed in proteins under native
conditions. In the other limiting case, kch << kc], opening events rarely result in exchange.

This case is called the EX2 type exchange where Eq. 8 simplifies to:

k

__02_ __ _
kex _ k z xkch —K0pkch (10)
C

where KOp is the equilibrium constant for opening and closing.

Based on the above equations, H-D exchange can report on the equilibrium (EX2)

and kinetic (EXl) features of the protein. In the native folded protein, the protection

27

factor, P, can range from 1 to 1010 or even higher. Residues with small P values are
generally associated with local breathing movements. Residues with large P are the
slowest exchanging residues and are associated with global unfolding. Therefore,
comparing the P values of backbone amides in homologous mesophilic-
hyperthermophilic proteins can give per-residue information on the relative

conformational stability of the two proteins.

1.7 Work presented in this thesis

The motivation for this thesis has been to verify the validity of the hypothesis that
enhanced thermostability in proteins implies enhanced structural rigidity. Since many
hyperthermophilic enzymes have reduced or no activity at temperatures far below their
physiological growth temperatures (i.e., 20°C-40°C), it is thought that their reduced
activity is a consequence of their enhanced structural rigidity. The emphasis of this
project has been to use techniques such as NMR and molecular dynamics (MD)
simulations that enable site-specific measures of flexibility and can access a wide time

scale range, to answer the following questions that stem from the hypothesis:

1. Is protein rigidity at low temperatures a consequence of stability at high

temperature?

2. Is reduced activity in hyperthermophilic enzymes a consequence of increased

structural rigidity?

28

In this work, I describe studies investigating the relationship between dynamics,
activity, and stability using adenylate kinase (AK) from the mesophile Escherichia coli
(ECAK) and the hyperthermophile Thermotoga neapolitana (TNAK) as models.
Adenylate Kinase was chosen as the model enzyme because of the following criteria:
(i) it is a well studied enzyme and AKs from a number of organisms have been
biochemically characterized; (ii) many crystal structures of AKs from several organisms
particularly E. coli are known, both free form and in complex with various substrates and
inhibitors (63); (iii) AKs are small monomeric enzymes, which make them accessible to
NMR and MD simulations; (iv) AKs have two “lids” that close upon the substrates ATP
and AMP to catalyze phosphoryl transfer to produce ADP. Hence, AKs undergo large
conformational changes during substrate binding and product release; and (v) backbone
dynamics of ECAK measured using 15N NMR relaxation have been reported (50, 51, 58)
which facilitates comparative studies of dynamics between ECAK and TN AK. As a first
step in the project, TNAK was cloned, expressed, and its biochemical properties were
investigated. Chapter 2 presents the results of biochemical characterization of the
hyperthermophilic TNAK. The results show that while TNAK is highly stable, as is
expected of a hyperthermophilic protein, it is as active as ECAK at 30 0C. This property
of TNAK is unusual for a hyperthermophilic enzyme, and it makes TNAK a particularly
interesting example for this project. TNAK and ECAK, therefore, make a unique
hypertherrnophilie-mesophilic enzyme pair, unlike other examples studied so far, to

investigate the relationship between dynamics and stability and dynamics and catalysis.

Chapter 3 presents results from 15N NMR relaxation studies on TNAK and the

comparison of these results to those of ECAK. l5N NMR relaxation studies are used as a

29

probe of backbone dynamics in the ps to ns as well as in the us timescales. The dynamics
of the two enzymes are correlated to the observed differences in stability and activity.
Chapter 4 presents the results from H—D exchange studies on TNAK and ECAK that
report on the dynamics in the minutes to day’s timescale. These studies give insight into
the structural distribution of stability in ECAK and TNAK. Chapter 5 presents the results
from MD simulation studies of ECAK in complex with its substrates ATP and AMP. The
results highlight important substrate-protein interactions that explain the observed
substrate specificity. Interactions between charged residues in the binding site, the Mg2+
cation, and water molecules maintain the geometry and distances of the AMP a-
phosphate and ATP B- and y-phosphates in such a way as to support an associative
reaction mechanism for phosphoryl transfer. Results from ECAK MD simulations are
presented here as a starting point for further investigations comparing ECAK and TNAK
using this technique. A summary and perspective of results is given in chapter 6.
Potential future directions are also discussed, including MD simulations that will provide
complementary information on the structural and energetic bases of TNAK

thermostability.

30

References

10.

ll.

12.

Abragam, A. 1961. Principles of nuclear magnetism. Oxford University Press,
London.

Bai, Y. W., J. S. Milne, L. Mayne, and S. W. Englander. 1993. Primary
structure effects on peptide group hydrogen-exchange. Proteins - Struct. Funct.
Genet. 17:75-86.

Berezovsky, I. N., E. I. Shakhnovich. 2005. Physics and evolution of
thermophilic adaptation. Proc. Nat. Acad. Sci. USA. 102:12742-12747.

Bloch], E., R. Rachel, S. Burggraf, D. Hafenbradl, H. W. Jannasch, and K. O.
Stetter. 1997. Pyrolobus fumarii, gen. and sp. nov., represents a novel group of

archaea, extending the upper temperature limit for life to 113 °C. Extremophiles.
1:14-21.

Brock TD, B. K., Belly RT, Weiss RL. 1972. Sulfolobus - New genus of sulfur-
oxidizing bacteria living at low pH and high temperature. Arch. Microbiol. 84:54—
68.

Burrgraf, S., G. J. Olsen, K. O. Stetter, and C. R. Woese. 1992. A
phylogenetic analysis of Aquifex pyrophilus. Syst. Appl. Microbiol. 15:352-356.

Butterwick, J. A., J. P. Loria, N. S. Astrof, C. D. Kroenke, R. Cole, M. Rance,
and A. G. Palmer. 2004. Multiple time scale backbone dynamics of homologous
thermophilic and mesophilic ribonuclease HI enzymes. J. Mol. Biol. 339:855-871.

Cavagnero, S., D. A. Debe, Z. H. Zhou, M. W. Adams, and S. 1. Chan. 1998.
Kinetic role of electrostatic interactions in the unfolding of hyperthermophilic and
mesophilic rubredoxins. Biochemistry 37 :3369—3376.

Cavagnero, 8., Z. H. Zhou, M. W. W. Adams, and S. 1. Chan. 1998. Unfolding
mechanism of rubredoxin from Pyrococcus furiosus. Biochemistry 37:3377-
3385.

Chakravarty, S., and R. Varadarajan. 2002. Elucidation of factors responsible
for enhanced thermal stability of proteins: A structural genomics based study.
Biochemistry 41:8152-8161.

Colombo, G., and K. M. Merz. 1999. Stability and activity of mesophilic
subtilisin E and its thermophilic homolog: Insights from molecular dynamics
simulations. J. Am. Chem. Soc. 121:6895-6903.

Cordier, F., B. Brutscher, and D. Marion. 1996. Measurement of ”Ca-(CO)-

13C cross-relaxation rates in 15N-(IC)-'3C-labelled proteins. J. Biomol. NMR
7:163-168.

31

l3.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

Davies, G. J., S. J. Gamblin, J. A. Littlechild, and H. C. Watson. 1993. The
structure of a thermally stable 3-phosphoglycerate kinase and a comparison with
its mesophilic equivalent. Proteins: Struct. Funct. Genet. 15:283-289.

Dayie, K. T., G. Wagner. 1997. Carbonyl carbon probe of local mobility in 13C,
15N-enriched proteins using high-resolution nuclear magnetic resonance. J. Am.
Chem. Soc. 119:7797-7806.

Engelke, J., H. Ruterjans. 1995. Determination of 13CCl relaxation times in
uniformly l3C/15 N-enriched proteins. J. Biomol. NMR 5:173-182.

Fitter, J., R. Herrmann, N. Dencher, and A. Blume. 2002. The role of protein

dynamics for thermal adaptation of mesophilic and thermostable a-amylase.
Biophys. J. 82:304A-304A.

Fitter, J ., R. Herrmann, N. A. Dencher, A. Blume, and T. Hauss. 2001.
Activity and stability of a thermostable (at-amylase compared to its mesophilic
homologue: Mechanisms of thermal adaptation. Biochemistry 40: 10723-1073 1.

Gershenson, A., J. A. Schauerte, L. Giver, and F. H. Arnold. 2000.
Tryptophan phosphorescence study of enzyme flexibility and unfolding in
laboratory-evolved thermostable esterases. Biochemistry 39:4658-4665.

Grattinger, M., A. Dankesreiter, H. Schurig, and R. Jaenicke. 1998.
Recombinant phosphoglycerate kinase from the hyperthermophilic bacterium

Thermotoga maritima: catalytic, spectral and thermodynamic properties. J. Mol.
Biol. 280:525-533.

Grottesi, A., M. A. Ceruso, A. Colosimo, and A. Di Nola. 2002. Molecular
dynamics study of a hyperthermophilic and a mesophilic rubredoxin. Proteins:
Struct. Funct. Genet. 46:287—294.

Haney, P., J. Konisky, K. K. Koretke, Z. Luthey-Schulten, and P. G.
Wolynes. 1997. Structural basis for thermostability and identification of potential

active site residues for adenylate kinases from the archaeal genus Methanococcus.
Proteins 28:117—130.

Hendsch, Z. S., and B. Tidor. 1994. Do salt bridges stabilize proteins - A
continuum electrostatic analysis. Protein Sci. 3:211-226.

Hernandez, G., F. E. Jenney, Jr., M. W. W. Adams, and D. M. LeMaster.
2000. Millisecond time scale conformational flexibility in a hyperthermophile
protein at ambient temperature. Proc. Nat. Acad. Sci. USA 97 :3 166-3 170.

Hollien, J., and S. Marqusee. 1999. Structural distribution of stability in a
thermophilic enzyme. Proc. Nat. Acad. Sci. USA 96: 13674-13678.

32

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

Hollien, J., and S. Marqusee. 1999. A thermodynamic comparison of mesophilic
and thermophilic ribonucleases H. Biochemistry 38:3831-3836.

Hvidt, A., J. G., Linderstram-Lang K. 1960. Hydrogen exchange in proteins.
Ad. Prot. Chem. 21:287-386.

Fitter, J., and J. Heberle. 2000. Structural equilibrium fluctuations in mesophilic
and thermophilic a-amylase. Biophys. J. 79:1629—1636.

Jaenicke, R. 1996. G]yceraldehyde-3-phosphate dehydrogenase from

Thermotoga maritima: strategies of protein stabilization. FEMS Microbiol. Rev.
18:215-224.

Karshikoff, A., and R. Ladenstein. 1998. Proteins from thermophilic and

mesophilic organisms essentially do not differ in packing. Protein Engin. 11:867-
872.

Kashefi, K., and D. R. Lovley. 2003. Extending the Upper Temperature Limit
for Life. Science 301:934-934.

Kawamura, S., Y. Kakuta, 1. Tanaka, K. Hikichi, S. Kuhara, N. Yamasaki,
and M. Kimura. 1996. Glycine-15 in the bend between two a-helices can
explain the thermostability of DNA binding protein HU from Bacillus
stearothermophilus. Biochemistry 35:] 195-1200.

Kujo, C., and T. Ohshima. 1998. Enzymological characteristics of the
hyperthermostable NAD-dependent glutamate dehydrogenase from the archaeon
Pyrobaculum islandicum and effects of denaturants and organic solvents. Appl.
Environ. Microbiol. 64:2152-2157.

Kumar, S., and R. Nussinov. 2001. How do thermophilic proteins deal with
heat? Cell. Mol. Life Sci. 58:1216-1233.

Kumar, S., C. J. Tsai, B. Y. Ma, and R. Nussinov. 2000. Contribution of salt
bridges toward protein thermostability. J. Biomol. Struct. Dyn.:79-85.

Kumar, S., C. J. Tsai, and R. Nussinov. 2000. Factors enhancing protein
thermostability. Protein Engin. 13:179-191.

Lazaridis, T., I. Lee, and M. Karplus. 1997. Dynamics and unfolding pathways
of a hyperthermophilic and a mesophilic rubredoxin. Protein Sci. 6:2589—2605.

Li, W. T., R. A. Grayling, K. Sandman, S. Edmondson, J. W. Shriver, and J.
N. Reeve. 1998. Thermodynamic stability of archaeal histones. Biochemistry
37: 10563-10572.

33

38.

39.

40.

41.

42.

43.

45.

46.

47.

48.

Li, Z. G., S. Raychaudhuri, and A. J. Wand. 1996. Insights into the local
residual entropy of proteins provided by NMR relaxation. Protein Sci. 5:2647-
2650.

Linderstrom-Lang, K. 1958. Deuterium exchange and protein structure. In:
Neubreger A, (ed.) Symposium on protein structure. London: Methuen.

Lipari, G., and A. Szabo. 1982. Model-free approach to the interpretation of
nuclear magnetic resonance relaxation in macromolecules .1. Theory and range of
validity. J. Am. Chem. Soc. 104:4546-4559.

Lipari, G., and A. Szabo. 1982. Model-free approach to the interpretation of
nuclear magnetic resonance relaxation in macromolecules .2. Analysis of
experimental results. J. Am. Chem. Soc. 104:4559-4570.

Nelson, K. E., R. A. Clayton, S. R. Gill, M. L. Gwinn, R. J. Dodson, D. H.
Haft, E. K. Hickey, L. D. Peterson, W. C. Nelson, K. A. Ketchum, L.
McDonald, T. R. Utterback, J. A. Malek, K. D. Linher, M. M. Garrett, A. M.
Stewart, M. D. Cotton, M. S. Pratt, C. A. Phillips, D. Richardson, J.
Heidelberg, G. G. Sutton, R. D. Fleischmann, J. A. Eisen, 0. White, S. L.
Salzberg, H. O. Smith, J. C. Venter, and C. M. Fraser. 1999. Evidence for
lateral gene transfer between Archaea and Bacteria from genome sequence of
Thermotoga maritima. Nature 399:323-329.

Nicholls, A., K. A. Sharp, and B. Honig. 1991. Protein folding and association:
insights from the interfacial and thermodynamic properties of hydrocarbons.
Proteins: Sruct. Funct. Genet. 11:281-296.

Ogata, Y., E. I. Imai, H. Honda, K. Hatori, and K. Matsuno. 2000.
Hydrothermal circulation of seawater through hot vents and contribution of
interface chemistry to prebiotic synthesis. Orig. Life Evol. Biosph. 30:527-537.

Peng, J. W., and G. Wagner. 1995. Frequency spectrum of NH bonds in eglin c
from spectral density mapping at multiple fields. Biochemistry 34: 16733- 16752.

Peng, J. W., and G. Wagner. 1992. Mapping of the spectral densities of N-H

bond motions in eglin-c using heteronuclear relaxation experiments. Biochemistry
31:8571-8586.

Prevost, M., S. J. Wodak, B. Tidor, and M. Karplus. 1991. Contribution of the
hydrophobic effect to protein stability: analysis based on simulations of the Ile-
96-Ala mutation in barnase. Proc. Nat. Acad. Sci. USA 88: 10880—10884.

Russell R. J. M., J. M. C. Ferguson, D. W. Hough, M. J. Danson, G. L.
Taylor GL. 1997. The crystal structure of citrate synthase from the

hyperthermophilic archaeon Pyrococcus furiosus at 1.9 A resolution.
Biochemistry 36:9983-9994.

34

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

Schuler, B., W. H. Kremer, H. R. Kalbitzer, and R. Jaenicke. 2002. Role of
entropy in protein thermostability: Folding kinetics of a hyperthermophilic cold

shock protein at high temperatures using 19F NMR. Biochemistry 41:11670-
11680.

Shapiro, Y. E., E. Kahana, V. Tugarinov, Z. C. Liang, J. H. Freed, and E.
Meirovitch. 2002. Domain flexibility in ligand-free and inhibitor-bound
Escherichia coli adenylate kinase based on a mode-coupling analysis of 15N spin
relaxation. Biochemistry 41:6271-6281.

Shapiro, Y. E., M. A. Sinev, E. V. Sineva, V. Tugarinov, and E. Meirovitch.
2000. Backbone dynamics of Escherichia coli adenylate kinase at the extreme

stages of the catalytic cycle studied by 15N NMR relaxation. Biochemistry
39:6634-6644.

Somero, G. N. 1978. Temperature adaptation of enzymes: Biological
optimization through structure-function compromises. Ann. Rev. Ecol. Systemat.
9:1-29.

Stetter, K. O. 1996. Hyperthermophilic procaryotes. FEMS Microbiol. Rev.
18:149-158.

Stone, M. J., S. Gupta, N. Snyder, and L. Regan. 2001. Comparison of protein
backbone entropy and beta-sheet stability: NMR-derived dynamics of protein G
Bl domain mutants. J. Am. Chem. Soc. 123:185-186.

Szilagyi, A., and P. Zavodszky. 1995. Structural basis for the extreme
thermostability of D-glyceraldehyde-3-phosphate dehydrogenase from

Thermotoga maritima: analysis based on homology modeling. Protein Engin.
8:779-89.

Szilagyi, A., P. Zavodszky. 2000. Structural differences between mesophilic,
moderately thermophilic, and extremely thermophilic protein subunits: results of a
comprehensive survey. Structure 8:493-504.

Tomschy, A., R. Glockshuber, and R. Jaenicke. 1993. Functional expression of
D-glyceraldehyde-3-phosphate dehydrogenase from the hyperthermophilic
eubacterium Thermotoga maritima in Escherichia coli. Authenticity and kinetic
properties of the recombinant enzyme. Eur. J. Biochem. 214:43-50.

Tugarinov, V., Y. E. Shapiro, Z. Liang, J. H. Freed, and E. Meirovitch. 2002.
A novel view of domain flexibility in E. coli adenylate kinase based on structural
mode-coupling 15 N NMR relaxation. J. Mol. Biol. 315:155—170.

Usher, K. C., A. F. A. De la Cruz, F. W. Dahlquist, R. V. Swanson, M. I.
Simon, and S. J. Remington. 1998. Crystal structures of CheY from Thermotoga
maritima do not support conventional explanations for the structural basis of
enhanced thermostability. Protein Sci. 7:403-412.

35

60.

61.

62.

63.

65.

66.

67.

68.

69.

70.

71.

72.

Vieille, C., J. M. Hess, R. M. Kelly, and J. G. Zeikus. 1995. xylA cloning and
sequencing and biochemical characterization of xylose isomerase from
Thermotoga neapolitana. Appl. Environ. Microbiol. 61: 1 867—1875.

Vieille, C., and J. G. Zeikus. 2001. Hyperthermophilic enzymes: sources, uses,
and molecular mechanisms for thermostability. Microbiol. Mol. Biol. Rev. 65:1—
43.

Vogt G., S. Woell, P. Argos. 1997. Protein thermal stability, hydrogen bonds,
and ion pairs. J. Mol. Biol. 269:631-643.

Vonrhein, C., G. J. Schlauderer, and G. E. Schulz. 1995. Movie of the
structural changes during a catalytic cycle of nucleoside monophosphate kinases.
Structure 3:483-490.

Wilquet, V., J. A. Gaspar, M. van de Lande, M. Van de Casteele, C. Legrain,
E. M. Meiering, N. Glansdorff. 1998. Purification and characterization of

recombinant Thermotoga maritima dihydrofolate reductase. Eur. J. Biochem.
255:628-37.

Wittebort, R. J., and A. Szabo. 1978. Theory of NMR relaxation in
macromolecules - restricted diffusion and jump models for multiple internal
rotations in amino-acid side-chains. J. Chem. Phys. 69: 1722-1736.

Woese, C. R. 1987. Bacterial evolution. Microbiol. Rev. 51:221-271.

Woese, C. R., O. Kandler, and M. L. Wheelis. 1990. Towards a natural system
of organisms - proposal for the domains Archaea, bacteria, and eucarya. Proc.
Nat. Acad. Sci. USA 87 :4576-4579.

Wolf-Watz, M., V. Thai, K. Henzler-Wildman, G. Hadjipavlou, E. Z.
Eisenmesser, and D. Kern. 2004. Linkage between dynamics and catalysis in a
thermophilic-mesophilic enzyme pair. Nat. Struc. Mol. Biol. 11:945-949.

Xiao, L., and B. Honig. 1999. Electrostatic contributions to the stability of
hyperthermophilic proteins. J. Mol. Biol. 289: 1435-44.

Yamazaki, T., R. Muhandiram, and L. E. Kay. 1994. NMR experiments for the
measurement of carbon relaxation properties in highly enriched, uniformly l3C,
15N-labeled proteins - Application to ”Ca carbons. J. Am. Chem. Soc. 116:8266-
8278.

Yang, D. W., and L. E. Kay. 1996. Contributions to conformational entropy
arising from bond vector fluctuations measured from NMR-derived order
parameters: Application to protein folding. J. Mol. Biol. 263:369-382.

Yip, K. S. P., T. J. Stillman, K. L. Britton, P. J. Artymiuk, P. J. Baker, S. E.
Sedelnikova, P. C. Engel, A. Pasquo, R. Chiaraluce, V. Consalvi, R.

36

73.

74.

75.

Scandurra, and D. W. Rice. 1995. The structure of Pyrococcus jFuriosus
glutamate-dehydrogenase reveals a key role for ion-pair networks in maintaining
enzyme stability at extreme temperatures. Structure 3:1147-1158.

Zavodszky, P., J. Kardos, A. Svingor, and G. A. Petsko. 1998. Adjustment of
conformational flexibility is a key event in the thermal adaptation of proteins.
Proc. Natl. Acad. Sci. USA 95:7406-7411.

Zeikus, J. G., B. T. D. Brock.1971. Protein synthesis at high temperatures:
aminoacylation of tRNA. Biochim. Biophys. Acta. 228:736-745.

Zeng, L., M. W. F. Fischer, and E. R. P. Zuiderweg. 1996. Study of protein

dynamics in solution by measurement of 13C(,,-(CO)-13C NOE and (CO)-13C
longitudinal relaxation. J. Biomol. NMR 7: 157-162.

37

Chapter 2

Biochemical characterization of Thermotoga

neapolitana adenylate kinase

Research presented in this chapter was published as the following manuscript: C. Vieille,
H. Krishnamurthy, H. H. Hyun, A. Savchenko, H. Yan, and J. G. Zeikus. 2003.
Thermotoga neapolitana adenylate kinase is highly active at 30°C. Biochem. J. 372:577—

585.

38

Work presented in this chapter was a cooperative effort by C. Vieille, H. Krishnamurthy,
H. H. Hyun, and A. Savchenko. The T. neapolitana adenylate kinase (TNAK) gene was
cloned by A. Savchenko. H. H. Hyun standardized the expression and purification
protocols for TNAK. Studies of the effect of temperature on TNAK activity and stability
shown in Figure II-2 were also done by H. H. Hyun. C. Vieille performed the sequence
analysis of adenylate kinases summarized in Table II-l and Figure H-l. C. Vieille also
carried out the experiments presented in Figure II-5 showing the titration curves for
TNAK Zn2+ with PMPS. H. Krishnamurthy performed the experiments to determine the
effect of pH on TNAK activity and stability (Figure II-3), and the calorimetric and
circular dichroism spectroscopy studies of TNAK (Figure II-4 and Table II-2). The
kinetic parameters of TNAK at 30 °C (Table 3) and the Zn“ content of TNAK through

atomic emission spectroscopy (Table H-4) were also done by H. Krishnamurthy.

2.1 Abstract

The adenylate kinase (AK) gene from Thermotoga neapolitana, a hyperthermophilic
bacterium, was cloned and overexpressed in Escherichia coli, and the recombinant
enzyme was biochemically characterized. The T. neapolitana AK (TNAK) sequence
indicates that this enzyme belongs to the long bacterial AKs. TNAK contains the four
cysteine residues that bind Zn2+ in all Gram-positive AKs and in a few other
Zn2+-containing bacterial AKs. Atomic emission spectroscopy and titration data indicate

a content of 1 mol of Zn2+lmol of recombinant TNAK. The EDTA-treated enzyme has a

39

melting temperature (Tm =93.5 °C) 6.2 °C below that of the holoenzyme (99.7 °C),
identifying Zn2+ as a stabilizing feature in TNAK. TNAK is a monomeric enzyme with a
molecular mass of approximately 25 kDa. TNAK displays Km and Km values at 30 °C
identical with those of the E. coli AK at 30 °C, and displays very high activity at 80 °C,
with a specific activity above 8000 units/mg. The unusually high activity of TNAK at

30 °C makes it an interesting model to test the role of enzyme flexibility in activity.

2.2 Introduction

Proteins from hyperthermophilic organisms are highly thermostable and are optimally
active at high temperatures (often 90 °C and above). Most studies agree that stabilization
mechanisms (such as hydrophobic interactions, hydrogen bonds, and salt bridges) vary
from one protein to another and that no single molecular mechanism is responsible for
protein thermal stabilization. One hypothesis explaining the remarkable thermostability
of hyperthermophilic enzymes is that these enzymes have enhanced conformational
rigidity at low temperatures. According to this hypothesis, psychrophilic, mesophilic,
thermophilic and hyperthermophilic enzymes with sequence similarity have comparable
catalytic efficiencies (indicated by kclem values) at their respective optimal
temperatures, because optimal activity requires a certain degree of conformational
flexibility at the active site. Owing to their increased rigidity at low temperatures,
thermophilic and hyperthermophilic enzymes are only marginally active at these
temperatures, and they gain flexibility required for optimal activity only at higher

temperatures (1). Recent experimental results (e.g. amide hydrogen/deuterium exchange

40

and Fourier-transform infrared spectroscopy) and molecular dynamics simulations show
that results vary from one protein to another. Some thermophilic enzymes are less
flexible than their mesophilic counterparts (2—4), whereas others are equally, if not more,
flexible than their mesophilic counterparts (5). The limited amount of experimental
results and the type of data available (most of the experimental flexibility data are
available only for the protein as a whole, rather than at the residue level) do not allow us
to determine clearly whether rigidity at mesophilic temperatures is a key factor in protein
thermostability. The time scale and amplitude of the conformational fluctuations that

might lead to low thermostability are also not clear.

For these reasons, we decided to compare a new set of similar enzymes in terms
of flexibility and thermostability. We chose the monomeric enzyme adenylate kinase
(AK) because its small size (25 kDa) renders it accessible to NMR and molecular
dynamics studies. AK (EC 2.7.4.3) catalyzes the reaction MgATP + AMP 4: MgADP +
ADP. Because it is the only enzyme forming ADP from AMP, this ubiquitous enzyme is
essential for the maintenance of the cellular ATP pool, for ATP production and for all
macromolecular syntheses (6). Over the past 27 years, numerous studies (e.g. X-ray
crystallography, NMR spectroscopy and site-directed mutagenesis) have been performed
to characterize the ATP- and AMP-binding sites of AK, to identify its catalytic
mechanism and to understand the mechanisms underlying the conformational changes

occurring during catalysis (reviewed in [7]).

Four AK families have been described previously (8, 9). The short AKs (family 1)
represented by the mammalian cytosolic enzymes, and the long AKs (family 2),

represented by the bacterial and mitochondrial enzymes differ by a 20 to 30-amino-acid

41

residue insertion that is present in the long enzymes. This insertion creates a separate
domain that acts as a lid covering the active site (10, 11). Among the bacterial enzymes,
two subgroups of AKs can be differentiated by the presence or absence of a structural
Zn“ in the lid (1 1, 12). The four Zn2+ ligands (typically four cysteine residues or three
cysteine residues plus a carboxylic residue) are conserved in all the Zn2+-containing AKs.
(12). The third AK family represented by methanococcal and Sulfolobus acidocaldarius
AKs does not show significant similarity to the other two AK families, with the exception
of the P loop, a signature motif in nucleotide kinases (9,13,14). In contrast with the
cytosolic and bacterial AKs, which are monomeric enzymes, methanococcal and
sulfolobal AKs are trimeric (14). A fourth AK subfamily was described recently and it
accounts at least for the Mycobacterium capricolum enzyme (15): this subfamily consists
of short bacterial enzymes that are related to the long bacterial AKs more closely when

compared with the short mammalian cytosolic AKs.

Despite the fact that monomeric AKs are well-studied small ubiquitous enzymes,
and even though the Escherichia coli AK (ECAK) unfolds almost completely reversibly
(l6), monomeric AKs have rarely been studied in terms of therrnostabilization
mechanisms and in terms of their activity versus stability at high temperatures. Zhang et
al. (17, 18) noticed that inactivation of the muscle cytosolic AK always occurred at lower
temperatures and at lower urea concentrations compared with other significant
conformational changes. These authors suggested that the active site was more flexible
than the rest of the enzyme. They also suggested that with increasing temperature the
active site reached flexibility levels incompatible with catalysis, while the rest of the

enzyme was still folded. The stabilization provided by Zn2+ in Zn2+-containing AKs was

42

studied by site-directed mutagenesis. A mutant ECAK engineered to contain a Zn2+
cation is significantly more stable than the wild-type enzyme (19). This mutant ECAK is
still less stable than the Bacillus stearothermophilus AK (BSAK), indicating that the Zn
finger-like structure in BSAK is not its sole stabilizing feature. This observation has been
confirmed by two facts: (i) the B. subtilis AK, a Zn2+-containing AK, is significantly less

stable than BSAK, and (ii) it is not more stable than the wild-type ECAK (20).

No AKs have been characterized from hyperthermophilic bacteria. However, a
wealth of AK structure information (e. g. X-ray and NMR data for ECAK and BSAK) is
already available, which would facilitate comparative studies between mesophilic and
hyperthermophilic bacterial AKs. For these reasons, we decided to characterize
Thermotoga neapolitana AK (TNAK). In the present study, we Show that, with a Tm
close to 100 °C, TNAK is indeed a highly thermostable protein. The three enzymes
ECAK, BSAK and TNAK thus represent a nice set of similar enzymes whose dynamic
properties can be compared at the residue level using NMR and molecular dynamics
approaches. With kinetic parameters identical with those of ECAK at 30 °C, TNAK is
one of the rare hyperthermophilic enzymes that are highly active at mesophilic
temperatures. These intriguing catalytic properties of TNAK will make the comparison of
TNAK, ECAK, and BSAK in terms of flexibility, activity and thermostability particularly

interesting.

43

2.3 Materials and Methods

2.3.1 Bacterial strains and plasmids

T. neapolitana strain 5068 [(21) and Deutsche Sammlung von Mikroorganismen und
Zellkulturen catalogue, 1993] was used as the source of chromosomal DNA to construct
the genomic library. E. coli strains XLl-BlueMRF’ and XLOLR (Stratagene, La Jolla,
CA, USA.) were used as the host and the excision plating strains respectively for the
T. neapolitana genomic library- E. coli strain CV2, which expresses the thermosensitive
Pro87 —-) Ser AK mutant (22), was used to express the T. neapolitana genomic library.
Strain CV2 grows at 30 °C, but it cannot grow at 42 °C. Strain DH5a(Life Technologies,
Gaithersburg, MD, USA.) was used for subcloning experiments. Strains HB101(DE3)
and CV2(DE3) were built using the éDE3 lysogenization kit (Novagen, Madison, WI,
USA.) Strain HB101(DE3) was used to overexpress the recombinant TNAK. E. coli
strains were grown in Luria—Bertani (LB) or superbroth (SB) (23) medium, which
contained 100 pg/ml ampicillin or 25 pg/ml kanamycin when necessary. Plasmid pCR2.1
(TA cloning kit; Invitrogen, Carlsbad, CA, USA.) was used to clone PCR products.

Plasmid pET23a(+) (Novagen) was the expression vector.
2.3.2 Library construction and screening

T. neapolitana genomic DNA was extracted as described previously (24). This genomic
DNA was partially digested with Sau3A. Sau3A fragments (2—12 kb) were isolated on a
10—40% sucrose gradient and cloned into the BamHI/alkaline phosphatase treated ZAP

Express vector using the ZAP Express Predigested Gigapack II Gold Cloning kit

(Stratagene). Mass excision of the library from the phage vector was performed
according to the manufacturer’s instructions. E. coli strain CV2 was transformed with the

excised library and transformants were screened for growth at 42 °C.

2.3.3 Manipulation of DNA

Plasmid DNA purification, restriction analysis, subclonings and PCRs were performed
using conventional techniques (25, 26). Oligonucleotides used for PCRs and for
sequencing were synthesized by the Michigan State University Macromolecular Facility
(East Lansing, MI, U.S.A.). DNA was recovered from agarose gels with the Geneclean II
kit (BIO 101, La Jolla, CA, U.S.A.). DNA sequences were determined on both strands
using the ThermoSequenase radiolabelled terminator cycle sequencing kit (U.S.
Biochemical Corp., Cleveland, OH, U.S.A.). Sequencing data were analyzed using
version 8 of the Sequence Analysis Software Package of the Genetics Computer Group
(University of Wisconsin, Madison, WI, U.S.A.) (27). The GenBank® accession no. for

the sequence published in the present paper is AF494055.

2.3.4 Subcloning of the AK gene from T. neapolitana into an expression vector

The AK gene from T. neapolitana (adk) was amplified using plasmid pTNAKl as the
template and Oligonucleotides 5’-GCATATGATGGC'I'I‘ATCTGGTGTT'I‘C-3’ (where
sequence CATATG creates an NdeI site) and 5’-CGAA_'I_IQ”ITATCA'ITTATCACTCC-
ACCC-3’ (where sequence GAATTC creates an EcoRI site) as primers. The amplified

adk gene was cloned into plasmid pCR2.1, and its sequence was verified. The adk gene

was then subcloned into NdeI and EcoRI sites of the vector pET23a(+) to create plasmid

45

pTNAK2. Since pTNAK2 was not well maintained in E. coli, the kanamycin-resistant
Genblock (kanR; Amersham Biosciences, Piscataway, NJ, U.S.A.) was cloned into the
unique EcoRI site of pTNAK2. HBlOl(DE3)(pTNAK2::kanR) cultures were grown in

the presence of kanamycin.

2.3.5 AK purification

Fresh 4 ml HBlOl(DE3)(pTNAK2::kanR) precultures in LB kanamycin were used to
inoculate four 1 litre flasks of SB kanamycin, which contained 1 mM ZnClz. TNAK
production was induced with 1 mM isopropyl-B-D-thiogalactoside when cultures reached
an attenuance D600 of 1. After a 20 h induction, the cultures were centrifuged at 5000 g
for 10 min and resuspended in 150 ml of 50 mM Tris/HCl, pH 7.4 (buffer A). The
bacterial suspension was incubated at 37 °C for l h with 0.6 mg/ml lysozyme and a few
mg of DNase 1. Cells were then disrupted by two passages through a French pressure
cell. The soluble and insoluble fractions were separated by centrifugation at 20 000 g for
20 min. The insoluble fraction was resuspended in approx. 80 ml of buffer A. Both
fractions were heat-treated at 80 °C for 20 min, cooled on ice for 5 min, and centrifuged
again at 20 000 g for 20 min. Pellets containing heat-denatured proteins were discarded.
Supematants from the heat-treated soluble and insoluble fractions were pooled together,
and they were loaded (at a flow rate of 0.8 mllmin) on to an Affi-Gel Blue Gel column
(2.6 cmx30 cm; Bio-Rad Laboratories, Hercules, CA, USA), pre-equilibrated with
buffer A. After washing the column with 5 vol. of buffer A, proteins were eluted with a
500 ml linear 0—1 M NaCl gradient in buffer A. Fractions containing AK activity were

pooled, dialyzed against three batches of buffer A, then concentrated in an ultrafiltration

46

cell equipped with a 10 kDa molecular mass cut-off membrane (Amicon, Beverly, MA,
U.S.A.). The resulting 20 ml solution was loaded on to a Sephacryl-SIOO HR column
(2.4 cmx90 cm; Pharmacia, Uppsala, Sweden), pre-equilibrated with buffer A. Elution
was performed with buffer A at a flow rate of 50 ml/h. Fractions containing AK activity
were pooled, dialyzed against 20 mM Tris/HCl, pH 8.6 (buffer B), and concentrated by
ultrafiltration as described above. This sample was loaded on to a DEAE— Sepharose Fast
Flow (Amersham Biosciences) anion-exchange column (2.6 cmx30 cm; flow rate 1
mllmin) pre-equilibrated with buffer B. After washing the column with 5 vol. of buffer B,

TNAK was eluted with a 500 ml linear 0—1.2 M NaCl gradient in buffer A.

