. v 1 . .v z
.1... #1:. .1.“ n.

3 :3 5
3. iii ha

1...}! .11

. é. at...

“£5me

2...”...

3-!

hr... .4.
fig?»
.. . . . x In: a... "m.
miss“? .

M:

..
.5
- $357... ,

. E
. . . hrgmnfiflhr
. 3A— ”“1
. . , A u. b.
. , , . 3.21.99...
.1 . . SN...
3 . ,
. .L. .
u”...

2:1:7T
:3151
1.. .0. 1

 

 

 

 

 

§J 1 11.... .l' 15.
3...! ulrsvl.‘ A: I! n .3

9%.... ; éiyw

.0... A. i alt-$014

, ..I. ..f .. ,
an". k ks”... _

V Lérggg £4...ng

 

.
:1».

This is to certify that the
dissertation entitled

A CRYSTALLOGRAPHIC INVESTIGATION OF LIGAND
BINDING TO OVINE PROSTAGLANDIN ENDOPEROXIDE H
SYNTHASE-1

presented by

Christine A. Harman

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Biochemistry and Molecular
Biolggy

 

7/ 77/7/-

V Major Professor’s Signature
/' 2/31/05

Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

UBRARY ‘
Michig State
University

 

 

 

.---.—.—-_.-

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:lClRC/DaIeDue.indd-p.1

 

 

A CRYSTALLOGRAPHIC INVESTIGATION OF LIGAND BINDING TO OVINE
PROSTAGLANDIN ENDOPEROXIDE H SYNTHASE-l

By

Christine A. Harman

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Biochemistry and Molecular Biology

2005

ABSTRACT

A CRYSTALLOGRAPHIC INVESTIGATION OF LIGAND BINDING TO OVINE
PROSTAGLANDIN ENDOPEROXIDE H SYNTHASE-l

By

Christine A. Harman

Prostaglandin Endoperoxide H Synthase-l (PGHS-l) and -2 (PGHS-Z) also
known as Cyclooxygenase-l (COX-1) and —2 (COX-2) catalyze the committed step in the
formation of prostaglandins and thromboxanes, in which two molecules of oxygen are
incorporated into AA forming PGHz. Past kinetic studies have indicated that arachidonic
acid (AA) can exist in at least three catalytically competent conformations within the
COX active site each of which lead to either PGGZ, the productive product, 11-HETE and
15-HETE (abortive products). Two single mutations V349A and W387F were found to
create almost equal amounts of PGG2 and 11-HETE. Results discussed in Chapter 11
show that a V349A/W387F mutant of ovine PGHS-l produced predominantly ll-I-IETE.
The crystal structure of AA bound to V349A/W3 87F oPGHS-l was determined to 3.1 A.
The conformation of AA in this double mutant was observed to be distinctively different
than the conformation of AA observed in the native enzyme, providing structural
evidence that the conformation of AA occurring at the moment of hydrogen abstraction
pre-determines the product of the COX reaction. Further, the roles of residues V349 and
W387 were found to be crucial in the properly positioning atoms C9 and C11 of AA for
endoperoxide formation.

We also investigated the binding of Mead acid, a non-essential fatty acid, to
PGHS-l. Although Mead acid was found to be a poor substrate for PGHS-l, products

are formed and resemble abortive COX products indicating that Mead acid may undergo

a similar mechanism in which the C13 proS hydrogen is removed, the rate limiting step.
A crystal structure of Mead acid bound to PGHS-l was determined to 2.85 A; but only
weak density was observed for Mead acid in the COX active site. Simulated annealing
experiments were used to observe a series of Mead acid conformers. Two binding
populations, the “catalytically possible” and inhibitory conformations, were observed.
Six of ten Mead acid conformers were observed in “catalytically possible” conformations
suggesting that Mead acid is capable of binding in a conformation that would support
catalysis. However, the inhibitory conformers demonstrated that Mead acid binds in
manner that would not support catalysis suggesting that the missing (96 double bond in
Mead acid is important for positioning the proS hydrogen of C13 for potential
abstraction. The ability of Mead acid to act as an inhibitor for PGHS-l was tested and
the Ki of Mead acid was determined to be ~62 uM indicating that Mead acid is not an
efficient inhibitor of PGHS. The combination of these studies demonstrate that the (06
double bond plays an important role in facilitating binding and positioning of 20-carbon
fatty acids.

PGHS-l is inhibited ~ten times more efficiently by the (S)-enantiomers of the or-
substituted indomethacin ethanolamides. Crystal structures of an enantiomeric pair of
these inhibitors were determined to 2.7-2.85 A, in an effort to understand the structural
basis of this enantiomeric selectivity of PGHS-l. Comparisons of the structures reveal
there are two different binding modes of these inhibitors, in which the (S)-enantiomer is
able to utilize the side pocket region PGHS-l while the R-enantiomer cannot because of
the R-stereochemistry. In addition, the binding of the R-enantiomer appears to confer

local strain to the protein suggesting this enantiomer may have a faster dissociation rate.

A dedication in memory of my beautiful mother, F elipa Arm (Naya) Zucker

iv

ACKNOWLEGEMENTS

First round of thanks goes to my advisors Dr. Bill Smith and Dr. Michael
Garavito in which their guidance has been invaluable. I am so grateful to Dr. Smith for
his support and encouragement as well as for sharing his vast knowledge of eicosanoids
and fatty acid metabolism, and I admire him greatly for his excellent editing ability. Dr.
Garavito was a wonderful mentor and I have learned so much from our conversations
about crystallography and science. I cannot thank him enough for what he has done for
me over the years, and I feel extremely fortunate to have worked for such a brilliant and
creative individual.

Much thanks goes out to Mike Malkowski and Anne Mulichak, past members of
the Garavito lab, for taking me on my first synchrotron trip and for giving me my first
lesson on crystallography. I will never forget them. I would also like to thank other
members of the Garavito lab including Amy, Mike D, Nicole and Rachel for their
friendship, laughs and for making my experience a wonderful one. Members of the
Smith lab whom I would like to thank include Jill, Cindy, Jiayan, and Yeon-Joo for their
friendship and advice over the years. I also owe thanks to Dr. Dave DeWitt for keeping it
real (and for his stock pile of gummy bears) as well as past and present members of his
lab particularly Jeff Leipprandt for the many, many bioreactors he grew up and harvested
for me.

Of course, I would like to thank my family for their love and faith in me. And lastly, I
would like to thank my loving and supportive husband Ben and our sweet daughter Mia

both of whom I love with all my heart and could not have done this without them.

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................... viii
LIST OF FIGURES ........................................................................................................ ix
LIST OF ABBREVIATIONS ......................................................................................... xi
CHAPTER 1: LITERATURE REVIEW
Introduction ............................................................................................................ 1
Prostaglandin Biosynthesis ..................................................................................... 5
The Two Activities of PGHS .................................................................................. 7
Suicide Inactivation ................................................................................................ 9
Biochemical Comparisons of PGHS-1 and PGHS-2 .............................................. 11
Structure of PGHS-1 and PGHS-2 ........................................................................ 15
Arachidonic Acid Binding to the COX Active Site ............................................... 20
Substrate Specificity of PGHS-1 and PGHS-2 ...................................................... 23
Inhibition of PGHSs ............................................................................................. 28

CHAPTER II: CRYSTAL STRUCTURE OF ARACHIDONIC ACID BOUND
TO A MUTANT OF PROSTAGLANDIN ENDOPEROXIDE SYNTHASE-l
THAT FORMS PREDOMINANTLY ll-HPETE

Introduction .......................................................................................................... 32
Experimental Procedures ...................................................................................... 35
Results .................................................................................................................. 42
Discussion ............................................................................................................ 49
Acknowledgements .............................................................................................. 55
CHAPTER III: MEAD ACID BINDING TO THE COX SITE OF PGHS-1
Introduction .......................................................................................................... 56
Experimental Procedures ...................................................................................... 58
Results and Discussion ......................................................................................... 65

CHAPTER IV: STRUCTURAL BASIS OF ENANTIOSELECTIVE INHIBITION
OF COX-l BY (S)—a-SUBSTITUTED INDOMETHACIN

ETHANOLAMIDES

Introduction .......................................................................................................... 78
Experimental Procedures ...................................................................................... 82
Result and Discussion ........................................................................................... 84
Summary .............................................................................................................. 94
Acknowledgements .............................................................................................. 95

vi

CHAPTER V: FUTURE DIRECTIONS

Structural Studies of PGHS-1 .................................................................................... 96
Structural Studies of PGHS-2 .................................................................................. 100
REFERENCES ............................................................................................................ 104

vii

Table 1

Table 2

Table 3

Table 4

Table 5

LIST OF TABLES

Data collection and refinement statistics for the Co3+-protoporphyrin
V349A/W387F oPGHS-l/AA cocyrstal structure .......................................... 40

Distances between selected side chains atoms of COX active site
amino acids and carbons of AA bound to native and V349A/W3 87F

oPGHS-l. ..................................................................................................... 50
Refinement statistics of oPGHS-l/AA and oPGHS-l/MA complexes ........... 63
Comparative inhibition of native COX-1 and COX-2 by various

ot-substituted indomethacin ethanolamides. ................................................... 81
Summary of data collection and refinement statistics. ................................... 85

viii

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5
Figure 6
Figure 7

Figure 8

7 Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14
Figure 15

Figure 16

Figure 17

LIST OF FIGURES

The arachidonic acid cascade ......................................................................... 2

Biosynthetic pathway of prostanoids derived from arachidonic acid
catalyzed by PGHS. ...................................................................................... 4

Interdependence of the peroxidase and cyclooxygenase activities. ................. 8

The mechanism for conversion of arachidonic acid to PGG2 by the

cyclooxygenase activity of PGHS ............................................................... 10
Comparison of the PGHS-1 and PGHS-2 proteins. ...................................... 13
Crystal structure of ovine PGHS-1 ............................................................... 17
Orientation of POX site relative to the COX active site of PGHS ................. 19
Arachidonic acid bound within the COX active site of

Co3+-protoporphyrin oPGHS-l structure . ..................................... 21
Conversion of various fatty acids to 1,2,3-series prostanoids ........................ 24

Superposition of various fatty acid bound within the COX active site of
PGHS-1 ...................................................................................................... 26

Cyclooxygenase mechanism based on the crystal structure of
arachidonic acid bound to native ovine PGHS-ll. ......................................... 34

Thin layer chromatogram comparing the products formed from

[1-‘4C]AA by native, V349A oPGHS-l, or V349A/W387F oPGHS-l ......... 43
Stereo view of AA bound within the active site of V349A/W 3 87F

oPGHS-l ..................................................................................................... 46
Arachidonic acid conformers obtained from simulated annealing. ................ 48
Comparison of AA bound to native and V349A/W3 87F oPGHS-l .............. 51

View of electron density observed in the POX and COX site of native
PGHS-1. .................................................................................................... 66

Initial Fo-Fc electron density observed in the COX active site of
oPGHS-l/MA complex ................................................................................ 69

ix

Figure 18

Figure 19

Figure 20

Figure 21
Figure 22
Figure 23
Figure 24

Figure 25

Figure 26

Figure 27

Figure 28

Comparison of catalytically possible and inhibitory conformations
of mead acid obtained from simulated annealing ........................................ 70

Comparison of the modeled proS hydrogen positions of the catalytically
possible conformers obtained from simulated annealing. ............................ 72

Comparison of the modeled proS hydrogen positions of the inhibitory

binding conformers obtained from simulated annealing .............................. 73
Arachidonic acid saturation curve of PGHS ................................................ 75
The K,- plot of Mead acid ............................................................................. 76
Compound 8 bound in the COX active site ................................................. 87
Compound 9 bound in the COX active site. ................................................ 88
Comparison of the placement of the (R)-ethyl vs. (S)-ethyl group

in COX-1 side pocket ................................................................................. 90
Citrate bound in the POX site ..................................................................... 91
Compound 8 bound in the COX active site after the structures

of COX-1 complexes were superimposed ................................................... 93
Crystal of human PGHS-2 grown in the presence of AA. .......................... 101

PGHS
oPGHS- 1
hPGHS-2
COX

POX

AA

CPLAZ
SPLAZ

(03 or (n-3)

(D6 or (n-6)

20:4, (06
18:2, (06
20:5, (06
20:3, (06
20:3, 009
ll-HETE
ll-HPETE
15-HETE
15-HPETE
12-HHTre
9-HODE
13-HODE
EGF

MBD
NSAID
B-OG
CIOE6
Compound 8
Compound 9
TLC

PG

Tx

LIST OF ABBREVIATIONS

prostaglandin endoperoxide H synthase

ovine prostaglandin endoperoxide H synthase-l
human prostaglandin endoperoxide H synthase-2
cyclooxygenase

peroxidase

arachidonic acid

cytosolic 85kDa phospholipase A2

secretory phospholipase A2

designation in which double bond present three carbons in from acyl end
of fatty acid

designation in which double bond present six carbons in from acyl end
of fatty acid

arachidonic acid

linolenic acid

eicosapentonic acid

dihomo-y-linolenic acid

mead acid

1 1-hydroxy-5,8, 12. 14-eicosatetraenoic acid

1 1 -hydroperoxy-5,8, 1 2,14-eicosatetraenoic acid
1 5-hydroxy-5,8,1 1 ,13-eicosatetraenoic acid

1 5-hydroperoxy-5,8,l 1, 1 3-eicosatetraenoic acid
12-hydroxy-5,8, 10-hepadecatrienoic acid
9-hydroxy- 1 0-, 12-octadecadienoic acid

1 3-hydroxy-9, 1 l-octadecadienoic acid
epidermal growth factor

membrane binding domain

non-steroidal anti—inflammatory
n-ocyl-B-D-glucopyranoside
polyoxyethylene(6)decyl ether
indomethacin-(R)-or-ethyl-ethanolamide
indomethacin-(S)-a-ethyl-ethanolamide

thin layer chromatography

prostaglandin

thromboxane

maximum velocity of a reaction
Michaelis-Menton constant

dissociation constant for enzymes-inhibitor complex
inhibitor concentration for 50% inhibition
ethylenediamine tetraacetic acid

phosphate buffered saline

xi

CHAPTER I

LITERATURE REVIEW

Introduction

Prostaglandin endoperoxide H synthases-l and -2 (PGHS-1 and -2) also known as
cyclooxygenases (COXs) are bifunctional, heme-containing and membrane-bound
enzymes that catalyze the conversion of arachidonic acid (AA) to PGHz, the committed
step of prostanoid biosynthesis [1-3]. Prostanoids, which include prostaglandins and
thromboxanes belong to a large class of bioactive, oxygenated Clg-sz compounds called
eicosanoids. Eicosanoids comprising the prostanoids, the leukotrienes, and the epoxy
eicosatrienoic acids, are formed from u)3(n-3) and (06(n-6) fatty acids through three
major pathways in which each path is named after the enzyme that catalyzes the
committed step of that pathway (Figure 1) [4]. In mammals, the predominant fatty acid
precursor for the synthesis of prostanoids is AA, a 20 carbon (D-6 fatty acid, and thus

eicosanoids derived from AA are formed through the arachidonic cascade .

Prostaglandins, which are formed via the cyclooxygenase (PGHS) pathway
(Figure 1), are local acting, hormones that regulate a broad range of normal physiological
functions including blood pressure and body temperature regulation, renal retention,
smooth muscle contraction, vascular homeostasis, platelet aggregation and inflammation
[5-7]. In addition to these normal physiological functions, prostaglandins have also been
associated with various inflammatory and blood-clotting diseases [8], and other

pathologies such as colon and ovarian cancer [9, 10] and Alzheimer’s disease [8].

malnomc Am
(20:4. (06)

 

P611 SYNTHASE 202

(CYCLOOXYGENASEI 2., s-,s-,12-,onsuoxvemss 02 CYTOCIIIONENSO 02

EPOXYGENASES "ADP"

momma“: WW EKOSAWNOK “'05 5.6-. o.9-.11.12-.om,is-

 

IPGHzl (HpETBl epoxv acowmmoic
ACIDS (some) (or $13))
“// \ "venomrcosammoic mm”
ACIDS IHEIB) mmoxv
mammmfiixgsmms (no) \ EKOSATETMENOIC
(P602. PGEI. mammal w H20 westerners)
mom-um

/ \«r \W

mmomomeues Lumen moxmsruo
(woman) ms)

Figure 1 The arachidonic acid cascade. Arachidonic acid is metabolized by three
separate enzyme systems, which include the cyclooxygenase, lipoxygenase, and P450
epoxygenase pathways.

As prostaglandins are generally unstable, they are not stored by cells, but rather
are synthesized in response to an extracellular stimulus. For example, when a hormone
interacts with its receptor, a series of events is activated that in turn activates one or more
phosholipases A2, that then cleave AA from the phospholipids of the cellular membranes.
Free AA is then acted upon by PGHS producing, PGH2, the common precursor of
prostaglandins and thromboxanes.

In the early 1990’s, a second isoforrn of PGHS was discovered [11]. Both

isoforrns, termed PGHS-1 and PGHS-2 catalyze the two reactions involved in converting

AA to PGH2. These activities include a cyclooxyenase activity, which catalysis a bis-
oxygenation of AA to form the intermediate hydroxperoxide, PGG2 and a peroxidase
activity catalyzing a two-electron reduction of PGG2 to form the final product of the
enzyme, PGH2 (Figure 2). These reactions occur at two distinct sites in the enzymes;
however, they are functional coupled through a requirement of heme for both activities
[12, 13]. Through the action of specific terminal synthases, PGH2 is converted into an
array of biologically diverse lipids mediators, the series-2 prostanoids that include (PG)
E2, F2a, D2, 12, and thromboxane (Tx)A2.

Despite numerous similarities in structure and function, PGHS-1 and PGHS-2
differ in their patterns of expression, regulation, substrate utilization, and inhibition.
PGHS-1 is constitutively expressed in most tissues, and its activity is mainly associated
with producing prostaglandins needed for maintaining homeostasis or “housekeeping”
functions such as renal water retention, blood pressure regulation, gastric intestinal
motility. Expression of PGHS-2 is rapidly induced in response to mitogens, cytokines
and growth factors [14]. Other notable disparities between the isoforms include
variations in glycosylation, an 18-amino acid cassette present in PGHS-2 but not PGHS-
], and other subtle structural differences within the COX active site that contribute to
differences in substrate utilization and inhibition between the isoforms.

The PGHS enzymes are the target of non-steroidal anti-inflammatory drugs
(N SAIDs), which include compounds like aspirin, ibuprofen and naproxen. All NSAIDs,
classified by their ability to reduce pain, fever and inflammation are known to inhibit the

cyclooxygenase activity by interacting with the cyclooxgenase site thereby preventing the

Hormone Stimulatlon

_ , / okines, mit ns,
Cell membrane—+Phosphophollplds / ”grommets?

Lipase(s)

COOH

Arachidonic acid
Cyclooxygenase 1+ 2 02 _.

 

PGG
°°“ 2 PGHS
Peroxidase 1+ 2 e'

"1,,/=\/\/COOH

 
 

 

PGH2 _.

lOH

T/erminal lSynthases

5H P602

HO OH PG E2

 

Figure 2 Biosynthetic pathway of prostanoids derived from arachidonic acid
catalyzed by PGHS.

binding of AA. However, inhibition by aspirin occurs through a different mechanism:
the COX active site in PGHS-1 and PGHS-2 is chemically modified whereupon the
hydroxyl group of Ser530 is acetylated. The effect of aspirin acetylation differs between
the enzymes, bringing attention to subtle structural differences demonstrating that PGHS-
2 has a slightly larger active site. Crystal structures of both PGHS-1 and PGHS-2 in
complexes with various inhibitors and ligands have been determined and have further
contributed to our understanding of inhibition of the PGHS enzymes. The understanding
of the pharmacological differences between the isoforrns is what led to the development

of the PGHS-2 selective inhibitors.

Prostaglandin Biosynthesis

The biosynthesis of prostaglandins and thromboxanes occurs through a three step
process in which AA is released from cellular membranes, converted to PGH2 by the
PGHS-1 and -2 and then acted upon by specific synthases that convert PGH2 into an
array of prostaglandins or thromboxanes [3, 4]. AA is the major precursor for
prostaglandins and thromboxanes and is stored at the sn2 position of the
glycerophospholipids of the cellular membranes.

AA is liberated from the membrane phospholipids by the action of
phospholipases. There are several groups of phospholipases that participate in releasing
AA from the cellular membranes; these include the Can-independent phospholipase
(iPLA2), Ca+2-dependent cytosolic phospholipase (cPLA2) and the inducible, secretory
phospholipases (sPLA2); however, cPLA2 and the secretory PLA2s are regarded as the

principal phopholipases involved in a coordinated action to release AA for utilization by

the PGHS enzymes [15-17]. The Cad-independent PLA2 is mainly involved in cell
membrane remodeling [16].