2.3.6 Gel-permeation chromatography

The purified TNAK (0.8 mg in 0.3 ml of buffer A) was loaded on to a Sephadex G-75
column (1.0 cmx40 cm). Elution was performed with buffer A at a flow rate of 0.2

ml/min. The molecular mass markers used were from the Sigma kit for molecular masses

in the range 65—66 kDa (catalogue no. MW-GF-70; Sigma, St. Louis, MO, U.S.A.).

2.3.7 AK assays

TNAK activity in the direction of ADP formation was determined by a coupled assay at
38 °C. The reaction mixture (1 ml) consisted of buffer A containing 250 mM KCl, 4 mM
MgC12, 1 mM ATP, 1 mM AMP, 0.5 mM phosphoenolpyruvate, 0.25 mM NADH and
2.5 units each of pyruvate kinase (catalogue no. P-9136; Sigma) and lactate
dehydrogenase (L-2500; Sigma). The reaction was started by the addition of TNAK,
which was appropriately diluted with buffer A. NADH consumption was followed at

340 nm. At temperatures above 40 °C, TNAK activity was determined by an end-point

47

method. The reaction mixture (0.8 ml; buffer A containing 250 mM KCl, 4 mM MgC12,
1 mM ATP and 1 mM AMP) was incubated with TNAK for 2-5 min at the temperature
of interest, and then cooled on ice. The reaction mixture (0.6 ml) was then transferred to
0.4 m1 of buffer A, containing 250 mM KCl, 4 mM MgC12, 1.25 mM
phosphoenolpyruvate, 0.625 mM NADH, 0.25 mM P',P5-di(adenosine-5')
pentaphosphate (‘ApSA’) and 2.5 units of lactate dehydrogenase. After measuring the
absorbance at 340 nm, 2.5 units of pyruvate kinase were added and absorbance was

measured again.

To determine the kinetic parameters of TNAK, assays were performed in the
presence of 005—2 mM ATP or 0.025—1 mM AMP, using the end-point method
described above. One unit of enzyme corresponds to 1 pmol of product formed per min

under the given conditions.

To determine the effect of pH on TNAK activity, enzyme activity was measured
in 50 mM sodium acetate (pH 3.7—5.5), sodium phosphate (pKaz) (pH 6.0—7.0), Tris
(pH 6.5—8.0), glycine (pH 8.5—10.5) and sodium phosphate (pKa3) (pH 11.0—12.0) at
40 °C using the end-point method. All pH values were adjusted at room temperature
(25 °C), and the ApKalAt for the different buffers (28) were considered to calculate the
pH values at 40 °C. The assays were performed at a constant ionic strength of 220 mM.
The amount of KCl added to adjust the ionic strength I was based on the equation
I=Z(Z,2[C,]), where Z,- is the number of charges of one of the ion and [C1] its

concentration.

48

2.3.8 Kinetic stability assays

The kinetic thermostability of TNAK was determined by incubating the enzyme (at
protein concentrations of 0.55 mg/ml and 5.5 jig/ml) in 0.1 M sodium phosphate (pH 7.5)
at 75, 80, 90, 95, 100, 105, and 110 °C for different time periods. The stability of TNAK
at 5.5 [1ng was also tested at 90 °C in the presence of 1 mg/ml BSA. After heat
inactivation, samples were immediately cooled on ice, then diluted 1000 times with
0.1 M Tris (pH 7.4). TNAK residual activity was determined with 10 pl of these diluted
samples using the coupled spectrophotometric assay. Remaining activity (100%)

corresponded to 1050 units/mg at 38 °C.

To determine the effect of pH on the kinetic stability of TNAK, TNAK
(1.85 pg/ml) was incubated in 50 mM sodium acetate (pH 3.7—5.5), sodium phosphate
(pKaz) (pH 6.0—7.0), Tris (pH 6.5—8.0), glycine (pH 8.5—10.5) and sodium phosphate
(pKa3) (pH 11.0—12.0) [total ionic strength adjusted to 220 mM with KCl (calculated as

above)] for 1 h at 50 °C. The residual activity of TNAK was assayed in 50 mM Tris

(pH 7.4) using the coupled spectrophotometric method at 30 oC.

2.3.9 Microcalorimetry

Differential scanning calorimetry (DSC) was performed with an MC-2 microcalorimeter
(Microcal, Northampton, MA, U.S.A.). Scans were performed with 2 mg/ml TNAK
solutions in buffer A in the presence of different concentrations of high-purity guanidine
hydrochloride (GdnHCl; Pierce, Rockford, IL, U.S.A.). The reference cell contained

buffer A with the corresponding GdnHCl concentration. Thermal gradients were from 25

49

to 100 °C at a scanning rate of 1 °C/min. To minimize the background noise, a buffer-
buffer scan was subtracted from the buffer—TNAK scan before data analysis. DSC scans
of TNAK and the apoenzyme in the absence of the denaturant were performed at the
Glasgow Biological Microcalorimetry Facility (University of Glasgow, Glasgow, UK).
These scans were performed with 1.3 mg/ml TNAK in 50 mM Mops (pH 7.4), with a 1

°C/min thermal gradient from 25 to 125 °C.
2.3.10 Spectroscopic methods

Spectrophotometric titrations of Zn2+ in TNAK with p-hydroxymercuriphenyl sulphonate
(PMPS) were performed in a Beckman DU650 spectrophotometer equipped with a Peltier
system, using quartz microcuvettes (1 cm path length). The enzyme was extensively
dialyzed against 40 mM Hepes/NaOH (pH 7.2) (buffer C) before titration. Buffer C,
PMPS and 4-(2-pyridylazo) resorcinol (PAR) solutions were prepared using MilliQ
water. Release of Zn2+ from TNAK was followed after adding successive samples of
1-10 mM PMPS to 300 pl of lOpM TNAK in buffer C, which contained 0.1 mM PAR.
Absorbances at 250 and 500 nm were recorded 5 min after each PMPS addition. In each
experiment, the first sample cuvette (10 pM TNAK in buffer C plus 0.1mM PAR) was

used as the blank, so that absorbance values recorded thereafter were values of M250 and

M500. Initial values of A250 and A500 for other sample cuvettes were measured before
adding the first PMPS sample. These initial absorbance values were later subtracted from

the experimental readings to eliminate variations due to cuvette differences.

CD spectroscopy was performed on a JASCO J -170 spectropolarimeter (JASCO,

Easton, MD, U.S.A.) equipped with a lml quartz cylindrical cuvette. The sample

50

solutions contained 0.03 mg/ml TNAK in 10 mM Tris (pH 7.4) in the presence of
different concentrations of high-purity GdnHCl. Solutions were incubated overnight at
room temperature for equilibration. For each GdnHCl concentration, a CD spectrum was
recorded in the absence of the enzyme. This spectrum was subtracted from the enzyme

spectrum before data were analyzed.

2.3.11 Other analytical procedures

Protein concentrations were determined using the Bio-Rad protein assay kit (Bio-Rad
Laboratories) with BSA as the standard. SDS/PAGE was performed as described
previously (29). Proteins were visualized by Coomassie Blue staining. Elemental analysis
was performed by inductively coupled plasma-atomic emission spectroscopy (ICP-AES)
at the Animal Health Diagnostics Laboratory, Michigan State University. The dialysis

buffer was used as the negative control for ICP-AES.

2.4 Results

2.4.1 Cloning of the adk gene from T. neapolitana in E. coli

E. coli strain CV2 transformed with the T. neapolitana genomic phagemid library was
plated on LB ampicillin agar and incubated at 42 °C. One colony was obtained after a
2 day incubation. The phagemid (pTNAKl) extracted from this colony contained a 3 kb

insertion. Transformation of strain CV2 with pTNAKl allowed CV2 to grow at 42 and

51

45 °C. The cell extract of CV2(pTNAKl) showed six times more AK activity than the

cell extract of CV2 without the plasmid.

2.4.2 Nucleotide sequence of the adk gene from T. neapolitana

The pTNAKl insert was sequenced using vector- and insert-specific primers. Two open
reading frames, ORFl and ORF2, were identified. ORFl was truncated at its 5’-terminus.
The two ORFs overlap by one nucleotide (results not shown). A BLAST search using the
truncated peptide encoded by ORF 1 as the probe showed significant similarity to SecY, a
subunit of the SecAEGY preprotein translocase complex. A BLAST search using the
peptide encoded by ORF2 as the probe showed significant similarity to multiple AKs. A
potential ribosome binding site was identified 4—7 nt upstream of ORF2 (i.e. the adk
gene). No potential promoter could be identified, suggesting that the T. neapolitana adk

gene is co-transcribed with secY.

2.4.3 Sequence comparisons

As shown in Table II-l and Figure II-l, TNAK belongs to the family of long bacterial
AKs. With the exception of Thermotoga maritima AK (TMAK) (with which it is 93.2%
identical), TNAK is more similar to the archaeal Archaeoglobus fulgidus AK than to
other bacterial AKs (Table H-l). TNAK also shows significant similarity to another
archaeal AK, the Pyrococcus abyssi enzyme. These two archaeal AKs show much higher
similarity to the long bacterial AKs, in general, than to the AKs from the archaea
S. acidocaldarius and Methanococcus igneus (results not shown), suggesting that

archaeal AKs do not represent a phylogenetically homogeneous enzyme group.

52

Table “-1 Percent identity (percent similarity) of TNAK with selected prokaryotic AKs.

Scores were obtained using the GAP software. Sequences were obtained from Genbank®
and from genome sequences.

 

Enzyme *Aful Aaeo BSAK ECAK PDAK *PAAK *SAAK *MIAK
source
TNAK 60.7 48.5 44.7 46.7 41.1 37.7 31.5 21.4

(70.1) (60.2) (57.6) (56.7) (56.1) (53.0) (43.2) (34.1)

*Archaeal species.

AFAK: Archaeoglobus fulgidus AK; AAAK: Aquifex aeolicus AK; BSAK: Bacillus
stearothermophilus AK; ECAK E. coli AK; PDAK: P. denitrificans AK; PAAK:
Pyrococcus abyssi AK; SAAK: S. acidocaldarius AK; MIAK: Methanococcus igneus
AK.

53

.chEn +~cN E 3202: 822mm.— oEonho .... m: C Ame—3208 533 @3053 wEuEa
$32 5 8302: 8.668 4 .8: 30380me ASE use. mh< FEB mucofioafi 8358 83:3: $3668 vocab Ea vascuwxomn >20

82.258 v—< A355 3935 .m can :36
:8 .m .335 max—eon .< A836 3.2—38.8595on .m 53$ 2.22.: .< Axon: «53:33: .H .8 “SE—am? 7: «Saw.—

 

 

   

         

 

 

 

 

mmszZMHmaqqmmHmm QHmU DHm>%..HO...OMO%mMHHmmZZmeHQ m> xnmm
UqumAQ4m .9. Q>¥¢MxezwflmdmxmwwwHAm<Hzomwm> m> Houm
ZUQUH>NA>Om> mm...mdoHH..qu0...M&M>>mHam490mm>>m : MH Oomd
mdquqammHQdm mEQOm 2H2..mqu...momwma>quzoxzz>m z¢> oumm
mMmequmme mHQ-B Q>Q..>AHU...xxmwonqmmBZOM>>m mm> Hdwd
meSUHHMA>md QHUHE m..¥AHU...meNNQH>mOHMMAN>M mm> mosh
32 was a 8885

Hmmm m quwwuHOUMOm>¥mmme -EtmwomeHm mH m>mmmMBHwHQH¢>QAMHwmowqwzquH>Omm_;. >Qmm
Hmmo m qumwgoaomimmzng amommgm >H gHamDm>QmmED>zHo¢mxz. . . .deOmH _._. .A Hoom
>me m H>¥>O ....... mmmmZflx :.emoemmsz m mH>>mam>mmAd>mo>zqoxxqumoqdmdox>.Lm gm Oomd
Bdmz wqmowux00>wmxfimmmmH ._ dOUZMUHm B mEQ>Qom>QHmH>VQAMmUHDdAZqudm¢Od>.Lm .A oumm
Bmmx wqmwwoxnuxzamxmmd-HwafldwuszBm EH >>mmmm>>>zH>¢QHMMquzqzmaqdmd04A_Lm .A admd
Bmmz m >AM>MUQQUAWQmMmmAmszmHmOUMmUHm m8 O>>>Qmm>amn>¢dhqoxzommqmmoqmmd0d>_L. .A mock

C .1 i. a. 9
Hmzmm>..mmAMmHHAH BQQH1.wmmqwmmzmeamBm¢x.H H «HweqowmmBHmmd mfi£mflflmOquHZZ xnmm
AmGZMUQMOdemx>Ad ma 0d DMdozwqmmUmx.> HOmHUMszH Odo .MUflGAflUAAHHmE Houm
H>zom...xmm>mqudHHADQm 0m mxdzxoqmeo.xo>¢ H Hm>mUMMM4Amx¢ oxwdm.qum>AHz 0mm¢
Amozouonxmammm>HUHH>mom om OMdOAOamBUmM. Z HmmHowdd>HMm¢ .0M0fl0.qwzd>qzz oumm
HmomeQmOOAmMM>HO man 0% mxdxxoqmfioxd.>¢ HOmHGNMmm>mX< .Oxwdwkquthzz Hdw<
Hmoxmunzmmqmmx>>mmwwmom 0% mxHxxUAMQZmMK>H H HmmHOmeOmed J . .mUAm>Aw¢zz mosh

 

+9 9

54

Most residues that bind ATP and AMP in ECAK (10) are conserved in TNAK as
well as in the A. fulgidus and P. abyssi enzymes (Figure II-l). The ATP- and AMP-
binding residues of ECAK that are not conserved in TNAK correspond either to ECAK
residues that have one atom at a 4.5 A distance from ATP or AMP or to residues that
interact with the nucleotides through their backbone nitrogen or oxygen atoms. The four
Zn2+- liganding cysteine residues characterized in all AKs from Gram-positive bacteria
(30) and in a few other bacterial AKs (31, 32) are present in TNAK (Figure II-l) and in
TMAK (results not shown). These cysteine residues are also present in the A. fulgidus

and P. abyssi enzymes (Figure II-l).
2.4.4 TN AK overexpression in E. coli and purification

We first attempted to overexpress TNAK in strain BL21(DE3) (Stratagene). The
expression level of TNAK remained mediocre (approx. 5 mg/l of culture) even after
induction with isopropyl-B-D-thiogalactoside. When we tried expressing TNAK in strain
CV2(DE3), no overexpression was observed with or without induction. E. coli strain
HBlOl(DE3) gave satisfying results: TNAK was not expressed at significant levels in the
absence of induction, and approx. 50 mg/l TNAK was obtained after 20 h induction in SB

medium.

During TNAK purification, a significant fraction of the recombinant enzyme
(approx. 30%) was present in the pellet after centrifugation of the bacterial crude extract.
Most of this enzyme was recovered in solution after resuspension of the pellet in buffer
A, heat treatment at 80 °C and centrifugation. The heat treatment at 80 °C was a highly

efficient purification step, and it allowed us to obtain the enzyme preparation without any

55

contaminating ECAK that was initially present in the crude extract. Our purification
procedure yielded pure TNAK, as determined by SDS/PAGE (results not shown). The
purified TNAK consisted of a single subunit of 25 kDa (as determined by SDS/PAGE),
and it migrated as a 29 kDa protein during gel permeation on Sephadex G-75, indicating
that TNAK is a monomeric enzyme, a trait typical of long bacterial AKs. The purified

TNAK showed a specific activity of 1050 units/mg enzyme at 38 °C.

2.4.5 Effects of temperature and pH on TNAK activity and stability

The specific activity of TNAK, as determined by the end-point method, increased with
temperature to its maximum (8400 units/ mg) at 80 °C, and then it started to decrease
sharply (Figure II-2A). The Arrhenius plot of TNAK activity (ADP formation) against
temperature was linear between 30 °C and 70 °C (Figure II-2B), with activation energy of
49 kJ/mol. This value is of the same order of magnitude as the activation energies for the
reactions catalyzed by the muscle cytosolic AK (38 kJ/mol) (17), by ECAK (46 kJ/mol)
and by BSAK (73 kJ/mol) (30). The kinetic thermostability of TNAK was tested at two
different enzyme concentrations (Figures II-2C and II-2D). TNAK is significantly more
stable at high concentrations or in the presence of BSA than at low concentrations. At
both concentrations tested, inactivation is of first order, with activation energies of
185.6 kJ/mol at 0.55 mg/ml enzyme and 44.4 kJ/mol at 5.5 pg/ml enzyme. TNAK is
optimally active between pH 7.4 and 8.0. It is more than 70% active at pH 7.0 and 8.5
(Figure II-3A). The enzyme is optimally stable at pH 8.3 and retains 80% activity at pH

7 .4 and 9.3. TNAK is not at all stable at pH values below 5.0 (Figure II-3B).

56

 

 

 

 

 

 

 

 

 

 

      
        

 

 

 

 

 

 

9.5
a, 8000- A B _ 9
E
D u
v 6000- 01:1 ~85 1;
.é‘ a a?
> ._ 8 O
o :-
§ 4000- S
M . I P75 5
<5: 2000-
- 7
E—
0 I I I I I T I I I I I _ 6.5
30 40 50 60 70 80 90 100110 0.0026 0.0028 0.003 0.0032 0.0034
Temperature (°C) 1/T emperature (UK)
9. 5 [-
'3‘ 75°C 2
'8 ~45 3’.
cu ' a
v—1 0 2-
g 90 C+BSA _ 4 a
.2 a
o -3.5 c
.2. E.
«a: 99
T3 0- - 3 g.
v <
5 C 110°C 5.-
'l I I I I I I I I I I I 2-5 V
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

Time (min) Time (min)

Figure II-2 Effect of temperature on TNAK activity and stability

(A) Specific activities of TNAK (El), ECAK (9), B. subtilis AK (A) and BSAK (I) at
various temperatures (20, 30). (B) Arrhenius plot of TNAK activity. (C, D) TNAK

concentrations were 0.55 mg/ml (C) and 5.5 pg/ml (D). (C) Cl, 90 °C; 0, 100 °C; A,
105 °C; a, 110 °C. (D) [1.75 °C; 0, 90 CC + BSA (BSA concentrationzl mg/ml); 0,
80 °C; 0, 90 °C; A, 95 °C.

57

 

F
>

00
O
l .

 

A
O
. 1

Relative activity (%)
a
o
l

 

 

O
l
I
d
-
d
_

 

 

 

 

3 5 7 9 11

g 100-

; . B

:25: 80-

g .

'3 60-

== .

E

g 40-

g .

g 20-

3 J

m 0 I I I I
3 5 7 9 11

Figure II-3 Effect of pH on TNAK activity (A) and stability (B)

Buffers used: 0, sodium acetate; A, sodium phosphate (pKag); 0, Tris; E], glycine; 0,
sodium phosphate (pKa3).

58

2.4.6 TNAK stability studied by calorimetry and CD spectrosc0py

We did not detect any obvious melting transitions in our DSC experiments with TNAK in
buffer A at temperatures below 100 °C (the upper temperature limit for our MC-2
microcalorimeter). For this reason, we followed TNAK thermal unfolding in the presence
of GdnHCl. Melting was almost completely reversible at a GdnHCl concentration of
0.4 M, as indicated by the almost complete superimposability of the excess heat capacity
curves obtained in two consecutive scans (results not shown). The percentage of
refolding after cooling decreased with increasing GdnHCl concentration. At 2.6M
GdnHCl, almost no transition was seen during a second scan. For this reason, no
thermodynamic analysis of the data was possible. In Figure II-4A, Tm of TNAK is shown
as a function of GdnHCl concentration. A linear curve fit of the data (y=-10.216x+99.334
with r2 +0.99l) would predict a Tm of 99.3 °C in the absence of the denaturant. A second
degree polynomial fit of the data (y=-1.315x2 -6.231x+ 97.125 with r2 =0.996) would
predict a Tm of 97.1 °C in the absence of the denaturant. A confirmatory DSC scan of
TNAK in the absence of the denaturant was performed by the Glasgow Biological
Microcalorimetry Facility on a calorimeter that allowed thermal gradients to be run up to
125 °C. At a scan rate of l °C/min, TNAK (1.3 mg/ml) showed a Tm of 99.7 °C. The
melting transition observed in the first scan was not observed in the second scan,
indicating that under these conditions, TNAK unfolded irreversibly. This Tm value is
significantly above the values measured for ECAK, B. subtilis AK and BSAK (Table II-
2) (20, 30). It also corresponds to the Tm predicted by linear extrapolation of the melting

data obtained in the presence of GdnHCl.

59

 

 

 

 

100 B ._..
_ 2
- _N
4 S
- :1
A90- _E}_\
U ‘ a:
a, - 2‘.
a I - :3
F 80- ~21
.. t:
. - a.
- 5

70 I I I I I I I I I

 

 

 

 

 

1.52.530123456

O
.9
LII
p—s

GdnHCl concentration (M)

Figure “-4 TNAK unfolding in the presence of GdnHCl.

(A) Tm as a function of GdnHCl concentration, followed by DSC. TNAK concentration
was 2 mg/ml (1:1). For Tm in the absence of GdnHCl, the TNAK concentration was
1.3 mg/ml (I). The two lines represent the linear curve fit and the second degree
polynomial fit of the data (see the Results section). (B) TNAK (30 ug/ml) equilibrium
unfolding induced by GdnHCl at 25 °C as determined by the ellipticity value at 222 nm.

60

Table II-2 Tm values of ECAK (19), B. subtilis AK, BSAK (30), and TN AK

 

 

Enzyme Melting temperature Reference
W (°C)

ECAK 51.8 (19)
ECAK quadruple Cys

63.0 (19)
mutant containing Zn?”
B subtilis AK 50.7 (20)
BSAK 74.5 (30)
Apo-BSAK 67.0 (30)
TNAK 99.1 This study
apo-TNAK 93.5 This study

««««««

 

Table II-3 Comparison of ECAK, BSAK, and TNAK kinetic parameters.

 

 

 

Reference
VM(ATP, AMP)
Temperature
Enzyme (°C) ATP (11M) AMP (PM) (units/mg protein)*
ECAK 30 51 1,020 (30)
BSAK 37 36 288 (30)
TNAK 30 49.4 i 4.9 40.9 i 6.6 l 130 This study

*No S.D available for Vmax (ATP, AMP), because it is the mean value calculated from Vmax (ATP)

and Vmu (AMP).

The TNAK GdnHCl equilibrium unfolding profile obtained by CD spectroscopy
at 25 °C is similar to that of other long bacterial AKs, with the following exception:
GdnHCl concentration at the midpoint transition for TNAK (i.e. 3.1 M) is much higher
than that for ECAK, B. subtilis AK and BSAK (0.98, 1.38 and 2.45 M respectively) (20,

30).
2.4.7 Kinetic properties of TNAK

Because TNAK activity at temperatures between 30 and 45 °C is comparable with that of
ECAK (Figure II-2A), the kinetic parameters of TNAK were determined at 30 °C for
comparison with those of ECAK. As shown in Table II-3, the kinetic parameters of
TNAK in the direction of ADP formation were more or less identical with those of

ECAK at the same temperature, namely 30 °C.
2.4.8 Determination of the Zn2+ content of TNAK

The four cysteine residues involved in Zn2+ binding in other AKs (30) are present in
TNAK also. This suggests that TNAK also contains Zn“. The presence of Zn2+ in TNAK
was tested by ICP-AES and by titration with the strongly dissociating thiol specific
reagent PMPS. ICP-AES results (Table II-4) indicate that the recombinant TNAK
contains 1 mol of Zn2+lmol of enzyme. Treatments with 10 mM EDTA for 30 min at
60 °C were almost completely inefficient in removing Zn” from TNAK. EDTA
treatment for 1 h at 75 or 80 °C was necessary to deplete approx. 75% Zn2+ from TNAK
(Table H-4). Other treatments (i.e. with 10 mM phenanthroline or with 5M GdnHCl plus

10 mM

62

Table II-4 Zn2+ content in TNAK as determined by ICP-AES

 

Treatment

EDTA (10 mM)

Phenanthroline (10 mM)

GdnHCl (5M) + EDTA
(10 mM)

Zn2+ content in Remaining
TnAK (mol/mol) activity

 

Temperature Duration
(°C) (min)
— — 0.88
20 30 0.85
40 30 0.9 l
60 30 0.76
70 30 0.58
75 60 0.24
80 60 0.23
75 60 0.58
20 60 0.26

1%)

100

n.d.*

n.d.

n.d.

n.d.

n.d.

94.6

n.d.

n.d.

 

*n.d., not determined

63

EDTA) were not more efficient than the EDTA treatments. As has been observed with

BSAK (30), reducing the Zn2+ content of TNAK does not affect its activity (Table II-4).

We confirmed the ICP-AES results by titrating Zn2+ in TNAK with PMPS at
25 °C. Formation of the PMPS—thiol chromophore (followed at 250 nm; Figure II-5)
shows that PMPS is linearly incorporated into TNAK, up to approx. 4 mol of PMPS/mol
of TNAK. AA250 becomes constant once the ratio [PMPS]/[TNAK] >4. PMPS binding is
accompanied by the release of Zn“, which immediately forms (PAR)2Zn2+ complexes
that can be detected at 500 nm. Parallel with what is observed at 250 nm, Zn2+ is linearly
released from TNAK up to approx. 4 mol of PMPS/mol of TNAK (followed at 500 nm;
Figure H—S).With a molar absorption coefficient of 6.6x104 M'l- min-l for (PAR)2Zn2+ at
20 °C (33), a complete release of Zn2+ from TNAK (i.e. 6.58 11M) would give a AAsoo of
0.43. The value of AAsoo reached at the plateau (i.e. 0.42) corresponded to 0.98 mol of

Zn2+lmol of enzyme, a result that agrees with the ICP-AES results.
2.4.9 Thermostability of the ape-TNAK

A quadruple cysteine mutant ECAK capable of binding Zn2+ had a T.,, 11 °C above that
of the wild-type ECAK (18). The apo-BSAK had a Tm 7.5 °C lower than that of the holo
BSAK (26). These two results indicate that Zn2+ binding is a stabilizing feature in these
two AKs. To test the role of Zn2+ binding in the thermodynamic stability of TNAK, we
measured the Tm of apo-TNAK by DSC. To prepare the apoenzyme, TNAK was
incubated at 80 °C for 1 h in the presence of 10 mM EDTA. It was then extensively

dialysed against buffer A containing 2 mM EDTA and also against three batches of

 

 

 

 

0.12 0.5
- o 0 0 0
01 0 ~04
008- '3 El 1:1 o
g ”tn 1: U -0.3 a?»
a .
o 006 3
<1 r02
0.04J
0.02- ’0"
0 I I I I I O
0 2 4 6 8 10 12

[PMPS]/[TNAK]

Figure II-S: Titration of Zn“ in TNAK with PMPS
Cl, AA250 (‘OD250’); O: AA500 (‘OD500’). [PMPS] and [TNAK] were corrected for

dilution before calculating the [PMPS]/[TNAK] ratios. AA250 and AA500 values were
corrected for dilution.

65

metal-free buffer A. By DSC, the apoenzyme shows a Tm approx. 6 °C below that of the

holoenzyme (Table II-2).

2.5 Discussion

As expected from TMAK sequence (32), TNAK belongs to the long bacterial AKs. Since
some proteins that are known to be monomeric in mesophiles exist as multimers in
hyperthermophiles (34—36), and since some AKs are trimeric (14), it was important for us
to determine the quaternary structure of TNAK. The fact that TNAK is a monomeric
enzyme that is highly similar to ECAK and BSAK (Table II-l, Figure II-l) makes
ECAK, BSAK and TNAK a perfect set of similar enzymes whose dynamic properties can
be compared at the residue level to correlate thermostability and activity with

conformational dynamics.

Molecular mechanisms underlying the thermostability of TNAK are unknown.
TNAK contains a significantly smaller number of alanine and glycine residues when
compared with ECAK or BSAK (26 in TNAK versus 39 in ECAK and 40 in BSAK). Six
alanine and glycine residues in ECAK are substituted with charged residues in TNAK.
These six residues are located on the surface of ECAK [as visualized in the X-ray
structure of ECAK (PDB code lake); results not shown]. Of the seven bulky hydrophobic
residues of ECAK (i.e. Ile, Leu, Met or Val) substituted with charged residues in TNAK,
six are exposed to the solvent (results not shown). Although these and other similar, not

mentioned, observations suggest that TNAK might be stabilized by optimized

66

hydrophobic packing and by an increased number of surface salt bridges, these
stabilization mechanisms remain highly hypothetical in the absence of supporting

structural and experimental evidence (the crystallization of TNAK is in progress).

The similarity of TNAK to the archaeal A. fulgidus AK is an intriguing result.
Such a high similarity level suggests that the two genes originate from a common
ancestor. Although evidence exists suggesting that lateral gene transfer has occurred
between thermophilic bacteria and archaea (37), the adk gene might not be among the
genes that have been laterally transferred. In A. fulgidus, the adk gene is not preceded by
secY (38), as it is in T. neapolitana. In addition, the T. maritima adk gene does not belong

to one of the 15 genomic cluster regions that are most archaea-like (37).

As shown in Figure II-2(A) and Table 11-3, TNAK is as active as mesophilic and
thermophilic bacterial AKs between 30 and 65 °C. In particular, it is three times as active
at 30 °C as BSAK is at 37 °C. This result is highly uncommon. Only three other
hyperthermophilic enzymes have been characterized that are highly active at mesophilic
temperatures: (1) the T. maritima L-isoaspartyl O-methyltransferase (39) is optimally
active at 85 °C, with a specific activity that is at least ten times higher than that of
mesophilic enzymes at 25—37 °C. When compared with similar mesophilic enzymes, the
T. maritima L-isoaspartyl O-methyltransferase had the lowest Km values for each of the
methyl-accepting substrates examined at 37 °C. (ii) The T. maritima indoleglycerol
phosphate synthase (40) is twice as efficient as the E. coli enzyme at 25 °C. This higher
catalytic efficiency for the hyperthermophilic enzyme is due to a 70 times higher affinity

for the substrate, which compensates for a 33 times lower km, value. (iii) The T. maritima

67

phosphoribosyl anthranilate isomerase (Tm-PRAI) (41) is almost four times more
efficient at 25 °C than its E. coli counterpart, E. coli phosphoribosyl anthranilate
isomerase (Ec-PRAI). Although the kca, value of Tm-PRAI is almost ten times lower than
that of Ec-PRAI, the affinity of Tm-PRAI for its substrate is almost 44 times higher than
that of Ec-PRAI. For these three T. maritima enzymes, the enzyme activity is increased
due to a very high affinity for their substrate. Their km, however, remains typically one
order of magnitude below that of the mesophilic protein. In TNAK, both Km and kcat
values are identical with those of ECAK at 30 °C, making TNAK, again, a unique

hyperthermophilic enzyme.

Hyperthermophilic enzymes are usually less active than their mesophilic
counterparts at low temperatures. One explanation is that hyperthermophilic enzymes are
too rigid at low temperatures to be highly active. Their flexibility increases with
temperature, and reaches levels compatible with optimal activity only at temperatures
close to the growth temperature of their source organism. This hypothesis and the fact
that TNAK shows kinetic parameters at 30 °C identical with those of ECAK suggest that
the conformational flexibility of the active site of TNAK should be comparable with that
of the active site of ECAK at 30 °C. The small size of bacterial AKs (20—25 kDa) makes
possible the use of techniques such as NMR relaxation studies or hydrogen/deuterium
exchange coupled with NMR detection, to study flexibility at the atomic level. These
techniques will be used to compare the molecular dynamics of TNAK and ECAK

catalytic sites at 30 °C.

68

The facts that TNAK at 0.55 mg/ml is kinetically and thermodynamically stable at
90 °C and that the optimal temperature for TNAK activity is only 80 °C suggest at first

that TNAK reversibly inactivates between 80 and 90 °C. On the other hand, TNAK
enzymatic assays are typically performed using TNAK concentrations below 1 pg/ml.
These concentrations are lower than the concentrations used for the kinetic
thermostability assays or for DSC. The lower kinetic thermostability of TNAK at low
concentrations suggests that the decrease in its activity at temperatures above 80 °C is
due to an irreversible inactivation. This result is different from what Zhang et a1. (17)
concluded for the muscle cytosolic AK. These authors suggested that the active site was
more flexible than the rest of the enzyme, and that the active site reached flexibility levels

incompatible with catalysis with increasing temperature, while the rest of the enzyme was

still folded.

We observed a linear relationship between Tm and GdnHCl concentration
(Figure II-4A). It is interesting to note that the Tm values obtained in the absence of
GdnHCl correspond to the Tm values predicted by extrapolation from the melting data
obtained in the presence of GdnHCl. Although we cannot, at this time, explain this
correlation thermodynamically, we observed similar linear relationships between T.n
values and GdnHCl concentration for the Thermoanaerobacter ethanolicus secondary
alcohol dehydrogenase (42) and for the T. neapolitana xylose isomerase (V.
Tchemajenko, C. Vieille and JG. Zeikus, unpublished work). It is not surprising to
observe a higher GdnHCl concentration for the midpoint of the unfolding transition of
TNAK than for ECAK and BSAK. Although an increase in GdnHCl concentration at the

midpoint of the unfolding transition of the enzyme is not an absolute rule for

69

hyperthermophilic proteins, similar observations have been made for many of them

(43,44).

TNAK is kinetically unstable at pH values below 5.0. This result is different from
that observed with the E. coli and chicken AKs, which are stable under acidic conditions.
This property has been used for their purification: a crude cell extract is acidified,
neutralized, and then denatured proteins are removed by centrifugation (17, 45, 46). It
would be interesting to determine whether the acid-instability behavior of TNAK is due

to its thermostability requirements.

The native Paracoccus denitrificans AK and TMAK apparently do not contain
any an+ (12). To determine whether these two enzymes contained any Zn“,
P. denitrificans and T. maritima were grown in the presence of 65ZnClz; the AKs were
purified from the bacterial crude extracts and tested for radioactivity (12). At no point
have these two enzymes been tested for the presence of a metal other than Zn”. Since the
recombinant P. denitrificans AK expressed in E. coli contains an equal amount of Fe2+
and Zn2+ (31), we cannot exclude the possibility that the native P. denitrificans AK
contains Fe2+ rather than Zn”. The same possibility exists for the native TMAK.
Purification and characterization of these two native enzymes would be necessary to
determine whether they contain any metal at all. In any case, with a Tm of 93.5 °C, the
apo-TNAK remains stable and active at the temperatures compatible with T. neapolitana
growth (i.e. 75—80 °C). Thus although Zn2+ stabilizes TNAK, its presence in the enzyme

is not absolutely required for TNAK’s function in vivo.

70

As observed with Pyrococcus furiosus a-amylase (47), removing metals from
hyperthermophilic enzymes often requires harsh treatments. Here, a 1 h EDTA treatment
at 80 °C is not totally efficient in removing Zn2+ from TNAK. But in contrast with
P. furiosus a—amylase, the Zn2+-liganding cysteine residues of TNAK are accessible for

equimolar reaction with PMPS at 25 °C.