The mobilization of AA destined for PG synthesis is initiated when various
hormones interact with their receptors (e.g. bradykinin, angiotensin II, thrombin). This
triggers an increase in intracellular Ca+2 that activates cPLA2, a Ca+2 dependent
phospholipase. The activation of cPLA2 by Ca+2 leads to the translocation of cPLA2 to
the endoplasmic reticulcum (ER) and nuclear envelope (NE) where it cleaves AA from
the sn2 position of the phospholipids on the cytosolic surface of the membrane. cPLA2 is
the dominant phospholipase involved in liberating AA from the cellular membranes;
however, when the stimulus is enduring and intense the inducible, secreted sPLA2 begins
to contribute providing an additional release of AA for utilization by the PGHS enzymes,
particularly for utilization by PGHS-2, the inducible form [16]. Thus the combined
action of these phospholipases is believed to be one mechanism for providing AA for
immediate and prolonged stimulation for utilization by either constitutive PGHS-1 via
cPLA2 releasing AA during immediate stimulation or inducible PGHS-2 via sPLA2
releasing additional AA after prolonged stimulation [4, 16, 17]. The cPLA2
phopholipases are specific in cleaving AA from the membranes; however, the secretory
forms of PLA2 lack any specificity for either the phospholipid head group or the acyl
group at the sn2 position [4]; therefore, under some conditions during a stimulus it is
possible that PGHS is exposed to mixed pool of fatty acids other than AA. This mixed

fatty acid pool effect the type of prostanoids produced from these substrates by the PGHS

enzymes [4].

Once AA is released from the cytosolic surface of [the ER and NE via cPLA2, it
traverses the membrane becoming available as a substrate for the PGHS enzymes
converting AA to PGH2. The product PGH2 is further converted to a specific prostanoid
depending on the presence of specific terminal synthases within the cell. The prostanoids
are secreted from the cells via some carrier-mediated transporter [18, 19] [20] where they
interact in an autocrine or paracrine fashion with specific prostaglandin receptors that are
in turn linked to heterotrimeric G-protein coupled receptors [21, 22]. The binding of
prostanoids to their specific G-protein linked receptor, in turn, activates a cascade of
intracellular signaling events characterized by increasing cellular levels of CAMP, Ca”,

and phosphatidyl inositol [23, 24].

The Two Activities of PGHS

The cyclooxygenase of PGHS binds AA and incorporates two molecules of
oxygen in a regio— and stereo-specific fashion forming the intermediate, PGG2, while the
peroxidase catalyses a two-electron reduction of PGG2 to form PGH2, the final product
[25, 26]. Although the two reactions occur at physically separate sites, the two activities
are functionally coupled. A model for the functional link between the COX and POX
activity initially proposed by Ruf and coworkers (Figure 3) involves a branched chain
mechanistic model in which the heme within the peroxidase site must first undergo a two-
electron oxidation by a hydroperoxide (a single POX turnover) in order to initiate
catalysis at the COX site [12]. During the first POX turnover an oxyferryl

heme/protoporphyrin radical cation species (Compound I) is formed, which creates a

CYCLOOXY G ENASE

 

 

Pox SUICIDE
INACTIVATION / A \
20
I u 385 2 4.385
1mm 111 ””59 "°1 ”$.10.qu T"
[PPIX-Fo “CW”? \“385 /385
Intermediate IIA Pm" @381” [233 T"

”0:4 R°H\ cox SUICIDE

335 P662 INAC‘I'IVA‘I'ION

 

 

PPM-3’13Iyymn 351381” —> Mal-01 1y;
/Cy Inumodmn ll
0
Whey-01385
\ Compound II
V
PEROXIDASE

Figure 3 Interdependence of the peroxidase and cyclooxygenase activities. Ruf
model [12] of the peroxidase- dependent activation of the cyclooxygenase activity of PGH
synthase via formation of an intermediate tyrosyl radical (Tyr3 85) F e3+ PPIX, ferric 1ron
protoporphyrin IX (heme); ROOH,alky1 hydroperoxide; ROH, alcohol; AA, arachidonic
acid; Fe =O PPIX, oxyferryl heme. CompoundI, an oxyferryl group (Fe(IV)= 0) plus a
protoporphyrin IX radical cation, Intermediate 11, an oxyferryl plus a neutral
protoporhyrin IX plus a Tyr3 85 tyrosyl radical; Compound 11, an oxyferryl group plus a
neutral protoporphyrin IX.

tyrosyl radical on residue Y3 85 (Intermediate 11) located within the COX site of PGHS
[13]. The Y385 tyrosyl radical is required to initiate the COX reaction. Thus, once a
tyrosyl radical is formed and AA is bound within COX site, the COX reaction begins [27,

28].

The COX reaction begins by abstraction of the 13 proS hydrogen atom of AA by
Tyr385 radical producing an arachidonyl radical, the rate-limiting step [25, 29] (Figure
4). The radical migrates to position C11 of AA and is subsequently attacked by 02
forming an llR-peroxyl radical. The peroxyl radical on C11 attacks the double bond at
C9 forming an endoperoxide between C9 and C11. Formation of the endoperoxide and
radical migration to C8 leads to cyclopentane ring formation between C8 and C12,
requiring a reconfiguration of the substrate. The radical migrates to C15 and leads to an
attack by a second molecule of oxygen at C15 forming a C15 peroxyl radical. The C15
peroxyl radical then abstracts the hydrogen from Y385 to yield PGG2 while
simultaneously regenerating the tyrosyl radical for another cycle of COX catalysis.
PGG2 diffuses from the COX site and migrates to the peroxidase site and undergoes a
two electron reduction yielding PGH2, which is then utilized by downstream synthases to
produce a variety of prostanoids including PGEZ, PGIZ, PGDZ, and PGF2u and

thromboxanes each of which are responsible for a variety of biological functions.

Suicide Inactivation

An interesting and not very well understood characteristic of PGHS catalysis is
that of suicide inactivation. Each PGHS monomer exhibits a limited number of turnovers
ranging from 10 to 1300 depending on the conditions before COX activity disappears
[3 O]. The observed loss of activity is not due to total consumption of substrate or product
inhibition, but most likely involves formation of an active radical species that irreversibly
inactivates the enzyme [31, 32]. Both the COX and FOX activities of PGHS were found
to undergo suicide inactivation; however, the relationship between POX inactivation and

COX inactivation remains unclear [33, 34]. Kinetic and spectral studies have provided

\ M _'E'_'_., \ _,
+02

I coon II coon

MW 9 /

0‘ .’ \ — R .1 11 . _
o o
\ \

I“ coon 'V coon

Q’M RIO. 15’

o 2 +02 H OOH
+H ' P 662

V coon VI coon

Figure 4 The mechanism for conversion of arachidonic acid to PGG2 by the
cyclooxygenase activity of PGHS. Arachidonic acid (AA) is converted to PGG2
through sequence of steps I-VI by the cyclooxygenase activity of PGHS. The C15
peroxyl group (-OOH) of PGG2 (V1) is reduced to —OH by the activity of the peroxidase
of PGHS to form the final product PGH2 [25].

evidence that POX inactivation originates from Intermediate 11 characterized as
Fe(IV)=O and a protein tyrosyl radical, which is the intermediate that links POX activity
to COX activity (Figure 3) [30]. Another spectral intermediate termed Intermediate 111
was isolated during the POX self-inactivation process and appeared to be produced from

Intermediate II [30]. Intermediate 111 is believed to be the species, which leads to

10

inactivation; however, the chemical mechanism remains a mystery, but is believed to
involve a radical mediated process that possibly cross-links the enzyme and eliminates
activity.

Other studies have shown that COX and POX inactivation are distinct events and
that each process originates from different intermediates [35]. A Y3 85F mutant, which
cannot form Intermediate II is able to undergo POX inactivation at the same rate as
native; therefore, residue Y385 is not required for POX inactivation; however, is
presumed to be required for COX inactivation. Studies with the Y3 85F mutant suggest
that COX inactivation results from a different intermediate (Figure 3) [35]. This
intermediate is suggested to be like Intermediate 111 containing an oxylferryl heme and a
protein tyrosyl radical; however, the radical would be centered on a tyrosine other than
Y385 [35]. Although there are still many unanswered questions into the nature of the
intermediates responsible for inactivation of POX and COX, nonetheless, these radical

intermediates lead to irreversible loss of activity of the enzyme.

Biochemical Comparisons of PGHS-1 and PGHS-2
The primary sequence of PGHS-1 and PGHS-2 are very similar having 60-65%

sequence identity between them, and between the individual isoforms of different
species, which have approximately 80-85% identity [36]. Both PGHS-l and PGHS-2
have a signal sequence at the N-terminus, which is removed during processing to mature
enzyme, thus PGHS-l and PGHS-2 when in mature form contain 576 and 587 amino

acids, respectively [36].

ll

Although, PGHS-l and PGHS-2 are similar in primary structure, there are several
major differences within several regions of the primary sequence (Figure 5). At the N-
terminus of PGHS-2, the signal peptide sequence is truncated in comparison to PGHS-l,
thus to account for this 15 is added to the numbering of PGHS-2 for comparing to
equivalent residues within the isoforms since numbering starts at the N-terminus of
nascent protein. Additionally, there is an 18-amino acid insertion at the C-terminus of
PGHS-2, which is not present in PGHS-l. The function of this 18 residue insertion is not
yet known; however, it had been speculated that is might be involved in protein turnover
and protein-protein interactions [36]. Significant sequence differences exist within the
membrane-binding region of PGHS-1 and PGHS-2, which only share 38% identity
between the isoforms. However, in regions that involve catalysis there is significant
conservation of residues with a few exceptions being within the COX active site. Lastly,
PGHS-l and PGHS-2 are differentially glycosylated in that PGHS-l is unifome N-
glycoylated at three sites that include Asn68, Asn144, and Asn410. These sites are also
N-glycosylated in PGHS-2; however, there is a forth site that is variably glycoslylated at
Asn580 [37]. The variable glycosylation of PGHS-2 is readily observed as a doublet on
SDS-PAGE, representing multiple molecular weight species whereas PGHS-l appears as
a singlet [37, 38]. For expression of active protein both PGHS-1 and PGHS-2 require
glycosylation only at site Asn410. Since activity of purified PGHS-l and PGHS-2 is
retained when sugars are removed, glycosylation at Asn410 appears to be mainly
involved in proper folding of the protein [37]. Glycosylation at the variable fourth site in
PGHS-2 does not seem be as important in protein folding since an Asn580 mutant of

PGHS-2 yield fully active protein [37]. The role of the fourth glycosylation site is

12

PGHS-1 m "22°

/
l E EGF I M80 1 Catalytic I
/

Signal
sequence

59% 38% 70%

*
PGHS-2 A18? R10/6(120)_ w

I EGF l MBD l Catalytic 3
Signal /STEL

sequence 18 aa
Insert

 

PTEL

Figure 5 Comparison of the PGHS-1 and PGHS-2 proteins. The three structural
domains of the PGHS enzymes are shown as EGF (epidermal growth factor domain),
MBD (membrane binding domain), and the Catalytic domain, which contains the two
active sites of PGHS. The percentages represent the similarity within the individual
domains between the isoforms. In addition, the lined boxes represent insertions within
PGHS-1 and PGHS-2 that are not present in the other isozyme. The glycosylation sites
within each of the proteins are represented by circled residue numbers. PGHS-1
numbering for the glycosylated residues in PGHS-2 are shown in ()’s.

not known; however, it might play a role in PGHS-2 expression in the baculovirus system
in which PGHS-2 is expressed more successfully as compared to PGHS-1.
Other differences between the isoforms include some significant sequence

differences within the membrane binding domain and some minor differences within the

13

COX active site. Despite these differences, crystal structures of each isoform show that
the three-dimensional structure of the two isoforms is nearly superimposable [36, 39-42].

There are also similarities and differences between the isoforms with respect to
cellular localization and expression patterns. Both enzymes have been localized to the
lumenal surface of the endoplasmic reticulum and the inner and outer membranes of the
nuclear envelope with PGHS-2 being found to be more highly concentrated in the nuclear
envelope as compared to PGHS-1 [43]. Although the isoforms are generally localized in
similar membranes, they differ substantially in their patterns of expression. PGHS-1 is
expressed constitutively and the prostaglandins produced via PGHS-l are mostly
involved in housekeeping functions such as blood pressure regulation, and gastric
intestinal motility and protection. PGHS-2, however, is predominantly expressed by
induction via growth factors, cytokines and mitogens; however, with some exceptions
being in the brain and kidney where PGHS-2 has been found to be constitutively
expressed. The prostaglandins produced via PGHS-2 are known to be involved in
inflammation, pain and fever, in addition to being implicated in pathological conditions
that include rheumatoid arthritis, colon cancer [8, 9] and neurological disorders such
Alzheimer’s disease [44], thus PGHS-2 has been the isoform of much focus as a drug
target because of its association with these various pathologies. Interestingly, recent
studies have found that PGHS-1 rather than PGHS-2 appears to be a major regulator of
PGE; production in ovarian cancer cells lines [10], demonstrating that PGHS-1 may also
play a role in various cancers. PGHS-1, however, is more commonly associated with
blood clotting diseases, and has been the primary target of aspirin in providing anti-

thrombogenic and cardioprotective therapy [45, 46].

14

Despite significant differences in patterns of expression and physiological and
pathological roles, on a basic functional level the two isoforms are similar. Both bind
one molecule of iron (FeIII) heme (protoporphyrin IX) per monomer, which is required
for catalysis. Additionally, both utilize the same substrate with similar affinity (Km ~5-10
uM), and carry out the same catalytic mechanism to produce the same product [3]. There
are some exceptions that involve substrate affinity and rate of product formation under
conditions in which AA and lipid peroxide concentrations are low. In vitro studies have
shown that at concentrations of AA below 2 uM, PGHS-l appears to have a higher
apparent Km than that of PGHS-2 [15, 17, 47]. Other studies have demonstrated that
activation of COX activity in PGHS-1 requires ten times the concentration of lipid
peroxide than that of PGHS-2 at conditions of low AA concentrations [48]. Therefore,
under conditions of low peroxide and AA concentration activity of PGHS-2 would be
favored over PGHS-l [17] providing one mechanism in which PGHS-1 and PGHS-2 can

independently function in the same cell.

Structure of PGHS-l and PGHS-2

The first crystal structure of ovine PGHS-l complexed with the NSAID
flurbiprofen was published in 1994 by Picot et a1 [39], and since then has been followed
by many other structure complexes with various inhibitors [40, 41] and AA [42], and
other fatty acid substrates [49, 50]. PGHS-1 and PGHS-2 exist structurally and function
as homodimers. All published crystal structures show that the PGHS monomer consists

of three structural domains: the N-terminal EGF domain, a membrane binding domain

15

(MBD) and a large C-terminal globular catalytic domain housing the cyclooxygenase site
and the peroxidase site where heme binds (Figure 6).

The EGF domain consisting of the first 40 amino acids at the N-terrninus is
mainly involved in dimerization of PGHS in that portions of this domain between two
monomers create a substantial region of the subunit interface of the PGHS homodimer.
However, being a common domain in several families of membrane proteins, the EGF
domain of PGHS has been speculated to perhaps also serve a role in membrane
integration of the membrane binding domain of PGHS [36]. The membrane binding
domain of PGHS, continuing from the EGF domain, consists of approximately 48 amino
acids that fold into four amphipathic a helices. The helices lie in slightly angled
orientations relative to each other with one helix protruding up into the catalytic domain
creating a portion of the entrance to the COX active site. The outside portions of the
helices are composed of hydrophobic residues providing a hydrophobic surface that can
interact with the first leaflet of the membrane bilayer. The monotopic nature of PGHS
has functional significance for substrate binding in that when substrate is released from
the membrane phospholipids it transverses from the hydrophobic compartment of the
bilayer through a direct route to the COX site provided by the membrane binding domain
which anchors PGHS in the membrane.

Just above the membrane binding domain sits the large catalytic domain of
PGHS, which contains the two active sites (Figure 6). The cylcooxygenase site, where
AA binds, exists as a long, hydrophobic channel originating from the membrane binding

domain and tunnels up into the core of the catalytic domain. The crystal structure of

16

 

    

Catalytic
Domain

   

Membrane Binding Domain

   

Figure 6 Crystal Structure of Ovine PGHS-l. Panel A shows PGHS as a dimer. Panel
B shows the monomer structure, which is composed of three structural domains. The N-
terminal EGF (epidermal growth factor) domain, and the membrane binding domain are
shown in green and magenta. The large catalytic domain blue) posses the two active
sites. The COX site is shown with AA bound while the Co +-proptoporphyrin is bound
within the POX site where heme would normally bind. Images in this dissertation are
presented in color.

Co+3—protoporphyrin IX oPGHS-l/AA complex shows that AA binds within the COX
active site in an “L-shaped” configuration in which the carboxylate is anchored at the
mouth of the site through ionic interactions with residues Arg120 and Y355 while the
acyl portion of AA weaves up inside the channel making numerous hydrophobic
interactions with various residues along the channel [42].

At the top of the COX channel, the catalytic residue Tyr385 resides and is ~12.5
A from the heme bound within the peroxidase site (Figure 7). The peroxidase site, which
sits just above the cyclooxygenase channel is a solvent exposed cleft that binds one heme
per monomer of PGHS. The heme in each monomer is coordinated by two histidines
with His388 as the proximal ligand and Hi3207 as the distal ligand. The peroxidase site
of PGHS is more open and solvent accessible in and around the heme as compared to
other peroxidases such as mammalian myeloperoxidase (MPO); however, this is not
unusual considering this site in PGHS must accommodate the binding of large alkyl
peroxide substrates like PGG2 [36]. Comparisons of numerous structures obtained for
both PGHS-1 and PGHS-2 have revealed the impact of interesting structural differences
between the isoforms. Some major structural differences include an upward shifi of one
end of helix D of the MBD providing a larger opening within the MBD of PGHS-2. In
addition, a substitution of three amino acids (1523, 1434, and H513 in PGHS-1 vs. V523,
V434, and R513 in PGHS-2) located within and near the active site of PGHS bestows
PGHS-2 with a 20% larger active site. This is the major structural difference that has
been exploited in the development of PGHS-2 selective inhibitors such as Vioxx and

Celebrex.

18

 

Figure 7 Orientation of POX site relative to the COX active site of PGHS. Co”-
protoporhyrin (shown with protoporhyrin purple and Co3+ ion pink) represents the heme
position in the POX site. Dashed line represents a distance of ~12.5 A between the Co3+
ion and the hydroxyl group of Tyr 385 located within the COX active site. AA (orange)
is shown bound in the COX active site.

19

Arachidonic Acid Binding to the COX Active Site

The conversion of AA to PGG2, the product of the COX site, involves a series of
regio- and stereo-specific reactions. Since AA can assume up to 107 low energy
conformations [51], the binding conformation of AA is not only of interest in
understanding the feasibility of catalysis, but also for understanding how the proper
regio- and stereo-specific additions of oxygen and ring formations are achieved. Many
kinetic and structural studies with mutants of PGHS—1 and PGHS-2 have focused on
understanding the interactions between active site residues and AA and how these
interactions play a role in binding and positioning of AA, in addition to catalysis. The
crystal structure of the oPGHS-l/AA complex, in which Cog—protoporphyrin IX
replaces heme, provided the first look at how AA binds within the COX active site. This
protein/ligand complex revealed significant hydrophobic contacts between AA and the
residues lining the COX active site (Figure 8) further elucidating the results obtained
from kinetic analysis of native and mutant forms of PGHS.

The role of various residues of the PGHS-1 COX active site have been classified
according to their involvement in the binding and positioning of AA, in addition to
catalysis. The classifications were derived from differences observed in kinetic
parameters and product formation ratios between native oPGHS-l and various mutants of
PGHS-1 [52, 53]. Mutation studies of Tyr3 85, which is positioned near C13 of AA in the
Co3+-protoporphyrin PGHS-l/AA crystal structure (Figure 8), have clearly shown Tyr3 85

to be involved as the tyrosyl radical that abstracts the hydrogen from C13. Tyr348 and

20

 

Figure 8 Arachidonic acid bound within the COX active site of Coy-protoporphyrin
oPGHS-l structure (PDB bank ldiy). Dashed line represents approximately 2.9A
distance from the C13 proS hydrogen of AA. Arachidonic acid binds with an extending
“L-shaped” configuration with carboxylate group making several ionic interactions with
R120 and Y355 and hydrophobic interactions with various residues lining the COX
channel.

Gly533, which are not involved directly in catalysis, appear to be important in
positioning C13 of AA for hydrogen abstraction by Tyr385. In particular, Gly533, which
contacts the carbons C19 and C20 of AA as observed in the crystal structure, is
speculated to be significant in maintaining the position of the m-end of AA so as to
support catalysis.

Other residues shown to be significant in positioning AA include residues V349

and W387. Studies with various mutants of V349 including V349A and V349L

21

demonstrated that AA could exist in at least three distinct and catalytically competent
conformations occurring at the moment of hydrogen abstraction, the rate-limiting step of
the COX reaction [52]. Thus, the conformation of AA at the moment of hydrogen
abstraction, would dictate what product would be formed. In the case of native enzyme,
greater than 95% of PGG; would be formed; however, mutant enzymes such as V349A
and V349L produced significant amounts of llR-hydroxy-SZ, 8Z, 12E, l4Z-
eicosatetraenoic acid (1 l-HETE), and lSR-hydroxy-SZ, 82, Ill, 13E or 15S- hydroxy-
SZ, 82, 112, 13E (15-HETE) being abortive products in which only one molecule of
oxygen was inserted before the reaction aborted [52]. Various mutants of PGHS-1 and
PGHS-2 resulted in a significantly altered shift in the ratio of products formed as
compared to the native enzyme, thus demonstrating the important role of the residues
lining the COX active site in maintaining an optimal conformation of AA to produce
PGGZ. Mutants of PGHS-1 such as V349A and W387F were of particular interest
because each mutation alone displayed a similar product profile in which approximately
30-50% of both llR-HETE and PGGz were formed and no detectable amount of 15-
HETE was formed [52, 53].