71

2.6 References

1 Jaenicke, R. 1991 Protein stability and molecular adaptation to extreme conditions.
Eur. J. Biochem. 2022715—728

2 D’Auria,S., Herman, P., Lakowicz, J. R., Tanfani, F., Bertoli, E., Manco, G. and
Rossi, M. 2000 The esterase from the thermophilic eubacterium Bacillus
acidocaldarius: structural—functional relationship and comparison with the esterase
from the hyperthermophilic archaeon Archaeoglobusfulgidus. Proteins 40:473—481

3 Manco, G., Giosu‘e, E., D’Auria, S., Herman, P., Carrea, G. and Rossi, M. 2000
Cloning, overexpression, and properties of a new thermophilic and thermostable
esterase with sequence similarity to hormone-sensitive lipase subfamily from the
archaeon Archaeoglobus fulgidus. Arch. Biochem. Biophys. 373:182—192

4 Wrba, A., Schweiger, A., Schultes, V., Jaenicke, R. and Zavodsky, P. 1990
Extremely thermostable D-glyceraldehyde-3-phosphate dehydrogenase from the
eubacterium Thermotoga maritima. Biochemistry 29:7584—7592

5 Lazaridis, T., Lee, I. and Karplus, M. 1997 Dynamics and unfolding pathways of a
hyperthermophilic and a mesophilic rubredoxin. Protein Sci. 622589—2605

6 Glaser, M., Nulty, W. and Vagelos, P. R. 1975 Role of adenylate kinase in the
regulation of macromolecular biosynthesis in a putative mutant of Escherichia coli
defective in membrane phospholipid biosynthesis. J. Bacteriol. 123:128—136

7 Yan, H. and Tsai, M. D. 1999 Nucleoside monophosphate kinases: structure,
mechanism, and substrate specificity. Adv. Enzymol. Relat. Areas Mol. Biol. 73: 103—
134

8 Rusnak, P., Haney, P. and Konisky, J. 1995 The adenylate kinases from a
mesophilic and three thermophilic methanogenic members of the Archaea. J.
Bacteriol. 177:2977—2981

9 Ferber, D. M., Haney, P. J., Berk, H., Lynn, D. and Konisky, J. 1997 The
adenylate kinase genes of M. voltae, M. thennolithotrophicus, M. jannaschii, and M.
igneus define a new family of adenylate kinases. Gene 185:239—244

10 M'uller, C. W. and Schulz, G. E. 1992 Structure of the complex between adenylate
kinase from Escherichia coli and the inhibitor Ap5A refined at 1.9 A resolution. A
model for a catalytic transition state. J. Mol. Biol. 224: 159-177

11 Berry, M. B. and Phillips, G. N. J. 1998 Crystal structures of Bacillus
stearothermophilus adenylate kinase with bound Ap5A, Mg2+ Ap5A, and Mn2+ Ap5A
reveal an intermediate lid position and six coordinate octahedral geometry for bound
Mg2+ and Mn”. Proteins: Struct. Function Genet. 32:276—288

72

12

13

14

15

16

17

18

19

20

21

22

23

Gilles, A. M., Glaser, P., Perrier, V., Meier, A., Longin, R., Sebald, M., Maignan,
L., Pistotnik, E. and Barzu, O. 1994 Zinc, a structural component of adenylate
kinases from Gram-positive bacteria. J. Bacteriol. 176:520—523

Kath, T., Schmid, R. and Sch"afer, G. 1993 Identification, cloning, and expression
of the gene for adenylate kinase from the therrnoaeidophilic arehaebacterium
Sulfolobus acidocaldarius. Arch. Biochem. Biophys. 307 :405—410

Vonrhein, C., B"onisch, H., Sch"afer, G. and Schulz, G. E. 1998 The structure of a
trimeric archaeal adenylate kinase. J. Mol. Biol. 282: 167—179

Munier-Lehmann, H., Burlaeu-Miron, S., Craescu, C. T., Mantsch, H. H. and
Schultz, C. P. 1999 A new subfamily of short bacterial adenylate kinases with the

Mycobacterium tuberculosis enzyme as a model: a predictive and experimental study.
Proteins 36:238-248

Reinstein, J ., Vetter, I. R., Schlichting, I., Rosch, P., Wittinghofer, A. and Goody,
R. S. 1990 Fluorescence and NMR investigations on the ligand binding properties of
adenylate kinases. Biochemistry 29:7440—7450

Zhang, Y. L., Zhou, J. M. and Tsou, C. L. 1993 Inactivation precedes conformation
change during thermal denaturation of adenylate kinase. Biochim. Biophys. Acta
1164:61—67

Zhang, Y. L., Zhou, J. M. and Tsou, C. L. 1996 Sequential unfolding of adenylate
kinase during denaturation by guanidine hydrochloride. Biochim. Biophys. Acta
1295:239—244

Perrier, V., Burlacu-Miron, S., Bourgeois, S., Surewicz, W. K. and Gilles, A.
1998 Genetically engineered zinc-chelating adenylate kinase from Escherichia coli
with enhanced thermal stability. J. Biol. Chem. 273: 19097—19101

Perrier, V., Surewicz, W. K., Glaser, P., Martineau, L., Craescu, C. T., Fabian,
H., Mantsch, H. H., B‘arzu, O. and Gilles, A. M. 1994 Zinc chelation and structural
stability of adenylate kinase from Bacillus subtilis. Biochemistry 33:9960—9967

Belkin, 8., Wirsen, C. O. and Jannasch, H. W. 1986 A new sulfur-reducing,
extremely thermophilic eubacterium from a submarine thermal vent. Appl. Environ.
Microbiol. 51:1180—1185

Haase, G. H., Brune, M., Reinstein, J., Pai, E. F., Pingoud, A. and Wittinghofer,
A. 1989 Adenylate kinases from thermosensitive Escherichia coli strains. J. Mol.
Biol. 207:151—162

Dong, G., Vieille, C., Savchenko, A. and Zeikus, J. G. 1997 Cloning, sequencing,
and expression of the gene encoding extracellular a-amylase from Pyrococcus

furiosus and biochemical characterization of the recombinant enzyme. Appl. Environ.
Microbiol. 63:3569—3576

73

24 Goldberg, J. B. and Ohman, D. E. 1984 Cloning and expression in Pseudomonas
aeruginasa of a gene involved in the production of alginate. J. Bacteriol. 158:1115—
1121

25 Sambrook, J., Fritsch, E. F. and Maniatis, T. 1989 Molecular Cloning: A
Laboratory Manual, 2nd edn, Cold Spring Harbor Laboratory Press, Cold Spring
Harbor, NY

26 Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith,
J. A. and Struhl, K. 1993 Current Protocols in Molecular Biology, Greene
Publishing and Wiley-Interscience, New York

27 Devereux, J., Haeberli, P. and Smithies, O. 1984 A comprehensive set of sequence
analysis programs for the VAX. Nucleic Acids Res. 12:387—395

28 Dawson, R. M., Elliott, D. C., Elliott, W. H. and Jones, K. M. 1969 Data for
Biochemical Research, 2nd edn, Oxford University Press, Oxford

29 Laemmli, U. K. 1970 Cleavage of structural proteins during the assembly of the head
of bacteriophage T4. Nature (London) 227 :680—685

30 Glaser, P., Presecan, E., Delepierre, M., Surewicz, W. K., Mantsch, H. H., Barzu,
O. and Gilles, A. M. 1992 Zinc, a novel structural element found in the family of
bacterial adenylate kinases. Biochemistry 31:3038—3043

31 Perrier, V., Burlacu-Miron, S., Boussac, A., Meier, A. and Gilles, A. M. 1998
Metal chelating properties of adenylate kinase from Paracoccus denitrificans. Protein
Engin. 11:917—923

32 Miura, K., Inouye, S., Sakai, K., Takaoka, H., Kishi, F., Tabuchi, M., Tanaka,
T., Matsumoto, H., Shirai, M., Nakazawa, T. et a]. 2001 Cloning and

characterization of adenylate kinase from Chlamydia pneumoniae. J. Biol. Chem.
276: 13490—13498

33 Hunt, J. B., Neece, S. H. and Ginsburg, A. 1985 The use of 4-(2-pyridylazo)
resorcinol in studies of zinc release from Escherichia coli aspartate
transcarbamoylase. Anal. Biochem. 146: 150—157

34 Hess, D., Kruger, K., Knappik, A., Palm, P. and Hensel, R. 1995 Dimeric 3-
phosphoglycerate kinases from hyperthermophilic archaea. Cloning, sequencing and
expression of the 3-phosphoglycerate kinase gene of Pyrococcus woesei in
Escherichia coli and characterization of the protein. Structural and functional

comparison with the 3-phosphoglycerate kinase of Methanothermus fervidus. Eur. J.
Biochem. 233:227—237

35 Voorhorst, W. G., Eggen, R. 1., Luesink, E. J. and De Vos, W. M. 1995
Characterization of the celB gene coding for fl-glucosidase from the
hyperthermophilic archaeon Pyrococcus furiosus and its expression and site-directed
mutation in Escherichia coli. J. Bacteriol. 177:7105-7111

74

36

37

38

39

40

41

42

43

44

45

46

47

Dams, T., Bohm, G., Auerbach, G., Bader, G., Schurig, H. and Jaenicke, R. 1998

Homo-dimerie recombinant dihydrofolate reductase from Thermotoga maritima
shows extreme intrinsic stability. Biol. Chem. 379:367—371

Nelson, K. E., Clayton, R. A., Gill, S. R., Gwinn, M. L., Dodson, R. J., Haft, D.
H., Hickey, E. K., Peterson, J. D., Nelson, W. C., Ketchum, K. A. et a1. 1999
Evidence for lateral gene transfer between Archaea and bacteria from genome
sequence of Thermotoga maritima. Nature (London) 399:323—329

Klenk, H. P., Clayton, R. A., Tomb, J. F., White, 0., Nelson, K. E., Ketchum, K.
A., Dodson, R. J., Gwinn, M., Hickey, E. K., Peterson, J. D. et a1. 1997 The
complete genome sequence of the hyperthermophilic, sulphate-reducing archaeon
Archaeoglobusfulgidus. Nature (London) 390:364—370

Ichikawa, J. K. and Clarke, S. 1998 A highly active protein repair enzyme from an
extreme thermophile: the L—isoaspartyl methyltransferase from Thermotoga maritima.
Arch. Biochem. Biophys. 358:222—231

Merz, A., Kn"ochel, T., Jansonius, J. N. and Kirschner, K. 1999 The
Hypertherrnostable indoleglycerol phosphate synthase from Thermotoga maritima is

destabilized by mutational disruption of two solvent-exposed salt bridges. J. Mol.
Biol. 288:753—763

Sterner, R., Kleemann, G. R., Szadkowski, H., Lustig, A., Hennig, M. and
Kirschner, K. 1996 Phosphoribosyl anthranilate isomerase from Thermotoga
maritima is an extremely stable and active homodimer. Protein Sci. 5:2000—2008

Burdette, D. S., Tchemajencko, V. and Zeikus, J. G. 2000 Effect of thermal and
chemical denaturants on Thermoanaerobacter ethanolicus secondary-alcohol
dehydrogenase stability and activity. Enzyme Microbial Technol. 27 :1 1—18

Jaenicke, R. 1991 Protein stability and protein folding. Ciba Found. Symp. 161:206—
216; discussion 217—221

Beaucamp, N., Schurig, H. and Jaenicke, R. 1997 The PGK—TIM fusion protein
from Thermotoga maritima and its constituent parts are intrinsically stable and fold
independently. Biol. Chem. 378:679—685

Heil, A., Muller, G., Noda, L., Pinder, T., Schirmer, H., Schirmer, I. and von
Zabern, I. 1974 The amino—acid sequence of sarcine adenylate kinase from skeletal
muscle. Eur. J. Biochem. 43: 131—144

Tagaya, M., Yagami, T., Noumi, T., Futai, M., Kishi, F., Nakazawa, A. and
Fukui, T. 1989 Site-directed mutagenesis of Prol7 located in the glycine-rich region
of adenylate kinase. J. Biol. Chem. 264:990—994

Savchenko, A., Vieille, C., Kang, S. and Zeikus, J. G. 2002 Pyrococcus furiosus a-
amylase is stabilized by calcium and zinc. Biochemistry 41:6193—6201

75

Chapter 3

Backbone dynamics in a hyperthermophilic

adenylate kinase that is highly active at 30 °C

The research in Chapter 3 is presented in the form of a manuscript to be submitted soon.

76

3.1 Introduction

Hyperthermophilic archaea and bacteria thrive at temperatures above 80 °C. This
adaptation implies that enzymes from these organisms are stable and functional at high
temperatures. Indeed, enzymes from hyperthermOphiles have developed unique
structure-function properties of high stability and optimal activity at temperatures above
70 °C. Hyperthermophilic proteins have been intensively studied for over 30 years to
understand their stability mechanisms. Such understanding is required both to gain
knowledge of the physical principles underlying protein thermostability and because of
thermostable enzymes’ potential biotechnological applications. Therrnostable enzymes
are useful in industrial applications such as food and starch processing, detergent

manufacturing, production of high fructose corn syrup, and organic syntheses (48).

Identifying the factors contributing to the enhanced thermostability of proteins
from hyperthermophiles has been a long standing problem, and many studies have been
directed towards comparing pairs of hyperthermophilic and mesophilic homologous
proteins. Several mechanisms have been proposed to contribute to the greater stability of
hyperthermophilic proteins (49), the most cited ones being enhanced hydrophobicity (18),
better packing, shortening or deletion of loops (38), increased occurrence of proline
residues (18, 52), increased helical content, and increased number of hydrogen bonds
(H-bonds) and salt bridges (27, 38, 50, 55, 57). Among these, the only consistent trend is
shown by salt bridges and hydrogen bonds, whose numbers increase in a majority of
hyperthermophilic proteins (27). An indirect evidence for this trend is the finding that

proteins encoded by known hyperthermophile genomes contain significantly larger

77

numbers of charged residues than the proteins encoded by mesophile genomes. The
current understanding, though, is that a unifying mechanism for thermostability does not
exist (25), and that different hyperthermophilic proteins use different thermal adaptation

strategies.

Many hyperthermophilic enzymes are optimally active at elevated temperatures, and
they are often inactive at low temperatures (i.e., around 20 °C — 37 °C) (9, 54, 58). There
is also a surprising constancy in the catalytic rates of homologous mesophilic,
thermophilic, and hyperthermophilic enzymes at the respective optimal growth
temperatures of their source organisms. In other words, they maintain “corresponding
states” with respect to their activity, topology, solvation, etc. in their respective
physiological conditions (43). The reduced activity of hyperthermophilic enzymes at low
temperatures appears to support the hypothesis that hyperthermophilic proteins are more
rigid than their mesophilic homologues at ambient temperatures. It is thought that
enzyme motions required for activity become too slow and too restrained in
hyperthermophilic enzymes at low temperatures, and that these proteins gain the
flexibility required for optimal activity only at higher temperatures. Lack of activity at
low temperatures need not be a consequence of stability at high temperatures, though, as
shown by a number of laboratory-evolved thermostable enzymes that are catalytically
efficient at low temperatures (15). Also, there is no fundamental reason for linking
stability and rigidity, since increased flexibility should lead to thermodynamic

stabilization through increased conformational entropy of the folded state (28).

The dynamic properties of several hypertherrnophilie-mesophilic enzyme pairs have

been compared using a variety of techniques such as Fourier transform Infrared

78

Spectroscopy (58), fluorescence anisotropic decay (31), hydrogen-deuterium (H/D)
exchange (21), neutron scattering (13, 14), NMR spectroscopy (5, 20), and molecular
dynamics simulations (8, 16, 28). While several of these comparisons have found
evidence supporting the hypothesis that hyperthermophilic enzymes are more rigid than
their mesophilic counterparts (23, 28, 44, 46, 58), others have not (8, 12, 16, 20). Several
of the techniques mentioned above are limited by their ability to measure only average
global dynamics. Enzyme activity and stability involve different structural regions and
different timescales. Hence, flexibility comparisons in hyperthermophilic-mesophilic
enzyme pairs should use site-specific measures of dynamics spanning a wide timescale
range. The limited amount of experimental data plus the scarcity of residue-level data
available today do not yet allow us to clearly determine whether rigidity at mesophilic
temperatures is a key factor in protein thermostability. The time scale and amplitude of

the conformational fluctuations that might lead to low thermostability are also not clear.

The purpose of this study is to investigate the relationships between
conformational flexibility, thermostability, and enzyme activity by comparing the
backbone amide motions of adenylate kinases (AK) from the mesophile Escherichia coli
(ECAK) and from the hyperthermophile Thermotoga neapolitana (TNAK) using l5N
NMR relaxation. We chose to study AK because it is monomeric, and because its small
size (i.e., 25 kDa) renders it accessible to NMR and molecular dynamics studies.
Backbone dynamics of both the ligand-free and ligand-bound forms of ECAK studied

with ISN NMR relaxation have already been reported (41, 42, 47).

AK is a ubiquitous, multi-domain enzyme that catalyzes the reversible transfer of

the y-phosphate group from ATP to AMP to form ADP. AKs have been extremely well

79

studied both biochemically and structurally with close to 20 crystal structures of various
AKs present in the protein data bank. The crystal structure of ECAK shows three
domains: a core, a lid, and an AMP-binding (AMPbd) domain (34, 35). A conserved
loop, termed the P-loop, essential for nucleotide binding, is also present. In the absence of
any ligands, AK adopts an open form. Based on the structures of ligand-free AKs and
those of several AKs complexed with nucleoside monophosphates, nucleoside
triphosphates, and inhibitors, it has been inferred that the lid and AMPbd domains
undergo large conformational changes upon binding of the substrates, AMP, ATP, and of
the cofactor Mg“. The enzyme ternary complex adopts a closed form in which the
AMPbd domain closes over AMP and the lid closes over ATP, thereby expelling waters

to prevent ATP and AMP hydrolyses (40).

TN AK is 40% identical to ECAK, with most of the ECAK ATP and AMP ligands
(Figure III-1) being conserved in TNAK (35). TNAK also shares high sequence identity
with several other bacterial AKs whose structures with various ligands are known
(Bacillus subtilis AK [BSubAK, 46%], Bacillus stearothermophilus AK [BStAK, 45%],
and Bacillus globisparus AK [BGAK, 43%]) (2, 3). Biochemical characterization of
TNAK showed that it is optimally active at 80°C, and that it is a highly thermostable
enzyme with a melting temperature (Tm) of 99.1 °C (48), more than 47 °C higher than the
Tm of 51.8 °C measured for ECAK (37). To our surprise and in contrast to all other
characterized hyperthermophilic enzymes, the kinetic parameters of TNAK at 30 °C are

identical to those of ECAK (48). Figure III-2, showing the kcm values of ECAK, TNAK,

80

 

 

 

 

E -@re21re31
20 30

 

 

 

 

l 10 4O 50

ECAK —-MRIILLGAPGAGKGTQAQFIMEKYGIPQISTGDMLRAAVK-SGSELGKQAKDIMDAGK
TNAK MMAYLVFLGPPGAGKGTYAKRIQEKTGIPHISTGDIFRDIVKKENDELGKKIKEIMEKGE
1 a4 1 CD U

60 7O 80 90 100 110
ECAK LVTDELVIALVKERIAQEDCRNGFLLDGFPRTIPQADAMK----EAGINVDYVLEFDVPD
TNAK LVPDELVNEVVKRRLSEKDCEKGFILDGYPRTVAQAEFLDSFLESQNKQLTAAVLFDVPE
351$) U33) [13.7 [E9 1 a7 1

120 130 140 150 160 170

ECAK ELIVDRIVGRRVHAPSGRVYHVKFNPPKVEGKDDVTGEELTTRKDDQEETVRKRLVEYHQ
TNAK DVVVQRLTSRRICPKCGRIYNMISLPPKEDELCDDCKVKLVQRDDDKEETVRHRYKVYLE

Cl [ a8 1 CE 1 019 j

 

 

180 190 200 210
ECAK MTAPLIGYYSKEAEAGNTKYAKVDGTKPVAEVRADLEKILG----
TNAK KTQPVIDYYGKKGILKR ----- VDGTIGIDNVVAEVLKIIGWSDK

Figure III-1: Alignment of TNAK and ECAK sequences using ClustalW. Helices (bars)

and B-strands (arrows) based on the ECAK crystal structure and TNAK 3D model are
indicated above the sequence.

81

4500

 

 

 

4000 4

3500 -

 

3000 n

2500 4

kcat (3")

2000 -
.500 . AAAK
1000 - o 0 o 000

500~ o I o o

 

0 20 4O 60 80 100 120

Temperature (°C)

Figure III-2 Activity profiles of ECAK and AAAK (data kindly provided by Wolf-
Watz, M and Kern, D) and kcm values of TNAK at 30 °C and 80 °C (A).

82

and the AK from another hyperthermophile, Aquifex aeolicus (AAAK) at various
temperatures, highlights TNAK’s unusual behavior. l5N NMR dispersion experiments on
AAAK and ECAK have shown the slower lid-opening rate of AAAK (compared to
ECAK) to be responsible for its low activity at 20 °C (53). TNAK’s intriguing catalytic
properties make the comparison of TNAK and ECAK in terms of flexibility, activity, and
thermostability particularly interesting and novel. This comparison will also be the third
leg in the triangular comparison involving ECAK and AAAK. Using lsN NMR
relaxation, we address the question: Do the dynamics of TNAK at 30 °C explain in any
way why this enzyme is active at 30 °C? If the hypothesis that enhanced thermostability
is attained by enhanced structural rigidity is true, then one expects TNAK to present
globally enhanced rigidity with regions of local flexibility that allow activity at low

temperature S .

In the present work, we report the resonance assignments for TNAK in the free and
inhibitor-bound forms. We investigate the motional properties of TNAK backbone
amides in solution on multiple time scales, from ps to us. Dynamics parameters of TNAK
free and inhibitor-bound forms are obtained using 15 N relaxation data measured at 30 °C.
Results for TNAK are compared to the 30 °C 15N relaxation data for the free and

inhibitor-bound forms of ECAK (42).

83

3.2 Materials and Methods

3.2.1 Sample preparation

Uniformly labeled recombinant TNAK was overexpressed in E. coli BL21(DE3) cells
transformed with pTNAK2:sz (48). Cells were grown in M9 minimal medium (39)
containing (i) 1 g/l of > 98% [UM-ammonium chloride and 2 g/l of 99%
[13C6]-D-glucose for assignments and (ii) 1 g/l of > 98% [ISM-ammonium chloride and
unlabeled glucose for relaxation measurements. Protein expression was induced at an
OD600nm of 1.0-1.2 with 0.4 mM isopropyl-B-d-thiogalactopyranoside for 20 h at 37 °C.
The previously reported protocol for TNAK purification (48) was modified as described
below. Supernatant from the heat-treated soluble extract was fractionated with
ammonium sulfate at 50% saturation to precipitate impurities. The supernatant was
dialyzed against 20 mM Tris-HCl (pH 8.6) (buffer A) and loaded on a Q-sepharose Fast
Flow (Amersham Biosciences, Uppsala, Sweden) anion exchange column (2.6 x 30 cm,
flow rate 2 mllmin) pre-equilibrated with buffer A. After washing the column with five
volumes of buffer A, TNAK was eluted with a 400 ml linear 0.2—0.7 M NaCl gradient in
buffer A. Fractions containing TNAK were pooled, dialyzed against 20 mM Tris-HCl
(pH 8.0) (buffer B) and loaded again on the Q-sepharose column pre-equilibrated with
buffer B. TNAK was eluted with a 007 M NaCl gradient in 50 mM Tris-HCl, pH 7.4.
Purified TNAK was concentrated in an ultrafiltration cell equipped with a 10,000
molecular weight cut-off membrane (Amicon, Beverly, MA), dialyzed against milliQ

water, freeze-dried, and stored at -20 °C.

84

NMR samples of TNAK in the ligand-free form were prepared by dissolving an
appropriate amount of freeze-dried TNAK in 320 pl of 50 mM sodium phosphate buffer
(pH 7.0) containing 10 11M sodium azide, 6% DzO/94% H20, and protease inhibitor
cocktail (Roche, Indianapolis, IN). TNAK concentrations were 1.8 mM for assignment
experiments and 1.5 mM for relaxation experiments. For the inhibitor-bound form of the
enzyme, TNAK was dissolved in the same buffer, containing in addition, 8 mM MgClz
and an appropriate amount of the inhibitor P', PS—di(adenosine-5’) pentaphosphate
(Ap5A). Final concentrations of TNAK and Ap5A were 1.8 mM and 5 mM

(assignments) and 1.5 mM and 3 mM (relaxation), respectively.
3.2.2 NMR spectroscopy
Sequential assignments

Triple-resonance data were collected at 30 °C on a Varian Unity Inova 600 spectrometer
equipped with triple-resonance probes at 14.1 T. 15N-HSQC, gradient sensitivity-
enhanced triple resonance CBCA(CO)NH, 3D HNCACB, HNCA, and 15N—edited
NOESY-HSQC data were collected using established pulse sequences (17, 33, 56). Data
were processed using NMRPipe (11) and NMRView (24). A Gaussian window function
in the direct dimension and a sine window function in the two indirect dimensions were

applied. Linear prediction was used for the indirect evolution time periods.
Relaxation measurements

2D sensitivity-enhanced NMR experiments were used to measure l5N longitudinal

relaxation rates (R1), transverse relaxation rates (R2), and heteronuclear 15N-{lH}

85

steady-state NOEs at 30 °C and 14.1T (26). The T1, T2, and NOE data were acquired with
spectral widths of 2200 x 8000 Hz (F 1 x F2). The number of complex points was
160x2048 (t, x t2) for T1 and NOE and 132 x 2048 for T2. T1, T2, and NOE
measurements were performed with 16, 32, and 64 transients per :1 experiment,
respectively. Relaxation delays for R1 experiments ranged from 0 to 1.8 ms and included
nine unique time points. The experiments for three time points were repeated twice for
error estimation in the measured peak heights. For R2 measurements, nine unique time
points were used; with parametric delays ranging from 0 to 170 ms. Experiments were
repeated twice for three of the time points. NOE spectra were recorded with a 3-s
pre-delay for the proton-saturated spectrum and 6.5 s for the proton-unsaturated
spectrum. A 3.5-s saturation period was used for both TNAK and TNAK*Ap5A. All

spectra were processed using NMRPipe (11) and NMRView (24).

Data processing

Resonance signal intensities were quantified using NMRView. R1 and R2 relaxation rates
were obtained by fitting cross-peak intensities as a function of relaxation delay time to a
two-parameter mono-exponential decay function using the Levenberg-Marquardt non-
linear least squares fitting software CurveFit (A.G. Palmer, Columbia University).
Uncertainties in the fitted relaxation rates were estimated using a jackknife algorithm.
Experimental peak uncertainties were determined from the duplicate time points. Steady-
state NOEs were calculated as the ratio of cross peak heights of the saturated and
unsaturated protons. The average percentage errors for TNAK’s R1, R2, and NOE were
2.7%, 1.6%, and 5%, respectively. The average percentage errors for TNAK*ApSA’s R1,

R2, and NOE were 2.6%, 1.1%, and 4.6%, respectively.

86

Data analysis

The relaxation rates and NOEs were fitted to the Lipari-Szabo model-free formalism (6,
7, 29, 30) using the ModelFree 4.0 software (32, 36). Inertial tensor analyses of TNAK
and TNAK*Ap5A were performed using TNAK and TNAK*Ap5A homology models
(see Structure coordinates) as input PDB files. Hydrogen atoms were added using
insightII. The inertia tensor calculations were performed using PDBINERTIA (from the
ModelFree 4.0 package [32, 36]), and the hydrodynamic calculations were performed

using the HYDRO suite of programs (10).

The R2/R1 ratios of a subset of rigid NH vectors (having negligible components of
internal motions and/or exchange contributions to the 15N relaxation rates) were selected
for calculating the principal components and the orientation of the diffusion tensors of

TNAK and TNAK*Ap5A. Residues were selected if their NOE values were > 0.7 and
(<T2>'T2.n)/<T2>'(<T1>'Tl,n)/<Tl>< 1-5 SD (1)

where (T2) is average T2 and T2, n is the T2 value of residue n. SD is given as the standard
deviation of the function ((T2) - T2, n) / (T2) - ((Tl) - T1, n) / (T1) (45). Components of the
overall rotational diffusion tensor were determined using local diffusion approximation as
implemented in the Quadric Diffusion software by Palmer et a1 (32, 36). To choose the
best-fit model to describe the molecular rotational diffusion, F statistic testing was used

to compare isotropic, axially symmetric, and anisotropic diffusion models.

Following model selection, individual NH data were fitted to each of the

following five motional models as described (32). In the models described below, the

87

square of the generalized order parameter S2 is given as S2 = SfZSsz, where sz refers to
the amplitude of motions in the fast timescale (the corresponding internal correlation time
tf< 200 ps) and 8,2 refers to amplitude of motions on the slow timescale with an internal
correlation time I, (I, > ~200 ps). Thus, motions characterized by S2 encompass ps-ns
timescale motions. The value of S2 ranges from 0 to l, 0 indicating high flexibility of the
NH vector and 1 indicating a completely restricted NH vector. Model 1 assumes that
slow internal motions are negligible (S,2 = l) and that fast internal motions are very fast
(< 20 ps). These assumptions leave only the order parameter S2 to describe motions on
the fast time scale. Model 2 assumes only that the slow internal motions are slow (S,2 =
1). This assumption gives two parameters to be derived, namely 82 and an effective
internal correlation time, 1., (= If), for fast motions. Model 3 yields, in addition to $2, a R.,,
terms that describes us-ms time scale exchange contributions to R2. Model 4 yields S2
and 1., for fast internal motions, as well as Rex. Model 5 assumes only that If tends to 0,
which allows order parameters on two timescales: sz for fast internal motions and S,2
and the corresponding ‘1:5 for internal motions slower than 500 ps but faster than the

overall correlation time “cc. For models 1 through 4, 82 = sz (S,2 =1).

The difference between parallel and perpendicular components of the
l5N chemical shift tensor (A0”) was taken to be -l60 ppm, and the N-H internuclear
distance was assumed to be 1.02 A (22). Fitting procedures, model selection criteria, and
optimization procedures were performed as described (32). A model was selected by
comparing the value of the x2 function at a critical value, CL, of 0.1 with the simulated

critical value. F-statistical testing (with OL = 0.2) was used to determine if the more

88

complicated model fitted the data better for each spin. Monte Carlo simulations were
performed using 500 randomly distributed synthetic data to determine uncertainties in the
fitted modelfree parameters. The internal modelfree parameters for each residue were
further optimized simultaneously with the global correlation time. The procedure was

iterated till the results converged.

Structure coordinates

A 3D structure of TNAK is not available at this time. A comparison of bacterial AK
structures (i.e., ECAK*Ap5A, BGAK*Ap5A, BsubAK*Ap5A, and BstAK*Ap5A)
shows that the AK fold is highly conserved. The BGAK*Ap5A, BSubAK*Ap5A, and
BStAK*Ap5A structures are highly similar to each other (RMSD is within 1.6 A) while
they differ from the ECAK structure mainly in the position of the lid domain and in the
C-terminus. The lid domain is involved in opening and closing motions during catalysis
and is therefore, likely to be mobile in solution. The C-terminus in ECAK*ApSA consists
of an a—helix while in BGAK*Ap5A, BStAK*Ap5A, and BSubAK*Ap5A, the helix is

followed by a short stretch of residues that do not have a regular secondary structure.

With 40% sequence identity or more, homology modeling can produce structures
that are in many respects equivalent to a medium resolution crystal structure. 3D models
of TNAK in both the open and closed forms were built by comparative modeling using
the MODELER software (Figure III-3) (l). TNAK*Ap5A was modeled using the
structures of BSubAK*ApSA (pdb code 1P3J), BGAK*ApSA (pdb code 183G), and

BStAK*Ap5A (pdb code IZIN) as templates. BGAK, BSubAK, and BStAK share 43%,

89

Figure III-3: Ribbon diagrams of the homology-modeled structures of A)
TNAK (open form) and B) TNAK*Ap5A (closed form). The lid and (AMPbd)
domains and the P-loop are indicated in the open form. The figures were
generated using the program Ribbons. Zn2+ in the lid domain and Ap5A are
shown with a ball and stick representation.

90

Lid

 

91

46%, and 45% identity with TNAK, respectively, and all four AKs contain a structural
zinc in their lid domain. TNAK was modeled using the structures of ECAK (pdb code
4AKE) and BStAK*Ap5A (pdb code IZIN) as templates. ECAK is the only bacterial AK
whose structure is available in the open form. Because ECAK does not have the

structural zinc in the lid, BSAK*Ap5A was used to model only the lid domain of TNAK.

The 3D structure of AK consists of three domains and two pairs of bending
regions, one between AMPbd domain and core (i.e., core-AMPbd hinges) and the other
between lid and core (i.e., core-lid hinges). In ECAK, the AMPbd domain contains
residues 31-72, the lid contains residues 119-156 and the core is formed by residues 1-28,
80-112, and 173-214. The core-AMPbd hinges involve residues 29, 30, and 73-79 and the
core-lid hinges consist of residues 113-118 and 157-172 (19). The corresponding residues
in TNAK domains are (1) core: 1-30, 83-119, and 180-220 (ii) AMPbd domain: 33-75
(iii) lid: 126-163 (iv) core-AMPbd hinges: 31, 32, and 76-82; and (v) core-lid hinges:

120-125 and 163-179.

3.3 Results

3.3.1 Resonance assignments of TNAK

TNAK contains 220 residues, 10 of which are prolines that are not detectable in a 2D

lH-ISN HSQC (Figure HI-4A). 1H and 15N backbone resonances of 197 out of the 210

92

Figure III-4A: 600 MHz 'H-'5N HSQC spectra of TNAK at 30 °C

93

wdd N51

9:
an—
mm—
S.
cm—
mN.
a.
mm—
mm—
.3
cm—
a:
w:
:—
o:
n:
v:
m:
N:
:—
o:
2:
we.
2:
we.
mo—
3:

 

 

 

 

 

  

Eng 21
o n w o e—
il? p p p p p p p n b b > r - bl P L n > n P P u b p u h p p p p p
h 0
._ .m
.. on O
A .
1. 2. m: 0 S.
p MN
2: 0
t. 0 1%: no
. 3.. an
. sue a a 03:
t. :. E “W .0 m2 0
. 3 v: w or:
t we.
1 0
. _ _
i. .208 «N. e e
. mm a. no no
8. ’9 9
n2 0 $—
4 .m- o
n
w
d 5H— m
1 o o
. an R .8
4 0 Q
0 can so.
J . hm—
J
t o
h ; QM \
. 0 0 0
1.. we 3. c—
.. O
. o
1 m _ oom

 

 

 

94

non-proline residues were assigned in ligand-free TNAK. Five residues in the P-loop,
three in the lid domain, N-terminal residues M1 and M2, and C-terrninal K220 are among
the 13 residues whose correlations could not be detected in the HSQC spectrum. Among
the 197 residues with detectable resonances, eight pairs of residues (i.e., L5 and Q122,
Y178 and H30, 167 and V55, I22 and K50, A19 and V177, D155 and D187, L66 and
V185, and K58 and D166) generated peaks that overlapped with each other. With
TNAK*Ap5A, 193 peaks were seen in the HSQC spectrum (Figure III-4B) out of the
expected 210. Five of the 17 missing residues are located in the P-loop, two are the N-
terminal residues M1 and M2, and two are the C-terminal residues D219 and K220.
Three correlations from residues in the lid domain and four from residues in the AMPbd
domain are also missing. Of the 193 correlations that are detected in the HSQC spectrum,
nine pairs of residues (i.e., L75 and R91, L61 and R126, R38 and E57, K72 and L179,
V123 and 8218, K78 and L99, K53 and D117, A94 and 1204, and L48 and W217) and
three triplets of residues (i.e., K43, D187, and K213; K42, V67, and Q95; and H30, R74,
and K176) generated overlapping peaks. TNAK and TNAK*Ap5A HSQC spectra
differed considerably from each other (Figures III-4A and III-4B), underlying the fact
that large conformational changes are induced by inhibitor binding, as has already been
observed for ECAK (42). Strip plots of HNCA spectra showing the sequential walk for

residues 190-197 in TNAK and TNAK*Ap5A are shown in Figure III-4C.

a-helices can be readily detected in a 3D-NOESY-HSQC spectrum due to strong

NOEs between consecutive amide hydrogens of OL-helices. oc—helices could be easily

95

Figure 111-48: 600 MHz ‘HJSN HSQC spectra of TNAK*Ap5A at 30 °c

96

 

3:
mm—
mm.
.m—
cm—
aw—
wa—
FN—
on—
ma—
wu—
MN.
«N—
_N_
ew—

 

 

» e
.26...- 0. ¢
. at... 9 . a _ z

.- . _ & can.»