Interestingly these various mutants of PGHS remain relatively active in
metabolizing AA; however, mutations of residues such as V349 and W387 result in
aberrant forms of PGHS, which drastically change the course of the reaction and
therefore, the products that are formed from AA. It was speculated from the crystal
structure of the Coy-protoporphyrin oPGHS-l/AA complex that mutations at V349 and
W387 were somehow resulting in a conformation of AA (at the moment of hydrogen

abstraction) approximately half of the time that would prevent the formation of an

22

endoperoxide bridge between atoms C9 and C11 of AA. As a consequence, the reaction
would abort thus yield almost equal ratios of 1 1R-HETE and PGG2. Chapter II discusses
in detail the structural roles of residues V349 and W387 and how they influence the
conformation of AA bound within the COX site as well as how these residues play a role

in the formation of the endoperoxide group.

Substrate Specificity of PGHS-l and PGHS-2

Arachidonic acid is the preferred substrate for the PGHS enzymes and is the
prostanoid precursor that is converted into 2-series prostaglandins, the predominant series
of prostanoids found in blood plasma [54]. However, the l-series prostaglandins are the
most abundant form of prostanoids found in seminal fluid indicating that PGHS can
utilize dihomo-y-linolenic acid (DHLA; 20:3n-6) as a substrate [25, 26], which is found
at higher levels than AA in vesicular gland tissue (Figure 9) [55]. Other fatty acids in
addition to DHLA have been found to be oxygenated by PGHS some of which include
linoleic acid (LA; l8-2n-6) (Jr-linolenic acid (18:3n-3) and eicosapentaenoic acid (EPA;
20:5n-3) [56], and some less common fatty acids such as adrenic acid (22:4n-6) [57],
columbinic acid (5E, 92, 12Z-octadecatrienoic acid), and Mead acid (SZ, 82, 112-
eicosatrienoic acid) [58].

The ability of COX to oxygenate a variety of fatty acids would not be unusual
given that there is the possibility that PGHS can be exposed to a mixed pool of free fatty
acids taking into account that there are varying abundances of fatty acids within different
tissues and that during a stimulus various fatty acids can be cleaved from the cellular

membranes by forms of secretory PLA2 that lack specificity for a particular fatty acid [4].

23

Arachidonic acid (20:4,n-6)

  

Dihomo+linoienic acid (20:3.n-6) lPGHs Eicosapentaenoic acid (20:5,n-3)
'''''' m
_ _ I ’Wcoon __ _
leans 0.... ..... PGHS
PGH2

l 2-Seriee Prostanoids l
o (2 double bonds)

1-Series Prostanoide
coon 3-Serles Prostanoids
o (1 double bond) m o (3 double bond!)
, _ coon
mm” “0“ MW
H6 P652

—.

s
8

HO i s‘ .
= PGE “0
HO 1 HO P653

Figure 9 Conversion of various fatty acids to 1,2,3-series prostanoids.

EPA, for instance, is stored at the sn2 position like AA, and can be cleaved upon
activation of phosholipases A2 [3, 49, 59, 60]. Subsequently, EPA can then be converted
to PGH3, which can further isomerize to various prostanoids such thromboxane A3 and
POD; [3, 59] (Figure 9).

Although these fatty acids substrates are utilized at lower efficiency as compared
to AA, they can compete with AA and in some cases act as inhibitors preventing the
formation of the 2-series prostanoids being derived from AA. As previously mentioned,
EPA can be converted into 3-series prostanoids through PGHS pathway; however, it acts
more as a potent inhibitor of AA oxygenation rather than being an efficient substrate [3,

56, 59]. The inhibitory effects of EPA can serve the same benefit as aspirin by reducing

24

the formation of the pro-thrombogenic thromboxane A2 by platelet PGHS. Similarly,
linoleic acid, like EPA, is a poor substrate for PGHS that can also effectively compete
with AA in binding to PGHS, in addition, it can be oxygenated by PGHS forming the
monohydroxy acids 9-HODE and l3-HODE [56, 61]. Other fatty acids that act as
competitive inhibitors of PGHS include 18:1,n-9 and 22:6, n-3; however, unlike EPA and
linoleic acid are not oxygenated by PGHS [34, 56, 62].

Several crystal structures of PGHS-l complexed with various fatty acid substrate
including DHLA, EPA and linolenic acid have been published and help elucidate the
structural basis of substrate specificity observed from mutagenic studies by highlighting
active site residues which contribute to substrate binding and positioning for utilization
by PGHS-1. All structures to date with fatty acid bound to PGHS-1 show a similar
overall “L-shaped” binding orientation similar to what is observed for AA bound to
PGHS-1 where the carboxyate (atom C1) is anchored at R120 and Y355 and the (Io-end
weaves up into the channel [42]. However, each fatty acid having slight differences in
flexibility of the substrate due to differences in the number of carbons and double bonds
as well as double bond positions can have a major influence on the binding orientation
which in turn affects the ability of the fatty acid to be utilized as a substrate (Figure 10)
[42, 49, 50]. DHLA, a close analog of AA having only one less double bond, is utilized
most efficiently by PGHS-1 and PGHS-2 and binds within the COX site of PGHS-1 very
similarly to AA; however, the positions of carbons C2-C10 differ significantly
particularly, the position of the C8-C9 double bond. This altered position is a result of
increased flexibility due to a missing double bond between carbons C5-C6. Despite these

binding differences, atom C13 of AA remains properly positioned for hydrogen

25

 

Y385

ProS H

 

Figure 10 Superposition of various fatty acid bound within the COX active site of
PGHS-1. Arachidonic acid, DHLA, linoleic acid, and EPA are shown in orange, green,
magenta, and blue. The proS hydrogens of AA, DHLA, and LA are shown and labeled
with arrow. EPA is bound an inhibitory mode in that the proS hydrogen of EPA is
positioned in an opposite direction as compared to the other fatty acids shown.

26

abstraction and atoms C8-C10 are still positioned to support endoperoxide formation and
cyclopentane ring closure [50]. The structure of DHLA represents a catalytically
productive conformation of substrate bound; however, oPGHS-l/EPA complex revealed
more of an inhibitory binding mode. EPA, like DHLA, binds with an L-shaped
conformation having the carboxylate interacting with Arg120; however, the w-end of
EPA having an extra double bond (Cl7-C18) is more inflexible, thus binding within the
hydrophobic pocket is accommodated by compression near the carbonyl end.
Consequently, EPA binds with a strained orientation where atoms C3-C10 pack more
closely to one side of the COX channel (V349) resulting in a mispositioning of C13 with
respect to Tyr3 85, the catalytic residue of the COX site (Figure 10) [49]. Linoleic acid
which is used more efficiently than EPA but less than DHLA binds to PGHS-1 in a less
strained and catalytically competent binding mode despite having two fewer carbons than
AA, DHLA, and EPA, thus demonstrating further that slight differences in carbon
number, number of double bonds and their positions effect flexibility in turn changing the
ability to be used as a substrate or act as an inhibitor.

Substrate specificity studies with PGHS-2 have also shown that PGHS-2 can
utilize a variety of fatty acids to similar and in some cases with better efficiency than
PGHS-1. Our understanding of the binding of fatty acids to PGHS-2 is not as well
established as for PGHS-1. Most structure complexes of PGHS-2 are with various
inhibitors, which have tremendously increased our understanding of subtle differences
between PGHS-1 and PGHS-2, namely, particularly the larger active site of PGHS-2 and
how it may be utilized in binding of ligands as compared to PGHS-1. Some of these

structural differences have presented more questions in relation to the reason there are

27

two isoforms of PGHS. Recently, it was shown that PGHS-2 was capable of oxygenating
the endocannabinoid, 2-arachionylglyerol with similar efficiency, as AA while PGHS-l
could not [63]. In addition, PGHS-2 was able to convert 2-AG to PGH2-glycerol, which
can be further converted to series 2-prostanyl-glycerol derivatives in vitro and in vivo [64,
65] creating a unique group of prostaglandins possibly having different roles from
prostaglandins derived from AA. The ability of PGHS-2 to utilize 2-AG while PGHS-l
does not is of major interest and is thought to involve the larger active site of PGHS-2,

which is able to accommodate the binding of this larger fatty acid.

Inhibition of PGHSs

The PGHS enzymes are inhibited through the action of NSAIDs, which bind to
the COX active site preventing the binding of substrate while leaving the peroxidase
activity unaffected. The NSAIDs are a structurally diverse group of compounds that
inhibit the enzymes nonselectively, selectively or through chemical modification. The
largest group of the nonselective NSAIDs is the acidic cyclooxygenase inhibitors some of
which include ibuprofen, flurbiprofen, naproxen, indomethacin, diclofenate, and aspirin.
These “classical ” NSAIDs are structurally similar in that they are composed of either a
mono or bi-cyclic aromatic ring system and all possess a free carboxylate group.

Although these compounds are similar in structure they display markedly
different dynamics in inhibition of PGHS, and can be divided into groups based on their
ability to undergo either simple competitive or time-dependent inhibition. Inhibitors such
as ibuprofen and naproxen are characterized as freely reversible competitive inhibitors of

PGHS; however, other NSAIDs like flurbiprofen and indomethacin inhibit PGHS

28

through a time-dependent manner characterized by increased potency of inhibition as
preincubation time of enzyme with inhibitor increases, thus inhibition is only slowly
reversible [66]. Various studies hinted that the structural requirements for time-
dependent inhibition in the case of some of these inhibitors is attributed to a combined
presence of a halogen and a free carboxylate group [66, 67].

Crystal structures of various NSAIDs bound to PGHS-1 and PGHS-2
demonstrated that the carboyxlate group of acidic NSAIDs interacts with the guanidinium
group of Arg120. This interaction was not only shown to be involved in time-dependent
inhibition in PGHS-l, but was also shown to be essential for high affinity binding of
acidic NSAIDs and AA. Mutations of Arg120 to smaller more neutral residues have
rendered PGHS-l resistant to inhibition by NSAIDs and abolished binding of AA to
PGHS-l [67]. However, comparative mutation studies in PGHS-2 have shown that an
ionic interaction between Arg120 and a free carboxylate group was not as crucial for
binding of NSAIDs and AA [68]. Although most classical NSAIDs are non-selective
inhibitors, in general they seem to bind more tightly to PGHS-l as compared to PGHS-2
possibly because of the importance of the ionic interaction between the carboxylate group
of the inhibitor and Arg120 of PGHS-1.

Aspirin, unlike other NSAIDs, irreversibly inhibits PGHS-1 by chemically
modifying the enzyme through aceylation of Ser530, a residue located within the COX
site near Tyr385, the catalytic residue. Aspirin acetylates Ser530 in both PGHS-1 and
PGHS-2 and causes a complete loss in the production of prostaglandins; however, the
effects of acetylation are different between the enzymes. Aspirin acetylation in PGHS-l

completely blocks AA from accessing the COX site, thus PGHS-l cannot oxidize AA

29

and is rendered completely inactive. However, acetylated PGHS-2 is able to bind and
oxidize AA, only producing 15R-HETE as the major product with no detectable
prostaglandin products [69, 70]. The ability of PGHS-2 to still accommodate the binding
and utilization AA even after aspirin acetylation suggested that there were some slight
structural differences in the form of a larger binding pocket in PGHS-2.

Past studies have documented that the PGHS isozymes do differ in their
sensitivities to inhbition by NSAIDs, in addition, structural studies have clearly identified
subtle structural differences between the isoforms, which led to the development of series
of PGHS-2 selective inhibitors. The most distinctive structural feature between the
isoforms are several amino acid differences within the COX active site in which residues
Ile434 and Ile523 are replaced with smaller valines and HisSl3 in PGHS-1 is replaced
with arginine in PGHS-2. These differences not only increased the size, but also the
chemical environment of the PGHS-2 COX active site. The most structurally diverse
class of PGHS-2 selective inhibitors are the diarylheterocycles, which are characterized
as having a cis-stilbene frame fused to a variety of heterocyclic rings being of 4,5 or 6-
membered rings. The most critical feature of this class of inhibitors was found to be the
presence of a sulfonarnide or methysulfonamide group on one of the phenyl rings of the
stilbene framework, making this class of inhibitors larger than the “classical” NSAIDs
[71]. The diarylheterocycles were shown to bind to both PGHS-1 and PGHS-2;
however, the selectivity towards PGHS-2 is a result of time-dependent inhibition while in
PGHS-1 there was competitive binding, but being freely reversible. The structural basis
of this selectivity and time-dependent inhibition towards PGHS-2 was found to be residue

523 where mutations of Va1523 to isoleucine in PGHS-2 caused a loss in selectivity by

30

PGHS-2 selective inhibitors [72]. In addition, x-ray structure studies demonstrated that
the sulfonamide group of these inhibitors interacts with the side pocket region of PGHS-
2, and that having an extra methyl group on residue 523 prevented access to this side
pocket region [41].

Recently, modification of the classical NSAID, indomethacin, produced an
interesting series of inhibitors, the indomethacin ethanolamides. Inhibition studies of
PGHS-l and PGHS-2 with these variously modified inhibitors led to an interesting
observation of enantioselective inhibition of PGHS-1, which was not observed in PGHS-

2. The structural basis of this inhibition will be discussed in detail in Chapter IV.

31

CHAPTER II

CRYSTAL STRUTURE OF ARACHIDONIC ACID BOUND TO A MUTANT
OF PROSTAGLANDIN ENDOPEROXIDE H SYNTHASE-l THAT FORMS

PREDOMINANTLY ll-HPETIEIl

Introduction

Prostaglandin endoperoxide H synthase (PGHS) is a heme-containing, bi-
functional enzyme that catalyzes the conversion of arachidonic acid (AA) to
prostaglandin (PG) H2 (Figure 2), the immediate precursor to prostaglandins,
thromboxanes and prostacyclin. In mammalian tissues, there are two isoforms of PGHS
[3, 73]. PGHS-l is constitutively expressed and produces prostaglandins in response to
hormone stimulation mainly for regulating housekeeping functions. PGHS-2, the
inducible form, is expressed in response to mitogens, growth factors, tumor promoters,
and/or cytokines and produces prostaglandins associated with pain, fever, and
inflammation[3, 73].

Despite the differences in patterns of expression, PGHS-l and PGHS-2 are quite
similar both structurally [39-41] and mechanistically[74]. Both isoforms function as
dimers with each monomer having an epidermal growth factor domain, a membrane
binding domain, and a large catalytic domain. The catalytic domain of PGHS possesses
two distinct, but functionally connected active sites [34, 39-41]. These two sites include
the cyclooxygenase (COX) site, which exists as a long hydrophobic channel within the

core of the protein and binds AA [42] and nonsteroidal anti-inflammatory drugs [39-41],

32

and a more solvent exposed peroxidase (POX) site containing a heme moiety, which is
involved in reducing PGG2 formed at the COX site to PGH2, the final enzymatic product.

The crystal structure of native Co+3 -protoporphyrin IX oPGHS-l in a complex
with AA shows that AA binds within the COX active site having the carboxylate group
anchored at the mouth of the hydrophobic channel via electrostatic interactions with
Arg120 and Tyr355 [42]. The carbons of the AA chain weave into the COX active site
making forty-eight hydrophobic contacts with residues lining this channel. Based on the
crystal structure and knowledge of the stereochemical requirements of the COX reaction
[25], a structural sequence of catalytic events was proposed (Figure 11) [42]. The COX
reaction begins with the abstraction of the l3proS hydrogen of AA by a tyrosyl radical
formed at Tyr385 creating an arachidonyl radical centered at C13 [74]. Following a
rearrangement centering the radical on C11, an attack of molecular oxygen occurs to
form an llR-hydroperoxyl radical (Figure 11). At this point it is proposed that rotation
about the ClO-Cll bond moves the llR-hydroperoxyl radical in close proximity to C9 to
form the 9, ll-endoperoxide group; concurrently, rotation about the ClO-Cll bond
additionally brings C8 near to C12 to form the cyclopentane ring. Repositioning of C12
closer to C8 for ring formation also repositions atoms C13 to C20. The result is an
optimal positioning of C15 for a second attack by molecular oxygen. Additionally, with
the lS-hydroperoxyl radical now positioned in proximity to Tyr385, Tyr385 is poised to
donate a hydrogen to form the product of the COX site, PGG2, thus, simultaneously

reforming the Tyr3 85 radical and regenerating the enzyme for another round of catalysis.

33

E on or:

II-arachidanyl
radical formation

     

   

hydrogen abstraction

+ 02
cyclization step

 

012

  
  

C"

a @012

 

 

15 -peroxy P662
radical

 
 

Figure 11 Cyclooxygenase mechanism based on the crystal structure of arachidonic
acid bound to native ovine PGHS-l. Figure shown is from a published modified
version [75] of the originally published scheme [42].

Both PGHS isoforms exhibit some lipoxygenase activity. Small amounts of 11-
hydroperoxy—8Z, 12E, 14Z,-eicosatrienoic acid and 15-hydroperoxy-8Z, llZ, 13E-
eicosatrienoic are formed from dihomo-y-linolenic acid [25], and the corresponding 11-

and 15-HETEs are produced from AA [50, 52, 70]. Thus, the native enzyme is not 100%

efficient in converting AA to PGG2. Additionally, with native enzyme the Km values for

34

AA in forming ll-HPETE, 15-HPETEs and PGG2 are different for each product [52].
This suggests that AA can exist in the COX site in at least three catalytically competent
arrangements at the time of hydrogen abstraction to yield either PGG2, llR-HETE, or
lSR/S-HETE [52].

Extensive kinetic and product analysis of numerous mutants of the COX active
site have identified several residues that influence the relative proportions of PGG2 and
monohydroperoxide products formed from AA [52, 53, 76]. Two single amino acid
substitutions, V349A and W387F were of particular interest to us because relatively more
llR-HPETE is formed (35-50% of total products) without substantially altering the
kinetics of oxygenation [53]. Because each single mutation favors the formation of 11R-
HPETE approximately 50% of the time, we constructed and analyzed a V349A/W387F
oPGHS-l mutant. As described in this chapter, this double mutant produces
predominantly llR-HPETE. To investigate why this double mutant produces
predominantly llR-HPETE from AA, we determined the crystal structure of C03”-
protoporhyrin V349A/W387F oPGHS-l with AA bound in the COX active site.
Significant differences in the conformations of AA bound to native oPGHS-l vs.
V349A/W387F oPGHS-l mutant enzyme were observed. Our results provide insight
into how Val349 and Trp387 influence the conformation of AA and how these residues

are involved in directing the formation of the endoperoxide group.

Experimental Procedures

Materials

Arachidonic acid, llR-HETE and ISS-HETE were purchased from Cayman

Chemical Co. (Ann Arbor, MI), [l-I4C]-Arachidonic acid (40-60 mCi/mmol) was from

35

NEN Life Science Products (Boston, MA). 8121 insect cells and 100X Antibiotic-
Antimycotic were purchased from Invitrogen. Fetal bovine serum and HyQ-SF X insect
medium used for cell growth and protein expression were purchased from HyClone
(Logan, UT). Ni-NTA superflow resin was purchased from Qiagen, tween-20 used for
solubilization and purification was from Pierce Chemical Co., and n-octyl-B-D-
glucopyranoside (B-OG) used for crystallization was from Anatrace (Maumee, OH).
Flurbiprofen and hemin were purchased from Sigma (St. Louis, MO) and Co“-
protoporphyrin IX from Porphyrin Products (Provo, UT). Oligonucleotides used as
primers for mutagenesis and for sequencing were prepared by the Michigan State

University Macromolecular Structure and Sequencing Facility.

Preparation of V349A/ W38 7F oPGHS-I

Site-directed mutants were prepared with a pFastbac vector containing a cDNA
encoding native oPGHS-l with a hexahistidine tag [77] and mutagenic Oligonucleotides
using the Stratagene QuikChange Kit (Stratagene, La Jolla, CA). Mutations were
confirmed with sequence analysis performed by the DNA Sequencing Facility at

Michigan State University.

Generation of Recombinant Baculovirus and Expression of oPGHS-I Mutants

DHlOBac cells were transformed with pFastbac containing cDNA for either
V349A oPGHS-l or V349A/W387F oPGHS-l to obtain baculoviral DNA, which was
then isolated and used to transfect sf21 insect cells to produce recombinant baculovirus.
The recombinant baculovirus was amplified, and supematants of amplifications were

used as inoculum stocks. Sf21 insect cells grown in a 17 L bioreactor to an approximate

36

density of 1.7-2.0 x106 cells/mL in medium with 1% antibiotic and antimycotic were
infected with recombinant baculovirus at a multiplicity of infection of 0.01 at 27°C.

Three to four days after infection, cells were harvested by centrifugation.

Purification oPGHS-I Mutants from Sf21 Insect Cells

Cell pellets from three bioreactors were resuspended in 1X Phosphate Buffer
Saline (PBS), pH 8.0 containing 20 mM imidazole and were lyzed by sonication. To
solubilize PGHS from the membranes, Tween-20 was added to the lysate up to 1% (v/v)
which was then incubated at 4° C for 45 min with gentle agitation. Solubilized lysate
was centrifuged at 10,000 rpm at 4° C for 20 min, followed by ultracentrifugation at
45,000 rpm at 4°C for 90 min. The supernatant was carefully removed and incubated for
at least 3 hrs at 4°C with Ni-NTA resin equilibrated in 1X PBS, pH 8.0, 20 mM
imidazole, and 0.1% Tween-20. After this incubation period, the mixture of Ni-NTA
resin plus supernatant was poured into a column which was washed with four column
volumes each of IX PBS, pH 8.0, 20 mM imidazole, 0.1% Tween-20 and 1X PBS, pH
8.0, 300 mM NaCl, 20 mM imidazole, 0.1% Tween-20. Hexa—histidine-tagged PGHS
was eluted with 1X PBS, pH 8.0, 200 mM imidazole, and 0.1% Tween-20. Fractions
with the highest specific peroxidase activity [42] were pooled and concentrated using a
Millipore Ultrafree —15 spin concentrator (30-kDa molecular weight cut off) to a protein
concentration of 8-10 mg/ml. Detergent exchange and desalting of the concentrated
protein was performed by spinning protein over 5 mL of packed G-25 sephadex
(Amersham Pharmacia) equilibrated in 20 mM HEPES, pH 7.0, 20 mM NaCl, lmM
NaN3, and 0.5 % B-OG. Protein concentrations were determined using BCA protein

assays (Pierce, Rockford, IL).

37

Analysis of Products formed from Arachidonic Acid

Aliquots (equivalent to 150 COX units, where one unit is defined as 1 nmol of O2
consumed/min) of purified oPGHS-l native and/or mutants were incubated for 1 min at
37° C with 35 uM [1-14C]-arachidonic acid with and without 200 uM flurbiprofen in a
reaction volume of 200 pl. Radioactive products were extracted and separated by thin
layer chromatography as previously described [53]. Products were visualized by
autoradiography and quantified by liquid scintillation counting. Negative controls from
samples containing flurbiprofen were subtracted from values measured from the

corresponding samples not containing flurbiprofen.

Characterization of V349A/ W38 7F oPGHS-I

Kinetic parameters for V349A/W387F oPGHS-l were measured with a
cyclooxygenase assay by monitoring the initial rate of 02 uptake at 37° C using an
oxygen electrode [53]. A typical assay consisted of 3 mL of 100 mM Tris-HCI, pH 8.0, 1
mM phenol, 1 mM hemin, and 2-100 uM AA. Reactions were initiated by adding a
volume of enzyme equivalent to 150 units of Ni-NTA purified V349A/W3 87F oPGHS-l.
Crystallization and Data Collection

Purified apo-V349A/W387F oPGHS-l at approximately 8 mg/mL (0.11 mM) was
reconstituted with a two-fold molar excess of Co3+-protoporphyrin IX. Prior to setting up
sitting drop crystallization experiments, the protein was incubated in the presence of AA
at a five-fold molar excess over the protein concentration [78]. Protein was mixed 1:1
with buffer containing 0.68 M sodium citrate, 06-09 M LiCl, 1 mM NaN3 and 0.3% [3-
OG and equilibrated within a reservoir containing 0.68-0.9 M sodium citrate, 0.60-0.90

M LiCl and 1 mM NaN3. Crystals appeared after four weeks to several months. Crystals

38

were harvested with a nylon loop and transferred briefly into a cryo solution containing
0.9 M sodium citrate, 1.0 M LiCl, 0.15% B-OG, 1 mM NaN3 and 24% (w/v) sucrose.
Crystals were immediately flash frozen in liquid propane at -165°C. Data were collected
at beamline 191D of the Structural Biology Center (Advance Photon Source, Argonne
IL). The data from a single crystal was indexed and integrated using HKL2000 [79] and

then scaled together using SCALEPACK [79] (Table 1).

Structure Determination and Refinement

The crystals of V349A/W387F oPGHS-l with AA were essentially isomorphous
with crystals of oPGHS-l complexed with AA. Thus, the structure of V349A/W387F
oPGHS-l was determined by using the protein atoms of native oPGHS-l/AA (Protein
Data Bank code IDIY) as a phasing model for simple rigid body refinement utilizing
CNS version 1.1 [80], in order to correct for small changes in lattice parameters. After an
initial round of simulated annealing, positional and group B-factor refinement, the
resulting R and R free factors were 25.4% and 31.2%. Inspection of the initial 2Fo-Fc
(contoured at 10) and Fo-Fc (contoured at 2.50) electron density maps using the program
CHAIN [81] allowed the placement of Co+3-protoporphyrin IX, seven carbohydrate
residues, and three B-OG detergent molecules. Strong electron density was observed for
AA starting from the carboxylate group up to carbon C4. With subsequent rounds of
refinement, the improved phases led to the appearance of further electron density, which
allowed the additional placement of atoms C5-C12 of AA. No further electron density
was observed for AA. Although no density was observed for C13, the position of C13

was modeled based on the limited orientations it can adopt due to the stereochemical

39

Table 1 Data collection and refinement statistics for the Coy-protoporphyrin
V349A/W387F oPGHS-l/AA cocyrstal structure.

 

 

Item AA AA
Model building Simulated annealing

Spacegroup P6522 P6522

a(A) 181.88 181.88

b(A) 181.88 181.88

c (A) 103.37 103.37
Resolution (A) 20-3.1 20-3.1
No. of unique reflections (F >00) 20640 20640
ComP'eteneSS (%) 99.1 (99.8)a 99.1 (99.8)a
R merge. all data (%;I>00) 7.6 (32.0)a 7.6 (32.0)a
No. ofatoms in refinement 4573 4571
R factor (%; reflections >20 on F) 23.3 23.8
Free-R factor (%; reflections >20 on F) 30.8 31.1
Mean positional error (A) 0.43 0.43
r.m.s.d in bond length (A)b 0009 0:009
r.m.s.d in bond angle (°) 1.6 1.6
r.m.s.d in improper angle (°) 1.1 0.92
r.m.s.d in dihedral angle (°) 22.6 22.3
Average B factor all atoms 44.9 41.1

Protein 40.1 40.2
Arachidonic Acid 55.6 56.7

 

2]The values in parentheses represent the values in the last shell (3.2-3.1A)

b . .
r.m.s.d, root mean square dev1atlon

40

restrictions of the C11-C12 double bond and was included in the refinement, thus C1
(including the carboxylate) to C13 of AA was included in the refinement of the model.
Further refinement included the addition of 22 water molecules at positions that were
within 2.4-3.6 A of a hydrogen bond donor/acceptor and an electron density peak of 30
or greater. The R and R free values of the final model are at 23.2% and 30.8% (Table 1).
Evaluation with PROCHECK [82] showed that all non-pro and non-gly residues lay
within the most favored or allowed regions of the Ramachandran plot.

Since medium resolution diffraction data were used in this analysis, the following
two tests were performed to verify that changes observed in the AA conformation
between the structures of native and V349A/W387F oPGHS-l were significant. In the
first test, Fo-Fc difference electron density maps were calculated using the observed
structure factors and modified phase sets. As the crystals of native and mutant oPGHS-l
were isomorphous, a calculated phase set was created by combining the Co”-
protoporphyrin V349A/W3 87F oPGHS-l, excluding the observed AA atoms, with atoms
from the AA conformer observed in the refined model of native Co3+-protoporphyrin
oPGHS-l/AA complex (Protein Data Bank code IDIY). A similar method has been
utilized previously to verify observed differences between AA and several other fatty
acid substrates bound within the COX site [49, 50]. In areas where the carbon positions
differ between the two observed conformations of AA, we would expect to observe the
appropriate positive and negative difference density peaks, indicating a shift in atoms
positions.

The second test addressed the possibility that other conformations of AA could

exist and be represented by the density, given the broad extent of the electron density

41

observed at this resolution. A parallel series of simulated annealing refinements were
performed, using the observed structure factors and the V349A/W387F oPGHS-l
coordinates. Each series of simulated annealing refinements were done four times in
order to generate a family of structures from each starting point. The starting
conformation of bound AA differed, however, between the two parallel series of
refinement. In the first series, the AA conformer found in native oPGHS-l [42], was
used, while in the second series, the AA conformer derived from the refined
V349A/W387F oPGHS-l structure was used. If the contributions from the X-ray data for
the mutant were significant, it was expected that the AA conformer seen in the mutant
would be recovered in the first case and retained in the second. If not, both series of
refinements would yield a wide range of unrelated AA conformers. As expected, the
simulated annealing refinement starting from the AA conformer derived from the refined
V349A/W387F oPGHS-l structure generated a family of AA conformations which were
very similar to the starting conformer. In contrast, all four AA conformers generated by
using the native AA conformer as the starting point for refinement looked remarkably
similar to the AA conformation found in the V349A/W387F mutant. Of these four
conformers, one fit the observed electron density quite well and was used for further
structure refinement. Statistics of the refinement of the final model using the conformer

from simulated annealing are shown in (Table 1).

Results

Comparison of Products Formed flom Arachidonic Acid by Native and Mutant oPGHS-I.

There are considerable differences in the proportions of oxygenated products

formed from [1-14C]-arachidonic acid by native oPGHS-l, V349A oPGHS-l and

42

V349A/W387F oPGHS-l (Figure 12). Native enzyme produces predominantly products
derived from PGG2 (~95% including HHT and various prostaglandins), and small

amounts of l l-HETE (~2.5%) derived from ll-HPETE and 15-HETE (~2.5%).

Native V349A V349A/W387F
flu rbiprofen - - + - +

AA—v .-

15-HETE —*

l1-HETE —» - -

HHT—b

Prostaglandin a...
PI'OdUCtS ’

Figure 12 Thin layer chromatogram comparing the products formed from [1-
”ClAA by native, V349A oPGHS-l, or V349A/W387F oPGHS-l. [1-‘4C]AA was
incubated for 1 min at 37°C in the presence or absence of the cyclooxygenase inhibitor
flurbiprofen with 150 cycloxygenase units of native, V349A, or V349A/W387F oPGHS-
1 purified by chromatography on Ni-NTA. The samples were extracted, and the
radioactive products were separated by thin layer chromatography and quantified by
liquid scintillation counting as described under “Experimental Procedures”. The
locations of AA, monohydroxy fatty acid, and PG standards are noted. HHT, 12-
hydroxy-Sc,8c,10t-heptadecatrienoic acid.

43

In comparison, the V349/W387F double mutant produces predominantly ll-HETE

(>84%) and correspondingly. less PGG2 derived products (~12%) than native enzyme.

V349A oPGHS-l forms about 50% ll-HETE [53].

Kinetic Analysis of V349A/ W38 7F oPGHS-l

To determine if AA is utilized with similar efficiencies as native and
V349A/W3 87F oPGHS-l , the Km value for the oxygenation of AA was determined using
a cyclooxygenase assay. The K,,, for AA of V349A/W3 87F oPGHS-l was determined to
be 5 11M, a value very similar to that previously reported for native enzyme under similar

assay conditions [52, 53].

Arachidonic Acid Bound in the COX site of C03 +-pr0t0porphyrin V349A/W38 7F oPGHS-I

The electron density of the Co3+-protoporphyrin V349A/W387F oPGHS-l/AA
complex was of good quality for building in considerable amounts of non-protein
substructure (i.e. the Co3+-protoporphyrin IX, the carbohydrate groups linked to residues
Asn68, Asnl44, and Asn 410, and three B-OG detergent molecules bound within the
membrane binding domain). The overall structure which includes the epidermal growth
factor, the membrane binding and the catalytic domains of the Co3+-protoporphyrin
V349A/W387F oPGHS-l/AA complex is virtually identical to that of Co?”-
protoporphyrin oPGHS-l/AA complex having an r.m.s deviation of 0.3 A over all Cu
atoms. This indicates that the overall structure is relatively unchanged by the two
mutations (V349A and W3 87F) in the COX active site.

The F o-Fc electron density map allowed the placement of carbons C1 to C12 into

the electron density with some degree of confidence although no electron density was

44

observed for carbons C13-C20. Given the broad extent of the electron density observed
at this medium resolution, it was essential to verify that the observed differences in
conformation of bound AA between native and V349A/W387F oPGHS-l were
significant. In a test using Fo-Fc difference electron density map analysis (see
“Experimental Procedures” of this chapter), inspection of the electron density maps
calculated with AA from the native oPGHS-l/AA complex within the V349A/W387F
oPGHS-l active site showed a significant positive peak contoured at 30 representing the
correct position of atoms C3-C7 of AA bound within the mutant active site (data not
shown). In addition, a smaller negative peak contoured at 30 was observed representing
the misplaced positions of C2-C4 of the conformation of AA observed in the oPGHS-
l/AA complex (Protein Data Bank entry IDIY). The results of this test indicate that the
data are of good quality at this limited resolution and show that conformation of AA
bound within the V349A/W387F oPGHS-l mutant differs from that seen in the native
enzyme.

A second validity test, using a series of simulated annealing refinement (see
“Experimental Procedures of this chapter), generated a family of AA conformers that
were indistinguishable from the conformation of AA bound within the V349A/W387F
mutant. All conformers from this refinement series fit the observed density well. The
conformer, which best fit the electron density is shown in Figure 13 along with the
conformer of AA obtained by model building. Both conformers are bound in the COX
channel of the mutant enzyme with the carboxylate group oriented and stabilized at the
mouth of the channel by two ionic interactions with the guanidinium group of Arg120

and one with the phenolic oxygen of Tyr355 (Figure 13). From the carboxylate at C1, the

45

carbons of AA weave through the COX active site making hydrophobic contacts with
various side chains along the path to Tyr385. Continuous density was observed from the
carboxylate of AA to C12; however; no density was observed past C12. The placement
of C13, the site of initiation of catalysis via hydrogen abstraction, was based on the
stereochemical constraints imposed by C12 and the C11-C12 double bond. Although
there are minor shifts in the position of the C8-C9 and C11-C12 double bonds, both
conformers are strikingly similar particularly when comparing the carbon positions of the

carboxylate group up to C7.

  
     

   
 
  
        
 
   
 
   
    
   
 

d,

9
I

9
‘l
-

§
- .." ‘
A
1‘",

  
 
  
   

1

A‘ v \
‘3 s‘
v.‘ .35?
_ Q“:

1‘!
[It

  

i
’31., '3‘"
9.34"

  

 
 

 
         

‘8
’1'“.

Figure 13 Stereo view of AA bound within the active site of V349A/W387F oPGHS-
1. The simulated annealing omit map Fo—Fc density is contoured at 4.0 0 is shown in
green. Carbon atoms of AA, which were originally built into the electron density, are
shown in pink (atoms C-l-C12). Carbon atoms of AA obtained fiom simulated annealing
are shown in blue. Side chains of various amino acids that line the COX active site and
contact the substrate are shown in a color scheme where carbons are gray, oxygens are
red, nitrogens are dark blue, and sulfur is yellow. Figures presented in Chapter H were
created in part using the program SETOR [83].

46

The family of conformers, including the conformer shown in Figure 13 obtained
from the simulated annealing experiments described in experimental procedures section,
is shown in Figure 14. Aside from minor differences in atom positions, the four
conformations obtained from the simulated annealing refinement starting from the AA
bound to native enzyme are all markedly similar to that obtained by model building
(Figure 13). The simulated annealing experiments show that the carbon chain weaves up
into the active site with the same trajectory for all eight conformers. The results of
simulated annealing tests strongly suggest that the X-ray data contain a sufficient amount
of information to derive a unique family of AA conformers that differ in conformation
from that seen in the native oPGHS-l/AA complex. The conformers in Figure 14A are
shown with their (1) ends (C13-C20). While there is no observable density for this portion
of the substrate, the positioning of these atoms from simulated annealing refinement was
entirely guided by the non-bonded energy terms of the molecular dynamics. Although
there are no experimental data to validate the positions of these atoms, the conformation
of the AA in this region is constrained by energy minimized, non-bonded contacts. Thus,
the (0 ends of the AA conformers are in stereochemically reasonable conformations.
While having a limited level of confidence, it is interesting to note that the positions of
the proS hydrogens are generally found to cluster in a group near to that of AA found in

native enzyme. (Figure 14A).

Structural Comparison of Arachidonic Acid Bound to Native and V 349A/ W38 7F PGHS-1

To verify that the two mutations V349A/W3 87F did not alter the overall integrity
of the active site, all atoms of 17 side chains (excluding residues 349 and 387) within the

COX active site of native and double mutant were superimposed giving a root mean

47

 

Figure 14 Arachidonic acid conformers obtained from simulated annealing. A.
Family of conformers obtained from simulated annealing originated from the starting
conformation of AA bound to native oPGHS-l in the V349A/W387F protein. The
starting conformation of AA used in the simulating annealing, shown in orange, was used
to derive the family of conformers shown in blue. For comparison, carbon atoms of AA,
which was originally built into the electron density, are shown in pink. Oxygens of the
carboxylate are colored red. Modeled proS hydrogens are shown in purple. B. Family
of conformers obtained from simulated annealing originating from the starting
conformation of AA originally built into observed electron density in the V349A/W387F
protein. The family of simulated annealing conformers is showin in yellow, the starting
conformation used form simulated annealing is shown in pink, and for comparison the
conformation of AA bound in native oPGHS-l is shown in orange

48

square deviation of 0.4A. This indicates that the positions of the active site residues in
the crystal structure of the V349A/W387F oPGHS-l were not significantly perturbed by
the presence of the two mutations. The overall binding of AA within the COX active
sites of both the native and the double mutant is similar with respect to many of the key
ionic and hydrophobic interactions (Table 2).

Although the structure of AA bound to the V349A/W387F mutant is grossly
similar to that of AA bound to native enzyme, there are several significant differences.
Atoms C3-C6 of AA are shifted into the space created by the Val349Ala substitution
(Figure 15). In the native structure, Val349 contacts C3 and C4 (Table 2). In the mutant,
these interactions are lost, and C3 and C4 rotate and move into the space created by this
mutation. Rotations in this portion of the AA molecule result in large shifts of more rigid
regions of AA as evidenced by the significant downward shift of the C5-C6 double bond
(Figure 15). The differences in the positions of carbons between AA in the native
enzyme versus V349/W387F oPGHS-l structures are as great as 3.0 A when taking into
account coordinate error of 0.43A. There is also a significant difference in the location of
the C11-C12 double bond of AA, which is moved via rotation about the C9-C10 bond
toward the space created by the Trp387Phe mutation. In the native structure, Trp387
makes two contacts with C11 and one with C12 (Table 2). The loss of these contacts in

the double mutant contributes to displacement of the C11-C12 bond.

Discussion

Native PGHS-1 forms small amounts of llR-HPETE and 15R/S-HPETE in
addition to the major product PGG2. [52, 53]. The monohydroperoxides have little or no

biological activity and are considered to be abortive products that result from a failure of

49

Table 2 Distances between selected side chains atoms of COX active site amino
acids and carbons of AA bound to native and V349A/W387F oPGHS-l. The
distances listed between the side chain atoms of AA were abstracted from the native
oPGHS-l/AA crystal structure (Protein Data Bank IDIY) [42] or from the distances
observed between AA bound to V349A/W387F/AA oPGHS-l as reported here (Protein
Data Bank 1U67). NA, not applicable. The mean coordinate errors of the two structures

 

 

are both about 0.4A.

Atom Native oPGHS-I Distance V349A/W3 8 7F oPGHS-I Distance
ofAA (Residue/atom) (A) (Residue/Atom) (A)
C3 V349/C01 3.4 A349/CGl >4
C4 V349/CG] 3.1 A349/CB >4
C5 [523/CG2 3.3 1523/CG2 >4
C5 V349CB >4 A349CB 3.0
C6 [523/CG2 3.4 [523/CG2 >4
C6 V349CB >4 A349CB 3.8
C7 L352/CD2 3.9 L352/CG2 >4
Cl 1 L3 52/CD2 3.4 L3 52/C D2 >4
Cl 1 W387/C22 4.0 NA NA
Cll W387/CH2 3.4 NA NA
Cll NA NA F387/CZ 3.7
Cll L352/CD2 3.4 L352/CD2 3.8
C 12 L3 52/CD2 3.9 L3 52/CD2 >4
C12 Y348/CEZ 3.2 Y348/CEZ >4
C12 W387/CH2 3.7 NA NA
C12 NA NA F387/CZ 3.7
C12 NA NA F387/CE2 3.9

 

50

W387F%
l .
U

 

Figure 15 Comparison of AA bound to native and V349A/W387F oPGHS-l. A.
stereo view of the superposition of AA bound to the V349A/W387F oPGHS-l COX site
(Cl-C12 colored blue with the (1) end portion ((C-l3-C20) obtained from simulated
annealing; colored majenta) and AA bound to the native oPGHS-l COX site (colored
orange). Atoms of amino acids side chains are shown using a color scheme in which
carbons are gray, nitrogens are blue, oxygens are red, and sulfur is yellow. For clarity
only residues Trp387 and Val349 (both colored orange) of the COX active site native
PGHS-1 are shown. These latter amino acids were mutated to Phe387 and Alam,
respectively, which are shown in gray. B. A view of the COX active site rotated 180°
with the same color scheme as in A.

51

the enzyme to form the endoperoxide. In earlier studies, we provided kinetic evidence
that the nature of the oxygenated product formed by native PGHS-1 is determined by the
conformation of AA when the rate determining step in cyclooxygenase catalysis-
abstraction of the l3proS hydrogen [25], occurs [52, 53]. We performed the studies
reported here as a structural test of this concept. We developed a V349A/W387F mutant
of PGHS-1 that forms primarily ll-HPETE, determined the structure of AA in the COX
active site of the mutant and finally compared the structures of AA in the native and
mutant oPGHS-l.