.1 .......M-_. . - v _ 0 .00:.

. .- . .r. .. . B

e... new! .92 mm 2. .0
.1.. WW... 5— am \r‘mI.n. 5:: . a MN NZ...
. etfi a N ...m.w... mm 3.... 8.6: 8.

ulAikLAlLl

 

LLLA
V
- 1
Vi '
no
a
11

wdd NSL

a:
w:
:—
o:

w.
4e

we

a.

o n... .S .3.-

89
2

 

0

:—

97

W. a 3:.

n:
v:
m:
N:
:—
o:
8—
we
no—
2:
mo—

AlnlnlnlnlnlnlnlnlnlLlLJ
O
O

o
i
o
8?.

 

SN

«2

l l I I A I I l I

m3 0...

nlnl

 

 

 

$2.922» .23 x<zto sea 9% 62:01: 3:3...

N05. om; mm...— vm: mmpo Nos. 5; capo hmpm ompv— mm: V0: mmpo NmFx P05. oo—U
md m Pd (N OK no.5 Nu m fin 0.0 1m Pd MK PK OK MK Md Cd
_ p _ p _ b p _ _ _ _ _ _ P . _

 

 

 

1‘s

,-
1x
‘0

., I roe

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.Le

 

 

 

<mQ<nx<ZH

98

discerned from the NOESY-HSQC spectra of TNAK and TNAK*Ap5A, and they

correlated well with those predicted by the 3D models.

3.3.2 Relaxation data

Excluding residues with overlapping or weak correlation peaks, a total of 177 residues
were used for TNAK’s relaxation data analysis. For TNAK*Ap5A, 164 residues were
used to the analysis after excluding 27 residues with overlapping correlations and 2
residues with weak correlations. The R1, R2 and 15N-{IH} NOE values of TNAK and
TNAK*Ap5A residues are shown in Figures III-5A, III-5B, and III-5C, respectively. The
10% trimmed mean values of R1, R2, and lsN-{IH} NOE values of TNAK
(TNAK*Ap5A) are 0.93 :1: 0.023 s" (1.01 :1: 0.024 s“), 20.53 1 0.30 5“ (18.32 :1: 0.17 s"),
and 0.87 :1: 0.039 (0.81 :t 0.017), respectively. Low NOE values reflect the presence of
considerable motions on the fast timescale (ps to us). As expected from the movements of
the lid and AMPbd domains triggered by substrates or inhibitor binding in AKs, and as
expected from the low NOE values in the ECAK free enzyme, TNAK's lid and AMPbd
domains show NOE values that are lower than those of the rest of the protein.
Surprisingly, TNAK*ApSA’s lid and AMPbd domains also show low NOE values,
suggesting that fast timescale motions are not suppressed in the lid and AMPbd domains
upon inhibitor binding. This trend in NOE values is unlike that seen for ECAK, where the
NOE values become uniformly high throughout the backbone upon inhibitor binding

(42).

99

Figure III-5: Relaxation parameters of TNAK (1:1) and TNAK*Ap5A (O). (A) R]:
(B) R2; and (C) 15N-{ 1H} NOE. The location of TNAK secondary structures and
domains are indicated on top of the figures. The open rectangles specify the
location of the hinge regions framing the AMPbd and lid domains.

100

‘SN-{1H} NOE

 

 

 

 

 

 

100 150 200

 

 

 

 

 

 

 

 

! . w ‘ i
'n-G-ii i==m n
rll-‘,'l"!r'-'5E -.. I I U L.] .1.-7"- ‘I-P
n.3"III-‘rIi-I ‘ r'lld' 'tIin’inb'r"

.' 1" -L - I'e'"-_‘ ' "'s ‘51-'-

r 3""i f.l ‘— E, 1';-
._. I - .

I turf-1 .- "r - ' 1'

I nt- 1.- i
I' I‘
§ $0. I

I T 1 fi"

 

 

100 150 200

TNAK Residue No.

101

3.3.3 Molecular rotational diffusion from hydrodynamic calculations and NMR

data

The ratios of TNAK’s principal moments were 1.0: 0.88: 0.529. Modeling TNAK as a
prolate ellipsoid yielded the Perrin shape factors Fa: 0.834 and Fc=1.335. The R2/R1
ratios for 112 residues not subject to large amplitude fast internal motions or to slow
timescale conformational exchange (see Materials and Methods) were used to estimate
the components of the rotational diffusion tensor. The axially symmetric model agreed
significantly better (p < 0.025) with the experimental data than the isotropic model did
according to an F-statistic test (F = 18.4). The fitting improvement for the fully
anisotropic model relative to the axially symmetric model is not significant (F = 0.021).
The prolate model was selected as the best fit to the data. After selecting the appropriate
models for the individual NH vectors and optimizing the parameters in modelfree

calculations, the final 1,, was 14.88 :1: 0.021 us with a D11 ID 2 ratio of 1.2 t 0.01. These

values agreed well with the 1,, (15.26 ns) and DH ID 2 (1.26) values predicted from the

structure and molecular weight of TNAK using hydrodynamic calculations.

The inertia tensor for TNAK*Ap5A using the homology-modeled structure has
principal moments with ratio 1: 0.88: 0.75. A rotational correlation time, 1c of 13.16 ns
and a D|| /D2 ratio of 1.1 were calculated for TNAK*Ap5A with a molecular weight
mass of 24.8 kDa using hydrodynamic calculations. R2/R1 data of 110 spins selected
using the criteria described in Materials and Methods were used to estimate the diffusion
tensor parameters. The axially symmetric and the fully anisotropic tensors fitted the data

better than the isotropic tensor, but the two models could not be distinguished statistically

102

 

using an F-test. Hence, the simpler model of axially symmetric oblate model was chosen.

The final 17c and DH ID 2 values after optimizing the modelfree parameters were

13.56 i 0.03 and 0.87 i 0.009, respectively.

One must consider the possibility of protein self association at such high
concentrations (1.8 mM), which can adversely affect the values of 1c and the relaxation
parameters. We do not believe that aggregation occurred in the TNAK and TNAK*ApSa
NMR samples for the following reasons: (i) The TNAK and TNAK*Ap5A Tc values
(14.88 ns and 13.6 ns, respectively) are close to the estimated re based on hydrodynamic
calculations (15.26 ns for TNAK and 13.16 us for TNAK*Ap5A) and agree quite well
with the molecular size and anisotropy of the protein; (ii) similar “cc values were
calculated for ECAK and ECAK*ApSA (15.05 ns and 11.4 ns, respectively) from R2/R1
data (42). Since ECAK and TNAK are close in their molecular size and shape, an
agreement in their to values suggests monomers are predominant in both the TNAK and

TNAK*Ap5A samples.
3.3.4 Backbone internal dynamics

The relaxation parameters of 112, 9, 32, 3, and 21 TNAK residues were fit to motional
models 1, 2, 3, 4, and 5, respectively. Contrast this with ECAK where the majority of the
residues were fit to model 2 (with parameters 82 and re) and not model 1 (only 82),
indicating a higher rigidity for many residues in TNAK. For ECAK, 2, 109, 7, 48, and 21
residues were fit to models 1, 2, 3, 4, and 5, respectively. For TNAK*Ap5A the

relaxation parameters of 84, 27, 35, 3, and 15 residues were fit to models 1, 2, 3, 4, and 5,

103

respectively, while in ECAK*Ap5A, 66, 59, 39, 5, and 30 residues were fit to models 1,

2, 3, 4, and 5, respectively.

Pico -nanosecond timescale dynamics

ECAK and TNAK’s generalized order parameter (52) values are plotted as a function of
residue number in Figure III-6A, while Figure IH-6B compares ECAK*ApSA and
TNAK*ApSA’s S2 values. A consistent increase in S2 values is observed for most TNAK
residues compared to the corresponding ECAK residues indicating that the amplitude of
internal motions in the ps-ns timescale is larger in ECAK than TNAK. In contrast, this
difference in 82 values almost vanishes in the inhibitor-bound forms, ECAK*ApSA and
TNAK*Ap5A. Although the 82 values of TNAK residues are uniformly higher,
individual residues show significant deviation from this trend. Many of these residues are
located in TNAK’s AMPbd and lid domains. Residues in the region between 43 and 62 of
AMPbd have similar S2 values in ECAK and TNAK. The S2 values of several residues in
TNAK’s lid domain are comparable to (i.e., residues 161, 162, 165-169, and 171) or
lower than (i.e., residue 154) the 82 values of the corresponding residues in ECAK.
Residues 98, 109, and 182 in TNAK’s core show lower S2 values than the corresponding
residues in ECAK’s core. Interestingly, these residues are located in the loops of TNAK's
core. One other residue that shows a lower 82 value in TNAK than in ECAK is residue 82

situated in one of the core-lid hinges.

In TNAK*APSA also several residues have lower S2 values than the

corresponding residues in ECAK*ApSA. These residues are 44-46 and 64 in the AMPbd

104

Figure III-6: Order parameters, 82, at 30 °C. (A) TNAK (A) and ECAK (O); and
(B) TNAK*Ap5A (A) and ECAK*ApSA (o). ECAK data (42) are plotted as per
the sequence alignment shown in Figure III-1. TNAK secondary structures and
domains are indicated on top of the figure. The open rectangles specify the
location of the hinge regions framing the AMPbd and lid domains.

105

 

 

 

 

 

1L1

 

 

 

0

l

1144411111111111111114111111llllllllllllllllllll
l

l l I I I I l I I

20 40 60 80 100 120 140 160 180 200 220

 

0.7 -§
0.6 é
0.5 i
0.4 -:

r—

.-

r-

08 _H-
I

r.

)-

 

0.3 :

 

llllll

lllllllllllLllllJlllllllllLllllllllJALiJ lllllllll

e . 10 °'
‘0!

e to O
D: I‘
z:

{I "E

 

 

 

 

4060 80100120140160180200220

TNAK Residue No.

106

domain and residue 166 in the lid domain. In the core domain, residues 83 and 109 have

similarly low 82 values in ECAK*ApSA and TNAK*Ap5A.

Mean S2 values ((82)) for each domain and for each secondary structural element
in ECAK, ECAK*Ap5A, TNAK, and TNAK*Ap5A categorized as belonging to the lid,
AMPbd, or core domains, or to the different hinge regions are listed in Table III-1. A few
salient points that arise from Table HI-l are: (i) the (82) values of the core, AMPbd, and
lid domains are similar to one another in each of the four enzyme forms indicating similar
overall rigidity on the ps—ns timescale across the whole protein; (ii) the % increase in the
overall (S2) values of the three domains in TNAK (compared to ECAK) are similar
(between 7.7% and 9.5%), indicating again a uniform increase in the overall rigidity of
TNAK. In other words, each domain contributes to a similar extent to TNAK’s overall
rigidity; and (iii) in contrast, the overall (82) values of the three TNAK*Ap5A domains
are similar to those of ECAK*Ap5A, indicating similar levels of rigidity on the ps-ns

timescale for both enzymes.

At the level of individual secondary structures, significant differences between
ECAK and TNAK (82) values are seen only for B9 and (15 of the core, for 013 of AMPbd,
for the loop region of the lid, and for a7 (belonging to one of the 2 core-lid hinges). In all
these regions TNAK has a higher (82) value than ECAK. The (82) of ECAK*ApSA’s
AMPbd loops and core helix 0:8 are actually higher than that of the corresponding
regions in TNAK*Ap5A. Loops in the AMPbd domain have amongst the lowest (S2)
values of any secondary structural elements in both TNAK and TNAK*Ap5A. This

statement is true for ECAK as well, but not for ECAK*ApSA. Interestingly, the

107

 

Table III-1: Average S2 values, <Sz>, of the secondary structures in the various
domains of ECAK, TNAK, ECAK*Ap5A, and TNAK*Ap5A.

a

ECAK and ECAK*ApSA data are from (42)
b Standard deviation is in parentheses
C Number of residues included in the average
d Regions defined in ( 19)

6 B7 and B8 are not predicted in the TNAK and TNAK*Ap5A models

108

 

 

 

 

 

 

e 5 .8 888 m 3 .8 888 8. 38.8 .88 8. $88 $8 .3883
N _ A888 88.8 2 No.8 8.8 N. 35.8 88.8 N. 38.8 88.8 to 88.8882
8 88.8 888 8 88.8 888 8 38.8 888 8 38.8 $8 88 88:2 2.28.8
mac—we.— owE:
. . e .
N 88 8 a8 8 38.8 N88 8 £8 8 N88 8 28.8 888 883 382.3728
888 _N88 888 8888 .155
N :88 88 NN 88.8 38 8 :88 88 8N 88.8 888 883
e - N 888 e - N 888 88
e - N 888 e - _ 888 a
N 88 8 28.8 888 N 888 8 $88 $8 88
m 38.8 38 8 88.8 888 N 28.8 38 8 :88 888 88 8:
288 N88 N88 8888 8:36
8 § .8 58 8_ 88.8 38 8 88.8 888 8 88.8 888 .883
8. 88.8 888 : 88.8 888 8 98.8 888 : 8.8.8 888 so
8 :88 888 8 88.8 888 8 88.8 888 8 38.8 888 8
8 :88 88 8. $88 38 8 388 N88 2 38.8 888 N8 8882... m
88 N88 :88 $88 =Ee>o
.8 :88 88 .8 888 N88 88 :88 88 8 88.8 58 883
8 88.8 38 2 €8.88 38 : 28.8 88 N_ 88.8 888 88
8 :88 N88 : 38.8 888 8 £88 88 8. 88.8 888 88
N :88 888 8 A88 888 8 28.8 888 8 38.8 888 we
8 3.8.8 888 8 38.8 888 v 88.8 888 8 88.8 888 :8
8 3.8.8 88 8 88.8 888 8 28.8 888 8 38.8 $8 88
8 88.8 38 8 38.8 58 8 3.8.8 888 8 38.8 888 a
8 :88 888 8 388 N88 8 38.8 888 e :88 888 N8 28
8 38.8 $8 8 A88 N88 8 38.8 «88 8 A 5.8 888 a
232:3
.52 pAva 02 8Ava UZ 8Ava 82 mAva Davaooom 58an
£38.29 “8.83.8288 822.8 .8288

 

significant increase seen in the ECAK’s AMPbd loop (82) value (from 0.79 in ECAK to
0.91 in ECAK*ApSA) upon inhibitor binding is not seen between TNAK and
TNAK*Ap5A. The AMPbd loop (32) of TNAK*Ap5A (0.85) remains similar to that of
TNAK (0.87), indicating similar dynamic characteristics of this loop in both forms. The
standard deviation associated with TNAK’s AMPbd loop (52) value is quite high (0.16).
It comes from the particularly low S2 values of residues 45 and 46, and indicates
substantial ps-ns timescale motion for these residues. In ECAK*ApSA and
TNAK*Ap5A, but not in ECAK and TNAK, the loops belonging to the lid-core hinges
have amongst the lowest (S2) values. Notably, these hinge loops become more flexible
upon inhibitor binding in TNAK ((Sz) decreases from 0.91 in TNAK to 0.85 in

TNAK*Ap5A) than in ECAK ((82) is 0.87 in ECAK and 0.86 in ECAK*ApSA).

As in ECAK, TNAK’s AMPbd domain features the highest density of residues
whose motions are best described by model 5, featuring motions in the slow (ns) and fast
(sub-ns) timescales. In TNAK, another cluster of residues that show model 5 type
motions is located in the lid and the hinge region following the lid. This cluster of

residues does not show model 5 type in TNAK*Ap5A.
Microsecond timescale dynamics

Microsecond to ms timescale motions are identified by the parameter Rex, which
represents exchange contribution to T2. Rex indicates conformational exchange between
conformational states that sense different chemical environments. Rex is directly

proportional to the chemical shift difference between the exchanging species.

110

Residues showing exchange contributions are illustrated in Figure III-7A (TNAK
and ECAK) and Figure III-7B (TNAK*Ap5A and ECAK*ApSA). In addition to the R.,,
values shown in Figures III-7A and III-7B, the dynamics of several TNAK and
TNAK*Ap5A P-loop residues could not be analyzed due to large line-widths attributable
to exchange broadening. Forty five ECAK residues vs. 22 TNAK residues show
exchange contributions at 30 °C. This difference is almost completely eliminated in the
inhibitor-bound enzymes where only 17 ECAK*ApSA residues show us timescale
motion vs. 14 in TNAK*Ap5A. Residues with R.,, > 1.5 s'l are listed in Table III—2 for
ECAK, TNAK, ECAK*Ap5A, and TNAK*Ap5A. The predominant differences in us
timescale motions between TNAK and ECAK lie in the core domain and in the loops of
core-lid hinges. Specifically, the core B-strands have no exchange contribution in TNAK,
while 3 of the 4 ECAK core B-strands have Rex values above 1.5 5". Interestingly, in both
TNAK and ECAK, helices a8 and (19 are quite mobile with several residues showing us
timescale dynamics. Helix a9 forms the C-tenninus in ECAK, while in the TNAK model
it stretches from residues 204—215. The C-terminal residues 216—220 do not form any
regular secondary structure in TNAK. This is seen also in BGAK*ApSA and
BsubAK*Ap5A. The reason for the presence of a large number of residues with RC),
values in the C-terminal region of both enzymes is not clear, although this could be
simply because of a greater mobility of an unanchored C-terminus. Also of particular
note is the presence of several residues with exchange contributions in ECAK and
TNAK’s core domain loops. Upon inhibitor binding, ECAK’s core is significantly

rigidified in the us timescale, as also observed at the ps-ns timescale. Twenty of the 22

residues in ECAK’s core B-strands and a—helices that had R.,,x values in ECAK do not

111

Figure III-7: Exchange contributions, Rex (5") of (A) TNAK (A) and ECAK (O),
and (B) TNAK*Ap5A (A) and ECAK*ApSA (O) at 30 °C. ECAK data (42) are
plotted as per the sequence alignment shown in Figure III-1. TNAK secondary
structures and domains are indicated on top of the figure. The open rectangles
specify the location of the hinge regions framing the AMPbd and lid domains.

112

Rex (8'1)

 

 

 

 

 

 

AMPbd Lid

1o '

A
8 T
s i
4t E § § i i i

l . 2
23 Pi 99;: it

. ‘-

-Zr L81 1 ttr.t.ttt
1 o 20 40 60 80100120140160180200T220
0.B i

. l

6 §

1— §

§
.11 I1
2£§Q§ g Q If I I
t E: § 5

0% f ‘
'21"“1'”'ifllli'lili"1'i““l""i“"i‘H‘iHHi‘H'l‘

0 20 4o 60 80 100120140160180 200 220

TNAK Residue No.

113

Table III-2: Residues with R.,, values > 1.5 s'1 in the various domains of ECAK,
TNAK, ECAK*Ap5A, and TNAK*Ap5A.

aECAK and ECAK*ApSA data are from (42).
bB8 is not predicted in the TNAK and TNAK*Ap5A models.
CRegions defined in (19)

114

 

8: .N: - 3.— .NE we. do— .no— 480— .Mm: .N8: to

 

 

 

8N. - 8N. 8: .3. 8.8 8888.888 888:. 88-8.8.8
- 88 .N8 .om E. . moot— omcofie owE; 38.—2.88-200
8... 88. .NN. 8... N... .NN. 8888
8- - 8- .8. . 88
- - - 88. 88
- 8N. 88. 8N. 88
8N. .N. .8: 8N. .N. .8N. 88 8...
- - 88 ..8 .8 .88 .88 8.8
- - - 8... 8.8
- N8 88 . N.8 8882...
88N .N8N .:. .88 88.88 .8. .: .8. .8 88. .88 ..8 .8N .8N SN .88. .88. .5. 88..
N.N .8.N .88N - EN .8.N N.N .88N N.N .8.N .88N .88N .88N .88N .88N .N8N 8.8
88. - 88. .88. 88. .88. .N8. .88. 88.88. .88. 88
- - 88 88 .88 .N8 8.8
- - - .8. 88
- 88. - 88. .88. .8
- .8 - 88
o . - m a Boo
mo: 50 mo: 30
88:283— 8. m 88:283. .8. a 9.3935

<88<su<ze 888838.58. 8.82... 88.58. 58.88:

115

have Rex values in ECAK*ApSA. In contrast, the core domain in TNAK*Ap5A shows
similar rigidity as in TNAK. The core loops in ECAK*ApSA and TNAK*APSA have
several residues showing us timescale dynamics but the individual residues showing R.,x

values are not the same in the free and bound-forms of the enzymes.

As would be expected from the involvement of this domain in conformational
changes, several residues of ECAK and TNAK’s AMPbd domains show us timescale
chemical exchange. While no residue in ECAK’s core-AMPbd hinges shows us
timescale dynamics, only one residue in the corresponding region in TNAK (residue 77)
shows us timescale motions. Note that the AMPbd domain and its hinges become more
rigid on the us timescale with only a single residue from these regions (#32) showing any
chemical exchange in TNAK*Ap5A. In contrast, several residues in ECAK*ApSA’s

core-AMPbd hinges show significant mobility.

ECAK and TNAK’s lid domains also feature several residues with exchange
contribution. In the inhibitor-bound forms of ECAK and TNAK also, several residues
(some of them the same as in the free forms) show significant Rex values. The notable
difference between ECAK*ApSA and TNAK*Ap5A lies in the core-lid hinges. Whereas
in the ECAK core-lid hinges, there are 8 residues with Rex values, there are none in
ECAK*ApSA. In contrast the two residues that show R.,x values in TNAK’s core-lid
hinges (residues 124 and 172), also show Rex values in TNAK*Ap5A. Note that lid

opening has been shown to be the rate-limiting step in ECAK catalysis (53).

116

3.4 Discussion

Much effort has been spent in the last two decades in investigating the relationships
between flexibility, stability, and activity in hyperthermophilic enzymes. These
investigations have been prompted by the fact that most hyperthermophilic enzymes are
inactive at ambient temperatures (30 0C). It is clear from these studies that the energy
landscape of enzymes is complex, with motions that span picoseconds to seconds and that
vary spatially. ECAK and TNAK make a unique mesophilic-hyperthermophilic enzyme
pair to study dynamics in relation to stability and activity owing to the unusually high
activity of TNAK at 30 0C. We have reported here the difference in the motional
properties of ECAK and TNAK over multiple timescales as determined by 15N relaxation

measurements acquired at 30 °C.

The differences in ECAK and TNAK motions are seen in the NOE data which
primarily reflect high frequency (ns) local motions. The high frequency motions in
ECAK’s AMPbd and lid domains are largely suppressed upon inhibitor binding. In
contrast, high frequency motions in TNAK*ApSA’s AMPbd and lid domains are as
prevalent as in TNAK. This difference in high frequency local motions in ECAK*ApSA
and TNAK*Ap5A is likely to reflect the difference in the two enzymes motional

properties in relation to activity.

Overall, TNAK is uniformly more rigid than ECAK at 30 °C in the ps to us

timescale (Figure III-8). Increased 82 values in TNAK, particularly in the core, are

117

Figure III-8: Backbone ps-ns motions (S2 values) summarized for TNAK and
ECAK open forms. Residues with insufficient data are colored gray. ECAK data
are from (42). Figures were generated using InsightH. Images in this dissertation
are presented in color.

Figure III-9: us timescale dynamics in ECAK and TNAK backbones open forms
as per the scale shown. Residues go from rigid (dark blue) to flexible (yellow).
Residues with insufficient data are colored gray. ECAK data are from (42).
Figures were generated using InsightII.

118

 

 

K
A
N
T

 

ECAK

 

 

 

119

indicative of increased enthalpic contribution to the thermodynamic stability of the native
state (5), since the formation of enthalpically favorable interactions is expected to quench
ps-ns timescale motions. The stability of TNAK’s core is further evidenced by the lack of
us timescale exchange in this region unlike what is observed in ECAK’s core
(Figure III-9). A few residues in ECAK and TNAK core loops have low S2 values in the
free and inhibitor-bound forms. Perhaps, the low S2 values of residues in the core loops
of all four enzyme forms reflect a compensatory effect for the loss in conformational
entropy in the rest of the core. ECAK*ApSA’s core is substantially more rigid than
ECAK’s core in the us timescale also. Together, these results indicate a higher
thermodynamic stability of the inhibitor-bound form of ECAK. Interestingly, the (82) of
the secondary structures in the core of TNAK*Ap5A are slightly lower than the
corresponding (82) in TNAK indicating a small increase in flexibility upon inhibitor
binding. If significant, this flexibility increase could reflect an inhibitor binding-induced

entropic contribution to TNAK*Ap5A stability.

Residues with low 32 values in the TNAK and TNAK*ApSA’s AMPbd domains
are coincidental with the dynamics of this region during catalysis. The loops in
ECAK*ApSA and TNAK*ApSA’s core-lid hinges have low (82) values. Although the
role of ps-ns timescale motions in slower functional processes is still not understood,
these high frequency motions may facilitate the occurrence of slower motions (4).
Therefore, the low (82) values of the core-lid hinge loops in only the complex forms may
have important implication for the lid-opening rates in the two enzymes. As pointed out
by Wand (51), it is not necessary for the timescale of functionally relevant motions to be

restricted by or to be correlated with functional rate constants.

120

Most of the residues in TNAK’s lid and its hinge regions that show exchange
contributions also show significant exchange contributions in TNAK*Ap5A. In contrast,
while ten residues show significant R.2x values in the core-lid hinge regions in ECAK ,
these regions contain no residues with Rex values in ECAK*ApSA. An 15N relaxation
dispersion study comparing a thermophilic AK from Aquifex aeolicus (AAAK) and
ECAK found that lid opening for product release was the rate—limiting step in the two
enzymes (53). AAAK, like other thermophilic enzymes is nine-fold less active than
ECAK at 20 OC. AAAK’s lower catalytic activity at 20 °C was fully accounted for by the
slower lid opening rate measured at the same temperature. The fact that TNAK is highly
active at 30 °C suggests that the significant exchange contributions observed in
TNAK*ApSA’s core-lid hinges reflect dynamics related to lid movements in catalysis. It

would be interesting to measure lid opening rates in TNAK to confirm this postulate.

The results from this study show that TNAK, a multi-domain protein in which
catalysis-related motions mainly involve hinge movements, is simultaneously highly
active at low temperatures and stable at high temperatures by localizing flexibility to the
hinge regions. The domains themselves are rigid without compromising activity at low
temperatures. Also, localizing flexibility to the hinges does not endanger the stability of
the protein. Thus, proteins showing catalytic hinge movements or a catalytic site between
well defined, independent domains may be more amenable to rationally engineering both

thermostability and high activity at low temperatures than single-domain proteins.

121

3.5 Conclusion

The increased S2 values of TNAK indicate increased rigidity consistent with TNAK’s
increased stability. The lack of us timescale dynamics in the TNAK core also indicates
increased stability of this enzyme. The nature of the change in us timescale dynamics
from TNAK to TNAK*Ap5A indicate that lid opening may not be restricted in this

hyperthermophilic enzyme, consistent with its activity at ambient temperatures.

122

3.6

10.

11.

References

Sali, A., and T. L. Blundell. 1993. Comparative protein modelling by satisfaction
of spatial restraints. J. Mol. Biol. 234:779-815.

Bae, E., and G. N. Phillips Jr. 2004. Structures and analysis of highly
homologous psychrophilic, mesophilic, and thermophilic adenylate kinases. J.
Biol. Chem. 279:28202-28208.

Berry, M. B., and G. N. J. Phillips. 1998. Crystal structures of Bacillus
stearothermophilus adenylate kinase with bound Ap5A, Mgzl Ap(5)A, and Mn“
Ap(5)A reveal an intermediate lid position and six coordinate octahedral

geometry for bound Mg2+ and Mn”. Proteins: Struct. Function, Genet. 32:276—
288.

Brooks III, C. L., M. Karplus, and B. M. Pettitt. 1988. Proteins: A Theoretical
Perspective of Dynamics, Structure, and Thermodynamics. In I. Prigogine and S.
A. Rice (ed.), Advances in Chemical Physics. John Wiley & Sons, New York.

Butterwick, J. A., J. P. Loria, N. S. Astrof, C. D. Kroenke, R. Cole, M. Rance,
and A. G. Palmer. 2004. Multiple time scale backbone dynamics of homologous
thermophilic and mesophilic ribonuclease HI enzymes. J. Mol. Biol. 339:855-871.

Clore, G. M., P. C. Driscoll, P. T. Wingfield, and A. M. Gronenbom. 1990.
Analysis of the backbone dynamics of interleukin-1B using two-dimensional

inverse detected heteronuclear 15N-lH NMR spectroscopy. Biochemistry 29:7387-
7401.

Clare, G. M., A. Szabo, A. Bax, L. E. Kay, P. C. Driscoll, and A. M.
Gronenbom. 1990. Deviations from the simple two-parameter model-free

approach to the interpretation of nitrogen-15 nuclear magnetic relaxation of
proteins. J. Am. Chem. Soc 112:4989-4991.

Colombo, G., and K. M. Merz. 1999. Stability and activity of mesophilic
subtilisin E and its thermophilic homolog: Insights from molecular dynamics
simulations. J. Am. Chem. Soc. 121:6895-6903.

D'Amico, S., J. C. Marx, C. Gerday, and G. Feller. 2003. Activity-stability
relationships in extremophilic enzymes. J. Biol. Chem. 278:7891-7896.

de la Torre, J. G., and V. A. Bloomfield. 1981. Hydrodynamic properties of
complex, rigid, biological macromolecules: theory and applications. Q Rev.
Biophys. 14:81-139.

Delaglio, F., S. Grzesiek, G. W. Vuister, G. Zhu, J. Pfeifer, and A. Bax. 1995.
NMRPipe: a multidimensional spectral processing system based on UNIX pipes.
J. Biomol. NMR 6:277-293.

123

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

Fitter, J., and J. Heberle. 2000. Structural equilibrium fluctuations in mesophilic
and thermophilic a-amylase. Biophys. J. 79: 1629-1636.

Fitter, J., R. Herrmann, N. Dencher, and A. Blume. 2002. The role of protein

dynamics for thermal adaptation of mesophilic and thermostable a-amylase.
Biophys. J. 82:304A-304A.

Fitter, J., R. Herrmann, N. A. Dencher, A. Blume, and T. Hauss. 2001.

Activity and stability of a thermostable a-amylase compared to its mesophilic
homologue: Mechanisms of thermal adaptation. Biochemistry 40: 10723-1073 1.

Gershenson, A., J. A. Schauerte, L. Giver, and F. H. Arnold. 2000.
Tryptophan phosphorescence study of enzyme flexibility and unfolding in
laboratory-evolved thermostable esterases. Biochemistry 39:4658-4665.

Grottesi, A., M. A. Ceruso, A. Colosimo, and A. Di Nola. 2002. Molecular
dynamics study of a hyperthermophilic and a mesophilic rubredoxin. Proteins
46:287—294.

Grzesiek, S., and A. Bax. 1992. Correlating backbone amide and side-chain
resonances in larger proteins by multiple relayed triple resonance NMR. J. Am.
Chem. Soc. 114:6291-6293.

Haney, P., J. Konisky, K. K. Koretke, Z. Luthey-Schulten, and P. G.
Wolynes. 1997. Structural basis for thermostability and identification of potential

active site residues for adenylate kinases from the archaeal genus Methanococcus.
Proteins 28:117-130.

Hayward, S. 2004. Identification of specific interactions that drive ligand-

induced closure in five enzymes with classic domain movements. J. Mol. Biol.
339:1001-1021.

Hernandez, G., F. E. Jenney, Jr., M. W. Adams, and D. M. LeMaster. 2000.
Millisecond time scale conformational flexibility in a hyperthermophile protein at
ambient temperature. Proc. Natl. Acad. Sci. USA 97:3166-70.

Hiller, R., Z. H. Zhou, M. W. Adams, and S. W. Englander. 1997. Stability
and dynamics in a hyperthermophilic protein with melting temperature close to
200 °C. Proc. Natl. Acad. Sci. USA 94:11329—11332.

Hiyama, Y., C. H. Niu, J. V. Silverton, A. Bavoso, and D. A. Torchia. 1988.
Determination of N-15 Chemical-Shift Tensor Via N-lS-H-2 Dipolar Coupling in
Boc-Glycylglycy1[N-15]Glycine Benzyl Ester. J. Am. Chem. Soc. 110:2378-
2383.

Hollien, J., and S. Marqusee. 1999. Structural distribution of stability in a
thermophilic enzyme. Proc. Natl. Acad. Sci. USA 96: 13674—13678.

Johnson, B. A., and R. A. Blevins. 1994. NMRView: A computer program for
the visualization and analysis of NMR data. J. Biomol. NMR 4:603-614.

124

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

Karshikoff, A., and R. Ladenstein. 1998. Proteins from thermophilic and

mesophilic organisms essentially do not differ in packing. Protein Engin. 11:867-
872.

Kay, L. E., P. Keifer, and T. Saarinen. 1992. Pure absorption gradient enhanced
heteronuclear single quantum correlation spectroscopy with improved sensitivity.
J. Am. Chem. Soc 114:10663-10665.

Kumar, S., C. J. Tsai, and R. Nussinov. 2000. Factors enhancing protein
thermostability. Protein Engin. 13: 179- 191 .

Lazaridis, T., I. Lee, and M. Karplus. 1997. Dynamics and unfolding pathways
of a hyperthermophilic and a mesophilic rubredoxin. Protein Sci. 6:2589—2605.

Lipari, G., and A. Szabo. 1982. Model-free approach to the interpretation of
nuclear magnetic-resonance relaxation in macromolecules .1. Theory and range of
validity. J. Am. Chem. Soc 104:4546-4559.

Lipari, G., and A. Szabo. 1982. Model-free approach to the interpretation of
nuclear magnetic-resonance relaxation in macromolecules .2. Analysis of
experimental results. J. Am. Chem. Soc 104:4559-4570.

Manco, G., E. Giosue, S. D'Auria, P. Herman, G. Carrea, and M. Rossi. 2000.
Cloning, overexpression, and properties of a new thermophilic and thermostable

esterase with sequence similarity to hormone-sensitive lipase subfamily from the
archaeon Archaeoglobusfulgidus. Arch. Biochem. Biophys. 373: 182-192.

Mandel, A. M., M. Akke, and A. G. Palmer, III. 1995. Backbone dynamics of
Escherichia coli ribonuclease HI: Correlations with structure and function in an
active enzyme. J. Mol. Biol. 246: 144-163.

Muhandiram, D. R., and L. E. Kay. 1994. Gradient-enhanced triple-resonance

3-dimensional NMR experiments with improved sensitivity. J. Magn. Reson.
Series B. 103:203-216.

Muller, C. W., G. J. Schlauderer, J. Reinstein, and G. E. Schulz. 1996.
Adenylate kinase motions during catalysis: An energetic counterweight balancing
substrate binding. Structure 4: 147- 156.

Muller, C. W., and G. E. Schulz. 1992. Structure of the complex between
adenylate kinase from Escherichia coli and the inhibitor Ap5A refined at 1.9 A
resolution - a model for a catalytic transition-state. J. Mol. Biol. 224: 159-177.

Palmer, A. G., M. Rance, and P. E. Wright. 1991. Intramolecular motions of a
zinc finger DNA-binding domain from Xfin characterized by proton-detected

natural abundance carbon-13 heteronuclear NMR spectroscopy. J. Am. Chem.
Soc. 113:4371-4380.