Previous studies identified Val349 and Trp387 as playing particularly significant
roles in guiding formation of the 9,11 endoperoxide group [52, 53, 76]. Substitutions at
either site yield mutant enzymes that produce 30-55% llR-HPETE [53]. The fact that
the V349A/W387F mutant described here forms mainly ll-HPETE suggests that in
native PGHS-1 Val349 and Trp3 87 function coordinately to facilitate PGG2 formation.

The structure of AA in the cyclooxygenase site of V349A/W387 oPGHS-l was
observed to be significantly different than that of AA in native PGHS-1. Both AA
structures appear to be catalytically competent in that the distances between the phenolic
oxygen of Tyr385 and the l3proS hydrogen of AA are estimated to be within about 3 A
in both structures; however, it should be noted that the 13- proS hydrogen modeled into
the AA structure in V349A/W387F oPGHS-l is more distant than that in the native
oPGHS-leA complex. Because native and V349A/W387F oPGHS-l form primarily

PGG2 and llR-HPETE, respectively, the results of our structural studies are consistent

52

with the overall model that the product formed by PGHS-1 is determined by the
conformation of AA in the cyclooxygenase site when the rate limiting step in the reaction
occurs.

Comparisons of the AA structures observed in the native and mutant enzymes and
modeling the structure of a putative 11-hydroperoxyl radical into the COX sites suggest
how Trp387 and Val349 might participate in forming the endoperoxide group. In the
native enzyme there are interactions between Trp387 and both C11 and C12 that are
eliminated when Trp387 is mutated to phenylalanine. The interactions between Trp387
and C11 and C12 may (a) constrain the orientation of the C11-C12 bond thereby
positioning C11 optimally with respect to C9 and (b) after oxygen insertion at C11, guide
rotations of the llR-hydroperoxyl group and the CIO-Cll bond to move the
hydroperoxyl group towards C9 for an R side attack. Mutation of Trp387 to
phenylalanine eliminates the steric interactions involving C11 and C12 that direct
formation of the endoperoxide group. We suggest that with the mutant enzyme, changes
in the rotational freedom about both the C10-C11 bond and the bond between C11 and
the peroxyl oxygen allow the ll-hydroperoxyl intermediate to assume unproductive
conformations, where the attack by the ll-hydroperoxyl group on C9 is unfavorable and
no endoperoxide can be thus formed.

The Val349Ala mutation further contributes to the altered orientation of the C8-
C13 portion of AA in the V349A/W387F mutant. Past studies [49, 50, 52, 53, 76] have
documented the influence of Val349 on product formation and the stereochemistry of O2
insertion. The role of Val349 in supporting formation of the endoperoxide group is

interesting because the effect is exerted over a relatively long distance. In native oPGHS-

53

1, Val349 protrudes into the COX channel and the CG] atom of Val349 contacts C3 and
C4 of AA (Table 2) restricting movement of the flexible C1-C5 region at the carboxyl
end of AA [42, 53]. When Val349 is mutated to alanine, C3 and C4 of AA rotate into the
space created by the mutation and the C5-C6 double bond moves into the newly available
space. These results are consistent with the idea that in the native enzyme, Val349 acts as
a structural guide to position the carboxyl end of AA [53]. The shifts in the positions of
atoms C3-C6 of AA in the V349A/W387F result in a significant change in the final
orientation of the C11-C12 bond that, in turn, permits movement of atoms C9-C12 of AA
into the extra space created by the Trp387Phe mutation (Figure 15). Thus, in the native
enzyme the positioning the carboxyl end of AA by Val349 must facilitate proper
positioning of the atoms C8-C13 for hydrogen abstraction and endoperoxide formation.
In summary, residues Val349 and Trp387 in native oPGHS-l are crucial for
ensuring that AA adopts the proper conformation so that the reaction proceeds on course
to produce PGG2 [53]. We have found that V349A/W387F human PGHS-2 also
produces predominantly ll-HPETE suggesting that these two residues play similar roles
in PGHS-2 (unpublished data). The structural evidence provided by our present study
supports the hypothesis that the nature of the products formed in the COX reaction
depends on the conformation of AA at the instant of abstraction of the l3proS hydrogen.
The conformation of AA in the COX site of native oPGHS-l differs significantly from
the conformation of AA in the COX site of V349A/W387F oPGHS-l. This additional,
albeit indirect evidence supports the conclusion that the conformation of AA observed in

the native enzyme [42, 53] is that which leads to PGG2 formation.

54

Acknowledgements

The use of the Advanced Photon Source was supported by the United States
Department of Energy, Basic Energy Sciences, Office of Energy Research, under
Contract W-31-109-Eng-38. This work was supported in part by Program Project Grant
P01 GM57323 and by R01 HL56773-06 and R01 GM68848 from the National Institutes
of Health.
1This work was previously published in the Journal of Biological Chemistry as an article
entitled “Crystal Structure of Arachidonic Acid Bound to a Mutant of Prostaglandin
Endoperoxide Synthase-l That Forms Predominantly 11-Hydroperoxyeicosatetraenoic
Acid, Harman, C.A., Rieke, C.J., Garavito, RM, and Smith, W.L. J. Biol. Chem. (2004)

279:42429-42935.

55

CHAPTER III

MEAD ACID BINDING TO THE COX SITE OF PGHS-1

Introduction

Much of what we know about the substrate specificity of the PGHS enzymes is
from kinetic and crystallographic studies with essential fatty acids [52, 56] [42, 49, 50],
which are discussed in Chapter 1. Essential fatty acids, which include linoleate (18:2,
(116), EPA (20:5, (113), DHLA (20:3, (116) and AA (20:4, (116) must be obtained from
dietary means in mammals that lack desaturases capable of incorporating double bonds at
positions at or beyond the (06 position. Thus, fatty acids that possess double bonds at 3
or 6 carbons in from the methyl end (tn-end) of the fatty acid are essential.

Lack of essential fatty acids in the diet leads to a condition called essential fatty
acid (EFA) deficiency, which is accompanied by a compensatory accumulation of mead
acid (SZ, 82, Ill-eicosatrienoic acid; 20:3, (1)9), in cellular membranes. The degree of
deficiency can be estimated by the ratio of Mead acid to AA in the phoshoplipids [54, 84,
85]. The pathophysioloical changes associated with EF A deficiency include salt-
dependent hypertension [86], inhibition of chronic inflammation [87] and impaired
reproductive and epidermal functions [88]. Although these physiological conditions have
mainly been attributed to the reduction in the biosynthesis of PGs because of the low
availability of AA in the cellular membranes, it has been suspected that Mead acid, a
non-essential fatty acid, might act as a substitute for AA [58, 89, 90]. It may therefore be
possible that this fatty acid and/or its metabolites contribute to some symptoms of EPA

deficiency.

56

Although Mead acid is a non-essential fatty acid that accumulates in times of EF A
deficiency, it is found as the major eicosanoic fatty acid component in normal cartilage
[91] and plays a structural role in maintaining necessary membrane fluidity [89]. Despite
these mainly structural roles, Mead acid is believed to have a limited metabolic purpose.
Interestingly, studies have found that Mead acid can, in fact be metabolized by 5-
lipoxygenase to leukotrienes, and by several others enzymes, which include arachionate
lZ-lipoxygenase, and cytochrome P450, strongly suggesting that Mead acid could
substitute for AA [58, 92-94].

The PGHS enzymes have been shown to be capable of utilizing various fatty
acids albeit with lower efficiency in vitro [56]. Several studies investigated this
possibility with Mead acid as well and found it to be a very poor substrate for the PGHS
enzymes; however, trace amounts of products including ll-HETE was shown to be
formed from Mead acid by PGHS [58, 90]. The nature of these products demonstrated
that Mead acid undergoes a similar reaction mechanism as observed with AA, in which
hydrogen abstraction occurs at C13. However, prostaglandin products are not formed
from Mead acid. Instead the products resemble that of “abortive” COX reaction
products, which suggests that Mead acid cannot support cyclopentane ring formation
probably because it lacks the (116 double bond [90]. This may also explain the low
catalytic efficiency observed for Mead acid as substrate for PGHS in that there is no bis-
allylic system centered around C 13 (missing (116 double bond), making hydrogen
abstraction at C13 more difficult [90, 95]. This is interesting considering that Mead acid
is very similar to AA, with the exception of the lack of the (116 double bond. This

highlights the significance of 20-carbon essential fatty acids, which have this crucial (1)6

57

double bond. The majority of the PGHS-1 crystal structures complexed with fatty acids
have been with essential fatty acids. These essential fatty acid complexes have
exemplified the importance of how the number and position of double bonds result in
optimal positioning of the fatty acid for catalysis and thus its ability to be utilized as a
substrate. Kinetic and crystallographic studies of PGHS with essential fatty acids have
not only shown how these fatty acids bind and undergo oxygenation by PGHS, but also
that several of them including DHLA and EPA (both 20 carbons in length) may support
endoperoxide formation and cyclopentane ring closure i.e. prostaglandin formation [49,
50, 56, 59]. In an effort to determine why Mead acid, which lacks the (116 double bond is
such as poor substrate for PGHS, we sought to determine a crystal structure of PGHS-1
complexed with Mead acid. This structural study attempts to elucidate whether 1) the
lack of the (116 double bond in Mead acid influences the binding and positioning of mead
acid for catalysis, thus making it a poor substrate or 2) the lack of (116 double bond is

more significant in providing the chemical environment for catalysis and cyclization.

Experimental Procedures

Materials

Arachidonic acid and Mead acid were purchased from Cayman Chemical Co.
(Ann Arbor, MI). S121 insect cells and 100X Antibiotic-Antimycotic were purchased
from Invitrogen. Fetal bovine serum and HyQ-SFX insect medium used for cell grth
and protein expression of protein used in AA complex were purchased from HyClone
(Logan, UT). Ni-NTA superflow resin was purchased from Qiagen, and n-octyl-B-D-
glucopyranoside (B-OG) used for crystallization was purchased from Anatrace (Maumee,
OH).

58

Expression of Recombinant Native Ovine PGHS-1 with C leavable His- Tag

Constructs of PGHS in pFastbac and preparation of baculovirural stocks were
previously prepared by Dr. Chong Yuan, in the laboratory of Dr. William Smith. The
constructs contained in pFastbac vector have a cDNA encoding native oPGHS-l and an
rTEV (recombinant tobacco etch virus) [96] protease cleavable 8X-histidine tag.
DHlOBac cells were transformed with pFastbac vectors to obtain baculoviral DNA,
which was then isolated and used to transfect sf21 insect cells to produce recombinant
baculovirus. The recombinant baculovirus was amplified, and supematants of
amplifications were used as inoculum stocks. Sf21 insect cells were grown in IL shaker
flasks to an approximate density of 1.7-2.0 x106 cells/mL in medium containing 1%
antibiotic and antimycotic and were then infected with recombinant baculovirus at a
multiplicity of infection of 0.1 at 27°C. Three to four days after infection, cells were

harvested by centrifugation.

Purification of Recombinant Native Ovine PGHS-1 with Cleavable His-tag and Native

Ovine from Ram Seminal Vesicles

Purification of recombinant ovine PGHS-1 from 3121 cells was performed
according to a modified procedure developed by Dr. Chong Yuan. Cell pellets from 4L
of cells were resuspended in Buffer A (composed of 20 mM Tris-HCl pH 8.0 and 100
mM KCl), homogenized and then lyzed by sonication. To solubilize PGHS from the
membranes, C10E6 was added to the lysate up to 0.8% (v/v) which was then incubated at
4° C for 45 min with gentle agitation. Solubilized lysate was centrifuged at 15,000 rpm
at 4° C for 20 min, followed by ultracentrifugation at 45,000 rpm at 4°C for 2 hrs. The

supernatant was carefully removed and incubated for at least 3 hrs at 4°C with Ni-NTA

59

resin (Qiagen) equilibrated in Buffer B (composed of 20 mM Tris-HCI, pH 8.0, 300 mM
KCl, 5 mM imidazole, 5% glycerol, and 0.1% C 10E6). After this incubation period, the
mixture of Ni-NTA resin plus supernatant was poured into a column which was washed
with four column volumes of Buffer B (composed of 20 mM Tris-HCl, pH8.0, 300 mM
KCl, 5 mM imidazole, 5% glycerol, and 0.1% CIOEé), Buffer C (composed of 20 mM
Tris-HCI pH8.0, l M KCl, 20 mM imidazole, 5% glycerol, 20mM galactose, and 0.1%
CloEb), and Buffer D (composed of 20 mM Tris-HCl pH8.0, 100 mM KCl, 50 mM
imidazole, 5% glycerol, and 0.1% C10E6). The 8X-histidine-tagged PGHS was then
eluted with Buffer E (composed of 20 mM Tris-HCI, pH8.0, 100 mM KCI, 250 mM
imidazole, 5% glycerol, and 0.1% C10E6). Fractions with the highest specific peroxidase
activity [42] were pooled and concentrated using a Millipore Ultrafree -15 spin
concentrator (30-kDa molecular weight cut off) to volume 250 11L. To cleave 8X-
histidine tag, protein was diluted to 10 mL with Buffer F (composed of 20 mM Tris-HCl,
pH8.0, 100 mM KCl, 250 mM imidazole, 5% glycerol, and 0.1 C10E6) to which 2,000
units of rTEV (Invitrogen) was added and allowed to incubate for >12hrs. To remove the
rTEV protein and histidine tag from PGHS protein, partially purified mixture was loaded
onto a second Ni-NTA, and was washed with two column volumes of Buffer B
(composed of 20 mM Tris-HCl, pH 8.0, 300 mM KCl, 5 mM imidazole, 5% glycerol,
0.1% C 10E6)- PGHS protein lacking the histidine tag was collected in the Buffer B flow
thru while the rTEV protein and histidine tag were retained on the column. Purified
PGHS was concentrated to ~50 111 and the buffer was exchanged into 1X PBS 0.0.4% [3-
OG with a final volume being ~100-250 11L of the purifed protein. Protein

concentrations were determined using the BCA protein assay (Pierce, Rockford, IL).

60

Native ovine PGHS-1, which was to be used for crystallization trials was also purified

from ram seminal vesicles as previously described [78].

Crystallization and Data Collection of oPGHS- I/MA and oPGHS- I/AA Complexes.

Apo-PGHS-l obtained from either sf21 cells or ram seminal vesicles was
concentrated to least 6-10 mg/mL and prepared for crystallization trials by adding 1.5
molar excess of Co3+-protoporphyrin IX to reconstitute the inactive halo-enzyme. Prior
to setting up crystallization experiments, protein was incubated with fatty acid substrates
(AA or Mead acid (MA)) at approximately five times the molar concentration of protein.
Using the sitting drop vapor diffusion method, 3 11L of protein was mixed with 3 11L of
buffer composed of 0.64 M sodium citrate, 0.3-0.9 M LiCl, 0.3% (w/v) B-OG, and 1 mM
NaN3, and equilibrated within reservoir containing 0.68-0.88 M sodium citrate, 0.3-0.6 M
LiCl, and 1 mM NaN 3. Crystals appeared within two weeks to several months. Crystals
were harvested, briefly soaked in cryosolution containing 1.0 M sodium citrate, 1.0 M
LiCl, and 0.5% B-OG and 1M sodium malonate as a cryoprotectant, and then were
immediately flash frozen in liquid N2 (-196°C). Data from both Mead acid and AA
protein complexes were collected at beamline S-ID at COM-cat (Argonne National Lab,

Argonne, IL). Data from a single crystal of both complexes were indexed and integrated

using DENZO [79] and then sealed with SCALEPACK [79].

Structure Determination and Refinement of oPGHS- I/MA and oPGHS- I/AA Complexes

Both oPGHS-l/MA and oPGHS-l/AA datasets were refined with CNS (version
1.1) [80] in which the protein portion of native oPGHS-l (Protein Databank entry 1DIY)

was used as starting model for rigid body refinement. With each refinement one round of

61

simulated annealing proceeded rigid body refinement, which was then followed by
iterative cycles of minimization, B-factor refinement (bgroup), and model building.
Various non-protein portions of the model (i.e. heme, sugars and detergents molecules)
were included as the phases improved during the refinement process. Insufficient
electron density for Coy-protoporphyrin and substrate was observed in the oPGHS-l/AA
structure, and thus these groups were not included in the refined model. It should be
noted that a second dataset of this complex was collected at the same time from the same
crystallization experiment. This dataset was also analyzed to see if the electron density
for Coy-protoporphyrin and mead acid improved. Unfortunately, similar results were
obtained with this dataset, thus further refinement was not performed.

Minimal electron density for Mead acid in the oPGHS-l/MA structure was
observed making it difficult to confidently position Mead acid within the density. As a
result Mead acid was not included in the refined model. There was enough electron
density for portions of Mead acid, however to conduct simulated annealing experiments
described later in this section. Although both models are incomplete, the current R and
Rffee for the oPGHS-l/MA and oPGHS-l/AA complexes are currently 25.1%& 30.1%

and 26.6% & 31.0% (Table 3).

Simulated Annnealing Experiments with the oPGHS-I/MA Complex

Because there was enough density observed in the COX active site to
indicate some degree of Mead acid binding, several trials of simulated annealing (from
CNS version 1.1) were performed in order to obtain conformers that would reasonably fit

within the observed electron density. The simulated annealing experiments were

62

Table 3 Refinement Statistics of oPGHS-l/AA and oPGHS-l/MA Complexes

 

 

 

 

SCALING oPGI-IS-l/AA oPGHS-l/MA
Space group p6522 p6522
Unit cell a=b, c (A) 183.12, 104.07 182.46, 104.26
# of unique reflections 26,448 24,030
Resolution (A) 30-2.75 30-2.85

R merge, all data (%; I > 0) 8.8 (25.3) 8.3 (44.4)
Completeness (%) 97.1 (76.0)a 98.5 (81.2) 3
Average I/0 15 18.1
REFINEMENT CNS CNS

# reflections in refinement 21,301 21462

# of reflections in working set 20293 20422

R & Free R Factor (%) reflections > 10 on F 26.6 & 31.0 25.1 & 30.1
Included Co3+-protoporhyrin in refinement no yes
Included fatty acid substrate in refinement no no

 

 

aRepresents values observed in last resolution shell.

performed as follows: Mead acid was placed in the COX active site of the PGHS-l/MA
complex and experiments were conducted using the observed structure factors of the

oPGHS-l/MA complex. The starting conformation of Mead acid in these experiments

63

was analogous to the original conformation of AA observed in the native PGHS-1 crystal
structure; however, the chemical nature of Mead acid is different in that the (116 double
bond (C14-C15 bond in AA) is missing. Ten trials of simulated annealing were
performed creating a comparison of ten separate conformers. Interestingly, the
conformers were remarkably different from each other within various regions of Mead
acid; in addition, only two of the ten conformers obtained possessed almost identical

conformations.

Ki studies of Mead acid with Native oPGHS-I

The K, of Mead acid was determined by measuring the Km of AA (concentration
curve of 2, 5, 25, 50, 100 11M) at fixed concentrations of 0, 2, 10, and 25 11M of mead
acid. Cyclooxygenase assays were performed at 37°C by monitoring the initial 02 uptake
using an oxygen electrode as previously described in [52] and in “Experimental
Procedures” of Chapter II. An exception to these procedures was that the enzyme and
Mead acid were added together to assay buffer and the reaction was initiated with AA.
The apparent K,,, of AA at each concentration of Mead acid was calculated from non-
linear regression of AA saturation curves and was plotted against mead acid
concentration to obtain the K, of mead acid.

The [CM value for Mead acid was also determined by measuring the rate
reduction of PGHS at constant AA concentration with increasing amounts of Mead acid.
The experiments were conducted at a constant concentration of 50 11M AA and sequential

increasing Mead acid concentrations until an ~50% reduction in the rate was observed.

64

Results and Discussion

Analysis of the oPGHS-l/AA Complex

Native oPGHS-l/AA structure was determined to 2.75 A resolution. Well-
defined density (contoured to 10) was observed for the protein portion in which various
side chains and carbonyl groups of the protein backbone were easily distinguishable. In
addition, high quality of density was observed for some but not all of the sugar residues,

which include two sugars at Asn68, two sugars at Asnl44, and two sugars at Asn410.
Although these portions of the structure were clearly observed, density for the Co3+-

protoporhyrin and AA were not as well defined being only sparingly observed at contour
levels below 2.50 (Figure 16).

In previous structures, significant and well-defined density has been observed for
COM-protoporphyrin, particularly the Co3+ ion; therefore, the lack of density in this
region of the structure was surprising and an indication of low occupancy possibly as a
result of insufficient reconstitution of Co3+-protoporphyrin. Several reasons could
explain the lack of binding of Co3+-protoporphyrin. First, Co3+-protoporphyrin does not
bind as well as heme, which is required for PGHS activity. In addition, the Co“-
protoporphyrin stock used for these crystallization studies could have been oxidized
therefore, could not bind to PGHS efficiently. The combination of these two possibilities
could explain the low occupancy of Co3+-protoporphyrin in this structure. Density in the
COX active site, where AA normally binds was also very minimal. As with previous
structures, density in the vicinity of Arg120 and Tyr355 was usually observed. In some

cases, density originating from the mouth of the active site up into the channel was also

65

 

g R120

Figure 16 View of electron density observed in the POX and COX site of native
PGHS-1. The top figure shows the observed electron density within the POX site at two
contour levels of 2.50 (green) and 3.00 (magenta) and superimposed within the site for
comparison is the Co +-protoporphyrin from the original PGHS-1 structure (ldiy).
Bottom figure shows the electron density observed in the COX site also shown at two
contour levels (same as in A). AA from the original native PGHS-l/AA complex (ldiy)
was superimposed for comparison in the COX active site

observed. In this case, no density was observed near the mouth; however, a slight
elongated bulge of density (contoured to 2.50) was observed further up in the channel,

possibly representing carbons C3-C6 (Figure 16). The lack density for AA again

66

indicates low occupancy, which is possibly a consequence of insufficiently bound Co3+-
protoporphyrin. Studies have indicated that with apo-PGHS-l there is a lower affinity for
ligands as compared to the halo form of PGHS.

Because the electron density for both Co3+-protoporphyrin and AA was
insufficient, these two significant portions of the structure were not included; therefore,
specific structural information could not be obtained from this structure complex.
Although this model could not serve as a comparison to the oPGHS-l/MA structure,
there were several major advances made in the process of obtaining this structure with
respect to purification and crystallization. This was the first time that crystallization trials
were performed with PGHS in which the 8X-histidine tag was removed. In this series of
experiments, removal of the histidine tag helped facilitate faster crystallization of PGHS
since crystals of PGHS having the histidine tag removed appeared in 1-2 weeks as
compared to PGHS still possessing the histidine tag appearing after 2-4 month or longer.

There is the possibility that the reduced time for crystallization could be a result
of increased purity of PGHS. This construct of PGHS-1 encoded a longer histidine tag,
an 8X-histidine tag, in this case, as compared to the previous constructs with a 6X-
histidine tag. A longer histidine tag may be more beneficial in that it can enhance the
binding i.e. by binding more tightly, thus allowing a more stringent washing of the
column which in turn would more effectively remove impurities before elution of protein.
In this case, it would be difficult to determine if the decreased time to grow crystals is a
result of more pure protein or removal of the histidine tag. Several experiments thus far
suggest that removal of the histidine tag may play the greater role; however, purer protein

may also assist in improving conditions for crystal growth.

67

Analysis of the oPGHS— I/Mead acid Complex

The structure of oPGHS-l/MA complex was determined to 2.85A. Well-defined
density was observed for the protein portion of the structure as well as most of the non-
protein substructure including heme and several sugars residues. Unlike the native
PGHS-I/AA structure, density for the Co3+-protoporphyrin group in the native PGHS-
l/MA complex was complete and very well defined with a large density peak for Co3+
ion observable to a contour level up to 50 indicating Co3+-protoporphyrin is bound with
high occupancy. Initial Fo-Fc density contoured to 2.5-30 within the COX active site
indicated the presence of bound Mead acid; however, the density was not continuous up
through the COX channel. A large peak of density was observed near the mouth in
addition to a second larger elongated peak reaching up into the COX channel to
approximately Tyr385 (Figure 17). AA from the original native PGHS-l/AA structure
was used as a guide in fitting Mead acid into the density with the carboxylate group being
placed into the peak of density observed at the mouth as a starting point fitting the
remaining carbon chain into the portion of density observed further up into the channel.
Although electron density was observed within the COX active site indicating Mead acid
was bound, at best the density only showed possible placement of atoms C1-C10.
However, placement of atoms C2-C10 was difficult to do with confidence. This
difficulty in placing Mead acid in that density is not unusual given that there are probably
many conformers that exist and that all or most of these conformers could reasonably fit
the electron density observed for mead acid. To remedy this situation of multiple

conformers that would reasonably fit the observed electron density, simulated annealing

68

experiments similar to those discussed in Chapter II were performed. These experiments

provided a means to utilize the x-ray information observed for the carboxylate portion

Y385 Y385

1523

 

Figure 17 Initial Fo-Fc electron density observed in the COX active site of oPGHS-
1/MA complex. Electron density is shown at two contour levels of 2.50 (green) and
3.00 (magenta).

of Mead acid and combine it with molecular dynamics, which would treat the portions of
Mead acid having no observed structural information. Ten trials of simulated annealing
experiments were performed as described in the “Experimental Procedures” section of
this chapter, and ten conformations were obtained. The conformers overall appear to
grossly fit the limited electron density; however, some conformers have a better overall

fit than others (Figures 18).

69

W387 m5 W387 Y385

 

1523

 

Figure 18 Comparison of catalytically possible and inhibitory conformations of
mead acid obtained from simulated annealing. A. Comparisons of the six catalytically
possible conformers of mead acid obtained from simulated annealing within Fo-Fc
electron density shown at two contour levels 2.50 (green) and 3.00 (magenta). Of the six
conformers, two have almost identical conformations (green and yellow). B.
Comparison of the four inhibitory conformers of Mead acid obtained fi‘om simulated
annealing shown within the Fo-Fc density (same density as shown in A).

70

There are significant differences observed between the conformers. With all the
conformers, the carboxyl group fit well within the small peak of density observed near
the mouth of the COX active site; however, extreme deviations start to occur reaching
atoms C2-C4, which is one region missing electron density. The trajectory of these
atoms is crucial in how the remaining atoms in the chain will be oriented. These
simulated annealing experiments reveal that there is much variability in these carbon
positions, which may account for the lack of well-defined density for this region of Mead
acid. The same may be true for the (o-end (atoms C14-C20) of Mead acid. Mead acid
would have more flexibility within its w-end as compared other fatty acids such as
DHLA, which has the Cl4-C15 double bond, and EPA having an addition double bond at
position C17-C18. Interestingly, the mead acid conformers obtained from simulated
annealing reflect this concept, in which significant variations occur near or at the C14-
C15 bond position resulting in extreme variation in position of the C13 atoms, and its
modeled in proS hydrogen. Comparing the position of the modeled proS hydrogens of all
the conformers, two groups of conformers were observed as being those in a catalytically
possible conformation and those in an inhibitory conformation. Of the ten conformers,
six are those of the catalytically possible conformers (including conformers 1, 3, 4, 6, 8,
9), while the remaining four conformers (conformers 2, 5, 7, 10) are considered to be of
inhibitory binding mode appearing unable to support the abstraction of the C13 proS
hydrogen of mead acid.

In comparing the positions of the modeled proS hydrogens of the catalytically
possible conformers, the proS hydrogens, ranging in distances of 2.7-3.2A from Tyr385,

the catalytic residue, appear to be in reasonable positions for potential abstraction (Figure

71

19). Interestingly, the carboxyl end and co-end of each of the “catalytically possible”
conformers do vary significantly; despite these variations in the carboxyl end of mead
acid, the trajectory of the carbon chain up to C13 of each conformer eventually aligns

with Tyr385 for the potential abstraction of the proS hydrogen at C13.

 

Figure 19 Comparison of the modeled proS hydrogen positions of the catalytically
possible conformers obtained from simulated annealing. Despite having significant

differences in atom positions within the carboxyl half and (D'Cfld of Mead acid, the proS
hydrogens are oriented towards Tyr385 within distances ranging from 2.8-3.3A.

When comparing how the two groups of conformers fit the observed electron density
within the COX site, the catalytically possible group of conformers appears to fit the
density better than the inhibitory conformers (Figure 18). The electron density, therefore,
seems to reflect an average conformer, which may be more similar to average of the
catalytically possible conformers. However, lack of density for atoms near or at atom
C13, in addition to there being so much variation observed within the carboxyl end of all

the conformers, this suggestion is just speculation.

72

The inhibitory conformers differ significantly from the catalytically possible
conformers particularly in areas near atoms C13-C15 of mead acid. Comparison of the
carbon positions between the catalytically possible conformers and the inhibitory
conformers reveal that the effects of missing the (116 double bond (C14-C15 double bond)
is more pronounced with the inhibitory conformers being reflected by the extremely
altered position of C13. The modeled proS hydrogens at C13 of three of the four
inhibitory conformers are positioned in such as way as to not support potential hydrogen
abstraction since the proS hydrogens of these conformers are oriented in opposing

directions with respect to the position of Tyr385 (Figure 20).

Y385 Y385

 

Figure 20 Comparison of the modeled proS hydrogen positions of the inhibitory
binding conformers obtained from simulated annealing. The proS hydrogens are
shown in very altered positions relative to Tyr385 demonstrating these binding modes are
unlikely to support catalysis. Conformer shown in orange (conformer 7) is observed as
being significantly different than the other three conformers.

73

Interestingly, this type of binding mode was captured in the crystal structure of EPA
bound to PGHS [49]. Unlike the other inhibitory conformers, the modeled proS
hydrogen at C13 of conformer 7 is somewhat oriented up towards Tyr3 85; however, it is
too far removed from the Tyr3 85 for potential abstraction (Figure 20).

These structural results are consistent with the fact that Mead acid is a poor
substrate for PGHS, being that it cannot be metabolized very efficiently. Trace amounts
of products are formed and were found to resemble that of abortive COX products such
as ll-HETE, thus suggesting that Mead acid undergoes a similar sequence of catalytic
steps beginning with the abstraction of the proS hydrogen, the rate limiting step. The
simulating annealing experiments reflect this in demonstrating that Mead acid can bind a
majority of the time in a conformation such that abstraction of the C13 proS hydrogen is
possible. Interestingly, the simulated annealing experiments also revealed that Mead acid
could bind in a mode that would not support catalysis (i.e. bind within the active site in
an inhibitory conformation). This opened up the possibility that Mead acid could actively
compete with AA in binding and may act as an inhibitor of PGHS. EPA is known to be a
potent inhibitor of PGHS. This was further evidenced by the crystal structure of the
PGHS-l/EPA complex in which EPA was captured bound in an inhibitory manner with
the proS hydrogen at C13 being oriented in the opposing direction of Tyr385 making
potential abstraction unlikely [49]. The Ki of Mead acid was measured to test the ability
of Mead acid to act as an inhibitor of PGHS-1. The Ki was determined to ~62 11M
(Figures 21 & 22), in addition, the IC50 of Mead acid was found to be ~100 11M at 50 11M
AA. These results indicate that mead acid can compete with AA and inhibit PGHS

activity to some degree. However, when comparing the Ki of Mead acid to the Km of AA

74

determined to be 9 11M with no Mead acid present (Figure 21), in addition to the IC 50
values ranging from 0.3-50 11M of various NSAIDs [71] and the indomethacin

ethanolamides [97] described later in Chapter IV, it appears that Mead acid

 

 

 

 

 

 

 

140 ~
120 - ;/,J---e-"”""
E I
E
E 100‘
Q)
S
3 80' Equation:
3 l y=P1*x/(P2 + x)
8 .0-
g R2= 0.93862
75 40-
g P1=137.1691:6.616
P2= 9.107 x 1.430
20-
0 ' 1 ' 1 ' W 0 1 r I 1
0 20 100

Arachidonic Acid (11M)

Figure 21 Arachidonic acid saturation curve of PGHS. Shown is one of four non-
linear regression fits (from Origin) of an AA saturation curve of PGHS at 0 11M Mead
acid. All data, including at least two measurements for each concentration of AA along
the curve was included in the non-linear fit. P1 and P2 represent Vmax and Km, both of
which were varied in the non-linear regression fit (Vmax =137.2 and Km=9.1 at zero mead
acid concentration). Non-linear regression fits of AA saturation curves of PGHS in the
presence of 5, 10, and 25 11M were plotted similarly in Origin. With the regression of
each of these plots, the P1 (Vmax) was fixed at 137.2 while varying P2 (Km) to obtain the
apparent Km of AA at various Mead acid concentrations. To determine Ki of Mead acid
the apparent Km were replotted against Mead acid concentration and results are shown in
Figure 22.

75

y = 0.1376x + 8.4518

  

 

R2=0.9116
E
x
H
C
0
'1" 1
& 1
“ 1
1
1
111111. 1.1111-11111111.1e1i1 . .111
-70 -60 -50 -4o -30 -20 -1o 0 10 20 30
Mead Acid (11M)

Figure 22 The K; plot of Mead acid. The apparent K," values obtained from similar
saturation curves shown in Figure 21 were plotted against the Mead acid concentration.
The K,- represented by the x-intercept was determined to be ~62 11M. The equation of the
regression and correlation coefficient (R2) are displayed in the graph.

is not a very potent inhibitor of PGHS. This is interesting given that Mead acid, although
a poor substrate appears to bind to the COX site so how can it not be a potent inhibitor.
One explanation is that Mead acid just does not bind as well as AA. This is evidenced by
observing the variability in carbon atoms positions of ten conformers obtained from the
simulated annealing experiments. In addition, the limited density observed for mead acid
in the COX active site possibly indicates low occupancy of Mead acid within the COX

site. As a result of increased flexibility within the (o-end (atoms Cl4-C20) of Mead acid

76

as compared to other fatty acids such as EPA or DHLA, a stable binding conformation
may not be able to be achieved. The density observed for the carboxylate group of Mead
acid near that mouth is shifted from the position of the carboxylate group of AA in
PGHS-l/AA complex. The consequence of this difference in the placement of the
carboxylate group of Mead acid is a loss of the interaction between the carboxylate group
of Mead acid and guanidium of Arg120 being a significant interaction for high affinity
binding of AA and other fatty acids as well as various NSAIDs with a free carboxylate.
With the loss of this interaction as a result of increased flexibility within the (D-el'ld, Mead
acid may not be able to stabilize an interaction with Arg120, thus preventing it from
binding as efficiently as AA.

These studies with mead acid demonstrate that the (116 double bond found in
essential fatty acids appears to not only play an important role in the positioning the C13
proS hydrogen of 20 carbon fatty acids with respect to Tyr385, but appears to be
significant in achieving high affinity binding of essential fatty acids to PGHS-l.
Although six of the ten conformers obtained from simulated annealing indicate that it is
possible for Mead acid to bind in a catalytically possible conformation allowing for
potential hydrogen abstraction at C13, a combination of the kinetic studies with Mead
acid and the limited electron density observed in the COX active site suggest that Mead
does not bind as efficiently as the essential fatty acids AA, DHLA, and EPA to PGHS-1,

thus may explain why this fatty acid is a poor inhibitor as well as a substrate of PGHS-l.

77

CHAPTER IV

STRUCTURAL BASIS OF ENANTIOSELECTIVE INHIBITION OF COX-1 BY (S)-

a-SUBSTITUTED INDOMETHACIN ETHANOLAMIDES

Introduction

Non-steroidal anti-inflammatory drugs (N SAIDs) provide therapeutic benefit
through the direct inhibition of the COX enzymes; however, the anti-inflammatory and
analgesic properties of NSAIDs are mainly a result of inhibition of COX-2. Although
non-selective inhibition of COX-1 by aspirin can be beneficial in reducing the risk of
mortality from cardiovascular disease, the use of aspirin and other NSAIDs has been
associated with undesirable gastrointestinal side effects such as stomach ulcers and
intestinal bleeding [71, 98]. The predominant role of COX-2 in pain and inflammation
spawned the development of selective inhibitors, which target only the COX-2 isoform in
order to achieve analgesic and anti-inflammatory properties with minimal associated
gastrointestinal toxicity. However, cardiovascular side effects have recently been
documented for at least certain COX-2 inhibitors (eg. rofecoxib, valdecoxib).

Structural and functional studies comparing COX-1 and COX-2 have identified
some general structural features of NSAIDs binding to the COX active site, in addition to
highlighting subtle biochemical differences between the isoforms that have been
exploited in the development of COX-2 selective inhibitors [39-41]. The entrance to the
COX active site is made up in part by a bundle of four amphipathic helices, which leads
to a constriction comprised of residues Arg120, Tyr355, and G1u524. The COX active

site then opens into a long hydrophobic channel that extends deep into the core of the

78

catalytic domain of the protein. Sitting above the constriction and aside the main
hydrophobic channel is a smaller, amphipathic region referred to as the “side pocket”. A
major difference in the active site of the isoforms arises mainly as a result of a valine
substitution in COX-2 for isoleucine in COX-l at position 523. This small change
provides enough additional space to allow better access to the side pocket of COX-2.
The active site of COX-2 can therefore accommodate the extra bulk of the
sulfonamoylphenyl or methylsulfonylphenyl moieties of COX-2 selective inhibitors,
particularly those of the diarylhetercyclics (celecoxib, rofecoxib), by placing them in the
side pocket.

The potential for interaction between moieties of the drug and residues within the
side pocket of COX-2 also seems to be responsible for a phenomenon known as time-
dependent inhibition by COX-2 selective NSAIDs. Although diarylheterocycles
competitively inhibit COX-2 better than COX-1, their lack of time-dependent inhibition
behavior towards COX-1 further enhances their selectivity towards COX-2 [99, 100].
The structural explanation for the lack of time-dependent inhibition towards COX-1 is the
extra methyl group of isoleucine at position 523, which restricts access of COX-2
inhibitors to the side pocket in COX-1.

An addition feature of COX-2 selective inhibitors, which may contribute to their
low affinity towards COX-1, is that they lack a free carboxylate group. Several crystal
structures of COX-1 and COX-2 have shown that NSAIDs with a free carboxylate moiety
form an ionic interaction with Arg120, which is located at mouth of the COX active site
[39, 41, 42, 101]. This interaction between the carboxylate of acidic NSAIDs and

Arg120 appears to be essential for substrate binding to COX-l, as site-directed

79

mutagenesis studies of COX-1 have demonstrated that the binding of AA and acidic
NSAIDs was greatly perturbed when Arg120 is mutated to smaller uncharged residues
[67]. In contrast, the equivalent mutations of Arg120 seemed to suggest that electrostatic
interactions with this residue in COX-2 were not at all critical for catalysis or NSAID
inhibition [68].

Modification of the carboxylate group of the non-selective NSAID indomethacin
generates a wide range of ester and amide adducts yielding an array of COX-2 selective
inhibitors [102]. The majority of these indomethacin derivatives were shown to be potent
and selective for COX-2; however, an exception was observed with the a-substituted
indomethacin ethanolamides series of derivatives [97]. The (R)/(S)-or-substituted
indomethacin ethanolamides both demonstrated potent inhibition of COX-2; however, the
(S)- a-substituted indomethacin ethanolamides inhibited COX-1 as effectively as COX-2.
Interestingly, the chiral preference observed with COX-1 inhibition was consistently
observed with (S) enantiomers across a wide range of (It-substitutions (Table 4).
Mutagenic studies in COX-2 were unable to identify residues around the COX active site
that contributed to the observed chiral selectivity. To understand the structural basis of
this chiral discrimination in COX-l , we determined the crystal structures of ovine COX-1
complexed with the a-ethyl substituted enantiomeric pair of these indomethacin
ethanolamides. These structures not only provide intriguing insight into the structural
basis for the stereo selective binding of the (S)-or-substituted indomethacin ethanolamides
to COX-1, but also provides an illustration of how COX-1 can bind a non-carboxylate

containing inhibitor with high affinity.

80

Table 4 Comparative inhibition of native COX-1 and COX-2 by various a-
substituted indomethacin ethanolamides. Data shown in table was obtained from
previously published study by Kozak et al. [97]

3,, R
HN B
O O
/ \
N
O

 

 

CI
ovine COX-l human COX-2
Compound ID or-substituent IC50 (uM) ICSO (11M)

6 (R)-CH3 33 0.17
7 (S)-CH3 0.59 0.27
8 (R)-CH2CH3 59 0.10
9 (S)-CH2 CH3 0.35 0.28
14 (R)-CH2CH2SCH3 51 0.25
15 (S)-CH2CH2$CH3 2-3 1-0

16 (R)-C6H5 47 0.44
I 7 (S)-C6H5 <1.3 0.085

 

81

h..-

Experimental Procedures

Materials

Detergents C10E6 used for solublization and n-octyl-B-D-glucopyranoside (B-OG)
used for purification and crystallization was purchased from Anatrace (Maumee, OH).
Hemin was purchased from Sigma (St. Louis, MO). Compound 8 (Indomethacin-(R)-a-
ethyl-ethanolamide) and compound 9 (Indomethacin-(S)-a-ethyl-ethanolamide) used in
crystallization experiments were synthesized according to the procedures described in
Kozak et al. [97]. The nomenclature as compounds 8 and 9 refers to their earlier

designation [97].

Crystallization of native ovine C OX- I/inhibitor complexes

COX-l used for crystallization was purified from ram seminal vesicles as previously
described [78]. Protein was prepared for crystallization trials by adding 1.5 molar excess
of hemin to reconstitute the protein. Inhibitors at a concentration of 100 11M of either
compound 8 or compound 9 were then added to protein and left to incubate for at least 10
min before setting up crystallization experiments. Using the sitting drop vapor diffusion
method, 3 (AL of protein was mixed with 3 11L of buffer composed of 0.64 M sodium
citrate, 0.3-0.9 M LiCl, 0.3% (w/v) B—OG, and 1 mM NaN3, and equilibrated within
reservoir containing 0.68-0.88M sodium citrate, 0.3-0.6 M LiCl, and 1 mM NaN3.

Crystals appeared within 2-3 weeks

Data Collection of 0C OX- 1/C0mp8 and oCOX-I/Comp9 complexes

Crystals were harvested, briefly soaked in a solution containing 1.0 M sodium

citrate, 1.0 M LiCl, and 0.15% (w/v) B-OG, and 1 mM sodium malonate as a

82

cryoprotectant. Crystals were then flash frozen in liquid nitrogen at -165°C. Data
collection on crystals of COX-l/drug complexes was performed at beamline 5-ID at
COM-cat (Argonne National Lab, Argonne, IL). Data from a single crystal of COX-
l/complexed with compound 8 (COX-l/Comp8) was indexed and integrated using
DENZO [79] and then sealed with SCALEPACK [79]. Datasets collected from two
crystals of COX-l/complexed with compound 9 (COX-1/Comp9) were indexed and

integrated separately and then merged together during scaling using SCALEPACK.