Perrier, V., S. Burlacu-Miron, S. Bourgeois, W. K. Surewicz, and A. M.
Gilles. 1998. Genetically engineered zinc-chelating adenylate kinase from

125

 

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

Escherichia coli with enhanced thermal stability. J. Biol. Chem. 273:19097-
19101.

Russell, R. J., J. M. Ferguson, D. W. Hough, M. J. Danson, and G. L. Taylor.
1997. The crystal structure of citrate synthase from the hypertherm0philic
archaeon Pyrococcusfuriosus at 1.9 A resolution. Biochemistry 36:9983-9994.

Sambrook, J., F. E. Fritsch, and T. Maniatis. 1989. Molecular cloning: A
laboratory manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor,
NY.

Schulz, G. E., C. W. Muller, and K. Diederichs. 1990. Induced-fit movements
in adenylate kinases. J. Mol. Biol. 213:627—630.

Shapiro, Y. E., E. Kahana, V. Tugarinov, Z. C. Liang, J. H. Freed, and E.
Meirovitch. 2002. Domain flexibility in ligand-free and inhibitor-bound
Escherichia coli adenylate kinase based on a mode-coupling analysis of N-15 spin
relaxation. Biochemistry 41:6271-6281.

Shapiro, Y. E., M. A. Sinev, E. V. Sineva, V. Tugarinov, and E. Meirovitch.
2000. Backbone dynamics of Escherichia coli adenylate kinase at the extreme
stages of the catalytic cycle studied by 15N NMR relaxation. Biochemistry
39:6634-6644.

Somero, G. N. 1978. Temperature adaptation of enzymes: Biological

optimization through structure-function compromises. Ann. Rev. Ecol. System.
9:1-29.

Svingor, A., J. Kardos, I. Hajdu, A. Nemeth, and P. Zavodszky. 2001. A better
enzyme to cope with cold - Comparative flexibility studies on psychrotrophic,
mesophilic, and thermophilic IPMDHs. J. Biol. Chem. 276:28121-28125.

Tjandra, N., S. E. Feller, R. W. Pastor, and A. Bax. 1995. Rotational diffusion
anisotropy of human ubiquitin from 15N NMR relaxation. J. Am. Chem. Soc
117:12562 - 12566.

Tsai, A. M., T. J. Udovic, and D. A. Neumann. 2001. The inverse relationship
between protein dynamics and thermal stability. Biophys. J . 81:2339-2343.

Tugarinov, V., Y. E. Shapiro, Z. Liang, J. H. Freed, and E. Meirovitch. 2002.
A novel view of domain flexibility in E. coli adenylate kinase based on structural
mode-coupling ”N NMR relaxation. J. Mol. Biol. 315: 155—170.

Vieille, C., H. Krishnamurthy, H. H. Hyun, A. Savchenko, H. Yan, and J. G.
Zeikus. 2003. Thermotoga neapolitana adenylate kinase is highly active at 30 °C.
Biochem. J. 372:577-585.

Vieille, C., and J. G. Zeikus. 2001. Hyperthermophilic enzymes: sources, uses,
and molecular mechanisms for thermostability. Microbiol. Mol. Biol. Rev. 65:1—
43.

126

50.

51.

52.

53.

54.

55.

56.

57.

58.

Vogt, G., S. Woell, and P. Argos. 1997. Protein thermal stability, hydrogen
bonds, and ion pairs. J. Mol. Biol. 269:631-643.

Wand, A. J. 2001. On the dynamic origins of allosteric activation. Science
293:U1-U1.

Watanabe, K., Y. Hata, H. Kizaki, Y. Katsube, and Y. Suzuki. 1997. The
refined crystal structure of Bacillus cereus oligo-1,6-glucosidase at 2.0 A

resolution: structural characterization of proline-substitution sites for protein
therrnostabilization. J. Mol. Biol. 269: 142-153.

Wolf-Watz, M., V. Thai, K. Henzler-Wildman, G. Hadjipavlou, E. Z.
Eisenmesser, and D. Kern. 2004. Linkage between dynamics and catalysis in a

thermophilic-mesophilic enzyme pair. Nature Structural & Molecular Biology
11:945-949.

Wrba, A., A. Schweiger, V. Schultes, R. Jaenicke, and P. Zavodsky. 1990.
Extremely thermostable D-glyceraldehyde-3-phosphate dehydrogenase from the
eubacterium Thermotoga maritima. Biochemistry 29:7584-7592.

Xiao, L., and B. Honig. 1999. Electrostatic contributions to the stability of
hyperthermophilic proteins. J. Mol. Biol. 289: 1435-1444.

Yamazaki, T., W. Lee, C. H. Arrowsmith, D. R. Muhandiram, and L. E. Kay.
1994. A suite of triple-resonance NMR experiments for the backbone assignment
of 15N, 13C, 2H labeled proteins with high sensitivity. J. Am. Chem. Soc
116:11655-11666.

Yip, K. S. P., T. J. Stillman, K. L. Britton, P. J. Artymiuk, P. J. Baker, S. E.
Sedelnikova, P. C. Engel, A. Pasquo, R. Chiaraluce, V. Consalvi, R.
Scandurra, and D. W. Rice. 1995. The structure of Pyrococcus furiosus
glutamate-dehydrogenase reveals a key role for ion-pair networks in maintaining
enzyme stability at extreme temperatures. Structure 3:1147-1158.

Zavodszky, P., J. Kardos, Svingor, and G. A. Petsko. 1998. Adjustment of
conformational flexibility is a key event in the thermal adaptation of proteins.
Proc. Natl. Acad. Sci. USA 95:7406-7411.

127

Chapter 4

Stability and flexibility in a mesophilic and a

hyperthermophilic adenylate kinase

The research in Chapter 4 is presented in the form of a manuscript to be submitted soon.

128

4.1 Introduction

Proteins from hyperthermophiles are stable and active at high temperatures, conditions in
which their mesophilic homologues denature. Since the three-dimensional (3-D) structure
of mesophilic and hyperthermophilic homologous proteins are very similar and since
their sequence similarity typically ranges between 40% and 85%, the mechanisms
responsible for the dramatic stabilization of hyperthermophilic proteins at high
temperatures is not readily apparent. To understand the molecular mechanisms that
confer thermostability, families of homologous mesophilic-hyperthermophilic proteins
have been compared both by biochemical experimentation as well as through
computational methods (6, 16, 22, 23, 41, 44). These studies have proposed several
mechanisms that could contribute to the stability of hyperthermophilic enzymes at high
temperatures. These mechanisms include increased numbers of hydrogen bond (H-bond)
interactions (45), presence of salt-bridge networks (23, 33, 48, 49), deletion or shortening
of loops, and better packing (33), to name a few. Attempts have been made to introduce
specific mutations, as predicted by these studies, in mesophilic enzymes to increase their
thermostability to the level of their hyperthermophilic homologues (8, 20, 21, 42).
Although such mutations have often increased mesophilic enzymes’ thermostability,
achieving the level of thermostability of their hyperthermophilic counterparts through
rational design has been impossible. This lack of success can be attributed to a lack of
understanding the extent to which each interaction contributes to thermostability and,
most importantly, how these stabilizing interactions are distributed in the

hyperthermophilic protein.

129

 

One hypothesis explaining the remarkable stability of hyperthermophilic enzymes
is that these enzymes have enhanced conformational rigidity at low temperatures. Indirect
evidence for this hypothesis comes from the fact that many hyperthermophilic enzymes
are inactive at low temperatures (20-40 0C) but have similar levels of catalytic efficiency
as their mesophilic homologue at the optimal growth temperature of their source
organism. It is argued that enzymes require a certain degree of conformational flexibility
for optimal activity. The presumed rigidity of hyperthermophilic enzymes is used to
explain their inactivity at low temperatures, and it is thought that these enzymes require
high temperatures to gain the requisite flexibility for activity (17). This hypothesis
thereby links local conformational fluctuations and global thermodynamic stability. It
implies that an increase in conformational flexibility decreases stability of the native
folded state. According to this hypothesis, stability and activity are fundamentally linked,
and gain in one would result in a loss of the other. However, as argued by Lazaridis et a1
(24), protein rigidity and stability do not have to be correlated per se. Conformational
flexibility in proteins can span several timescales, from picoseconds (ps) to seconds or

more and could be spatially localized.

Native state hydrogen-deuterium (H-D) exchange experiments have become an
important tool to study conformational flexibility in proteins. H-D exchange experiments
allow one to compare the structural distribution of stability in hypertherrnophilic-
mesophilic enzyme pairs. Comparing the backbone amide protection factors in the two
proteins can also provide insights into the stabilities of individual sites in the proteins. H-
D exchange studies with varying denaturant concentrations or temperature are a common

technique to determine the thermodynamics of protein stability (2, 4). Several studies

130

q

have shown that a subset of the most slowly exchanging residues exchange only during
global unfolding events. The free energy of exchange of these residues corresponds to the
free energy of unfolding (1, 3). H-D exchange also has emerged as a tool for determining
the kinetics of protein conformational changes as well as the global unfolding kinetics (9,
39, 40). Several studies have used H-D exchange experiments to compare mesophilic-
hyperthermophilic enzyme pairs. H-D exchange rates in these proteins have been
measured using a variety of techniques such as Fourier transformed Infrared
Spectroscopy (FTIR) (26, 50), mass spectrometry (25, 26), and NMR (12, 19) . Although
results from several of these studies suggest that the hyperthermophilic enzyme is more
rigid than its mesophilic homologue, such studies have primarily used techniques such as
FT IR that provide a structurally averaged measure of amide exchangeirates. NMR
methods for following H-D exchange, on the other hand, allow the full potential of site-
resolved H-D exchange to be realized. In an H-D exchange study of
Thermus thermophilus and Escherichia coli RNase H (TtRNase and EcRNase,
respectively) using NMR, it was found that the general stability distribution was similar
for TtRNase and EcRNase. Region-specific AG of unfolding increased by approximately
41-42% going from EcRNase to TtRNase, suggesting that each region of the protein
contributes the same relative stabilization in TtRNase. The authors also showed that
under conditions where the two enzymes have similar global stability (i.e., 0 M
guanidine-HCl for EcRNase and 0.88 M Guanidine-HCl for TtRNase), values of region-
specific AG of unfolding are very similar for the 2 proteins (14). Conversely, an H-D
exchange comparison study of Pyrococcus furiosus and Clostridium pasteurianum

rubredoxins demonstrated a more spatially localized increase in rigidity in the

131

hyperthermophilic protein with some regions, in fact, exhibiting higher flexibility than

the mesophilic protein (12, 13).

In the work presented here, we use native state H-D exchange to compare the
rigidity distribution in the hyperthermophilic Thermotoga neapolitana adenylate kinase
(TNAK) and the mesophilic Escherichia coli adenylate kinase (ECAK) through
determination of residue-specific protection factors (P) in the two proteins. Adenylate
kinase (AK) is an ubiquitous enzyme that catalyses the transfer of ATP’s y-phosphate to
AMP to form ADP. The ECAK structure reveals three domains: the core, the lid, and the
AMPbd domains (27, 28). Crystal structures of ECAK in the free form and in complex
with various substrates and inhibitors also show that the lid and AMPbd domains are
significantly displaced upon ligand-binding (35, 36, 46). These two domains close over
the substrates, presumably to protect them from hydrolysis, to adopt the ‘closed’
conformation. Following phosphoryl transfer, the lid and AMPbd domains open up to

release the products, and the enzyme adopts the so-called ‘open’ conformation.

Although TNAK and ECAK share 40% sequence identity and the active site
residues are conserved, TNAK is a highly thermostable enzyme with a melting
temperature (Tm) of 991°C, more than 47°C higher than the Tm of 518°C measured for
ECAK (43). TNAK, though, is an unusual hyperthermophilic enzyme in that it is as
active as ECAK at 30 °C (43). Since a crystal structure of TNAK is not available, we
modeled TNAK’s structure in the free form as well as in the ligand-bound form using the
software MODELLER 8v] (1, 10). The models of TNAK and TNAK bound to the
inhibitor, diadenosine pentaphosphate (i.e., AP5A) (referred to as TNAK*Ap5A) are

shown in Figure IV-l. The domains in ECAK are defined as AMPbd: residues 31-72; lid:

132

Figure IV-l: Ribbon diagrams of the homology modeled structures of A) TNAK
(open form) and B) TNAK*Ap5A (closed form). The lid and AMP binding
(AMPbd) domains are indicated in the open form. The figures were generated
using the program Ribbons. Zn“ in the lid domain and Ap5A are shown with a
ball and stick representation.

133

 

134

residues 119-156; core-AMPbd hinges: residues 29, 30, and 73-79; core-lid hinges: 113-
118 and 157-172; and core: rest of the residues (11). The domain limits used here are
different from those described by (36), where the respective AMPbd and lid domains are
described as residues 30-59 and 122-159. The definitions used here seem to better match

the ECAK crystal structures.

TNAK domain definitions are based on sequence alignment with ECAK and on
the model are (i) AMPbd: residues 33-75; (ii) lid: residues 126-163; (iii) core-AMPbd
hinges: residues 30, 31, and 76-82; (iv) core-lid hinges: residues 120-125 and 164-179;

and (v) core: rest of the residues.

To obtain comprehensive information on the flexibilities of ECAK and TNAK in
the two forms and at various timescales, we performed l5N NMR relaxation analyses of
TNAK and TNAK*Ap5A as well as H-D exchange analyses of ECAK, ECAK*Ap5A,
TNAK, and TNAK*Ap5A. 15N NMR relaxation data for ECAK in the open and closed
forms have been reported previously (37, 38) and the ECAK and TNAK 15N NMR
relaxation results are compared in Chapter 3. To complete the flexibility mapping of
ECAK and TNAK at various timescales, we present a comparison H-D exchange data for

ECAK, ECAK*Ap5A, TNAK, and TNAK*Ap5A at 30 °C.

135

a?"

4.2 Materials and Methods

4.2.1 Protein samples

TNAK was over-expressed in E. coli BL21(DE3) cells containing the expression plasmid
pTNAK2:sz (43). TNAK was uniformly lsN labeled by growing the cells in M9
minimal medium (34) containing lg/l of > 98% [ISM-ammonium chloride. Protein
expression was induced at an OD600nm of 1.0-1.2 with 0.4 mM isopropyl-B-d-
thiogalactopyranoside (IPTG) for 20 h at 37°C. The previously reported protocol for
TNAK purification (43) was modified as described below. Supernatant from the
heat—treated soluble extract was fractionated with ammonium sulfate at 50% saturation to
precipitate impurities. The supernatant was dialyzed against 20 mM Tris-HCl (pH 8.6)
(buffer A) and loaded on a Q-sepharose Fast Flow (Amersham Biosciences, Uppsala,
Sweden) anion exchange column (2.6 x 30 cm, flow rate 2 mllmin) pre-equilibrated with
buffer A. After washing the column with five volumes of buffer A, proteins were eluted
with a 400 ml linear 0.2—0.7 M NaCl gradient in buffer A. Fractions containing TNAK
(as determined by OD280nm and SDS-PAGE) were pooled, dialyzed against 20 mM
Tris-HCl (pH 8.0) (buffer B) and loaded again on the Q-sepharose column pre-
equilibrated with buffer B. TNAK was eluted with a 0-0.7 M NaCl gradient in 50 mM
Tris-HCl, pH 7.4. Purified TNAK was concentrated in an ultrafiltration cell equipped
with a 10,000 molecular weight cut-off membrane (Amicon, Beverly, MA), dialyzed
against milliQ water whose pH was adjusted to the experimental value to maintain the

ionization state of the protein, freeze-dried, and stored at -20°C.

136

The E. coli adk gene was amplified from E. coli K12 genomic DNA and cloned
into the pET24a(+) Ndel and Xhol sites to yield plasmid, pECAKl. Once cloned, the
E. coli adk gene was sequenced by the MSU Genomic Technology Support facility to
verify that no mutation had been introduced during the amplification step. Uniformly
15N-labeled ECAK was obtained by growing E. coli BL21(DE3) (pECAKl) cells in M9
minimal medium (34) containing 1 g/L 15N-ammonium chloride (> 98%) as the sole
nitrogen source. ECAK expression was induced by adding 0.4 mM IPT G at an OD600nm
of 1.0-1.2. The cells were harvested after 20 hours of induction at 37 °C and lysed by two
passages through a French pressure cell. The soluble and insoluble fractions were
separated by centrifugation at 20,000 x g for 20 min. ECAK was purified as described
(31) with the modification that the protein was eluted from an Affi-gel Blue gel affinity
column (Biorad, Hercules, CA) using a 0—2 mM NaCl gradient in 50 mM Tris-HCl, pH
7.4. For the ion-exchange chromatography, Q-sepharose Fast Flow (Amersham
Biosciences, Uppsala, Sweden) column was used instead of DEAE. The purified protein
was extensively dialyzed against de-ionized water with the pH adjusted to the

experimental value, lyophilized, and stored at -20 °C.
4.2.2 NMR sample preparation and data collection

The lyophilized protein samples were dissolved in 50 mM sodium phosphate buffer in
D2O at the desired pH to initiate exchange. The buffer was prepared by mixing
appropriate amounts of sodium diphosphate and sodium monophosphate in D2O to obtain
the desired pH* (i.e., the pH meter reading in D2O without adjusting for the deuterium
isotope effect). The final NMR sample contained 2 mM protein (TNAK or ECAK) and

10 11M sodium azide in 320 111 deuterated buffer. To prepare the inhibitor-bound forms of

137

 

the proteins (ECAK*ApSA or TNAK*Ap5A), Ap5A and MgC12 in D2O were also added
to the mixture at 5 mM and 8 mM final concentrations, respectively. The following
procedure was followed while setting up an exchange experiment to minimize the dead
time: (i) a mock sample identical to that used in the exchange experiment was used to
tune the NMR probe and adjust the shim settings; (ii) all acquisition parameters were
determined using the mock sample; (iii) the exchange experiments were initiated by
adding the lyophilized protein in 320 1.11 of 50 mM sodium phosphate buffer in D2O
pre-heated to 30 °C. Ap5A and MgCl2 were added immediately to prepare the inhibitor-
bound protein samples; (iv) the solution was then briefly mixed and centrifuged to
remove insoluble protein, and the sample was quickly transferred to a shigemi tube;
(v) the lock signal was monitored until it stabilized, shim settings were readjusted with
the exchange sample, and data acquisition was initiated. The temperature for the
exchange experiments was 30 0C. The exchange experiments were performed at two
different pH values for each protein conformation. The pH* values measured at the end
of the exchange experiments were 7.0 and 7.6 (TNAK), 7.1 and 7.8 (ECAK), 7.0 and 7.5
(TNAK*Ap5A), and 7.0 and 7.6 (ECAK*ApSA). For each exchange experiment, the
start time was taken to be the time of protein-buffer mixing. The dead time of the
experiment (i.e., from the time that the protein and buffer were mixed to the start of
spectrum recording) was approximately 4-5 min. The last spectrum recorded was

approximately 46 hours after initiating the exchange.

The 1H-ISN HSQC spectra were collected on a Varian Unity Inova 600 MHz
spectrometer equipped with a triple resonance probe. Spectral widths were 8,000 Hz and

1,900 Hz in the 1H and 15N dimensions, respectively. With 32 x 1956 (t. x t2) complex

138

points and four transients per t. increment, the total acquisition time for each spectrum

was 5 min and 18 sec.
4.2.3 H-D exchange rate analysis

Data were processed using NMRPipe (7) and NMRView (18). A Gaussian window
function in the direct dimension and a sine window function in the indirect dimension
were applied. Phase and baseline corrections were used in all cases. The HSQC spectra
were zero-filled to 2048 x 512 data points in lH and 15 N dimensions, respectively. The
H-D exchange rates were quantified using the intensity of the 1H-15 N amide cross-peaks
in a series of HSQC spectra recorded at different times. For each amide, i, the time
dependency of peak intensity was fit to a single exponential decay function equation,
I(t) = 10(t) + I, exp(-k¢xt), using the program CURVEFIT (Arthur Palmer group, Columbia
University). CURVEFIT uses a least-squares fitting algorithm to yield the exchange rate,
kex. The intrinsic rate constant (Bai factor), kch, for each amide proton at 30 0C and a
specific pH was predicted from the protein sequence using the program HXPrep
(courtesy Zhonggyi Zhang), which employs the procedure described in (Bai et a1, 1993).
The calculated kch values and the measured kc, values were used to calculate the

protection factors. For the theory underlying EXl and EX2 exchange mechanisms, see (2,

4, 47).

139

4.3 Results and Discussion

4.3.1 The exchange mechanism at the experimental pH* values

Native state H-D exc hange in TNAK and ECAK at 30 °C was followed by collecting
lH-ISN HSQC NMR spectra at different time points for 46 h. Exchange experiments were
performed at two pH values for each protein. TNAK is most stable between pH 7.0 and
pH 8.5, and its activity at pH 6.5 drops to 50% of its optimal activity at pH 7.5.(43). For
these reason, we chose 7.0 and 8.0 as the target experimental pH values for the two
enzymes. The final pH* of the deuterated samples were 7.0 and 7.6 for TNAK and 7.1
and 7.8 for ECAK. Twenty four of the ECAK amide protons with measurable exchange
rates showed EX2 exchange at 30 °C, with higher exchange rates at pH* 7.8 than at pH
7.1. The ratio of exchange rates for these amide protons at pH* 7.8 and pH* 7.1 was
consistent with a ten-fold increase in exchange rate for every pH unit increase (i.e., the
pH-dependent increase expected for EX2 exchange mechanism). Fourteen ECAK amide
protons showed similar exchange rates at the two pH* values, indicating that these
protons likely exchange via the EXl mechanism. Similarly, twenty TNAK amide protons
did not show any difference in their exchange rates, within error, at pH* 7.0 and pH* 7.6
indicating an EXl exchange mechanism for these protons. All other TNAK amide
protons with measurable exchange rates (i.e., 23) showed EX2 exchange in these
conditions. The pH dependence of exchange rates was also determined for ECAK*ApSA
and TNAK*Ap5A. The pH* values of the deuterated samples were 7.0 and 7.5 for
TNAK*Ap5A and 7.0 and 7.6 for ECAK*ApSA. As for the ligand-free proteins, 19 (out

of 47) ECAK*ApSA amide protons and 28 (out of 41) TNAK*Ap5A amide protons did

140

not show any pH-dependent exchange rate variation. Since H-D exchange is
base-catalyzed, EX2 is generally the dominant exchange mechanism between pH 4.0 and
pH 7.0 (3, 19, 29). The EXl mechanism is usually observed at alkaline pH values
(usually above pH 9.0), where kch exceeds kc]. Since the experiments reported here were
performed between pH 7.0 and 8.0, some residues are likely to show a mixture of EX]

and EX2 mechanisms or predominantly EXl mechanism of exchange.

4.3.2 Determination of P values

For both EXl and EX2 mechanisms, the exchange rate can be measured by fitting a
single exponential function to peak intensities, and the protection factors P can be

calculated using equation [1]:

P: 18./k... = Isa/18.: UK... 111

where k0p and kc] are the rate constants for conformational opening and closing in the

scheme [2] (2, 4, 47):

k
0p kch

Closed 4: Open —) Exchanged [2]
CI

The P values reported and discussed in this work were calculated from the exchange data
collected at pH* 7.0. P factors were calculated for only those amide protons with
measurable exchange rates, since some exchanged within the dead time of the experiment
while some others remain unexchanged during the time course of the experiments (~46

hours). Exchange data for 75, 70, 60, and 93 residues in ECAK, TNAK, ECAK*Ap5A,

141

and TNAK*Ap5A, respectively, could not be obtained. The missing exchange data can
be attributed to two reasons: (i) not all residues could be detected in the HSQC spectrum
during the initial NMR assignments of ECAK (202 assigned [37]), ECAK*ApSA (191
assigned [37]), TNAK (197 assigned [Chapter 3]) and TNAK*Ap5A (193 assigned
[Chapter 3]); (ii) since HSQC spectra for the exchange experiments were recorded for
only 5 min, resolution was compromised and several crosspeaks overlapped. H-D

exchange rates for the overlapped peaks were not quantified.

The percentages reported in this paragraph were calculated based on the total
number of residues having exchange data. Overall, TNAK has a significantly higher
percentage of unexchanged residues (20%) than ECAK (12%). Conversely, a smaller
percentage of TNAK residues (51%) exchange in the dead time of the experiments than
in ECAK (64%). If unexchanged residues are indicative of high conformational rigidity,
and if very fast exchange indicates high flexibility, these results suggest that TNAK is
more rigid than ECAK. Although the increase in percentage of protected amides in
ECAK*ApSA relative to ECAK is modest (only 3%), the percentage of unexchanged
amide protons increases significantly (by 7%) from ECAK to ECAK*Ap5A, implying
increased rigidity for inhibitor-bound form of ECAK. The increase in the percentages of
unexchanged amides (2%) and amides with some protection factor (3%) from TNAK to
TNAK*Ap5A are smaller than between ECAK and ECAK*ApSA. Still, they suggest that
TNAK*Ap5A is at least as rigid, if not more rigid, than TNAK. Not surprisingly,
TNAK*Ap5A is overall more rigid than ECAK*ApSA (54% of the amides are protected
in TNAK*Ap5A vs. 46% in ECAK*ApSA). The following sections discuss in detail the

distribution of protected residues in the four enzyme forms.

142

4.3.3 The slowest exchanging amide hydrogens are in the ECAK and TNAK cores

The log of P values computed for the ECAK and TNAK residues with measurable amide
proton exchange rates are shown in Figures IV-2A and IV-2B. Residues whose amide
protons exchanged within the dead time of the NMR experiments were arbitrarily given a
logP value of 0.1, indicating rapid exchange, to distinguish them from amide protons
whose exchange rates could not be quantified. Residues whose amide protons did not
exchange within the time period of the experiment were arbitrarily assigned a logP value

of 7.

Figures IV-2A and IV-2B show that more amide protons in TNAK than in ECAK
are unexchanged or have some level of protection from exchange. The domain
distribution of protected amides in all four enzyme forms is summarized in Table IV -1. In
ECAK, 59% of the core residues show some level of protection, while only 21.4% and
3% of the AMPbd and lid residues, respectively any protection. This clear demarcation
between a relatively rigid core and very flexible AMPbd and lid domains is absent in
TNAK, where residues with moderate to high protection are distributed more uniformly
throughout the protein. Table IV-l suggests that not all regions of the protein contribute
to the higher stability of TNAK to the same extent. Indeed, the percentages of ECAK and
TNAK core amides which show some level of protection are very similar, indicating
similar rigidity in the two enzyme cores. The increase in rigidity occurs predominantly in
TNAK’s AMPbd and lid domains, which show 2.2-fold (AMPbd) and l8-fold (lid) more

amide protons that are protected compared to ECAK.

The slowest exchanging amide protons in native proteins are typically involved in

143

Figure IV-2: Sequence dependence of log of protection factors, P (I) in (A)
ECAK (B) TNAK (C) ECAK*ApSA and (D) TNAK*Ap5A. S2 values (0) for
ECAK and ECAK*ApSA (35) and for TNAK and TNAK*Ap5A (Chapter 3) are
shown. Residues that exchanged within the experimental dead time have been
assigned an arbitrary logP of 0.1 to distinguish them from residues whose
exchange rate could not be quantified. Residues that did not exchange during the
time course of the experiments have been assigned an arbitrary logP value of 7. A
cartoon indicating the location of secondary structures and domains is indicated
above the data.

144

 

0.4

0.4

0 20 40 100 120 140 160 180 200 22

o o 0 °

0
O
O-‘NQAUIOSV

0 20 40 60 80 100 120 140 160 180 200

AMPbd Ltd

.... .o n D - 0°05

0 oz}, o . - a gym _ -

o . e

. & o 3 § 0° 00° 0 o l I. 0 O o o I 0d) I . $ 3 $0
0 . 0° 0 . on . N . u
0 ' o 00“.: °° a
' O O
o o
o .
8
o

O

o
o o

n

OOANQAUICDV

00

8°

0

°o

o

o

(P

‘b

u

o

0 o

o
.60

—:'
fly
- 0"
O-‘NW-bmmfl

0 20 40 60 80 100 120 140 160 180 200

AMPbd Lid

(9° (5‘9 a: '0 ° - . 8'6 “I” 0 o°°°d>o$ °

' 012:8 °.°‘b° ’°°°°°:' 15%| ‘5‘” °: . %°°.,°°°° k-f’? °. 6
.' 0 ° c

3

2

1

0

20 40 60 80 100 120 140 160 180 200 220
Residue No.

145

logP

logP

logP

logP

 

Table IV -1 Distribution of protected amides in the free and inhibitor forms of ECAK and
TNAKa.

 

 

 

 

ECAK TNAK ECAK*ApSA TNAK*Ap5A

A 13 A B A B A B

Core 59 16.6 56 22.5 56.5 29 6O 29
AMPbd 21.4 3.6 41 13.8 25 9.3 37 14
Lid 3.7 0 46 21.4 51.6 3.2 61 22

 

aPercentages were calculated based on the number of residues in each domain that have exchange data

A: % residues with some protection
B: % unexchanged amide protons

 

well-ordered H-bond networks, and/or they are buried and have restricted solvent access.
Exchange at these amides typically takes place only during global unfolding events.
Although the percentage of protected amides are similar in ECAK and TNAK’ core, the
only suggestion that TNAK’s core is more stabilized than ECAK’s core comes from
comparison of the percentage of unexchanged amide protons. Only 17% of ECAK core’s
amide protons are unexchanged compared to 23% in TNAK’s core. Of the twelve ECAK
amide protons that did not exchange during the time course of the experiment eleven are
located in the core domain. Of these eleven arrride protons, ten are in the B strands, and
one (Ile26) is in a loop. The only unexchanged amide proton not belonging to the core
(Va168) is located in helix 01.4 of ECAK’s AMPbd domain. Using the ECAK crystal
structure (pdb code: 4AKE), we computed the best possible H-bond network in the
ECAK free form using the Optimal H-bonding Network module of the software WHAT
[F (15, 32). The WHAT IF results indicate that all unexchanged ECAK amide protons are

involved in H-bonds. Particularly, the core amide protons form main chain H-bonds with

146

other core residues, highlighting the importance of these interactions in the stability and

rigidity of the core domain.

Of the 30 TNAK amide protons that did not exchange, 21 belong to core domain
residues. Although, as in ECAK, a majority of these 21 residues belong to the core
B-strands, eight reside in the core loops and a—helices, indicating an additionally
rigidified core domain in TNAK. Each of these 21 core amide protons is involved in at
least one H-bond with the main chain of another core residue. Some additionally

participate in H-bonds with side chains.

That TNAK’s lid domain is much more rigid than ECAK’s lid is evidenced not
only by the percentage of protected amides but also from the number of unexchanged
amides in this domain. There are five unexchanged amide protons in TNAK’s lid while
there are none in ECAK’s lid. TNAK’s lid domain has four cysteines (i.e., Cysl33,
Cysl36, Cy8153, and Cy8156) that bind a Zn2+, forming a Zn-finger-like structure. The
unexchanged TNAK lid residues that are highly protected include Cysl33, Lysl35,
Cysl36, Tyr140, and Cys158. Of these residues, only Cys 133, Tyr 140, and Cys 158 are
involved in H-bonds according to our WHAT IF analysis. In contrast, the corresponding
ECAK residues, which are also involved in H-bonds are not protected. The fact that these
TNAK amide protons remain unexchanged could also be a consequence of solvent
inaccessibility. The presence of these unexchanged residues indicates a highly stabilized
lid domain in TN AK relative to ECAK. In ECAK, the Zn-finger-like structure is replaced
by an H-bond network that may not be as stabilizing as the Zn-finger. The structural role
of Zn2+ as a stabilizing factor has been studied previously by site-directed mutagenesis by

creating a quadruple cysteine mutant ECAK (Ca-ECAK) capable of binding Zn”.

147

Calorimetric studies showed that C4-ECAK is significantly more stable (Tm increases
from 51.8 °C to 63 0C) than the wild-type ECAK (5, 30). Correspondingly, removal of
Zn2+ from TNAK reduces its melting temperature by nearly 6°C to 935°C (43).
Therefore, while Zn2+ is not the sole contributor to TNAK’s thermostability, it may play

an important role in stabilizing the lid.

In addition to the core and lid domains, TNAK’s AMPbd domain also has four
residues (Arg38, Asp46, Val71, and Leu75) belonging to helices 0:2, (13, and 0.4 that have
unexchanged amide protons. These amide protons are involved in the same kind of
H-bonds as the corresponding residues in ECAK. In spite of this similarity, only the
amide proton of ECAK Val68 exchanges slowly. Moderate to no protection at all is
observed for the other three corresponding ECAK AMPbd amide protons (i.e., Arg36,
Ser43, and Ile72). The results from the H-D exchange analyses of ECAK and TNAK are

summarized in Figures IV-3A and IV-3B.
4.3.4 Inhibitor-binding stabilizes ECAK’s AMPbd and lid domains

Figures IV-2C and IV-2D show the plots of residue number vs. logP values for
ECAK*ApSA and TNAK*Ap5A. Again, the logP values of unexchanged amide protons
are set to 7 and those of amide protons that exchanged within the dead time of the

experiments are set to 0.1.

Compared to ECAK, ECAK*ApSA has 10% more residues that show protection
against hydrogen exchange. A total of 25 residues in ECAK*ApSA show no exchange

during the time course of the experiment. The core in ECAK*ApSA is

148

Figure IV-3 Protection factors summarized for (A) ECAK, (B) TNAK, (C)
ECAK*Ap5A, and (D) TNAK*Ap5A backbones as per the scales shown.
Residues that did not exchange within the time course of the experiments are
colored blue while those that exchanged within the dead time of the experiments
are colored red. Residues whose exchange rates could not be quantified are
colored grey. Figures were generated using insight H.

149

QM NM a.” “N N.N Q— Q; —.— NO
0 I

 

(O 9.0

 

 

 

ow om am am ma ow o... A. 30
.. . 4., .

A

 

v..o one
a w

 

150

somewhat more rigidified with 29% of amide protons unexchanged, compared to ~17%
in ECAK. All the core residues whose amide protons were unexchanged in the free form

continue to be protected in the bound form. In addition, two amide protons (Arg206 and

Ala209) in the C-terminal core helix (org), are also unexchanged.

The most predominant difference between ECAK and ECAK*ApSA is in the lid
domain (Table IV-l). While all but one ECAK lid residues exchange during the dead
time of the experiment, ~52% of ECAK*ApSA lid residues have some level of
protection, and Arg13l remains unexchanged. The Arg131 amide proton has one extra H-
bond with His126 that is absent in ECAK. Only 3 of 16 partially protected amides have
H-bonds that are absent in the free enzyme form. The enhanced protection of these
residues could be because they are buried upon inhibitor binding, thereby limiting solvent

accessibility.

In contrast to the lid domain, the AMPbd domain shows only marginal
stabilization in ECAK*ApSA and this stabilization is limited to a few residues in helices
a2 and a4. Residues between 40 and 63 exchange within the dead time of the experiment
indicating the high flexibility of this region in the inhibitor-bound form also. The increase
in the percentage of protected amides in the AMPbd domain is due to the unexchanged
amide protons of Asp33, Val68, and Arg71. It is interesting to note that none of these
amides form H-bonds with Ap5A, and that many of their intra—molecular H-bonds are
already present in ECAK. Their protection in the inhibitor-bound form could result from

an increase in the stabilization of the AMPbd domain that limits solvent accessibility.

151

Overall, the inhibitor-bound form of ECAK has protection factors for more residues than

the free form (Table IV-l) indicating increased conformational stability of ECAK*ApSA.