Structure Determination and Refinement with CNS

The COX-1 structures complexed with compound 8 or compound 9 were
determined using native COX-1 structure (Protein Databank entry IDIY) as a phasing
model for rigid body refinement utilizing CNS version 1.1 [80]. After an initial round of
simulated annealing, several iterative cycles of positional and group B-factor refinement
and model building in which sections of model including several sugar molecules,
detergent molecules, in addition to fitting inhibitors were performed. To further complete

the model, water molecules were included and were followed by additional refinement.

Structure Determination and Refinement with REFMAC

The COX-l/Comp9 and COX—l/Comp8 datasets were also refined with
REF MAC version 5.0 in the CCP4 program suite [103]. REFMAC refinement for the
COX-l/Comp9 dataset started with the protein portion of COX-l/Comp9 structure
refined by CNS (Protein Databank entry lDIY). The COX-1/Comp9 structure was
initially put through TLS refinement [104] and then restrained positional refinement.

Each round of refinement included 10 cycles of TLS at fixed B-factor of 40, followed by

83

20 cycles of restrained refinement (which included positional refinement followed by B-
factor refinement). A single TLS group consisting of protein residues 33-580 was used
for the TLS portion of refinement. Iterative cycles of TLS/restrained refinement were
followed by cycles of map calculation and model building. As refinement progressed,
the TLS group was expanded to include protein, in addition, to heme, sugars, detergent
molecules and compound 9 as they were included in the model. After adding heme, six
saccharide groups, and a glucopyranoside head group from B-OG, and approximately 20
waters, the R and Rfrcc converged at 23.9% and 29.7%. The same refinement protocol
was applied to the COX-1/Comp8 dataset to complete the comparison of the two

refinement programs (Table 5).

Result and Discussion

Comparison of CNS vs. REF MAC Refinement for C OX-I/inhibitor complexes

There was a significant difference in the R factors obtained from CNS as compared to
REFMAC with the COX-l/Comp9 and COX-1/Comp8 structures (Table 5). Although,
the crystallographic R factors differ noticeably, the electron density maps obtained from
both refinement programs were virtually identical, especially with respect to the electron
density observed for the inhibitor in the active site. The shape of the electron density was
identical, and the inhibitor orientation determined in CNS refinement fit easily into the
density obtained from refinement with REFMAC. The significant difference in the
refinement statistics between CNS and REFMAC may not be unusual given that each

program utilizes different methods and algorithms to calculate bulk solvent and other

84

E

Table 5 Summary of data collection and refinement statistics.

 

 

 

 

 

 

SCALING COX-1/Comp8 COX-l/Comp9
Spacegroup P6522 P6522
(a)A 181.72 181.40
(b)A 181.72 181.40
(c) A 104.09 103.40
Resolution A 2.85 2.70
No. of unique reflections 24,207 28,072
Completeness 97.4 (82.9)“ 99.9 (100) 1
11mg, 7.8 (24.1) " 9.7 (33.5) “
REFINEMENT CNS REFMAC CNS REFMAC
No. of used reflections 22932 22275 23511 26478
No. of atoms in refinement 4598 4641 4575b 4653 b
R-factor (%) 24.2 21.7 28.3 24.1
Free R (%)c 29.6 26.1 33.5 29.2
Mean positional error (A) 0.45 0.49 0.54 0.58
r.m.s.d. bond length (A) 0.008 0.013 0.013 0.014
r.m.s.d. bond angle (°) 1.60 1.78 1.76 1.56
Average B-factor
protein 68.9 72.6 54.2 49.3
inhibitor 95.3 87.1 87.0 55.2
citrate 73.5 94.7 NA NA

 

 

 

aValues in parentheses represent values in highest resolution shell
I’The number of atoms included in refinement differ since some atoms were

not included in CNS refinement.
cFree R was calculated with 5% of reflections set aside as the test set which is

parameters during refinement. This is particularly true for the TLS

(translation/libration/screw) refinement procedure in REF MAC [104, 105], which better

treats the static and dynamic disorder arising from rigid body displacement than

traditional B-factor refinement.

85

Binding of Compound 8 vs. Compound 9 in the COX site

The electron density maps obtained from both CNS and REF MAC refinement of
the COX-l/Comp8 structure shows that compound 8 binds in an orientation that closely
resembles the way the parent compound, indomethacin, binds to the enzyme [41, 101]:
the chlorobenzyl group is orientated up to the apex of the channel by Y3 85, the methoxy
group of the indole ring points towards the COX-2 side pocket, and the ethanolamide
group is positioned towards the mouth of the channel near R120 and Y355 (Figure 23).
The hydroxyl group of the ethanolamide moiety of compound 8 makes a hydrogen bond
with the guanidinium group of R120 (Figure 23). The electron density for compound 9,
the (S)-stereoisomer, in the COX-1/Comp9 structure, obtained from refinement with
either CNS and REFMAC, clearly shows an elongated stretch of electron density within
the side pocket region of the COX active site defined by residues Gln192, His90, L517,
Phe518, Ile523 (Figure 24A). This contrasts with what is seen with the COX-1/Comp8
complex. The orientation of compound 9 that best represents the observed electron
density is with the chlorobenzyl moiety orientated towards the mouth, the methoxy group
points towards the top of the channel by Y385, and the ethanolamide group is positioned
in this side pocket region (Figure 24 A&B). The ethanolamide group of compound 9
makes many interactions with the hydrophilic and hydrophobic residues of the side
pocket with the hydroxyl group making hydrogen bonds with His90 and Gln192 and
several hydrophobic interactions with Phe518 and Ile523 (Figure 24 B). An additional
hydrophilic interaction is observed between the guanidium group of Arg120 and the
carbonyl group of the chlorobenzyl moiety oriented at the mouth of the active site. The

extra density that is observed in the side pocket of the COX-1/Comp9 complex is the

86

major difference observed in the electron density maps between the two complexes thus

indicating two different and contrasting modes of binding for Compound 8 and 9.

 

Figure 23 Compound 8 bound in the COX active site. A. Stereo view of compound 8
bound in the COX active site with simulated annealing omit map difference density
(blue) contoured to 40. Various residues within the active site are shown with carbons
for compound 8 shown in orange, oxygen red, nitrogen blue, and chlorine magenta. B.
Stereo representation of compound 8 bound in the COX site in the same manner as the
parent compound indomethacin; the chlorobenzyl group oriented up towards the top of
the channel, the methoxy group on the indole ring points towards the side pocket (L517,
F518, 1523, Q192, S516), and the ethanolamide group with R—ethyl substitution sits next
to R120 and Y355 at the mouth of the active site. Carbons atoms of compound 8 are
shown with same color scheme as in (A). The hydroxyl group of ethanolamide group
makes a hydrogen bond with R120, while the R-ethyl group is positioned just outside the
mouth of the active site.

87

 

Figure 24 Compound 9 bound in the COX active site. A. Stereo view of compound 9
fitted into simulated annealing omit map difference density (blue) contoured to
40. Carbons atoms of compound 9 are colored green with heteroatoms colored same as
in Figure 23. B. Compound 9 bind much differently than 8: the chlorobenzyl group is
oriented towards the mouth of the active site, the methoxy group of indole ring is oriented
up into the top of the channel, and the ethanolamide group with the S-ethyl substitution
lies within the side pocket region of the channel. The hydroxyl group of the
ethanolamide is oriented near the hydrophilic region (cyan) of the side pocket making
hydrogen bonds with residues Q192 and H90 while the S-ethyl group is oriented into the
more hydrophobic region (pink) of the side pocket making Van der Waals intereactions
with 1523 and F518.

88

Evaluation of the Structural Basis for Stereoselectivity

The inhibitor complexes suggest that the distinctly different binding modes for
compound 8 and compound 9 arise as a result of the ability of one enantiomer to optimize
binding interactions within the side pocket. The ethanolamide moiety of compound 9,
which is able to access the side pocket in COX-1, utilizes both the hydrophilic and
hydrophobic properties of the COX-1 side pocket; the hydroxyl group of the
ethanolamide makes hydrogen bonds to residues Glnl92 and His90, while the ethyl
group makes several hydrophobic contacts with residues Ile523, Phe518, and Leu517
(Figure 24 B). The binding mode of compound 9 is analogous to the way COX-2
selective inhibitors bind to COX-2 suggesting that the side pocket in COX-1 is accessible
depending on the nature of the substitutents present in the inhibitor. Structure-activity
studies, which systematically evaluated various substitutions at the or-position of
indomethacin ethanolamides, found that altering the hydroxl group influenced the chiral-
dependent selectivity towards COX-1 to some degree [97]. When the hydroxyl group
was replaced with either a methoxy or methyl group the chiral dependence changes so
that the (R) enantiomers were more potent against COX-l than the corresponding (S)
enantiomers. Collectively, these results suggest that the inability of the (R) enantiomers to
inhibit COX-1 was not merely due to an increase in unfavorable steric interactions, but
also dependent on the loss of favorable binding interactions. This is demonstrated in the
crystal structures presented here in that an elaborate network of interactions is needed to
favor the more stable binding mode of compound 9. Conversely, the binding mode of
compound 8 observed in the crystal structure displays a quite different set of interactions,

where none occur within the side pocket. Instead, the ethanolamide moiety of compound

89

8 is oriented towards the mouth of the active site, making one hydrophilic interaction
between the hydroxyl of compound 8 and guanidium group of Arg120 (Figure 23 B).

To visualize why compound 8 does not adopt the same binding mode as
compound 9, compound 8 was modeled in COX active site by positioning the
ethanolamide moiety within the side pocket (Figure 25 A). With the hydroxyl group of

the ethanolamide positioned to make potential ionic interactions with Q192 and H90, the

 

L517

0192

H90

   

Figure 25 Comparison of the placement of the (R)-ethy1 vs. (S)-ethyl group in COX-
1 side pocket. A. Stereo View of compound 8 modeled in the COX active site as
compound 9 would bind. The R-ethyl group of 8 (orange) is positioned so that the
hydroxyl group is in an orientation where it could hydrogen bond with Q192 and H90. B.
Stereo of compound 9 bound within the COX active site for comparison of the position of
the ethyl group.

90

R-ethyl group was unable to take advantage of similar hydrophobic interactions observed
for compound 9 (Figure 25 B) because the ethyl group is oriented in the opposite
direction of the S-ethyl group of compound 9. In addition, it became clear that the R-
stereochemistry made it more difficult to avoid steric clashes particularly with the protein

backbone between residues 352 and 355.

Anomalous ligand interactions in the COX-I/Compound 8 Structure

During inspection of the COX-1/Comp8 structure, significant Fo-Fc electron
density (contoured to 40) was observed near the heme within the peroxidase active site.
The shape of the density appeared to resemble citrate, which is present in the
crystallization buffer. After including citrate in the model refinement, the refined density

from CNS showed that citrate fit well (Figure 26). Citrate is bound with one of its

 

Figure 26 Citrate bound in the POX site. Stereo view of citrate bound in the POX site
of the COX-1/Comp8 structure with simulated annealing omit map density (blue)
contoured to 30. Citrate (carbons yellow, oxygen red) is straddled between heme (carbon
gray, oxygen red, nitrogen blue, iron rust) and Lys222 (carbon cyan, nitrogen blue).
Residues proposed to interact with PGG2, the substrate of the peroxidase are shown as
spheres. Site of trypsin cleavage is near residue R277.

91

carboxylate groups interacting with one of the propionate groups of heme. A second
interaction involves the amino group of Ly5222 interacting with another one of the
carboxylate groups of citrate (i.e. citrate is straddled between the heme and Lys222)
(Figure 26). An additional hydrophilic interaction occurs with the hydroxyl group of
citrate and the same propionate group of the heme. Structure refinement using REFMAC
did not improve the electron density around the citrate in the peroxidase active site as
markedly as refinement with CNS. Nonetheless, both refinement experiments clearly
suggest that citrate weakly binds to the peroxidase active site as suggested by the
observed electron density and temperature factors. The potential relevance of citrate
binding to the peroxidase active site is that Lys222 is located in the immediate vicinity of
several residues (Lys211, Gln289, and Va1291) proposed to interact and possibly stabilize
binding of PGG2, the substrate of the peroxidase reaction (Figure 26) [106, 107].

What makes the observation of citrate binding interesting is that it was not
observed in the peroxidase site of the COX-1/Comp9 structure, despite the fact that the
crystallization conditions and crystal handling protocols were identical in both cases.
Thus, are there structural differences in inhibitor binding, which may favor binding of
citrate in one case and not the other? Previous studies have shown that binding of
inhibitors to the COX site can conribute to stabilizing POX activity [108]. Several crystal
structures of different COX-l/ligand complexes (including NSAIDs and fatty acid
substrates) were superimposed onto the COX-1/Comp9 and COX-l/Comp8 structures.
All of the active site residues displayed only small, and not statistically significant,
changes in conformations; however, in the COX-1/Comp8 structure, Arg120 adopts a

significantly altered conformation in order to accommodate the binding of compound 8.

92

Unlike the more extended conformation seen in the COX-1/Comp9 structure and most
other COX-1 crystal structures, Arg120 is found in a more kinked configuration in the
COX-1/Comp8 structure (Figure 27). This structural readjustment of Arg120 results in a
slight widening of the mouth to the COX active site suggesting that the binding of
compound 8 may cause local strain, which may, in turn, be manifested in a faster

dissociation rate for compound 8. This model for why compound 8 is a weaker inhibitor

Y385
VV387

F518

L517

3530

E192

 

Figure 27 Compound 8 (orange) bound in the COX active site after the structures of
COX-1 complexes were superimposed. The green residues represent the positions
observed in the COX-1/Comp9 structure and the gray residues show their positions in the
COX-1/Comp8 structure. The only residue to show a significantly altered conformation
is R120: in the COX-1 complex R120 is extended, while it is kinked in the COX-
1/Comp8 complex. This conformational change upon binding of compound 8 may be the
result of some local strain. One consequence of this adjustment of R120 conformation is
that the mouth of the active site becomes wider, suggesting that compound 8 may have a
faster dissociation rate as compared to compound 9.

93

than compound 9 is currently being tested. The observation of citrate binding to the POX
site in the COX-1/Comp8 structure, but not in the COX-1/Comp9 structure makes the
altered conformation of Arg120 more intriguing. Past studies have shown that inhibitor
binding can protect the COX-1 enzyme from cleavage of an exposed peptide loop located
near the POX site by trypsin (Figure 26) [109, 110]. Further studies have revealed there
is differential protection from trypsin cleavage depending on the ligand binding in the
COX site suggesting that the binding of different ligands may induce different extents of
conformation changes, resulting in varying sensitivities to protease cleavage [111]. Thus,
we speculate that the strained binding of a weak COX-1 inhibitor, like compound 8,
promotes citrate binding to the peroxidase active site, while the tighter binding mode of
compound 9 disfavors citrate binding. However, the mechanism for transmitting the
subtle structural differences occurring in the POX site arising from ligand binding to the

COX active site is unknown.

Summary

The COX-1 complexes of compounds 8 and 9 provide some initial clues about the
physical parameters that define the modes of binding and chiral selectivity of these or-
substituted indomethacin ethanolamides. Several unanswered questions still remain: (a)
why (S)-a-substituted ethanolamides with quite bulky R-groups bind just as well as the S-
stereoisomers with less bulky R-groups, (b) how bulky an R-group can the COX-1 side
pocket accommodate, and (c) would the binding mode of the (S)-or-substituted
compounds shift if the R-group gets too bulky? The crystal structure of the COX-
1/Comp9 complex clearly shows there is enough space for the ethyl group, but this may

not be the case for larger moieties.

94

A recent computer modeling study [112] investigated the basis of stereo-selective
binding using several indomethacin ethanolamides possessing the larger isopropyl
substituent. Although, the binding modes obtained for the (S)-series of inhibitors were
not similar to that seen in the COX-1/Comp9 structure, they did reveal that the molecule
having the R-isopropyl group had to maintain a constrained conformation in order to
achieve the same binding interactions as the S-enantiomer. This directly supports what
was observed in the crystal structure in that the binding of compound 8 induces strain
within the protein as evidenced by altered conformation of Arg120.

However, the physical basis for the chiral selectivity of COX-1 for a wider range
of R-groups may turn out to be much more complicated than simple steric hindrance. For
example, the strong interactions between the ethanolamide group of compound 9
compensates for the lack of a strong interaction between a carboxylate group and Arg120
and Tyr355 being necessary for high affinity binding in COX-l. Crystal structures of
COX-1 complexed with or-substituted indomethacin ethanolamides having larger R-
group substitutions may provide further information about interactions that define the

different binding modes within the COX active site.

Acknowledgements

All data presented here was collected at COM-CAT (Sector 32 beamline) at Advanced Photon
Sources (Argonne, IL), and thus we would like to thank Joe Brunzelle of LS-CAT for his
assistance during our time at COM-CAT. This work was supported in part by the United States
Department of Energy, Basic Energy Sciences, Office of Energy Research, under Contract W-31-
109-Eng—38 and by grants R01 HL56773, P01 GM57323, and R01CA89450 from the National

Institute of Health.

95

CHAPTER V

FUTURE DIRECTIONS

Structural Studies of PGHS-1

Although there are many reported structures of PGHS-1 that are invaluable in
providing an understanding of the catalytic mechanism, product formation, substrate
specificity and inhibition, there are still more protein/ligand complexes of PGHS-1 that
are of interest for future studies. Several interesting mutants yet to be studied structurally
are those favoring formation of 15-HETE, another abortive product that is produced in
minor amounts by native PGHS-1 and PGHS-2. Residues within the hydrophobic pocket
of the COX site that contact the (Io-end of AA appear to be involved in proper positioning
of the w-end of AA thereby influences PGG2 formation. When these residues are
mutated product formation is shified to favor 15-HETE; therefore, it would be interesting
to study these mutant proteins structurally. With the reproductive success of obtaining
high quality crystals of native and mutant PGHS complexed with various ligands, this
process has shown to be applicable in pursing further structures of AA and other fatty

acids bound to mutants of PGHS.

The conformation of arachidonic acid bound to a mutant that make predominantly 15-

HE TE

We know that native PGHS enzymes are not 100% efficient in forming PGG2
from AA as observed by formation of minor amounts of the abortive products ll-HETE

and 15-HETE, in addition, the conformation of AA at the moment of hydrogen

96

 

abstraction will favor a particular product. With a crystal structure of AA bound to a
mutant of PGHS that forms predominantly ll-HETE (Chapter II), a different
conformation of AA was observed and increased our understanding of how ll-HETE is
formed. Mutation studies have revealed that there are residues that when mutated will
favor formation of lS-HETE. How PGHS forms 15-HETE is not well understood
structurally; however, possible mechanisms explaining lS-HETE formation have been
proposed and focus on the positioning of AA in the active site as well as how this
influences the stereochemical addition of oxygen at C15.

Oxygen addition at C15 is of much interest because the stereochemistry of oxygen
addition at the C15 position has shown to be more variable with mutants of PGHS-1 that
favor lS-HETE formation as compared to oxygen addition at C11 with mutants favoring
ll-HETE, being exclusively the R-enantiomer. Although native PGHS-l forms
exclusively ISS-HETE, various mutations have not only influenced the products formed
but also the stereochemistry in which a mixture of R/S 15-HETE is formed with the 15S-
HETE enantiomer being the predominant enantiomer. Understanding the stereochemistry
of oxygen addition a C15 in PGHS-1 is even more intriguing when considering the case
of aspirin acetylated PGHS-2, in which the 15R-HETE enantiomer is formed exclusively
[70, 113]. Thus, alterations of particular residues within the COX active of both isoforms
result in not only favoring the formation of lS-HETE, but can also change the
stereochemistry of oxygenation at this position of AA.

Although many elegant experiments have led to a basic understanding of the
residues involved in influencing the shift from PGG2 to lS-HETE in both PGHS-1 [53]

and PGHS-2 [76, 113, 114], the structural basis of how oxygenation first occurs at C15

97

and how these residues influence the stereochemistry is not clear in either PGHS-1 or
PGHS-2. There are two single mutations in PGHS-1 that have been shown to shift the
product profile to favor formation of 15-HETE from AA having both the R/S
stereochemistry. These mutations include L534A, favoring formation of ~50% 15S-
HETE and V349L, favoring formation of ~30% 15-HETE with 70% S to 30% R [53].
Potentially, a V349l/L534A mutant of PGHS-1 may form predominantly lS-HETE, and
therefore, could be used for structural studies. A crystal structure of AA bound to this
mutant could demonstrate the adopted conformation AA that leads to 15-HETE, in
addition to gaining some structural information about the influence of residues on the

stereochemistry of oxygen addition at C15 of AA.

Arachidonic acid binding to G533A PGHS-1

The residue G533 lies at the distal end of the cycloxygenase channel where the
C0 carbon abuts up against the last two carbons of the 0) end of AA [42]. Mutation of
G533 to alanine in ovine PGHS-1 and human PGHS-2 result in loss of COX activity;
however, POX activity is retained [53]. Interestingly, this same mutant in murine PGHS-
2 still has both COX and POX activity [115]. Residue G533 having a drastic effect on
catalysis when mutated is far removed from the site of catalysis, thus it is thought from
structural analysis that this residue influences the positioning of the w-end of AA
allowing for optimal placement of C13 for abstraction. However, it is unclear if catalysis
is lost as result of a misalignment of C13 upon binding of AA or if the mutation at G533
simply prevents binding of AA. A crystal structure of AA bound to oPGHS-l G533A
mutant would be of major interest for two reasons. This structure would not only

demonstrate that AA is binding to the G533A mutant, but could also show the binding

98

conformation of AA and in turn show the significance in positioning of the m-end and

how this would effect catalysis.