In TNAK*Ap5A, the percentage of amide protons that show any protection from
exchange only marginally increases compared to TNAK (~5%). Several of TNAK’s core
residues remain unexchanged in TNAK*Ap5A as well, due to many of the same H-bond
interactions that are preserved in both conformational states. Even though fewer residues
have exchange data in TNAK*Ap5A (127 residues) than TNAK (150 residues), the
protection distribution is highly similar in the cores of TNAK and TNAK*Ap5A. While
the AMPbd domain is not additionally stabilized in TNAK*Ap5A, the lid domain is more
rigid in TNAK*Ap5A than in TNAK. Except for Cys136, none of the newly protected
amides forms any new (i.e., not found in TNAK) intra—molecular H-bonds nor do they

form H-bonds with Ap5A.

For both the AMPbd and lid domains, the distribution of unexchanged amide
protons is different in the two TNAK forms. For example, lid residue 158 that was
unexchanged in TNAK is only marginally stabilized in TNAK*Ap5A, and residue 137
that has no protection against exchange in TNAK is fully protected in TNAK*Ap5A.
Similarly, the AMPbd residue 62 that has no protection in TNAK is fully protected in
TNAK*Ap5A. None of TNAK*ApSA’s unexchanged lid and AMPbd residues seem to

H-bond with the inhibitor.

Taken together, the results show that while ECAK is significantly rigidified upon
inhibitor binding, the same is not true for TNAK. The only common factor in these two

enzymes is the significant increase in the rigidity of the lid domains, especially for

152

ECAK. The AMPbd domain is only marginally stabilized in ECAK, while in TNAK
there is almost no change upon inhibitor binding. In both enzymes, but particularly in
ECAK, several AMPbd amide protons exchange within the dead time of the experiments

indicating considerable local fluctuations in this domain.

Taken together, the results show that not only is TNAK not uniformly stabilized
compared to ECAK, the two enzymes are not uniformly stabilized upon inhibitor binding.
In the free form, the lid and AMPbd domains are considerably more rigid in TNAK than
ECAK while in the closed forms of both enzymes; the lid domain becomes more rigid
upon inhibitor binding. Results from the H-D exchange analysis of ECAK*ApSA and

TNAK*Ap5A are summarized in figures IV-3C and IV-3D.

4.3.5 Hinge regions

The AMPbd—core hinge regions consist of nine residues of which 5, 6, 8, and 4 residues
have exchange data in ECAK, TNAK, ECAK*Ap5A, and TNAK*Ap5A, respectively.
The lid-core hinge regions consist of a total of 22 residues and 13, 16, 14, and 13 of these
residues have exchange data in ECAK, TNAK, ECAK*Ap5A, and TNAK*Ap5A,
respectively. As seen from Figure IV-2, amide protons of 3 out of the 5 residues with
exchange data in ECAK’s AMPbd-core hinge residues and 4 out of the 6 residues with
exchange data in TNAK’s AMPbd-core hinges have no protection from exchange (i.e.,
have a logP value of 0.1 in Figure IV-2). The remaining residues in this region in both the
enzymes have logP values less than 3.0. In the lid-core hinges, the amide protons of 10
out of 13 residues in ECAK and 11 out of 16 residues in TNAK exchange within the dead

time of the experiments. As was noted for the AMPbd domain, amide protons of the

153

remaining residues in the lid-core hinges of both enzymes have logP values less than 3.0.
In the inhibitor-bound form of ECAK, 5 out of 8 amide protons in the AMPbd-core
hinges and 8 out of 14 amide protons in the lid-core hinges have no protection from
exchange. Corresponding numbers for TNAK*Ap5A are: 1 out of 4 in the AMPbd-core
hinges and 8 out of 13 in the lid-core hinges. Interestingly in both ECAK*ApSA and
TNAK*Ap5A, a single residue (Ile 29 in ECAK*ApSA and Ile31 in TNAK*Ap5A) in
the AMPbd-core hinge remains unexchanged during the experiments. This Be is H-
bonded to an Asp (Aps84 in ECAK*ApSA and Asp 87 in TNAK*Ap5A) that is
important for coordinating Mg2+.Since fast exchanging residues indicate high flexibility,
especially if not solvent exposed, the high density of unprotected residues in the hinges
indicate high flexibility for the AMPbd-core and lid-core hinge regions in both ECAK

and TNAK in the free and inhibitor bound forms.
4.3.6 Hydrogen exchange and 15N NMR relaxation

The backbone dynamics of TNAK and TNAK*Ap5A in the ps-ns timescale at 30 °C have
been characterized using 15N NMR relaxation rate measurements (Chapter 3). Backbone
dynamics of ECAK and ECAK*ApSA at 30 0C studied with 15N NMR relaxation have
been reported previously (38). They were used for comparison with the TNAK and
TNAK*Ap5A results (Chapter 3). On the ps-ns timescale, TNAK is uniformly more rigid
than ECAK, particularly in the core domain. This rigidity difference is abolished in
ECAK*ApSA and TNAK*Ap5A, due a rigidity increase in ECAK upon inhibitor
binding. Although TNAK is more rigid than ECAK, several residues in the TNAK
AMPbd and lid domains exhibit lower than average 82 values consistent with the higher

flexibility of these domains (Chapter 3). To determine if the trends for 82 values and the

154

trends for P values from hydrogen-exchange coincide, the l5N NMR relaxation order
parameters and the logP values from amide hydrogen exchange for ECAK, TNAK,
ECAK*Ap5A, and TNAK*Ap5A are plotted in Figures IV-2A, IV-2B, IV-2C, and

IV-2D, respectively.

Comparing the 15N relaxation dynamics and H-D exchange data on ECAK and
TNAK brings up some interesting differences in the two enzyme cores. The average
15N relaxation order parameters ((82)) show that ECAK and TNAK’s cores are only
marginally more rigid in the ps-ns timescale than the rest of the proteins. The <52) values
of the core, AMPbd, and lid domains increase by a similar amount (8.5%, 7.7%, and
9.5%, respectively) going from ECAK to TNAK, suggesting a uniform stabilization
across the protein fold of the hyperthermophilic enzyme in the ps-ns timescale. The H-D
exchange data, on the other hand, show that ECAK’s core is significantly more rigid than
the rest of the protein with the rigidity decreasing in the order core > AMpbd >> lid.
Moreover, TNAK’s core is similar in rigidity to that of ECAK in the 2 ms timescale. The
only suggestion that TNAK’s core is more rigid than that of ECAK resides in the fact that
22.5% of TNAK core amides remain unexchanged, against only 16.6% in ECAK’s core.
TNAK’s increased rigidity is most significant in the lid domain, followed by the AMPbd
domain. Hence, unlike on the ps-ns timescales, the increase in rigidity of TNAK on the 2
ms timescale is localized rather than uniform. In the us timescale (determined by the
exchange contribution (Rex) to the transverse relaxation rate [T2]), many more residues in
ECAK’s core, particularly the B-strands, have Rex values than in TNAK’s core, indicating

a higher rigidity of the TNAK core in the us timescale.

155

It is evident from both the l5N NMR relaxation and H-D exchange data that
ECAK’s AMPbd domain is characterized by local fluctuations. For example, the S2
values of several residues (i.e., for residues 43-45, 61, and 62) are lower than average.
This domain also contains contiguous residues showing model 5-type, sub-ns to ns
timescale motions ([38], not shown]. In the H-D exchange data also, ECAK residues 30
through 65 are completely unprotected, their amide protons exchanging within the dead
time of the exchange experiments, indicating exchange through local fluctuations.
Although overall, TNAK’s AMPbd domain is more rigid than ECAK’s AMPbd domain,
it also has several residues showing low 82 values (i.e., 32, 43, 46, 48, and 59-62), the S2
value being similar in both enzymes in many cases (Chapter 3). The H-D exchange data
corroborates these results because several amide protons in the region between 33 and 65
in TNAK’s AMPbd domain are unprotected or show only moderate protection. Hence

ECAK and TNAK’s AMPbd domains feature dynamics in ps to 2 ms timescales.

ECAK’s lid domain, on the other hand, does not show any significant trend in S2
values. Several residues, though, show Rex values indicating significant us timescale
dynamics. The lid domain is also the least protected from exchange in ECAK indicating
pervasive local fluctuations in the 2 ms timescales. Residues in TNAK’s lid also show
Rex values, although these are fewer in number in TNAK than in ECAK. H-D exchange
data show the TNAK lid to be more rigid than ECAK’s lid (likely due to the presence of
Zn”), although several residues in this region show only moderate protection. Hence,
ECAK and TNAK’s lids are likely to be characterized by motions mainly in timescales

greater than us.

156

4.4 Conclusions

Here we have compared the conformational stability of a mesophilic and a
hyperthermophilic AK using H-D exchange monitored by NMR. Protection factors
derived from the exchange rates have been used as a measure of conformational
dynamics in the two proteins. The core domains of ECAK and TNAK are similarly stable
as indicated by the percentage of protected amides in the two proteins. The increase in the
rigidity of TNAK, compared to ECAK is not uniform across all domains. Instead the lid
and AMPbd domains are proportionally more rigid in TNAK. This result is different from
the one obtained from H-D exchange analyses of EcRNase and TtRNase, where the
relative stability increase of all TtRNase subdomains/substructures is the same (14).
These results reinforce the fact that stabilization mechanisms vary from one protein
family to another, and that they are not necessarily uniformly distributed around the
structure. The dynamics in ECAK and TNAK domains also are not similar across all
timescales. While the AMPbd domain in both enzymes shows fluctuations ranging from
ps to 2 ms timescale, motions in 2 us timescales are predominant in the lid domains of
ECAK and TNAK. The pattern of rigidity increase in TNAK is also different at different
timescales. Whereas it is uniformly more rigid than ECAK in the ps-ns timescales, only
its lid and AMPbd domains are more rigid in 2 ms timescales. The presence of Zn2+ in
TNAK’s lid domain plays an important role in the local stability of that domain. In
essence, the increased rigidity of TNAK is not uniform across all the domains nor is it

uniform at all timescales.

157

The fact that the lid and AMPbd domains are rigid structures in TNAK does not
preclude them from having flexible regions and flexible hinges. This result is important
in light of the fact that TNAK is as active as ECAK at 30°C. The hinge regions are
important for the movement of the lid and AMPbd domains, and the domains themselves
can be rigid without compromising domain closing and opening. TNAK, therefore, is an
excellent example of a hyperthermophilic enzyme where rigidity for stability and
flexibility for activity are partitioned in a manner that makes it simultaneously active at

low temperatures and stable at high temperatures.

Although the protection factors are used here as reports of the conformational
stability of ECAK and TNAK, the energetics of the stability of the two enzymes have not
been determined. Future directions of this research include measurement of dependence
of H-D exchange rates in ECAK and TNAK on denaturant concentrations and
temperature. These measurements can be used to fully determine residue-specific free

energies of unfolding.

158

4.5

10.

11.

12.

References

Sali., A., and T. L. Blundell. 1993. Comparative protein modelling by
satisfaction of spatial restraints. J. Mol. Biol. 234:779-815.

Arrington, C. B., and A. D. Robertson. 2000. Kinetics and thermodynamics of
conformational equilibria in native proteins by hydrogen exchange. Methods
Enzymol. 323: 104-124.

Arrington, C. B., and A. D. Robertson. 1997. Microsecond protein folding
kinetics from native-state hydrogen exchange. Biochemistry 36:8686-8691.

Bai, Y. W., J. J. Englander, L. Mayne, J. S. Milne, and S. W. Englander.
1995. Thermodynamic parameters from hydrogen exchange measurements.
Methods Enzymol. 259:344-356.

Burlacu-Miron, S., V. Perrier, A. Gilles, E. Pistotnik, and C. T. Craescu.
1998. Structural and energetic factors of the increased thermal stability in a

genetically engineered Escherichia coli adenylate kinase. J. Biol. Chem.
273:19102—19107.

Chakravarty, S., and R. Varadarajan. 2002. Elucidation of factors responsible
for enhanced thermal stability of proteins: A structural genomics based study.
Biochemistry 41:8152-8161.

Delaglio., F., S. Grzesiek, G. W. Vuister., G. Zhu., J. Pfeifer., and A. Bax.
1995. NMRPipe: a multidimensional spectral processing system based on UNIX
pipes. J. Biomol. NMR 6:277-293.

Eidsness, M. K., K. A. Richie, A. E. Burden, D. M. Kurtz, and R. A. Scott.
1997. Dissecting contributions to the thermostability of Pyrococcus furiosus
rubredoxin: B-sheet chimeras. Biochemistry 36: 10406-10413.

Ferraro, D. M., and A. D. Robertson. 2004. EXl hydrogen exchange and
protein folding. Biochemistry 43:587-594.

Fiser, A., R. K. Do, and A. Sali. 2000. Modeling of loops in protein structures.
Protein Sci. 9: 1753-1773.

Hayward, S. 2004. Identification of specific interactions that drive ligand-
induced closure in five enzymes with classic domain movements. J. Mol. Biol.
339:1001-1021.

Hernandez, G., F. E. Jenney, Jr., M. W. Adams, and D. M. LeMaster. 2000.
Millisecond time scale conformational flexibility in a hyperthermophile protein at
ambient temperature. Proc. Natl. Acad. Sci. USA 97:3166-3170.

159

13.

14.

15.

l6.

l7.

18.

19.

20.

21.

22.

23.

24.

25.

26.

Hernandez, G., and D. M. LeMaster. 2001. Reduced temperature dependence

of collective conformational opening in a hyperthermophile rubredoxin.
Biochemistry 40: 14384-14391 .

Hollien, J., and S. Marqusee. 1999. Structural distribution of stability in a
thermophilic enzyme. Proc. Natl. Acad. Sci. USA 96: 13674-13678.

Hooft, R. W. W., C. Sander, and G. Vriend. 1996. Positioning hydrogen atoms
by optimizing hydrogen-bond networks in protein structures. Proteins: Struct.,
Funct., Genet. 26:363-376. '

Jaenicke, R. 1996. Glyceraldehyde-3-phosphate dehydrogenase from
Thermotoga maritima: strategies of protein stabilization. FEMS Microbiol. Rev.
18:215-224.

Jaenicke, R. 1991. Protein stability and molecular adaptation to extreme
conditions. Eur. J. Biochem. 202:715—728.

Johnson, B. A., and R. A. Blevins. 1994. NMRView: A computer program for
the visualization and analysis of NMR data. J. Biomol. NMR 4:603-614.

Kahsai, M. A., E. Martin, S. P. Edmondson, and J. W. Shriver. 2005. Stability
and flexibility in the structure of the hyperthermophile DNA-binding protein
Sac7d. Biochemistry 44: 13500-13509.

Kawamura, S., Y. Kakuta, 1. Tanaka, K. Hikichi, S. Kuhara, N. Yamasaki,
and M. Kimura. 1996. Glycine-15 in the bend between two tat-helices can
explain the thermostability of DNA binding protein HU from Bacillus
stearothermophilus. Biochemistry 35:1195-1200.

Kawamura, S., 1. Tanaka, N. Yamasaki, and M. Kimura. 1997. Contribution
of a salt bridge to the thermostability of DNA binding protein HU from Bacillus

stearothermophilus determined by site-directed mutagenesis. J. Biochem.
121:448-455.

Kumar, S., and R. Nussinov. 2001. How do thermophilic proteins deal with
heat. Cell Mol. Life Sci. 58:1216—1233.

Kumar, S., C. J. Tsai, and R. Nussinov. 2000. Factors enhancing protein
thermostability. Protein Engin. 13:179-191.

Lazaridis, T., 1. Lee, and M. Karplus. 1997. Dynamics and unfolding pathways
of a hyperthermophilic and a mesophilic rubredoxin. Protein Sci. 6:2589—2605.

Liang, Z. X., T. Lee, K. A. Resing, N. G. Ahn, and J. P. Klinman. 2004.
Thermal-activated protein mobility and its correlation with catalysis in
thermophilic alcohol dehydrogenase. Proc. Natl. Acad. Sci. USA 101: 9556-9561.

Liang, Z. X., I. Tsigos, T. Lee, V. Bouriotis, K. A. Resing, N. G. Ahn, and J.
P. Klinman. 2004. Evidence for increased local flexibility in psychrophilic

160

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

alcohol dehydrogenase relative to its thermophilic homologue. Biochemistry
43:14676-14683.

Muller, C. W., G. J. Schlauderer, J. Reinstein, and G. E. Schulz. 1996.

Adenylate kinase motions during catalysis: An energetic counterweight balancing
substrate binding. Structure 4: 147-156.

Muller, C. W., and G. E. Schulz. 1992. Structure of the complex between
adenylate kinase from Escherichia coli and the inhibitor Ap5A refined at 1.9 A
resolution - a model for a catalytic transition-state. J. Mol. Biol. 224:159-177.

Neira, J. L., P. Sevilla, M. Menendez, M. Bruix, and M. Rico. 1999. Hydrogen
exchange in ribonuclease A and ribonuclease S: Evidence for residual structure in
the unfolded state under native conditions. J. Mol. Biol. 285 :627-643.

Perrier, V., S. Burlacu-Miron, S. Bourgeois, W. K. Surewicz, and A. M.
Gilles. 1998. Genetically engineered zinc-chelating adenylate kinase from
Escherichia coli with enhanced thermal stability. J. Biol. Chem. 273:19097-
19101.

Reinstein, J ., M. Brune, and A. Wittinghofer. 1988. Mutations in the nucleotide
binding loop of adenylate kinase of Escherichia coli. Biochemistry 27:4712—
4720.

Rodriguez, R., G. Chinea, N. Lopez, T. Pons, and G. Vriend. 1998. Homology
modeling, model and software evaluation: three related resources. Bioinformatics
14:523-528.

Russell, R. J. M., J. M. C. Ferguson, D. W. Hough, M. J. Danson, and G. L.
Taylor. 1997. The crystal structure of citrate synthase from the hyperthermophilic
archaeon Pyrococcusfuriosus at 1.9 A resolution. Biochemistry 36:9983-9994.

Sambrook, J., E. F. Fritsch, and T. Maniatis (ed.). 1989. Molecular cloning: A
laboratory manual., 2nd edition ed. Cold Spring Harbor Laboratory Press, Cold
Spring Harbor, NY.

Schlauderer, G. J., K. Proba, and G. E. Schulz. 1996. Structure of a mutant
adenylate kinase ligated with an ATP-analogue showing domain closure over
ATP. J. Mol. Biol. 256:223—227.

Schulz, G. E., C. W. Muller, and K. Diederichs. 1990. Induced-fit movements
in adenylate kinases. J. Mol. Biol. 213:627—630.

Shapiro, Y. E., E. Kahana, V. Tugarinov, Z. C. Liang, J. H. Freed, and E.
Meirovitch. 2002. Domain flexibility in ligand-free and inhibitor-bound
Escherichia coli adenylate kinase based on a mode-coupling analysis of 15N spin
relaxation. Biochemistry 41:6271-6281.

Shapiro, Y. E., M. A. Sinev, E. V. Sineva, V. Tugarinov, and E. Meirovitch.
2000. Backbone dynamics of Escherichia coli adenylate kinase at the extreme

161

39.

40.

41.

42.

43.

45.

46.

47.

48.

49.

50.

stages of the catalytic cycle studied by 15N NMR relaxation. Biochemistry
39:6634-6644.

Sivaraman, T., C. B. Arrington, and A. D. Robertson. 2001. Kinetics of
unfolding and folding from amide hydrogen exchange in native ubiquitin. Nat.
Struct. Biol. 8:331-333.

Sivaraman, T., and A. D. Robertson. 2001. Kinetics of conformational

fluctuations by EXl hydrogen exchange in native proteins. Meth. Molec. Biol.
168:193-214.

Szilagyi, A., and P. Zavodszky. 1995. Structural basis for the extreme
thermostability of D-glyceraldehyde-3-phosphate dehydrogenase from
Thermotoga maritima: analysis based on homology modelling. Protein Engin.
8:779-789.

Tomschy, A., G. Bohm, and R. Jaenicke. 1994. The effect of ion-pairs on the
thermal stability of D-glyceraldehyde 3-phosphate dehydrogenase from the
hyperthermophilic bacterium Thermotoga maritima. Protein Engin. 7: 1471-1478.

Vieille, C., H. Krishnamurthy, H. H. Hyun, A. Savchenko, H. Yan, and J. G.
Zeikus. 2003. Thermotoga neapolitana adenylate kinase is highly active at 30°C.
Biochem. J. 372:577—585.

Vieille, C., and J. G. Zeikus. 2001. Hyperthermophilic enzymes: sources, uses,
and molecular mechanisms for thermostability. Microbiol. Mol. Biol. Rev. 65:1—
43.

Vogt G., W. S., and Argos P. 1997. Protein thermal stability, hydrogen bonds,
and ion pairs. J. Mol. Biol. 269:631-643.

Vonrhein, C., G. J. Schlauderer, and G. E. Schulz. 1995. Movie of the
structural changes during a catalytic cycle of nucleoside monophosphate kinases.
Structure 3:483-490.

Wildes, D., and S. Marqusee. 2004. Hydrogen-exchange strategies applied to
energetics of intermediate processes in protein folding. Methods Enzymol. 380:
328-349.

Xiao, L., and B. Honig. 1999. Electrostatic contributions to the stability of
hyperthermophilic proteins. J. Mol. Biol. 289:1435-1444.

Yip, K. S. P., T. J. Stillman, K. L. Britton, P. J. Artymiuk, P. J. Baker, S. E.
Sedelnikova, P. C. Engel, A. Pasquo, R. Chiaraluce, V. Consalvi, R.
Scandurra, and D. W. Rice. 1995. The structure of Pyrococcus furiosus
glutamate dehydrogenase reveals a key role for ion-pair networks in maintaining
enzyme stability at extreme temperatures. Structure 3:1147-1158.

Zavodszky, P., J. Kardos, Svingor, and G. A. Petsko. 1998. Adjustment of
conformational flexibility is a key event in the thermal adaptation of proteins.
Proc. Natl. Acad. Sci. USA 95:7406-7411.

162

Chapter 5

Molecular dynamics simulation of Escherichia coli

adenylate kinase complexed with its substrates

Research presented in this chapter was published as the following manuscript:

H, Krishnamurthy, H. F. Lou, A. Kimple, C. Vieille, and R. I. Cukier. (2005).
Associative mechanism for phosphoryl transfer: A molecular dynamics simulation of
Escherichia coli adenylate kinase complexed with its substrates. Proteins-Structure

Function and Bioinformatics 58:88-100.

163

5.1 Abstract

The ternary complex of Escherichia coli adenylate kinase (ECAK) with its substrates
adenosine monophosphate (AMP) and Mg-ATP, which catalyzes the reversible transfer
of a phosphoryl group between adenosine triphosphate(ATP) and AMP, was studied
using molecular dynamics. The starting structure for the simulation was assembled from
the crystal structures of ECAK complexed with the bisubstrate analog diadenosine
pentaphosphate (AP5A) and of Bacillus stearothermophilus adenylate kinase complexed
with APSA, Mg2+‘ and 4 coordinated water molecules, and by deleting one phosphate
group from AP5A. The interactions of ECAK residues with the various moieties of ATP
and AMP were compared to those inferred from NMR, X-ray crystallography,
site—directed mutagenesis, and enzyme kinetic studies. The simulation supports the
hypothesis that hydrogen bonds between AMP’s adenine and the protein are at the origin
of the high nucleoside monophosphate (NMP) specificity of AK. The ATP adenine and
ribose moieties are only loosely bound to the protein. while the ATP phosphates are
strongly bound to surrounding residues. The coordination sphere of Mg“. consisting of 4
waters and oxygens of the ATP [3- and y-phosphates, stays approximately octahedral
during the simulation. The important role of the conserved Lysl3 in the P loop in
stabilizing the active site by bridging the ATP and AMP phosphates is evident. The
influence of Mg“, of its coordination waters, and of surrounding charged residues in

maintaining the geometry and distances of the AMP u-phosphate and ATP [3- and

y—phosphates are sufficient to support an associative reaction mechanism for phosphoryl

transfer.

164

5.2 Introduction

Adenylate kinases (AKs) are small, monomeric phosphoryl transferases that catalyze the
reversible transfer of a phosphoryl group from adenosine triphosphate (ATP) to
adenosine monophosphate (AMP) by nucleophilic attack on the y—phosphate of ATP (1).

The net reaction is schematized as
Mg2+ -ATP + AMP <::> Mg3+-ADP + ADP (Scheme 1)

AKs are composed of 3 domains: the core, LID, and AMP—binding (AMP-bd) domains
(Figure V—l). In the absence of substrates, AKs are in an open form. In the presence of
ATP, AMP, and Mg“, the LID and AMP-bd undergo major conformational
rearrangements, resulting in the enzyme closing to form the ternary complex and
expelling waters to prevent ATP and AMP hydrolysis (2). The best-known AK families
are the short AKs represented by the mammalian cytosolic enzymes, and the long AKs
represented by bacterial and mitochondrial enzymes. The long AKs contain a 20- to 30-
residue insertion inside the LID domain (3, 4). Among the short AKs, only the porcine
cytosolic AK has been crystallized in the absence of substrate [Protein Data Bank (PDB)
code: 3ADK] (5). On the other hand, the long AKs from Escherichia coli (ECAK),
Bacillus stearothermophilus, Saccharomyces cerevisiae, and bovine mitochondria have
been crystallized with various substrate—inhibitor combinations or with the 5 phosphate
bisubstrate analog, diadenosine pentaphosphate [i.e.. AP5A] (6—15). The crystal structures
of the E. coli and bovine mitochondrial AKs in the absence of any ligands are also known
(10, 16). ATP binds to the catalytic site as the MgZIATP complex. in which the

catalytically essential Mg“ is coordinated to ATP's B- and -y-phosphates and to 4 water

165

molecules. Thus, the functional catalytic assembly is a ternary complex. In this work, we
focus on ECAK, since a wealth of kinetic (l7), NMR (18—22), and mutagenesis (23—28)
data are available for this enzyme. Crystallography can provide structures of enzymes
complexed with products, with inhibitors that resemble substrates, and even with
inhibitors that model transition states. In this work, we create a model of the ECAK —
Mg-ATP - AMP ternary reaction complex based on the ECAK - AP5A structure
[PDB code: lAKE] (6) and on the B.stear0therm0philus AK (BSAK) structure
crystallized with AP5A and Mg2+ [PDB code: 1ZIO] (ll). Constructing a reaction
complex model allows one to study the stabilization of a possible catalytically competent
complex that is more realistic than a model inferred from a crystal structure containing
substrate mimics. Thus, the ternary complex model will be used as the starting point of a
molecular dynamics (MD) simulation that can assess how ATP and AMP are maintained
in a catalytically competent configuration with the aid of the surrounding protein, the
Mg“ ion, and its 4 coordinating waters. The highly charged complex suggests a focus on
strong hydrogen bonds (H-bonds) and salt bridges that can serve the purpose of
maintaining a catalytically relevant geometry. Two simulations of AKE have appeared
previously. One, a 300-ps study of ECAK complexed with AP5A, was carried out in the
gas and solvated states (29). A number of residues that would be candidates for strong
interaction with a Mg2+ ion were suggested in this study. The other simulation used a
weighted masses MD method to explore the nature of the open conformation of
apo—ECAK that could be reached starting from a number of closed form (with different
bound substrates) crystallographic structures (30). The similarities and differences

between various experimental conclusions based mainly on studies using AP5A, and the

166

simulation of the ECAK - Mg—ATP ' AMP structure are analyzed. While there are some
differences in details, the suggested distinction between the high specificity of AKs for
the AMP substrate and the lower specificity for ATP is supported by the simulation.
Particular attention is devoted to investigating the interactions of the conserved Lysl3
(the numbering refers to ECAK) located at the end of the P loop, with the AMP and ATP
phosphate groups. The simulation shows that Lysl3 does play an important role in
substrate stabilization by bridging the AMP and ATP phosphates. There is great interest,
in general. in the mechanism of phosphoryl transfers, whether it is an associative or a
dissociative mechanism, these being extremes of a continuum of mechanisms
qualitatively defined by the phosphorous—phosphorous distance, which increases in going
from an associative to a dissociative mechanism (25, 31—34). There is evidence (32) that
a dissociative mechanism is favored in solution, while an associative mechanism is
favored in enzymes. The nature of the transition state (or reaction intermediate) also is an
issue. with the most frequently suggested geometry invoking a pentavalent phosphorous
(25, 31—34). As we discuss, the simulation carried out herein maintains the AMP and
ATP terminal phosphates at a distance that supports an associative phosphoryl transfer

mechanism.

167

5.3 Methods

5.3.] Starting structure

The 1.9 A“ resolution crystal structure of the ECAK - Ap5A complex (7)
[PDB code: 1AKE] was used as the starting point for constructing an ECAK - AMP -
MgZIATP ternary complex, which will be denoted as 1AKE*. The 1AKE PDB file
contains the coordinates of the 2 enzyme-inhibitor complexes (A and B) present in the
crystal asymmetric unit. We used complex B to build 1AKE*. AP5A (Figure V-l) is
considered to be a bisubstrate mimic of the reactant state. since ECAK catalyzes the
reaction of ATP and AMP to produce two ADPs (Scheme 1). ATP and AMP coordinates
were obtained by eliminating AP5A’s 5—PO4 group (see Figure V-l). While the 1AKE
structure provides the AP5A coordinates needed to create well-positioned ATP and AMP
molecules within the protein, it does not contain the catalytic Mg2+ and its 4 waters of
coordination. To create the starting structure for the MD simulation, we imported the
coordinates of Mg2+ and its coordinated waters from the crystal structure of the BSAK -
Mg2+-AP5A complex [PDB code: 1210] (11). The 1210 and 1AKE structures were
superimposed by .fitting 1ZIO’s AP5A molecule onto that of 1AKE. lZIO’s Mg2+ and its
4 coordination waters were then merged with the ECAK ' AMP - ATP structure to create
1AKE*. This structure contains Mg2+ positioned between nonbridging oxygens of ATP’s

B—and ’y-phosphates and the 4 Mg2+ coordinating waters. as displayed in Figure V-1.

168

 

Figure V-l. Structures of 1AKE (left) and 1AKE* (right). ECAK is shown in ribbon
representation. AP5A, ATP, and AMP are in stick representation; the Mg2+ atom and its 4
coordinating waters are in ball—and~stick representation. ATP and AMP were created by
deleting APSA’s 8—PO4. Mg“ and its coordinating waters were imported from the 1210
structure.

169

 

5.3.2 Force field

The GROMOS96 force-field was used for the MD simulations (35). This force field is a
united atom force field with explicit polar hydrogens, and hydrogens on the
phenylalanine, tryptophan, and tyrosine rings. The GROMOS96 force field contains
parameters for all protein residues, as well as for ATP, but it does not include parameters
for AMP. Parameters for AMP were created based on the ATP parameters (35) The force
field is identical up to atom 23 (A05* as numbered on ATP). The parameters from 32
(AFC) to 36 (AH3PG) in ATP are then used for AMP’s monophosphate group. The
ionization states of all residues were set to reflect a neutral pH. Histidines were assumed
to be neutral, with the N5 atom protonated, and the lysine. arginine. aspartate, and
glutamate residues were assumed to be charged. To assess whether this charge
assignment is reasonable in the binding pocket, where numerous positively charged
residues bind the substrate phosphates, we adopted the following procedure. The
accessible surface area (ASA) of each residue was evaluated using the Surface Racer
program (36), for the residues in the 1AKE* complex and in the 1AKE* structure in the
absence of the substrates. For each residue, the ASA in the presence of substrate was
subtracted from the ASA in the absence of substrate. The nonzero ASA difference values
then single out those residues that are buried at the protein—substrate interface. The
charged residues buried at the protein—substrate interface, Lysl3, Arg36, Asp84, Arg88,
Argl l9, Arg123, Arg156, and Asp158, have a net charge of +4. These charges are almost
compensated, through various interactions discussed below, by the net charge of +2 from
ATP (-3, with 1 terminal oxygen protonated), AMP (-l), and Mg” (+2). A look at the

charged residues buried at the protein—substrate interface shows that, with the exception

170

of Arg123. all the ionizable groups on positively charged residues are engaged in charge—
charge interactions with the substrate phosphates. All these charge—charge interactions
are stable throughout the 3 ns simulation period (see below). The side-chain of Asp84 is
engaged in stable H bonds with 2 of the waters coordinated to Mg”, namely, 219 and
221. Arg123 forms a salt bridge with Asp159, which is also stable during the simulation.
Note that Asp159, along with its neighboring Asp158 at the C-terminal end of the LID
domain, are conserved in all AKs implicating a crucial role for them in AKs. These 2
aspartates form salt bridges with Arg123 and Arg156 side-chains only in the substrate
bound form. Muller and Schulz (7) suggested that salt bridges Arg123—Asp159 and
Arg156—Asp158 (formed with the aid of ATP and AMP, which help in positioning the
side—chains) are instrumental in the conformational change leading to LID closure. Thus,
the conventional ionization states assumed in this study are appropriate, since almost all
of the charged groups buried in the 1AKE* complex have their charges satisfied either

through protein—substrate, or protein—protein interactions.
5.3.3 Molecular dynamics

The MD simulation was carried out using CUKMODY (37), a code designed for the
efficient simulation of proteins and other large solutes. A combination of a cell index
method with linked lists (38) and a Verlet neighbor list (39) is used to provide linear
scaling with the number of atoms in the pair list routine, essential for the large systems
considered here. For the Verlet neighbor list, the outer distance was r, = l 1.84 A and the
inner distance was r... = 10.42 A. The pair list was updated whenever any atom moved a
distance greater than 0.5*( r, - n), leading to updates roughly every 25 steps. Electrostatic

interactions are evaluated using the charge-group method. to be consistent with the

171

parametrization of the GROMOS force field. The SHAKE algorithm (39) was used to
constrain bond lengths permitting a 2 fs timestep. Periodic boundary conditions were
used. The simulation was carried out at constant number. volume. and temperature
(NVT), with velocity scaling to control the temperature at 303 K. The runs were all
performed on a PC with dual 1.6GHz AMD Athlon processors. A 1 ns simulation takes

about 2 weeks on a single processor.
5.3.4 Equilibration

The 1AKE* system was equilibrated according to the following protocol. A simulation
box of side 59.19 A was filled with an equilibrated sample of 6912 waters, and the
1AKE* structure, totaling 2162 atoms, was centered in the simulation cell. The waters
that overlapped any 1AKE* atom were discarded if r0} < 0'0}, where r”,- is the distance
between water oxygen (atom 0) and protein atom (atom j), and 0'0}- is the van der Waals
distance parameter. This procedure eliminated 1470 waters. One—body forces, with force
constant k = 30 kcal/mol/Az. were used on all protein and ligand atoms for the initial
50,000 steps. This procedure allowed the waters to equilibrate with the protein and with
each other. The van der Waals length (0) and well depth (e) values for the water oxygen-
Mg2+ interactions were designed to maintain coordination throughout the simulations. To
enforce the initial water coordination of Mg“, the parameters listed in Table V-l were
used to equilibrate the system and then modified as indicated for the duration of the

simulation.

After the initial 50,000 steps with one-body forces on, the one—body force

constant was linearly reduced to 0 over 30,000 steps. The system was then run for

172

Table V-l. Water-Mg“ Van der Waals parameters for the Mg2+ coordination waters
during the MD equilibration and simulation periods.

 

 

 

 

Equilibrate with Run with
0. (a) (A) e (a) (K) 0. (a) (A) 6. (a) (K)
Water 218 2.16137 50,0000 2.07364 54,2733
Water 219 2.07814 20000.0 2.07364 54,2733
Water 220 1.88086 20,0000 2.07364 54,2733
Water 221 2.19517 50,0000 2.07364 54,2733

 

(”The van der Waals interaction between atoms i and j is

12 6
V =4e,-j [(qj/aij) —(r,-j/a,-j) :l where (7,-J- =(0',- +aj)/2 and 8,}- = sig-

173

another 20,000 steps to make sure that the system’s strain was sufficiently reduced to
initiate unconstrained dynamics. A 0.5-us trajectory was run at 273 K, after which the
temperature was linearly brought to 303 K over 10,000 steps. After the temperature had

reached 303 K. the MD production trajectory proceeded for 3 ns.
5.3.5 H-bond—salt Bridge analysis

Interactions of AMP and ATP with 1AKE* residues were analyzed through H—bond
analysis. Donor and acceptor atoms within a maximum distance of 3.5 A and a minimum
angle of 120° were considered to form H-bonds. The data were analyzed every 0.8 ps.
Interactions between a charged phosphate group of the substrates and the side-chain of a
positively charged residue (arginine or lysine) were considered to be salt bridge—like.

Oppositely charged groups within 4.5A of each other were considered to be saltbridged.

5.4 Results and Discussion

5.4.1 Validation of simulation

The radius of gyration of the protein increases very slowly during the first 740 ps. with an
average of 17.18 i 0.09 A during that period and, subsequently, suddenly increases and
stabilizes to an average of 17.60 i 0.11 A during the remainder of the simulation,
indicating that the overall dimension of the protein is stable. (The radius of gyration of
1AKE in the crystal structure is 16.635 A.) The RMSD (i.e., root-mean-square deviation

from the starting 1AKE* structure) for all Ca’s increases slowly from 1.5 to 3.1 A during

174

the first 800 ps of the production trajectory; then, between 1 100 and 1300 ps. it increases
more sharply to finally plateau between 4.0 and 5.0 A during the rest of the simulation.
The RMSD for the Cats of the core domain (residues 1—30, 60—120, and 160—214)
increases very slowly to 2.3 A during the first 1 ns of the production trajectory, then
plateaus between 2.1 and 2.75 A before finally increasing after 2800 ps. The RMSD for
the Ca’s of the core plus the AMP-bd domains (residues 1—120 and 160-214) increases
during the first 800 ps at the same rate as the RMSD for all Ca’s, then oscillates between
2.5 and 3.4 A during the next 2000 ps. before finally increasing after 2800 ps. These
results indicate that the deviation from the crystal structure is not distributed uniformly
along the sequence, and that most of the large deviations (in particular after the first 1200
ps) occur in the LID domain. To compare residue fluctuations in 1AKE* with those in the
1AKE crystal, RMSF values (i.e., positional root-mean square fluctuations from the
average structure) were calculated for all C0L carbons as RMSF" = ((rn - (r,,))2)l‘/’, with (r,,)
being the average position of the Ca carbon of residue it over time, and r,, being the
position at every step. Simulation RMSF values are compared with 1AKE
crystallographic RMSF values in Figure V—2. Crystallographic RMSF values were
calculated from the B values using the equation RMSF = \/3B/81t2. Regions with low
RMSF values (e.g., helices (x4 and (13, and strand [34) in the simulation coincide well with
the regions with low crystallographic RMSF values. In contrast, the regions showing high
simulation RMSF values and high crystallographic RMSF values do not always match

each other. Residues 41—48 (the loop between helices or: and (x3), and 132-140 (the loop

between strands [3L2 and 8L3) show high simulation RMSF values that were not expected

175

Figure V-2: Comparison of 1AKE crystallographic B-factors (O) and simulation
RMSF values (0) for all Cu atoms. Residues involved in crystal contacts
(+ signs) are indicated at the bottom of the graph. Secondary structure elements

are shown as published (7). AMP-bd and Lid domains are indicated by two-
headed arrows.

176

 

Simulation RMSF (A)

 

 

 

 

 

 

 

. E

.. m.—
, m4

N...

. mé

 

 

Pi ... .. ......
........... ...... .. . ....fi....
..«‘ ... ... . .
N .1. .{ «...Jm .. at... p .. .....‘ . .*v. .’.....~.:
...... . ...; 5:... 3...... ......3
Vi. .... . . .
m. .
C. .
N .
1m m w. H.
WNW...”

(v) asw‘a orqdmfioumsflo

 

177

from their average crystallographic RMSF values. Residues 74—79 (the loop
between helix 01.4 and strand [53) show the opposite trend: Although they have the
highest crystallographic RMSF values in 1AKE (1.72 i 0.17 A for 74—79 Ca’s
against an average B factor of 1.001 i 0.19 for all Ca’s), these residues show
simulation RMSF values (2.16 i 0.49 A) only slightly above the average for the

protein (1.93 i 0.94 A). The low crystallographic RMSF values associated with
residues 42—48 in 1AKE might be due to the fact that residues 44. 45, 47, and 48
are involved in crystal contacts (7). In the simulation, the absence of crystal
contacts results in a larger conformational freedom for this region of the protein.
In this respect, it is interesting to note that residues 46—48 adopt different
secondary structures in 1AKE complexes I and II (7). Miiller et al. (10) later
noticed that residues 44—48 are part of helix or; in the apo-enzyme (PDB code:
4AKE). These crystallographic observations suggest that this region is highly
susceptible to conformational changes, and they agree with the large RMSF
values observed for residues 41—48 in our simulation. Because residues 74—79
have the highest crystallographic RMSF values in 1AKE. and because these
residues have lower crystallographic RMSF values in the apo-enzyme, it was
suggested that this loop is part of an energy counterweight in ECAK (10). The
low RMSF values observed for residues 74—79 in our simulation do not support
this energy counterweight hypothesis. Another possibility is that the high
crystallographic RMSF values of residues 74 79 are an artifact of crystallization.
In 1AKE, this loop is involved in crystal contacts: Glu75 forms a H-bond and salt

bridge with Asp76 and Arg78 of another molecule, respectively (7). These

178

interactions might force this loop into a conformation that is not 100% occupied
or that would not be favored in solution. Some of these differences may be
attributable not only to the different conditions involved in crystalline versus
liquid states but, of course, also to the 1AKE* model, where AP5Ais replaced by
the ATP-AMP-Mg2+-4(HgO) complex. The highly charged ATP and AMP
phosphate chains and Mg2+can exert significant forces that extend far into the
protein.Thus, differences can arise from this replacement.The integrity of the
ATP-AMP-Mg2+4(HzO) complex as the simulation proceeds is well maintained.
as displayed in Figure V-3. The Mg2+ keeps its ligation to the ATP y— and B-PO4
oxygens, and the 4 waters are coordinated to the Mg”, forming an octahedral
coordination sphere. The distance between the ATP 'y-PO4 and AMP a—PO4 is
reasonable for a catalytically competent configuration, and residues thought to be

key for catalysis are positioned close to the reactant complex, as we detail below.

The stereochemical quality of 5 snapshots of 1AKE* taken between 900 and 1500

ps was checked using the software PROCHECK v 3.5.4. The results show that 74—79%

of the residues are in the most favored regions of the Ramachandran plot, and l7—23%

are in the additional allowed regions (around 97% total). These numbers are comparable

to values obtained from solution structures of proteins from NMR. These observations

suggest that the protein does not go through disallowed dihedral conformations during the

simulation.

179

Figure V-3: The ATP-AMP-Mg2+-4(H20) complex and several of the (charged)
residues that are within 7A radius of Mg2+ from a snapshot around 2.5 ns. The key
catalytic residues LYS13, ARG156, and ASP84, are displayed. Residues SER30,
GLY14, GLY12, and GLYIO are also within this radius but are not displayed for
clarity.

180

 

“size. 1 56

\

 

181

5.4.2 Substrate specificity

5.4.2.1 AMP binding site

AKs, in general, are known to have higher specificity for their nucleoside monophosphate
(NMP) substrate than for their nucleoside triphosphate (NTP) substrate.40—43 Extensive
mutagenesis studies and structural data have shown that AKs control AMP specificity
through several interactions with highly conserved residues (24, 44, 45). The H—bonding
pattern between AMP and ECAK was analyzed over the complete 3 ns simulation
(Figure V—4). Figure V-S shows all the interactions of AMP and ATP with ECAK
residues that are present during the 3 ns simulation. In the 1AKE crystal structure of the
ECAK - AP5A complex (7), AMP’s adenine interacts with ECAK through 5 strong
H-bonds (2 H-bonds to backbone atoms and 3 H-bonds to side-chains) and several other
less polar interactions. A strictly conserved glutamine residue, Gln92, plays a critical role
both in catalysis and AMP specificity (43, 44, 46) through H-bonds with AMP’s adenine
N6 the orientation of Gln92’s carboxamide is fixed by additional H-bonds to a backbone
amide (from Arg88) and a conserved carboxylate (Asp61). In the simulation, plots (not
shown) of the H-bonds between Asp61, Arg88, and Gln92 versus time show that Gln92’s
side-chain is indeed fixed by very stable H—bonds. Specifically, Gln92zNE2 H-bonds with
Asp61 side-chain oxygens ODl and OD2, and G1n92:OE1 H-bonds with Arg 88:N. Two
other H-bonds between Thr31:OGl, Gly85zO, and the adenine N6 that are stable over the
trajectory also help in discriminating against guanosine monophosphate (GMP), which
lacks the NH2 group. Additionally, Thr31:OGl forms a stable H-bond with AMP’s

adenine N7. Other H-bonds like adenine N6 to Phe86zN and adenine N7 to Gly322N are

182

Figure V-4: Hydrogen bonds and salt bridges between the protein and ATP and between
the protein and AMP during the course of the simulation. Donor and acceptor atoms
within a maximum distance of 3.5 A and a minimum angle of 120° were considered to
form H-bonds. Charged donor and acceptor groups within a maximum distance of 4.5 A
were considered to form salt bridges (indicated by a *). H-bond and salt bridge analysis
was performed every 0.8 ps (i.e., 3750 time points). Shown are only the H-bonds and salt
bridges that are present at least 1.3% of the time (i.e., in at least 50 time points). In the
designation of ATP and AMP, R denotes Ribose, A denotes Adenine and a, B and 7
denote phosphates.

183

 

:AN6
:R02
:a03
2002
:R03
:aOZ
:002

092:081
R167zNE
R88:NH2 *
R88zNH2 *
R88:NH2
R88:NH1 *
K13:NZ *
T31:OG1
R167zNH2
R167:NE
R88zNH1
Q92:NEZ
K13:NZ *
T31:OG1
R88:NH1 *
K13:NZ *
R88:NE
R88:NH1 *
R88:
K57:
G56:
G85:

iibéooo§
N
"

8170:032

R88zNE
El70:OE1
R167:O
E170:OE1

 

184

ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:
ATP:

ATP:
ATP:
ATP:
:RO3

ATP

R156zNH2
R156:NH2
K13zNZ *
K132NZ *
R156:NH2
K13:NZ *
R156zNH1
R123:NH1
R123zNH2
R123zNHl
T15:
G14:
G14
K13
G12
All:
G10:
GlO:
K200:O
T15:OG1
016zNBZ
016:0E1
V132:O
016:NE2
K13:NZ *
0158:0D2
V202:N
R156:NH1
R119:NH1
H134zNDl
G198:O
G122:O
R119:NH2
R119zNH1
R119:NH1
R123:NH1
Y133:O
R156:NE
H134zN
T15:OG1
R123:NH2
T15:OG1
R119zNH2
R156:NH1
G12:N
R156zNH1
T199:O
R124:N
K13:N *
R119:O
K13:N
R156zNH2
H134:ND1
R123zNH1
R156zNH1
R119:O
R124:N
G122:O
O
O

ZZzééézz

R124:
R123:

(D
D
D

Time (ps

0001

0091

M

000

0096

 

0008

cflV

00.0... .5 09050 35.0.. 0;. .05 30505 0. 2090.0 0.... ..0 N
0002500 50... 0305 06503.3 .600 :0 0926 T 05 :52. 02.050 0.0 as}. 0:. .:.< .2 F .0 .0 00m .3... 038....
202.20. 055050. 0:. 0550000 256835 “6:02 w: m 50:2. 05 @550 2.0.0.0 £569.35 ”£060.26
5:00”. .... 059”. 50.... 00620.0 0.02, 060 003.35... mc-m 05 @550 a. .:.< 05... 3:53. x<Um

0002500 0cm 2. 052 0c. .0358. x<Um 0002500 356235 0055 5.... 0:0 0:00-: 053m ”m.> 0.52....

0.32
2...... 0 .2....22
#20 1.0\ 2.:<
_ _ 2.20
ollo 2.2.20.
\ /~:2..w_:z..22 $2422.85.
0 0 e
/o\ _ ...0 .m. -0
roolualolelorfiro»
n
m z __ __ __
Z/v0\ ///~O O O O 25.0
m \ _ fl a a £252..
0/ __ 2.20
/ .0 \.z 2 22 2.22
.z\ /. \ 220 2.10
a 0 .z :0 ~22...
0.2.2:. _ z .0022.
1%...
0.80. m

$2 222......
.20 «00.2.... . .
82.52.02.621 \E0 $0.20
1.0 1.0
3|... 02......
N22...
\ / $2.32
0 0
/O\ _ -OK
fOOrlLalO-
m
m z __ 2
\z/..0\ /~0 0 N22...
.0 _ a 82......
// m__ .2 N22...
52\ 0/®0\
00.5 _
2%:
8.2.2.0
2...... <
0.20
$0.80

.0058.

185

on and off during the simulation (Figure V-4). Another residue, Arg88. forms transient
H-bonds with the AMP adenine N6 and N7, as well as with AMP’s 0t-PO4, and with the
AMP ribose O3 (Figure V—4). These bonds are possible because Arg88’s side-chain is
positioned parallel to the 1: rings of AMP’s adenine. In 1AKE (9), Arg88 occupies 2
different conformations in the 2 molecules of the asymmetric unit. In the BSAK - AP5A
complex (11). Arg88 adopts yet another configuration, suggesting that specific
interactions between Arg88 and AMP are not what make this residue essential in AMP
binding. Arg88 is highly conserved in AKs, and its function in substrate binding and/or

catalysis is the subject of ongoing discussions.

Only 1 good (2.68 A) H-bond between AMP ribose and LysS7zO is seen in the
1AKE crystal structure. Interactions between LysS7zO and Gly56:O and the ribose are
present only during the first 500 ps of the simulation (Figure V-4). However, a H-bond
between the side-chain of Arg167 (involving NHl, NH2, or NE) and the ribose is present
during the entire simulation (Figure V-4). In 1AKE, Arg167 is within H-bonding distance
of the fourth (5) phosphate in AP5A. Examination of snapshots from the simulation show
that the side-chain of Arg167 moves toward the ribose moiety early in the simulation.
The difference in Arg167 orientation between the 1AKE crystal structure and our
simulation may be due to the presence in 1AKE of the extra phosphate in AP5A. In
1AKE, Glu170‘s side-chain points away from the ribose. Starting after 700 ps of our
simulation, Glul70 (OEl and OE2) forms a stable, double H-bond with AMP ribose O2
and O3 (Figure V-4). Whether this double H-bond is significant for AMP is open to

question, because Glul70 is replaced by an alanine or a valine in the

186

B. stearothermophilus, Thermotoga neapolitana, yeast, and bovine heart mitochondrial

AKs.

Several arginines (Arg36, Arg88, Argl l9, Arg123, Argl31, and Arg167) are
situated in the active site cleft surrounding AP5A’s phosphate groups in 1AKE,
presumably to help stabilize the negatively charged phosphates. Some of these arginines
have been shown to participate in phosphate binding and catalysis (7, 47—50). In 1AKE,
APsA’s c-PO4 (that corresponds to AMP’s (It-P04) is well fixed through interactions with
R362NH2, R882NH2, and R1562NH2, the latter interaction being weaker than the first
two (7). AP5A’s c-PO4 sits between the head groups of Arg88 and Arg36. Chemical
modification of ECAK with phenylglyoxal completely inactivates the enzyme (24).
Reinstein et al (24) showed that this complete inactivation is obtained by modification of
a single arginine: Arg88. Mutant R88G is 1000 times less active than the wild-type
ECAK, the Michaelis constant (K.,) values for ATP and AMP increasing up to 5-fold and
85-fold, respectively. The simulation results show that R882NH2 and R882NE have very
stable interactions with AMP’s Ot-PO4. These interactions are salt bridge—like considering
the charges of the groups involved; hence, they are much stronger than H-bonds. In a
previously published 300-ps simulation of the ECAK - AP5A complex. the interactions
between Arg88 and AP5A’s e-PO4, analyzed as H-bonds, were not preserved (29). In the
1AKE* simulation, H—bonds involving Arg36zNI-ll and NH2, and Glul70:OEl and 0E2
appear after 500 ps. The Arg36 H-bonds are to the ribose. whereas in the 1AKE crystal
structure, Arg36 H-bonds an e-PO4 oxygen. It seems that the removal of AP5A’s 6—PO4 to
create ATP and AMP (with the resulting loss of 2 negative charges) weakens some of the

charged and polar interactions between phosphates and arginines and rearranges the

187

H-bond and salt bridge—like interactions. The presence of Mg2+ can also influence the
distribution of charged interactions between the protein Arg residues and the phosphate
groups. Permanent H-bonds are formed between AMP’s a-PO4 and the Lysl3zNZ.
Lysl3, situated in the P loop of all AKs, has been described as the “invariant lysine" due
to dramatic loss of activity by chemical modification. In summary. interactions of Thr31,
Arg88, Gly85. and Gln92 with the adenine N6 are very persistent providing the
possibility of discrimination. Arg88 and Lysl3 are crucial for binding the (X-PO4 and

positioning it suitably to receive a phosphate group from ATP.
5.4.2.2 ATP binding site

The ATP-binding site is formed by residues in the P-loop (residues 7—15), part of the LID
domain, and residues in the C-terminal helix with its connecting loop region (residues
198—214). The ATP H-bonds found in the simulation are shown in Figure V-4, and the
stable interactions are indicated in Figure V-S. Similar to other AKs, ECAK is less
specific for ATP than for AMP (42), and most of the tight interactions between the
protein and Mg-ATP involve the ATP phosphates. In the ECAK - Ap5A complex, ATP‘s
adenine moiety is only loosely bound to the pretein through a single H-bond (7) between
Lys200:O and ATP’s adenine N6. This H bond leads to a preference for ATP over
guanosine triphosphate (GTP). The simulation shows (Figure V-4) that this H—bond
breaks and then reforms during the 3 ns, due to .uctuations in the position of ATP’s
adenine. The position of Lys200 itself does not deviate much from the crystal structure,
and the H—bond between Lys200:O and adenine N6 is formed whenever the adenine

moiety of ATP moves to the appropriate position.

188

In the crystal structure, the ATP adenine is parallel to the Argl l9 guanidinium
group, suggesting a cation—1t interaction (7). The backbones of Pro201 and Val202 are in
van der Waals contact distance of the adenine atoms. The Argl l9—ATP interaction is not
stable during the simulation, owing again to the large movements of ATP’s adenine. N7
of the adenine ring forms an H-bond with Argl l9:NHl and NH2. Although this bond is
relatively infrequent (occurring only about 8% of the time), it occurs on and off during
the simulation (Figure V-4), concurrent with the movement of the adenine ring. A similar
unstable H-bond is also seen between Val 202:0 and ATP:AN6. Two new H-bonds
between ATP:AN6 and either Tyr199:O or Glyl98:O also form more or less in
synchronization with the movement of the adenine ring. Taken together, none of the
H-bonds are strong enough to hold ATP’s adenine in place. Moreover, these interactions
(with the exception of the H-bond involving Lys200) are independent of the base (it can
be either adenine or guanine). While Lys200 is not conserved among the AKs, Argl 19 is
conserved in all known NMP kinases. These observations concur with experimental data
confirming that ATP adenine is not tightly bound to the enzyme, and that it can be
substituted with other NTPs such as GTP. In ECAK, the ATP adenine is quite solvent
accessible, and so is Argl 19: these observations probably lead to the extensive sampling
of conformation space observed for ATP adenine in our MD simulation. They might also
explain why the R128A mutation (equivalent to R119A in ECAK) only moderately

perturbed the kinetic parameters of the chicken muscle AKl enzyme (28).

Only one strong H-bond between the ATP ribose O3 and Tyrl33:O exists in the
crystal structures of ECAK - Ap5A and ECAK - AMP . AMPPNP. It is thought that this

H—bond plays a significant role in the closing of the LID over ATP, since it is one of the

189

few interactions between the LID and AP5A (2). The H-bond between Tyrl33 and ribose
is present only during the first 750 ps of our simulation. Even during this time, the
distance between Tyrl33:O and ribosezO3 is close to 3.5 A, making it a very weak
interaction and hence unlikely to be a driving force in LID closure. A weak H—bond
involving Hisl34zO and ATP's ribose 02 (at 3.48 A in 1AKE) is present until 1000 ps.
Since Hisl34 lies in the LID domain that shows considerable backbone flexibility during
the simulation (Figure V-2), this H-bond is present only during the first nanosecond. An
H—bond between ATP ribose O3 and Argl24:N appears after ~1290 ps. This H-bond is
fairly stable. occurring 78% of the time. Taken together, the ATP ribose, like the adenine,

is also not very tightly bound to the enzyme.

In contrast to ATP ribose and adenine, the ATP phosphates have extensive
contacts with the protein. The triphosphate-binding site is a highly charged pocket
formed by the LID and the core. It is lined with several arginines and the P loop (glycine-
rich), and it is often described as a giant anion hole (51). Phosphate binding, orientation,
and transfer are mediated by Arg36, Arg88, Arg123, Arg156, Arg167, and Lysl3, all
interacting with the P-loop backbone (9). As in the case of AMP, interactions between the
phosphate oxygens and the positively charged side-chains of arginines and lysines are salt

bridge—like in character and hence much stronger than H-bonds.

In 1AKE, a-PO4 shows 3 good H—bonds. It is connected in the P loop to Thr15:N
(at 2.85 A), Thr15zOGl (at 2.71 A), and to Glyl2:N (at 2.98 A). The or-PO4:02 is also
within H—bonding distance from Argl23zNH2 (at 2.91 A). Thr23 (Thr15 in ECAK) in

chicken muscle AKl displays strong nuclear Overhauser effects (NOEs) with the

190

adenosine moiety of Mg-ATP (52), although it is within H-bonding distance from the
ATP 0t-PO4 in the AKl crystal structure. Phosphate stereochemistry experiments in AKl
also suggest that Thr23 directly interacts with 0t-POi (53). This result agrees with the
1AKE structure, in which Thr15:OG1 forms 2 stable H-bonds with (it-P04. Based on
these observations, it was suggested that in AK], Thr23 plays a role in catalysis through
direct interaction with ATP (53). Although Thr23 is conserved in all known AKs, the
mild perturbations of mutant T23A’s kinetic parameters (km, and K,,,) are insufficient to
confirm a functional role for Thr23 (53). In our simulation, Thr15:OGl and Thr15:N
form permanent H-bonds with 0t-PO4 (Figures. V-4 and V-S) in accord with its suggested
catalytic role. Thr15 also forms transitory H-bonds with ATP adenine N6, N7, and N9, in
keeping with its proximity to ATP adenine, as seen in the solution structure of AKl and
in the crystal structure of the yeast AK - Mg-Ap5A complex (13). These H—bonds are

weaker. lasting for at most 500 ps.

The H-bond between Gly12:N and ATP’s Ot-PO.; seen in the 1AKE structure is
present only rarely during the simulation. The on—off nature of this H-bond in the
simulation originates from fluctuations in the position of the phosphate relative to Glyl2.
Instead, we observe a stable H-bond between ATP OL—PO4 and Glyl4:N throughout the
simulation. The salt bridge—like interaction with Arg123zNHl is present only for 100 ps,
with the side-chain of Arg123 moving away from ATP early in the simulation. The ATP
phosphate chain adopts a slightly different position in the equilibrated 1AKE* compared
to that in 1AKE. a-PO4 is farther away from Arg123 in 1AKE* than in 1AKE, leading to

a weaker salt bridge—like interaction. The side-chain of Arg123 is thus able to fluctuate

191

and move away from or-PO4 during the simulation. This side-chain. however, moves
within 4.5 A of ATP y-POi, thus creating another potential salt bridge interaction. Note
that in the 1AKE crystal structure, Arg123zNH1 is 3.94 A from a y—PO4 oxygen providing
a bifurcated situation that, when the extra AP5A P04 is removed, most likely leads to the
observed repositioning of Arg123 in the simulation. In our simulation, the charge
compensation for (x-POi is provided by its salt bridge interaction with Arg156zNHl,
which is present about 50% of the time (V- 4), rather than its salt bridge with
Arg123:NHl. Another H-bond not found in 1AKE forms between ATP (I-PO4 and
Lysl32N after 600 ps, and it is present during most of the remaining 2.4 ns. Argl l9:NHl
is within 4.5 A of or-PO4 until about 1750 ps. Even during this time, the interaction is

weak and intermittent.

In 1AKE, B-POi is very well connected to the P loop through as many as 4 strong
H-bonds with Alal lzN. Glyl2zN, Lysl3zN, and Glyl4:N, and through a salt bridge with
Lysl3zNZ. Almost all the H-bonds involving ATP B-PO4 have a strong presence
throughout the simulation (Figure V-4). The invariant P-loop lysine, Lysl3, forms a
permanent H-bond with Ala8:O during the simulation (not shown), thus acting as a
bridge between Ala8 and ATP phosphate chain. The stability of this H-bond underlines
the importance of the bridge in maintaining the integrity of the P-loop, since Ala8 and
Lysl3 are at opposite ends of the P-loop. Another H-bond between Lysl3zNZ and
Gly7:O is also present throughout the simulation, confirming the structural importance of

Lysl3. At no point is there an H-bond between the side-chain of Arg123 and B-POi as

192

seen in 1AKE. With a donor -acceptor distance of 3.3 A in 1AKE, the H-bond between

Argl23zNH2 and B-PO4 is weak even in the crystal structure.

The to-be-transferred y—PO4 is loosely bound, with fewer interactions with the
active site residues than or-PO4 and B-PO4 in 1AKE, only forming salt bridge-like
interactions with Lysl3zNZ and Arg1232NH2, and weakly interacting with Arg156zNH2.
In the BSAK . Mg-Ap5A complex, y—PO4 is within H—bonding distance of Lysl3, and
within 4.5 A of the side-chains of Arg123 (NHl and NH2) and Arg156. These H-bonds
have donor—acceptor distances of 3.3 A or more, indicating weaker interactions with the
protein as compared to ATP’s a-PO4 and B-PO4. In our simulation, the salt-bridges
between y—POi and Arg156zNHl and NH2 are present as in the crystal structure. The
interactions with Lysl3 and Arg123 are not stable in the simulation. They are replaced by
other interactions arising from the presence of Mg2+, and from the enhanced
conformational range available to y—PO4, which is no longer covalently attached to
APsA’s 5-PO4. y—PO4 is now free to rotate and, with its negative charges. it strongly
interacts with Mg“. An additional H—bond not seen in the crystal structures is seen in our
simulation between Gly10:N and y—PO4. This H-bond is stable throughout the simulation

and it is the only link between y—PO4 and the P loop.

In summary. we see about the same interactions between the ATP phosphate
chain and the P loop (main-chain and side-chains) as in the crystal structure, but we see
more and stronger interactions with Arg156 than with Arg123, for both B-POi and y-POi.

Without exception. all permanent H-bonds between ATP and the protein in 1AKE* in the

193

simulation involve the 3 ATP phosphate groups (Figures V-4 and V-5), the ATP adenine

and ribose moieties being only loosely bound to the protein.
5.4.2.3 Role and conformation of active site residues around Mg2+

Kinases require a divalent cation, usually Mg}, for catalysis. The ATP 13- and
y—phosphates are coordinated to Mg2+, and the true substrate for AKs is Mg2+-ATP. Mg2+
serves to neutralize the highly negative charge on the ATP triphosphate, thereby reducing
electrostatic repulsion of the transferred phosphate and increasing the efficiency of
nucleophilic attack. In AKs, conserved serine, threonine, or aspartate residues participate
in binding the Mg2+ cation, coordinating it either directly in the first coordination sphere

or indirectly through water molecules.

In porcine muscle AK - CoATP, AK - CoGTP, and AK - CoGDP, 3'P NMR
relaxation rates indicate that Co2+ is directly coordinated to only B-PO4 (GDP) or B-PO4
and y—PO4 [ATP and GTP] (54). In addition, a-PO4 is too far from Co2+ to be in the first
coordination sphere, but also too close to be in the second coordination sphere (not
enough room for a water molecule). These results suggest that ATP-Mg2+ binds AK as a
[3,7 bidentate complex rather than an 0t,B,Y tridentate complex. In the structure of the
BSAK - Mg2+-AP5A complex, Mg2+ is coordinated to the phosphates corresponding to
ATP’s [3— and y—phosphates, and to 4 water molecules numbered 300—303 ( l l).
Mg2+-AP5A is anchored to BSAK through several H-bond networks mediated by the 4
Mg2+-coordinating water molecules. Waters 300 and 303 help orient the a-PO4 of AMP.
Water 300 is also positioned to interact with Arg36 and Asp33, thus linking Mg2+-AP5A

to the protein. Water 302 is within H-bond distance of ATP‘s 3 phosphate groups,

194

reinforcing the linkage between the phosphate chain and Mg2+. Water 301 H-bonds to
Asp84zOD2 and Glyl4:N, and it is the only water that does not interact with the
phosphates. It is evident that this H-bond network is important for orienting the donor
and acceptor phosphates and aiding the ternary complex to proceed to the transition state.
Based on their 300 ps simulation results of ECAK - AP5A. Kern et al. predicted that 3
aspartate residues, Asp84, Asp33, and Aspl 10, are possible candidates for binding the
Mg2+ complex (29). Ser30, Thr3l, and Thr89 were also suitably placed and were
available to interact with Mgz‘”. Orientation of the side-chains of these residues and the

position of the phosphate chain left just enough room for the M g2+ ion.

In the simulation, Mg2+ remains coordinated to 4 waters (218—221) as dictated by
the force field (see the Methods section). The water—Mg“ distances are short in the
crystal structure, indicating that these waters are ligated to the ion. Then, one can choose
to modify the force field by inserting specific “covalent” bonds, or by using noncovalent
but relatively strong interactions (55), as we have done. Even though no such interactions
were introduced between Mg2+ and ATP 13- and y—phosphates, we find that throughout the
simulation Mg2+ is also coordinated to ATP’s B-PO4 through its 02 oxygen, and to ATP’s
y—PO4 through its 01 oxygen. As shown in Figure V-6, the coordination is close to
octahedral. Furthermore, the additional flexibility introduced by having ATP/AMP versus
APSA in the structure, as in the 1210 crystal structure. still preserves the octahedral

geometry.

A second coordination sphere for Mg2+ includes Asp84: ODl and OD2 that have

waters 219 and 221 (equivalent to waters 301 and 303 in BSAK) screening this direct salt

195

Figure V-6: Mg2+ coordination sphere. The four Mg2+-coordinated waters, the B-PO4zO2,
and the y-PO4:OI form an essentially octahedral first coordination sphere. The multi-
functional role of Lysl3 in anchoring ATP and AMP is evident. Asp84 is in the second
coordination sphere and is H-bonded to 2 of the Mg2+—coordinated waters. Arg156 is
almost in hydrogen bonding distance of the other two Mg2+-coordinated waters and may
be positioned to participate in catalysis. See text for further discussion. Distances are in
angstroms.

196

 

 

 

 

[hr 31‘?
r. 11.: L
in [1: ti

11611 0'3

£117.03

 

197

bridge—like interaction. The H-bonds between water molecules 219 and 221 and the side-
chain of Asp84 are very stable over the 3 ns trajectory (not shown), indicating the
importance of Asp84 in binding the Mg“ water complex. In AKl, mutation of the
corresponding Asp93 to alanine decreased the km, 650-fold (56). The D93A mutant
showed no structural perturbation as evidenced by NMR analysis, indicating a local
effect. Moreover, the AKl D93A mutant had a markedly lower affinity for Mg2+ (56).
Similar effects were also observed for the ECAK D84A mutant (26). Considering the fact
that Mg” plays a dual structural and chemical role in catalysis, these results suggest that
Asp84 is important in the structural role of Mg“. Specifically, in the AKl D93A mutant,

Mg2+ may be unable to orient the phosphate chain for the transfer (56).

Throughout the simulation, Asp84’s main- and side-chains form a stable network
of H-bonds. Besides the above—mentioned H-bonds with waters 219 and 221, Asp84
forms stable H-bonds with Ser30 and Thr31. The H-bond between Asp84 and Thr31
backbone atoms is particularly significant, since Thr3l is also stably linked to AMP’s
adenine. Thus, these interactions anchor AMP to the Mg2+ complex involving the ATP
phosphate chain. Asp84 interacts with Ser30 and lle29 through strong H-bonds and also
forms a persistent salt bridge with Lysl3zNZ in the P-loop. Additionally, Ser30 is by

itselfcoordinatcd to the Mg2+ ion through water molecule 219.

As in BSAK, water molecule 220 (water 302 in 1210), is within H-bonding
distance of ATP 0t-, [3. and y—phosphates, strengthening the Mg2+—phosphate complex
and properly orienting the ATP phosphate chain. Interactions not seen in BSAK but
transiently present in the 1AKE* simulation include H-bonds between Arg156 and waters

220 and 218. Arg156 has been implicated in the stabilization of the transition state.

198

Another rather weak H-bond is also seen between Glyl4:N and water molecule 219. A

corresponding H-bond is present in 1210 involving Glyl4 and water molecule 301.

Lysine, with its backbone amide, long and flexible side-chain, and charged head
group, is well-suited for a multifunctional role. These features are apparent in Figure V-6,
where Lysl3 stabilizes the active site by forming H-bonds and salt bridge—like
interactions with both ATP and AMP using its flexibility and multifunctionality. The

other interactions of Lys 1 3 are discussed in previous sections.
5.4.3 Catalytic Mechanism

The 1AKE* simulation can be used to suggest whether the mechanism of phosphoryl
transfer is associative or dissociative. at least from the perspective of geometrical
requirements. The consensus is that in enzymes in general (31—34, 57). and in AKs in
particular (25, 43), an associative mechanism is operative. The geometric criterion for an
associative mechanism is that the transition state has sufficiently short entering and
leaving group distances to support reasonably large fractional bond numbers, n. For
example, if an 8N2 mechanism were operative (n = 1/2) the P—O distance would be 1.91 A,
while if a fully associative mechanism were operative (n = 1), the P—O distance would be

1.73 A. the 11.0 single-bond distance.

In Figure V-7, the ATP y—P to AMP OL—P distance is plotted during the 3 ns
simulation time. For most of the time, this distance ranges between 4.5 A and 5.0 A,
about I A greater than the 3.8 A van der Waals P—P contact distance. In the absence of
the Mg2+ cation and Lys 13, with its ability to span ATP and AMP, it would be difficult to

support the close approach of these (negatively charged) phosphate groups. The P—P

199

 

 

PE_PG distance

 

 

8
< 7 — ‘ A —— ——
8
C 6 , _._ l t. _ - 7 fi, k a y _
2
g 5 .rffl'l'll It . m. r r r [I] m

4 ‘ 1 ,

0 500 1000 1500 2000 2500 3000
Time (ps)

Figure V-7: Distance between ATP y—P and AMP a-P atoms during the 3-ns simulation

200

 

 

distance suggests that an associative mechanism could be operative in AKE, as illustrated
in Figure V—8. The atoms ATP B—P, Lysl3zNZ and AMP Ot—P form an essentially
equilateral triangle with apex Lysl3zNZ and equal legs ATP B-P—Lysl31NZ and
Lysl3zNZ~AMP Ot-P. The atoms Lysl3zNZ, ATP B-P, ATP y—P, and AMP Ot—P form a
plane. where the ATP B-P—AMP (x-P distance is approximately twice that of the ATP [3
P—ATP y-P distance. Thus. there would be room to rotate the ATP y-PO4 into a line
formed by ATP B—P, ATP y—P, and AMP 0t—P and produce a trigonal bipyrimidal
phosphorane transition state, as suggested by Reinstein et al (25). In this configuration.
Lys132NZ would be positioned above the 3 apical oxygens, those in the plane
perpendicular to the ATP B—P, ATP 'y—P, and AMP a—P line. The Mg2+ (4H20) complex
is below the above-mentioned plane, with Mg“ equidistant from the oxygens (shown as
spheres) associated with ATP B—P and ATP 'y—P. From these geometrical considerations,
the role played by Mg2+ (4H20) and Lysl3 in maintaining the proximity and orientation

of the ATP and AMP phosphates appropriate for phosphoryl transfer is clarified.

5.5 Conclusions

MD simulation of the ECAK - Mg—ATP - AMP ternary reactant complex that we
constructed shows general agreement with conclusions drawn from the analysis of the
crystallographic and other studies of ECAK complexed with AP5A. The strong repulsive
interaction between the terminal, negative phosphate groups that results from the split of

AP5A into ATP and AMP is counteracted by the presence of Mg” and, presumably,

201

 

Figure V-8: Goemetry of M92+ association with its four coordinating waters, ATP and AMP
phosphates,and Ly513. Distances are in A.The displayed distances allow room for the ATP
y—PO4 to rotate to form an ATP B-P, ATP y—P and AMP a—P line and produce a trigonal bipy—
rimidal phosphorane transition state.-

202

some of the positively charged surrounding residues. The AMP adenine N6 strongly
interacts with the protein through H-bonds, which forms the basis for NMP specificity. In
contrast, and also in agreement with conclusions drawn from the crystallographic data
(6), the ATP adenine and ribose moieties are only loosely bound to the protein, while the
phosphates interact strongly with nearby residues. Octahedral coordination of the Mg2+
by 4 waters and by oxygens of ATP B- and ’y-phosphates is maintained throughout the
simulation. The conserved Lysl3 in the P loop bridges the ATP and AMP phosphates, a
phenomenon that relies on the lysine side-chain's unique properties of flexibility and
H-bond—salt bridge capability. In addition to Lysl3, the Mg“. its coordination waters.
and some surrounding charged residues maintain the AMP (x-phosphate and ATP B- and
y phosphates in a configuration suggesting that phosphoryl transfer occurs by an
associative mechanism in AK. The simulation results could provide a starting point for a
combined quantum mechanical—molecular mechanical simulation of the reaction

mechanism for phosphoryl transfer.

203

5.6

10.

ll.

12.

References

Walsh C. 1979. Enzymatic Reaction Mechanisms. San Francisco: W. H. Freeman
and company 978 p.

Schulz GE, Muller CW, Diederichs K. 1990. Induced-fit movements in
adenylate kinases. J Mol Biol 213:627—630.

Schulz GE. 1992. Induced-fit movements in adenylate kinases. Faraday Discuss
93:85—93.

Vieille C, Krishnamurthy H, Hyun HH, Savchenko A, Yan H, Zeikus JG.
2003. Thermotoga neapolitana adenylate kinase is highly active at 30°C.
Biochem J 372:577—585.

Dreusicke D, Karplus PA, Schulz GE. 1988. Refined structure of porcine
cytosolic adenylate kinase at 2.1 A resolution. J Mol Biol 199:359—371.

Miiller CW, Schulz GE. 1988. Structure of the complex of adenylate kinase

from Escherichia coli with the inhibitor P1,P5-di(adenosine-5'-)pentaphosphate. J
Mol Biol 202:909—912.

Miiller CW, Schulz GE. 1992. Structure of the complex between adenylate
kinase from Escherichia coli and the inhibitor Ap5A refined at 1.9 A resolution.
A model for a catalytic transition state. J Mol Biol 224:159—177.

Miiller CW, Schulz GE. 1993. Crystal structures of two mutants of adenylate
kinase from Escherichia coli that modify the Gly-loop. Proteins 15:42—49.

Berry MB, Meador B, Bilderback T, Liang P, Glaser M, Phillips GN, Jr.
1994. The closed conformation of a highly flexible protein: the structure of E. coli
adenylate kinase with bound AMP and AMPPNP. Proteins 19:183—198.

Miiller CW, Schlauderer GJ, Reinstein J, Schulz GE. 1996. Adenylate kinase
motions during catalysis: an energetic counterweight balancing substrate binding.
Structure 4: 147—156.

Berry MB, Phillips GNJ. 1998. Crystal structures of Bacillus stearothermophilus
adenylate kinase with bound Ap5A, Mg2+ Ap5A, and Mn?” Ap5A reveal an
intermediate lid position and six coordinate octahedral geometry for bound Mg2+
and Mn”. Proteins: Struct Function, Genet 32:276—288.

Abele U, Schulz GE. 1995. High-resolution structures of adenylate kinase from

yeast ligated with inhibitor Ap5A, showing the pathway of phosphoryl transfer.
Protein Sci 4: 1262-127 1.

204

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

Egner U, Tomasselli AG, Schulz GE. 1987. Structure of the complex of yeast
adenylate kinase with the inhibitor P1,PS-di(adenosine-5'-)pentaphosphate at 2.6
A resolution. J Mol Biol 195:649—658.

Diederichs K, Schulz GE. 1990. Three-dimensional structure of the complex
between the mitochondria] matrix adenylate kinase and its substrate AMP.
Biochemistry 29:8138—8144.

Schlauderer GJ, Proba K, Schulz GE. 1996. Structure of a mutant adenylate
kinase ligated with an ATP-analogue showing domain closure over ATP. J Mol
Biol 256:223—227.

Schlauderer GJ, Schulz GE. 1996. The structure of bovine mitochondrial
adenylate kinase: Comparison with isoenzymes in other compartments. Protein
Sci 5:434—441.

Sheng XR, Li X, Pan XM. 1999. An iso-random Bi Bi mechanism for adenylate
kinase. J Biol Chem 274:22238—22242.

Sinev MA, Sineva EV, Ittah V, Haas E. 1996 Domain closure in adenylate
kinase. Biochemistry.35:6425—6437.

Lin Y, Nageswara Rao BD. 2000. Structural characterization of adenine
nucleotides bound to Escherichia coli adenylate kinase. 1. Adenosine
conformations by proton two-dimensional transferred nuclear Overhauser effect
spectroscopy. Biochemistry 39:3636—3646.

Lin Y, Nageswara Rao BD. 2000. Structural characterization of adenine
nucleotides bound to Escherichia coli adenylate kinase. 2. 31P and 13C relaxation

measurements in the presence of cobalt(II) and manganese(II). Biochemistry
39:3647-3655.

Shapiro YE, Sinev MA, Sineva EV, Tugarinov V, Meirovitch E. 2000.
Backbone dynamics of Escherichia coli adenylate kinase at the extreme stages of
the catalytic cycle studied by (15)N NMR relaxation. Biochemistry 39:6634—
6644.

Shapiro YE, Kahana E, Tugarinov V, Liang ZC, Freed JH, Meirovitch E.
2002. Domain flexibility in ligand-free and inhibitor-bound Escherichia coli

adenylate kinase based on a mode—coupling analysis of N-15 spin relaxation.
Biochemistry 41:6271—628 1.

Reinstein J, Brune M, Wittinghofer A. 1988. Mutations in the nucleotide
binding loop of adenylate kinase of Escherichia coli. Biochemistry 27:4712—
4720.

Reinstein J, Gilles AM, Rose T, Wittinghofer A, Saint Girons I, Barzu O,
Surewicz WK, Mantsch HH. 1989. Structural and catalytic role of arginine 88 in

205

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

Escherichia coli adenylate kinase as evidenced by chemical modification and site-
directed mutagenesis. J Biol Chem 264:8107—8112.

Reinstein J, Schlichting I, Wittinghofer A. 1990. Structurally and catalytically
important residues in the phosphate binding loop of adenylate kinase of
Escherichia coli. Biochemistry 29:7451—7459.

Rose T, Glaser P, Surewicz WK, Mantsch HH, Reinstein J, Le Blay K, Gilles
AM, Barzu O. 1991. Structural and functional consequences of amino acid
substitutions in the second conserved loop of Escherichia coli adenylate kinase. J
Biol Chem 266:23654-23659.

Rose T, Brune M, Wittinghofer A, Le Blay K, Surewicz WK, Mantsch HH,
Barzu O, Gilles AM. 1991. Structural and catalytic properties of a deletion
derivative (D133-157) of Escherichia coli adenylate kinase. J Biol Chem
266:10781—10786.

Tsai MD, Yan HG. 1991. Mechanism of adenylate kinase: site-directed
mutagenesis versus X-ray and NMR. Biochemistry 30:6806—6818.

Kern P, Brunne RM, Folkers G. 1994. Nucleotide-binding properties of
adenylate kinase from Escherichia coli: a molecular dynamics study in aqueous
and vacuum environments. J Comput Aided Mol Des 8:367—838.

Elamrani S, Berry MB, Phillips GN, McCammon JA. 1996. Study of global
motions in proteins by weighted masses molecular dynamics: adenylate kinase as
a test case. Proteins 25:79—88.

Knowles JR. 1980. Enzyme-catalyzed phosphoryl transfer reactions. Ann Rev
Biochem 49:877—919.

Hollfelder F, Herschlag D. 1995. The nature of the transition-state for enzyme-
catalyzed phosphoryl transfer - hydrolysis of O-aryl phosphorothioates by
alkaline-phosphatase. Biochemistry 34: 12255—12264.

Matte A, Tari LW, Delbaere LTJ. 1998. How do kinases transfer phosphoryl
groups? Structure 6:413—419.

Lahiri SD, Zhang GF, Dunaway-Mariano D, Allen KN. 2003. The

pentacovalent phosphorus intermediate of a phosphoryl transfer reaction. Science
299:2067—207 1 .

van Gunsteren WF, Billeter SR, Eising AA, Hiinnenberger PH, Kriiger P,
Mark AE, Scott WRP, Tironi IG. 1996. Biomolecular Simulation: The
GROMOS96 Manual and User Guide. Ziirich: Vdf Hochschulverlag AG an der
ETH Ziirich.

206

 

37.

38.

39.

40.

41.

42.

43.

45.

46.

47.

48.

Tsodikov OV, Record MTJ, Sergeev YV. 2002. A novel computer program for
fast exact calculation of accessible and molecular surface areas and average
surface curvature. J Comput Chem 23: 600—609.

Cukier RI, Seibold SA. 2002. Molecular dynamics simulations of prostaglandin
endoperoxide H synthase-1: Role of water and the mechanism of compound I
formation from hydrogen peroxide". J Phys Chem B 106: 12031-12044.

Hockney RW, Eastwood JW. 1981. Computer Simulation Using Particles. New
York: McGraw-Hill. 540 p.

Allen MP, Tildosley DJ. 1987. Computer Simulation of Liquids. Oxford:
Clarendon Press. 385 p.

Tomasselli AG, Noda LH. 1980. Mitochondrial ATP:AMP phosphotransferase
from beef heart: purification and properties. Eur J Biochem 1032481—491.

Ito Y, Tomasselli AG, Noda LH. 1980. ATP:AMP phosphotransferase from
baker's yeast. Purification and properties. Eur J Biochem 105:85—92.

Saint Girons I, Gilles AM, Margarita D, Michelson S, Monnot M,
Fermandjian S, Danchin A, Barzu O. 1987. Structural and catalytic
characteristics of Escherichia coli adenylate kinase. J Biol Chem 262:622—629.

Yan H, Tsai MD. 1999. Nucleoside monophosphate kinases: structure,
mechanism, and substrate specificity. Adv Enzymol Relat Areas Mol Biol 73:103-
134.

Beichner S, Byeon I-JL, Tsai M-D. 1996. Structure-function relationship iof
adenylate kinase: Glu-101 in AMP specificity. In: Kaumaya PTP, Hodges RS,
editors. Peptides: Chemistry, Structure and Biology. Kingswinford, England:
Mayflower Sicentific Ltd.; p 721—723.

Okajima T, Tanizawa K, Yoneya T, Fukui T. 1991. Role of leucine 66 in the
asymmetric recognition of substrates in chicken muscle adenylate kinase. J Biol
Chem 266: 1 1442—1 1447.

Okajima T, Tanizawa K, Fukui T. 1993. Site-directed random mutagenesis of
AMP-binding residues in adenylate kinase. J Biochem 114:627—633.

Kim HJ, Nishikawa S, Tokutomi Y, Takenaka H, Hamada M, Kuby SA,
Unsugi S. 1990. In vitro mutagenesis studies at the arginine residues of adenylate
kinase. A revised binding site for AMP in the X-ray-deduced model.
Biochemistry 29:1107-1 1 1].

Yan HG, Dahnke T, Zhou BB, Nakazawa A, Tsai MD. 1990. Mechanism of
adenylate kinase. Critical evaluation of the X-ray model and assignment of the
AMP site. Biochemistry 29:10956—10964.

207

 

49.

50.

51.

52.

53.

54.

56.

57.

Yan HG, Shi ZT, T sai MD. 1990. Mechanism of adenylate kinase. Structural
and functional demonstration of arginine-138 as a key catalytic residue that
cannot be replaced by lysine. Biochemistry 29:6385—6392.

Dahnke T, Shi Z, Yan H, Jiang RT, Tsai MD. 1992. Mechanism of adenylate
kinase. Structural and functional roles of the conserved arginine—97 and arginine-
132. Biochemistry 31:6318—6328.

Dreusicke D, Schulz GE. 1986. The glycine-rich loop of adenylate kinase forms
a giant anion hole. FEBS Lett 208:301—304.

Byeon IJ, Yan H, Edison AS, Mooberry ES, Abildgaard F, Markley JL, Tsai
MD. 1993. Mechanism of adenylate kinase. 1H, 13C, and 15N NMR

assignments, secondary structures, and substrate binding sites. Biochemistry
32: 12508—12521.

Shi Z, Byeon IJ, Jiang RT, Tsai MD. 1993. Mechanism of adenylate kinase.
What can be learned from a mutant enzyme with minor perturbation in kinetic
parameters? Biochemistry 32:6450—6458.

Ray BD, Rosch P, Rao BD. 1988. 31P NMR studies of the structure of cation-
nucleotide complexes bound to porcine muscle adenylate kinase. Biochemistry
27:8669—8676.

Banci L. 2003. Molecular dynamics simulations of metalloproteins. Curr Opin
Chem Biol 7: 143—149.

Yan HG, Tsai MD. 1991. Mechanism of adenylate kinase. Demonstration of a
functional relationship between aspartate 93 and Mg2+ by site-directed

mutagenesis and proton, phosphorus-31, and magnesium-25 NMR. Biochemistry
30:5539—5546.

Mildvan AS. 1997. Mechanisms of signaling and related enzymes. Proteins-
Structure Function and Genetics 29:401-416.

208

Chapter 6

Conclusions and future directions

The popular hypothesis regarding the stability of hyperthermophilic enzymes at high
temperatures postulates that these enzymes have enhanced conformational rigidity in
their folded native state. The enhanced rigidity is responsible for the low catalytic activity
of these enzymes at low temperatures. To date, no other physical explanation regarding
the constancy in the maximal catalytic rates of hyperthermophilic and mesophilic enzyme
pairs at their respective source organisms’ optimal growth temperatures, has gained as
much attention. This chapter summarizes how TNAK may not conform fully to this
hypothesis, and how TNAK, and more enzymes like it should be studied to deepen our
understanding of the link between flexibility, stability, and activity.

The hypothesis implies that, as the temperature increases, the thermal fluctuations
in an enzyme increases, leading to increased catalytic activity up to the optimal rate that
can be measured. Beyond the temperature for optimal activity, the increased fluctuations

lead to instability in the enzyme structure, loss of activity, and global unfolding. This

209

 

simplistic view of conformational dynamics links local conformational fluctuations and
global stability and thereby links thermodynamic stability with maximal catalytic
activity. A practical outcome of this hypothesis would be that thermostability and
enhanced catalytic activity, desirable properties in many enzymes with industrial
applications, cannot be achieved. However, in contrast to this expectation, several
enzymes have been evolved in the laboratory for enhanced thermostability while
maintaining low temperature activity (2, 12), suggesting that the low activity of
hyperthermophilic proteins at low temperatures has its basis in evolution rather than in
any physico-chemical limitation. Moreover, TNAK which is as active as ECAK at 30°C
is itself a good example of a naturally evolved enzyme that contradicts this expectation of
a tradeoff between activity and stability. ECAK and TNAK therefore, make a unique and
interesting enzyme pair to study the interplay between flexibility, activity, and stability.
On the basis of physical principles, linking structural rigidity to global stability is
overly simplistic. Changes along the protein sequence can change flexibility and stability
in different ways. As an instance, it has been shown that salt-bridges can be
thermodynamically destabilizing at room temperature due to the desolvation penalty
associated with their formation (3). Introducing a mutation in a protein interior to form a
salt-bridge may result in increased rigidity but also cause thermodynamic instability. On
the other hand, a cavity-inducing mutation in the interior of a protein is likely to increase
flexibility and decrease stability. Flexibility depends on the energy landscape of proteins
near their native states. Proteins are complex systems with different modes of fluctuations
that span from ps to greater than ms timescales. These fluctuations contribute differently

to activity and stability. For example, hinge-bending domain movements as seen in AKs

210

need not be correlated to the stability of the domains themselves. Hence, flexibility for
activity is not necessarily coupled to flexibility for stability, unless active site residues are
involved in both. The first step towards understanding the interplay between flexibility,
stability, and activity is to recognize the range of motions, their amplitude and their
spatial distribution in the proteins under investigation. In this study, TNAK was used as a
model system to investigate dynamics at every residue, in the ps, ns, and us timescales as
well as longer timescales (in the order of minutes to days), to understand how dynamics

on these timescale relate to activity and stability.

6.1 Dynamics in ECAK and TNAK on the ps-ns and us timescales

l5N NMR relaxation was used to investigate ps to us and us timescale backbone
dynamics of TNAK in its open and closed conformations. These results were compared
to the results from a similar study on ECAK (10). The following observations were made
from the comparison:

(i) The average S2 values of the domains showed that no single domain was more
rigid or flexible that the others in both ECAK and TNAK, in their free and
complexed forms. In other words, overall, the core, the lid, and the AMPbd
domains have similar ps-ns timescale flexibility in all four enzyme forms.

(ii) TNAK is more rigid than ECAK on the ps-ns timescale and the increase in
rigidity is uniform across the entire protein.

(iii) Although having overall similar flexibility to the core and lid domains, the

AMPbd domain is singled out as having many residues lower 82 values, i.e.,

211

showing higher amplitude of ps-ns motions. This is true both for ECAK and
TNAK. Many TNAK residues in this region had S2 values comparable to the
corresponding residues in ECAK indicating similar flexibility as ECAK.

(iv) ECAK becomes significantly more rigid upon inhibitor binding. Specifically,
the average 32 values of the inhibitor-bound forms of ECAK and TNAK
become very similar.

(v) The core-lid hinge loops in ECAK*ApSA and TNAK*Ap5A have low

average 82 values compared to the rest of the protein.

 

(vi) On the us timescale also, TNAK is more rigid than ECAK. us timescale
motions are more pervasive in ECAK, present in the core, lid, and AMPbd
domains, and the hinges. In TNAK however, the core B-strands are

specifically excluded from having motions on this timescale. us timescale
motions are found only for residues belonging to the core helices and loops,
the AMPbd and lid domains, and the hinges of these domains.
While TNAK’s increased rigidity is attributable to increased enthalpic contributions from
intramolecular interactions, what does the higher flexibility of AMPbd, lid, and hinge
residues in the ps to ns timescale signify?

Analyses of order parameters in several enzymes have shown that ps-ns timescale
motions are modulated in the free enzyme, in the presence of substrates and inhibitors
and in different mutant forms. Some well studied examples include the HIV protease in
complex with various inhibitors (9). Low order parameters were observed for residues in
the loop preceding the flap that closes over the active site in an inhibitor-bound form of

HIV—protease. The authors suggested that the ps-ns timescale motion in these residues

212

could be important for substrate-binding and product release (7) Similar studies of
different complexes of E. coli dihydrofolate reductase (DHFR) representing different
stages of the catalytic cycle provide insight into how ps-ns timescale motions are
modulated at the different stages of the catalytic cycle and how these motions are related
to catalysis (8, 9). Motions on the ps-ns timescale of key loops in DHFR complexes were
implicated in dynamics related to substrate accessibility to the active site and in product
release. These examples clearly show that ps-ns timescale motions play an important role
in enzyme function, modulation, and substrate recognition. These observations therefore,
raise the question: what is the timescale of kinetically relevant motions? While it is
tempting to ascribe functional significance to only us-ms timescale motions since this
timescale often corresponds to the catalytic tum-over numbers, the timescale of
kinetically relevant motions need not be restricted by the timescale of the measured
catalytic rates. If a catalytic event is not strongly correlated to a functionally relevant
motion, then this motion can occur on a much faster timescale. In this case, several such
dynamic events may be required for a much slower catalytic transition to occur.

The facts that (i) several residues with low S2 values are found mainly in the
AMPbd residues of ECAK and TNAK (ii) that the average S2 values of the core-lid hinge
loops in ECAK*ApSA and TNAK*Ap5A are lower than the rest of the protein and (iii)
that residues in TNAK’s core-lid hinges in both the free and inhibitor-bound forms show
us timescale motions, suggest that these motions are perhaps linked to catalysis related
conformational transitions. Moreover, several TNAK residues in these regions have
similar 82 values as the corresponding ECAK residues. Taken together, the data suggests

that flexibility required for activity may be similar in ECAK and TNAK. Specifically, the

213

 

lid-opening rate for product release in TNAK may be comparable to that of ECAK. Since
lid-opening rate has been shown to be rate-limiting in ECAK catalysis (11), similar lid-
opening rates in TNAK and ECAK may explain their similar activities at 30°C. This then
would be in contrast to the result for another thermophilic AK, the Aquifex aeolicus AK
(AAAK), whose lid-opening rate (kop) was measured using l5N NMR relaxation
dispersion studies (11). It was found that the kop of AAAK was slower than the kOp for
ECAK at 20°C. The difference in the kop for the two enzymes explained well the
difference in their catalytic rates measured at 20°C.

In summary, the 15N NMR relaxation study of TNAK has shown that TNAK is
more rigid than ECAK, in accordance with the hypothesis. However, contrary to the
expectation of the hypothesis, it maintains high activity at low temperatures by localizing
flexibility to regions (such as hinges) that are important for inducing conformational
changes. Since it is the hinges that are important for motions for activity, localizing
flexibility to the hinges while keeping the lid and AMPbd domains rigid does not
compromise TNAK’s activity. Conversely, localizing flexibility in the hinges does not

compromise the stability of the protein.

Future directions

The 15N NMR relaxation method is only an indirect method for detecting us timescale
motions. Deviation of observed T2 values from the values predicted from T1 and NOE
values is used as evidence for motion in the us timescale and is described by the
parameter Rex. Since Rex is directly proportional to the chemical shift difference of the

exchanging nuclei in their different conformational states, absence of Rex does not mean

214

 

there is no conformational exchange for a particular residue. If the chemical environment
sensed by a nucleus in the different conformational states is similar, then no Rex value for
that residue will be observed within experimental error. Moreover, nothing can be said
about the rate of conformational exchange from R.,, values. It was speculated from the
results of the 15N NMR relaxation analyses, that TNAK’s lid-opening rate at 30°C may
be comparable to that of ECAK, since several residues in the hinge between TNAK’s lid
and core domains showed R.,, values. To confirm this theory, accurate values of kop and

kc] (lid-closing rate) can be obtained using 15N NMR relaxation dispersion experiments.

6.2 H-D exchange rate analyses of ECAK and TNAK

H-D exchange is a powerful tool to examine protein dynamics on longer timescales
(minutes-days) in the native state. To complement the ISN NMR relaxation measurements
of TNAK dynamics on the ps-us timescale, H-D exchange experiments were used to
investigate slower motional modes. H-D exchange rates of the open and closed forms of
ECAK and TNAK were measured and compared. The purpose of the H-D experiments
was to compare the distribution of structural rigidity in the two enzymes, in the two
conformational states. Protection factors, P, were used as a measure of stability and
residue-specific P values were calculated. The results show that:

(i) ECAK is not uniformly rigid. ECAK’s core is much more rigid than its

AMPbd and lid domains.
(ii) TNAK’s core is only somewhat more rigid than its AMPbd and lid domains
(iii) The increase in rigidity in TNAK is significant only for the lid and AMPbd

domains. TNAK’s core is only marginally more rigid than ECAK’s core.

215

 

(iv) Zn2+ likely plays an important role in stabilizing TNAK’s lid.

(v) In both ECAK and TNAK, the lid is significantly more stabilized upon
inhibitor binding than the other domains.

(vi) Although the AMPbd and lid domains of TNAK are more rigid than ECAK,
there are several residues in these domains and their hinges that exchange
within the dead time of the experiments indicating significant local
fluctuations.

These results indicate that TNAK is not uniformly more rigid than ECAK at all
timescales. While in the ps-ns and us timescales, TNAK is uniformly more rigid than
ECAK, in the slower timescales of H-D exchange (2 s) only TNAK’s lid and AMPbd are
more rigid than those of ECAK. This result highlights the overly simplistic nature of the
commonly accepted hypothesis. Proteins are characterized by motional modes on a wide
range of timescales. Hence one cannot talk about flexibility in general, as the hypothesis
does, but one must specify the timescale. Also, the hypothesis implies a uniform increase
in rigidity across the structure of a hyperthermophilic protein. While this is true for the
ps-ns timescale for TNAK, it is not true for timescales greater than seconds. Although
TNAK’s AMPbd and lid domains are more rigid than ECAK’s, there are regions in these
domains and their hinges that are characterized by local fluctuations. Flexibility for
activity and rigidity for stability are partitioned in TNAK in a manner that makes it
simultaneously highly active at room temperature and stable at high temperatures.

H-D exchange rates were used to calculate residue-specific AGHX values for Thermus
thermophilus and E. coli ribonuclease (TtRNase and EcRNase) (4). The results showed

that the pattern of stability distribution for the two enzymes were the same with AGHx

216

 

values of TtRNase residues being on average higher than those of EcRNase. Moreover,
each region of the protein contributes the same relative stability to each protein. This is in
contrast to TNAK, where the lid and AMPbd domains contribute proportionally more to
its stability. Therefore the mechanism of stabilization is different for different proteins.
Proteins have individual features of adaptation and hence it may not be possible to find a

stabilization mechanism that is common to all hyperthermophilic proteins.

Future directions

While H-D exchange results reported in chapter 4 provide insight into stability
distribution in ECAK and TNAK, it would be more useful to measure the energetics of
stability in the two enzymes. To obtain protein stability parameters such as free energy,
enthalpy, and entropy of unfolding, the dependence of H-D exchange rates on increasing
denaturant concentration or increasing temperature has to be determined.

As summarized in Chapter 1, exchange in proteins takes place through three
mechanisms. By the first mechanism, exchange occurs due to global unfolding. For
residues that undergo exchange due to global unfolding, the AGHX = AGunf at all
denaturant concentrations, [D], and AGHx varies linearly with [D]. The plot of AGHX vs.
[D] yields a slope called the m value. The m value is roughly proportional to the amount
of surface area exposed in the opening reaction. In native state H-D exchange
experiments, low [D] is used such that less than 1% of the protein molecules are
unfolded. The presence of denaturant simply modulates the equilibrium population of

folded and unfolded molecules. The stability of the protein (AGunf) under native

217

 

conditions is determined by extrapolating the AGHX value to zero concentration of the
denaturant.

Other than the slow exchanging amides, some amides exchange due to partial
unfolding mechanisms and their exchange rate is consequently faster than the slowest
exchanging amides. This is the second mechanism of exchange. For these residues, at low
[D], AGHX < AGunf and the m value is lower than the m value calculated for residues
exchanging via global unfolding. That is, these residues are less dependent on the
denaturant for exchange when [D] is low. As [D] increases, exchange in these amides
becomes dominated by global unfolding and the m values at high [D] become consistent
with the m values of the slowly exchanging amides. Generally, one can find clusters of
residues with similar AGHX, allowing one to identify specific structural elements that
unfold cooperatively.

Finally, for residues exchanging via the third mechanism, no dependence on [D]
is seen. These amides undergo exchange through non-cooperative events such as local
fluctuations involving single amides.

Analysis of exchange rates in ECAK and TNAK at different denaturant
concentrations or temperature will allow one to delineate residues in the exchange core
(the most slowly exchanging amides), structural elements that undergo cooperative, local
unfolding and residues that are involved in local fluctuations. The AGHX for the residues
in the two proteins can be measured and compared, giving us a more detailed view of the

level of structural stability level in different regions of ECAK and TNAK.

218

6.3 MD simulation of ECAK and TNAK

Like NMR, molecular dynamics simulation can provide details, at the atomic level, about
motions in a protein as a function of time. The rapid increase in computing power and
advances in potential energy functions describing the forces in a protein have reached a
point where protein simulations can provide important dynamic aspects that can
withstand critical examination by experimentalists Thus, simulations can now be
designed to answer specific questions about the dynamics of a model system, often more
readily than experimentation. Chapter 5 presented results from the MD simulation of
ECAK in complex with its substrates, AMP, ATP, and Mg”. These simulations were
carried out in collaboration with Dr. Robert. I. Cukier, Department of Chemistry,
Michigan State University. The results presented in this work showed the importance of
H-bonds and salt-bridges between AMP’s adenine and phosphate moieties and ECAK.
These bonds were shown to be the basis for the high specificity of AKs for AMP. The
results also showed ATP, AMP, Mg2+ and its coordinating waters, and the active site
side-chains to maintain an octahedral geometry throughout the simulation, thus lending
strong support to an associative mechanism for phosphoryl transfer.

The simulation results of ECAK complexed with AMP and Mg-ATP show good
agreement with experimental data, validating the fact that the AK system can be
accurately simulated. Several in silico experiments can now be designed to compare the
dynamics of ECAK and TNAK. Before TNAK simulations can be designed, crystal
structures of TNAK in the free and substrate-bound forms are necessary. The accuracy of

MD simulations depends on the quality of structural coordinates provided as input, and

219

 

structures obtained from homology modeling may not yield reliable results. Attempts to
crystallize TNAK in the free form and in complex with various substrates are currently

underway.

Future directions

With the crystal structure of TNAK in hand, simulations investigating the kinetic and
thermodynamic stabilities of ECAK and TNAK can be performed. This would provide
another means by which to compare the enzymes and provide a deeper understanding of
the interplay between flexibility, stability, and activity. The following two MD simulation
methods are of particular interest and are already being carried out on ECAK by Dr.
Cukier’s group:

1. Essential dynamics to study motions of AMPbd and lid domains in the two AKs.

2. Study the potential of mean force (PMF) along the reaction coordinate of

conformational change involving the AMPbd and lid domains.

Thermodynamic stability of a protein is characterized by an ensemble of degenerate
substates spanning a large conformational space (1). Due to thermal fluctuations, protein
molecules experience dynamical processes that allow transitions between these substates
non-periodically (5). These transitions are thought to be responsible for the wide
timescale-range of motions observed in proteins. MD simulations can capture many of
these motions, limited only by the length of the simulation. To extract the most important
motion modes, principal component analysis (PCA) or essential dynamics (ED) is used to
analyze the trajectories from an MD simulation. ED is a statistical method that can

identify slow motion modes in proteins and, thereby, facilitate the study of long timescale

220

 

dynamics. These motional modes are sufficient to describe most of the fluctuations in the
system under investigation. ED analyses of Pyrococcus furiosus and Desulfovibrio
vulgaris rubredoxins (PfRd and Dde, respectively) have revealed three main clusters of
conformations for the two proteins (6). These clusters are much better defined in PfRd
than in Dde, indicating the mesophilic Dde to be more flexible than the
hyperthermophilic PfRd. ED analysis can also be used to define a reaction coordinate for
conformational changes, which is required for another set of simulations called
constrained MD simulations described below.

The second method of simulation of interest, called constrained MD simulations,
uses the reaction coordinate for conformational closing/opening for ECAK and TNAK
obtained from ED analysis, to determine the PMF along the reaction coordinate. The aim
of this simulation is to calculate and compare the free energy barrier for conformational
opening/closing for ECAK and TNAK. For a thermophilic AK such as AAAK, which has
a much slower lid-opening rate than ECAK at 20°C, this simulation may reveal a higher
energy barrier to opening for AAAK than ECAK. TNAK, on the other hand, has similar
activity as ECAK at 30°C. It is expected, therefore, that the barrier to lid-opening for
ECAK and TNAK may be similar.

In conclusion, NMR investigation of TNAK dynamics has shown that TNAK is
able to maintain high activity at low temperatures and stability at high temperatures by
partitioning rigidity for stability and flexibility for activity in its structure in an intricate
manner in different timescales. At the same time, TNAK’s activity at 30°C is about 10
times lower than at 80°C, a fact that may still be attributed to its high rigidity at 30°C.

Although the prevailing rigidity hypothesis has existed for over 3 decades, and many

221

 

mesophilic-hypertherrnophilic enzyme pairs have been studied to test its validity, it is
only now that the hypothesis can be tested thoroughly. Developments in NMR and MD
simulation techniques have opened the possibility of a more careful analyses of the
dynamics in many more such protein pairs in atomic detail. Proteins have individual
features of adaptation and several examples have to be studied to deepen our

understanding of the role of flexibility in stability and activity.

222

6.4

10.

ll.

12.

References

Frauenfelder, H., S. G. Sligar, and P. G. Wolynes. 1991. The Energy
landscapes and motions of proteins. Science 254: 1598-1603.

Giver, L., A. Gershenson, P. O. Froskgard, and F. H. Arnold. 1998. Directed
evolution of a thermostable esterase. Proc. Natl. Acad. Sci. USA 95:12809-12813.

Hendsch, Z. S., and B. Tidor. 1994. Do salt bridges stabilize proteins? A
continuum electrostatic analysis. Protein Sci. 3:211-226.

Hollien, J., and S. Marqusee. 1999. Structural distribution of stability in a
thermophilic enzyme. Proc. Natl. Acad. Sci. USA 96: 13674-13678.

Krumhansl, J. A. 1987. Presented at the Proceedings in Life Sciences, Austin.

Lazaridis, T., 1. Lee, and M. Karplus. 1997. Dynamics and unfolding pathways
of a hyperthermophilic and a mesophilic rubredoxin. Protein Sci. 6:2589—2605.

Nicholson, L. K., T. Yamazaki, D. A. Torchia, S. Grzesiek, A. Bax, S. J.
Stahl, J. D. Kaufman, P. T. Wingfield, P. Y. S. Lam, P. K. Jadhav, C. N.
Hodge, P. J. Domaille, and C. H. Chang. 1995. Flexibility and function in HIV-
] protease. Nature Structural Biology 2:274-280.

Osborne, M. J., J. Schnell, S. J. Benkovic, H. J. Dyson, and P. E. Wright.
2001. Backbone dynamics in dihydrofolate reductase complexes: Role of loop
flexibility in the catalytic mechanism. Biochemistry 40:9846-9859.

Sawaya, M. R., and J. Kraut. 1997. Loop and subdomain movements in the

mechanism of Escherichia coli dihydrofolate reductase: Crystallographic
evidence. Biochemistry 36:586-603.

Shapiro, Y. E., M. A. Sinev, E. V. Sineva, V. Tugarinov, and E. Meirovitch.
2000. Backbone dynamics of Escherichia coli adenylate kinase at the extreme
stages of the catalytic cycle studied by 15N NMR relaxation. Biochemistry
39:6634-6644.

Wolf-Watz, M., V. Thai, K. Henzler-Wildman, G. Hadjipavlou, E. Z.
Eisenmesser, and D. Kern. 2004. Linkage between dynamics and catalysis in a

thermophilic-mesophilic enzyme pair. Nature Structural & Molecular Biology
11:945-949.

Zhao, H. M., and F. H. Arnold. 1999. Directed evolution converts subtilisin E
into a functional equivalent of therrnitase. Protein Engineering 12:47-53.

223