Indomethacin Ethanolamides/PGHS—I crystal structures

Chapter IV revealed the structural basis of how the stereochemistry of the
indomethacin ethanolamides can affect how these novel inhibitors bind and ultimately
effect the degree of inhibition of PGHS-1. Although some knowledge was gained from
crystal structures of an enantiomeric pair of these inhibitors, questions still remain as to
the physical basis for chiral selectivity of COX-1 with indomethacin ethanolamides with
a wider range of R-group substituents. The side pocket of COX-1 can accommodate an
ethyl substitution (Table 4, Compound 9), but possibly not larger R-groups (Table 4,
Compounds 14-17), thus is there may be a shift in binding modes that would explain how
these inhibitors with bulkier R-groups would bind. The results obtained from Chapter IV
do provide testable conclusions and hypotheses. Mutating residues that have been shown
to interact with the inhibitors, particularly Phe518, Leu517, and Glnl92 could highlight
their roles in stabilizing the binding modes of compound 9 via hydrogen bonding and
hydrophobic interactions. In addition, more detailed binding experiments comparing the
association/dissociation rates of compound 8 and 9 could clarify chiral selectivity based
on the binding dynamics of these inhibitors. Lastly, as mentioned in Chapter IV, a crystal
structure of an or-substituted indomethacin ethanolamide having a larger R-group
substitution bound to COX-1 could demonstrate a shift in binding modes when the R-
group gets too bulky for the side pocket. A good candidate R-group substitution for a

structure complex would be a phenyl group substitution (Table 4, Compound 17). This

99

group would be a good representative because of its larger size and limited flexibility,

thus electron density could be easily recognized within the active site.

Structural Studies of PGHS-2

There are have several structures of PGHS-1 bound to various fatty acids that
have been significant in elucidating the structural basis of binding and oxygenation of
these various fatty acids; however, there is very little known structurally about how these
fatty acids bind to PGHS-2. There have been at least two published crystal structures of
AA bound to PGHS-2; however, these structures have provided no clear structural
understanding of fatty acid binding because the AA was bound in an unproductive
conformation [116, 117]. The reason as to why AA was captured in this conformation is
interesting and may predict a future problem in obtaining fatty acid complexes with
PGHS-2, nonetheless, further structures are needed to understand the differences in

substrate specificity of the two isoforms.

Crystallization of human PGHS-2 complexed with fatty acids

Crystallization of human PGHS-l has 5 been a very difficult task; however,
several recent approaches have been successful in obtaining diffraction quality crystals of
human PGHS-2. Several constructs of PGHS-2 have been made (Dr. Chong Yuan in the
lab of Dr. Smith) in order to I) obtain highly purified protein and 2) increase ability of
protein to crystallize. The range of constructs made so far include an rTEV cleavable N-
terrninus 8X-histidine tag, and an rTEV cleavable 8X-histidine tag located just before 18-
amino acid (aa) cassette at C-terminus of PGHS-2. Another construct of interest is a

native PGHS-2 with a histidine-tag inserted just before the 18-aa cassette thus allowing

100

simultaneously cleavage of the histidine-tag and the C-terminus using the endogenous
trypsin cleavage site located near to the 18-aa cassette region of PGHS-2. The benefit of
these constructs is to use the histidine-tag for purification and then cleave off the histidine
tag to reduce possible problems in crystallization arising from the presence of the
histidine tag. In addition, the 18-aa cassette of PGHS-2, which is not crucial to catalytic
activity of PGHS, has never been observed crystallographically possibly because it is
very flexible and therefore may be a problem in crystallization. Thus, removal of the 18-
aa insert may improve the chances of obtaining crystals.

Recently, crystals have been obtained of a PGHS-2 construct having both the his-
tag and 18-aa cassette removed (Figure 28) from crystallization conditions that had
previously been know to crystallize PGHS-2 [118]. The crystals, being extremely fragile
diffracted to approximately 5-6A and currently a preliminary space group is being
determined. Future work will include reproducing these crystals and optimizing
conditions to reduce the time to grow and possibly increase the stability of the crystals so

that multiple substrate and inhibitor complexes can be obtained consistently.

Figure 28 Crystal of human PGHS-2 grown in the presence of AA.

101

 

Structures of human PGHS-2 with various fatty acid

Most of the structural information of fatty acid binding to PGHS was obtained
with PGHS-l mostly because of the robustness of PGHS-1 crystallization. Thus, once
reproducible crystallization conditions are established for PGHS-2 mirror structures of
various substrate complexes with PGHS-2 can be obtained for comparison.

Interesting differences in substrate specificity between the two isoforms have yet
to be understood on a structural level. As mentioned previously, only one crystal
structure of PGHS-2, the murine form, with the substrate bound has been reported. This
structure captures AA bound in an unproductive conformation in which the carboxylate
group was oriented up into the hydrophobic channel. The observation that AA was
captured in a “backwards” configuration within the COX active site of PGHS-2 might
indicate the difficulty in capturing a catalytically competent conformation of the
substrate. PGHS-2 is known to be the more promiscuous of the isoforms in utilizing
various fatty acids. PGHS-2 was able to selectively oxygenate 2-arachidonyl glycerol, an
endocannabinoid, with similar efficiency as AA while PGHS-l was not. The fatty acid,
2-arachidonyl glycerol, in comparison to AA, has a glycerol group in place of
carboxylate group (C1 attached via the 2 position of glycerol). This increases the size of
this fatty acid, thus it is suggested that PGHS-2 can utilize this fatty acid more efficiently
than PGHS-1 because it can bind within the larger PGHS-2 active site. Interestingly, this
substrate may prove to be useful. The extra bulk of the glycerol group, may force a
binding orientation most likely a productive conformation within the PGHS-2 active site.
Therefore, a productive conformation of a fatty acid may be obtained and at the same

time show how this interesting fatty acid binds within the PGHS-2 active site.

102

Crystal structures of PGHS-2 with indomethacin ethanolamides

With preliminary structural results of how this compounds bind to PGHS-1,
comparative structures with PGHS-2 would be very interesting. The crystal structures of
indomethacin ethanolamides bound to PGHS-1 described in Chapter IV reveal two
binding modes as a result of a separate set of binding interactions, which apparently
stabilize the inhibitors. Comparing the binding of these inhibitors, which inhibit PGHS-2
with similar and better efficiency could emphasize and validate similarities of binding
between PGHS-2 and PGHS-1. Specifically, it could provide a clearer understanding of
the binding mode of Compound 9 in PGHS-1 as well as understanding the binding mode

of Compound 8 in PGHS-2 and how it would compare to PGHS-l.

103

REFERENCES

1. Smith, W. L., Garavito, R. M. & DeWitt, D. L. (1996) Prostaglandin endoperoxide H
synthases (cyclooxygenases)-1 and -2, J Biol Chem. 271, 33157-33160.

2. Marnett, L. J., Rowlinson, S. W., Goodwin, D. C., Kalgutkar, A. S. & Lanzo, C. A.
(1999) Arachidonic acid oxygenation by COX-1 and COX-2 - Mechanisms of catalysis
and inhibition, J Biol Chem. 274, 22903-22906.

3. Smith, W. L., DeWitt, D. L. & Garavito, R. M. (2000) Cyclooxygenases: Structural,
Cellular, and Molecular Biology, Annu Rev Biochem. 69, 149-182.

4. Smith, W. L. & Murphy, R. C. (2002) The eicosanoids: cyclooxygenase,
lipooxygenase, and epoxgenase pathways in Biochemistry of Lipids, Lipoproteins, and
Membranes (Vance, D. E. & Vance, J. E., eds), Elsevier Science B.V., Amsterdam.

5. Oates, J. A., Fitzgerald, G. A., Branch, R. A., Jackson, E. K., Knapp, H. R. & Roberts,
L. J. ( 1988) Clinical Implications of Prostaglandin and Thromboxane A-2 Formation 2,
New Engl J Med 319, 761-767.

6. Oates, J. A., Fitzgerald, G. A., Branch, R. A., Jackson, E. K., Knapp, H. R. & Roberts,
L. J. (1988) Clinical Implications of Prostaglandin and Thromboxane-A2 Formation 1,
New Engl J Med 319, 689-698.

7. Smith, W. L. & Marnett, L. J. (1991) Prostaglandin Endoperoxide Synthase -
Structure and Catalysis, Biochim Biophys Acta. 1083, l-l 7.

8. Katori, M. & Majima, M. (2000) Cyclooxygenase-2: its rich diversity of roles and
possible application of its selective inhibitors, Inflamm Res. 49, 367-392.

9. Kargman, S. L. & al., e. (1995) Expression of Prostaglandin G/H synthase-l and -2
protein in human colon cancer, Cancer Res. 55, 2556-2559.

10. Kino, Y., Kojima, F., Kiguchi, K., Igarashi, R., Ishizuka, B. & Kawai, S. (2005)
Prostaglandin E2 production in ovarian cancer cell lines is regulated by cyclooxygenase-
] not cyclooxygenase-2, Prostaglandins, Leukot Essent Fatty Acids. 73, 103-111.

104

11. Fletcher, B. S., Kujubu, D. A., Perrin, D. M. & Herschman, H. R. (1992) Structure of
the mitogen-inducible TIS10 gene and demonstration that the TISlO-encoded protein is a
functional prostaglandin G/H synthase, J Biol Chem. 267, 4338-4344.

12. Dietz, R., Nastainczyk, W. & Ruf, H. (1988) Higher oxidation states of prostaglandin
H synthase. Rapid electronic spectroscopy detected two spectral intermediates during the
peroxidase reaction with prostaglandin G2, Eur J Biochem. 1 71, 321-328.

13. Tsai, A. L., Wu, G., Palmer, G., Bambai, B., Koehn, J. A., Marshall, P. J. &
Kulmacz, R. J. (1999) Rapid kinetics of tyrosyl radical formation and heme redox state
changes in prostaglandin H synthase-1 and-2, J Biol Chem. 274, 21695-21700.

14. Kujubu, D. A., Fletcher, B. S., Varnum, B. C., Lim, R. W. & Herschman, H. R.
(1991) TIS10, a Phorbol Ester Tumor Promoter-inducible mRNA from Swiss 3T3 Cells,
Encodes a Novel Prostaglandin Synthase/Cyclooxygenase Homologue, J Biol Chem. 266,
12866-12872.

15. Murakami, M., Kambe, T., Shimbara, S. & Kudo, I. (1999) Functional coupling
between various phospholipase A(2)s and cyclooxygenases in immediate and delayed
prostanoid biosynthetic pathways, J Biol Chem. 274, 3103-3115.

16. Fitzpatrick, F. A. & Soberrnan, R. (2001) Regulated formation of eicosanoids, J Clin
Invest. 107, 1347-1351.

17. Smith, W. L. & Langenbach, R. (2001) Why there are two cyclooxygenase isozymes,
JClin Invest. 107, 1491-1495.

18. Schuster, V. L. (1998) Molecular mechanisms of prostaglandin transport, Annu Rev
Physiol. 60, 221-242.

19. Nishio, T. & al., e. (2000) Molecular identification of a rat novel organic anion
transporter moatl, which transports prostaglandin D(2), leukotriene C(4), and
taurocholate, Biochem Biophys Res Commun. 275, 831-838.

20. Chan, B. S., Bao, Y. & Schuster, V. L. (2002) Role of conserved transmembrane
cationic amino acids in the prostaglandin transporter PGT, Biochemistry. 41, 9215-9221.

105

21. Boie, Y. & al., e. (1997) Molecular cloning and characterization of the four rat
prostaglandin E2 prostanoid receptor subtypes, Eur J Biochem. 340, 227-41.

22. Wise, H. (1999) Characterization of chimeric prostacyclin/prostaglandin D(2)
receptors, Eur J Biochem. 386, 89-96.

23. Sarrazin, P. & al., e. (2001) Characterization of the prostaglandin receptors in human

osetoblasts in culture, Prostaglandins Leukot Essent Fatty Acids. 64, 203-210.

24. Negishi, M., Sugimoto, Y. & Ichikawa, A. (1995) Molecular mechanism of diverse
actions of prostanoid receptors, Biochim Biophys Acta. 1259, 109-119.

25. Hamberg, M. & Samuelsson, B. (1967) On the mechanism of the biosynthesis of
prostaglandins E-l and F-l-alpha, J Biol Chem. 242, 5336-5343.

26. Hamberg, M. & Samuelsson, B. (1967) Oxygenation of unsaturated fatty acids by the
vesicular gland of sheep, J Biol Chem. 242, 5344-5354.

27. Shimokawa, T., Kulmacz, R. J., DeWitt, D. L. & Smith, W. L. (1990) Tyrosine 385
of prostaglandin endoperoxide synthase is required for cyclooxygenase catalysis, J Biol
Chem. 265, 20073-20076.

28. Tsai, A. L., Hsi, L. C., Kulmacz, R. J., Palmer, G. & Smith, W. L. (1994)
Characterization of the tyrosyl radicals in ovine prostaglandin H synthase-1 by isotope
replacement and site-directed mutagenesis, J Biol Chem. 269, 5085-5091.

29. Tsai, A. L., Kulmacz, R. J. & Palmer, G. (1995) Spectroscopic Evidence for
Reaction of Prostaglandin H Synthase-1 Tyrosyl Radical with Arachdionic Acid, J Biol
Chem. 270, 10503-10508.

30. Wu, G., Wei, C. H., Kulmacz, R. J., Osawa, Y. & Tsai, A. L. (1999) A mechanistic
study of self-inactivation of the peroxidase activity in prostaglandin H synthase-1, J Biol
Chem. 274, 9231-9237.

31. Smith, W. L. & Lands, W. E. M. (1972) Oxygenation of Polyunsaturated Fatty-Acids
during Prostaglandin Biosynthesis by Sheep Vesicular Gland, Biochemistry. 11, 3276-&.

106

 

32. Hemler, M. E. & Lands, W. E. M. (1980) Evidence for a Peroxide-Initiated Free-
Radical Mechanism of Prostaglandin Biosynthesis, J Biol Chem. 255, 6253-6261.

33. Markey, C. M., Alward, A., Weller, P. E. & Marnett, L. J. (1987) Quantitative
Studies of Hydroperoxide Reduction by Prostaglandin-H Synthase - Reducing Substrate-
Specificity and the Relationship of Peroxidase to Cyclooxygenase Activities, J Biol
Chem. 262, 6266-6279.

34. Marshall, P. J. & Kulmacz, R. J. (1988) Prostaglandin H synthase: distinct binding
sites for cyclooxygenase and peroxidase substrates, Arch Biochem Biophys. 266, 162-
170.

35. Song, 1., Ball, T. M. & Smith, W. L. (2001) Different Suicide Inactivation Processes
for the Peroxidase and Cyclooxygenase Activities of Prostaglandin Endoperoxide H
Synthase-1, Biochem Biophys Res Commun. 289, 869-875.

36. Garavito, R. M., Malkowski, M. G. & DeWitt, D. L. (2002) The structures of
prostaglandin endoperoxide H synthases-1 and -2, Prostaglandins Other Lipid Mediat.
68-69, 129-152.

37. Otto, J. C., DeWitt, D. L. & Smith, W. L. (1993) N-glycosylation of prostaglandin
endoperoxide synthases-l and -2 and their orientations in the endoplasmic reticulum, J
Biol Chem. 268, 18234-18242.

38. Barnett, J., Chow, J., Ives, D., Chiou, M., Mackenzie, R., Osen, E. & al., e. (1994)
Purification, characterization and selective inhibition of human prostaglandin G/H

synthase 1 and and expressed in the bacuolvirus system, Biochim Biophys Acta. 1209,
130-139.

39. Picot, D., Loll, P. J. & Garavito, R. M. (1994) The X-Ray Crystal-Structure of the
Membrane-Protein Prostaglandin-HQ) Synthase-1, Nature. 367, 243-249.

40. Luong, C., Miller, A., Barnett, J., Chow, J., Rarnesha, C. & Browner, M. F. (1996)
Flexibility of the NSAID binding site in the structure of human cyclooxygenase-2,
Nature Struct Biol. 3, 927-933.

107

41. Kurumbail, R. G., Stevens, A. M., Gierse, J. K., McDonald, J. J., Stegeman, R. A.,
Pak, J. Y., Gildehaus, D., Miyashiro, J. M., Penning, T. D., Seibert, K., lsakson, P. C. &
Stallings, W. C. (1996) Structural basis for selective inhibition of cyclooxygenase-2 by
anti-inflammatory agents, Nature. 384, 644-648.

42. Malkowski, M. G., Ginell, S. L., Smith, W. L. & Garavito, R. M. (2000) The
productive conformation of arachidonic acid bound to prostaglandin synthase, Science.
289, 1933-1937.

43. Spencer, A. G., Woods, J. W., Arakawa, T., Singer, 1. I. & Smith, W. L. (1998)
Subcellular localization of prostaglandin endoperoxide H synthases-1 and -2 by
immunoelectron microscopy, J Biol Chem. 273, 9886-9893.

44. McGeer, P. L. & McGeer, E. G. (1999) Inflammation of the brain in Alzheimer's
diease: implications for therapy, J Leuko Biol. 65, 409-415.

45. Patrignani, P. & al., e. (1994) Biochemical and pharmacological characterization of '
the cylcooxygenase activity of human blood prostaglandin endoperoxide synthases, J
Pharmacol Exp. Ther. 271, 1705-1712.

46. Patrono, C. (1994) Aspirin as an antiplatelet drug, New Engl J Med. 330, 1287-1294.

47. Shitashige, M., Morita, I. & Murota, S. (1998) Different substrate utilization between
prostaglandin endoperoxide H synthase-1 and -2 in NIH3T3 fibroblasts, Biochim Biophys
Acta. 1389, 57-66.

48. Chen, W., Pawelek, T. R & Kulmacz, R. J. (1999) Hydroperoxide dependence and
cooperative cyclooxygenase kinetics in prostaglandin H synthase-1 and-2, J Biol Chem.
274, 20301-20306.

49. Malkowski, M. G., Thuresson, E. D., Lakkides, K. M., Rieke, C. J., Micielli, R.,
Smith, W. L. & Garavito, R. M. (2001) Structure of Eicosapentaenoic and Linoleic Acids

in the Cyclooxygenase Site of Prostaglandin Endoperoxide H Synthase-1, J Biol Chem.
276, 37547-37555.

50. Thuresson, E. D., Malkowski, M. G., Lakkides, K. M., Rieke, C. J., Mulichak, A. M.,
Ginell, S. L., Garavito, R. M. & Smith, W. L. (2001) Mutational and X-ray
crystallographic analysis of the interaction of dihomo-gamma -linolenic acid with
prostaglandin endoperoxide H synthases, J Biol Chem. 276, 10358-10365.

108

51. Gund, P. & Shen, T. Y. (1977) Model for Prostaglandin Synthetase Cyclo-
Oxygenation Site and Its Inhibition by Antiinflammatory Arylacetic Acids, J Med Chem.
20, 1146-1152.

52. Thuresson, E. D., Lakkides, K. M. & Smith, W. L. (2000) Different catalytically
competent arrangements of arachidonic acid within the cyclooxygenase active site of
prostaglandin endoperoxide H synthase-1 lead to the formation of different oxygenated

products, J Biol Chem. 275, 8501-8507.

53. Thuresson, E. D., Lakkides, K. M., Rieke, C. J., Sun, Y., Wingerd, B. A., Micielli,
R., Mulichak, A. M., Malkowski, M. G., Garavito, R. M. & Smith, W. L. (2001)
Prostaglandin endoperoxide H synthase-l: the functions of cyclooxygenase active site

residues in the binding, positioning, and oxygenation of arachidonic acid, J Biol Chem.
276, 10347-10357.

54. Willis, A. L. (1987) in CRC Handbook of Eicosanoids: Prostaglandins and Related
Lipids (Willis, A. L., ed) pp. 3-46, CRC Press, Inc., Boca Raton, FL.

55. Lands, W. E. M. & Samuelsson, B. (1968) Phospholipid Precursors of
Prostaglandins, Biochim Biophys A cta. 164, 426-&.

56. Laneuville, O., Breuer, D. K., Xu, N. X., Huang, Z. H., Gage, D. A., Watson, J. T.,
Lagarde, M., Dewitt, D. L. & Smith, W. L. (1995) F atty-Acid Substrate Specificities of
Human Prostaglandin Endoperoxide-H Synthase-1 and Synthase-2 - Formation of 12-
Hydroxy-(9Z,13E/Z,15Z)-Octadecatrienoic Acids from Alpha-Linolenic Acid, J Biol
Chem. 270, 19330-19336.

57. Larsen, L. N., Dahl, E. & Bremer, J. (1996) Peroxidative oxidation of leuco-
dichlorofluorescein by prostaglandin H synthase in prostaglandin biosynthesis from
polyunsaturated fatty acids, Biochim Biophys Acta-Lipids Lipid Meta. 1299, 47-53.

58. Oliw, E. H., Homsten, L., Sprecher, H. & Hamberg, M. (1993) Oxygenation of

5,8,1 l-Eicosatrienoic Acid by Prostaglandin Endoperoxide Synthase and by Cytochrome-
P450 Monooxygenase - Structure and Mechanism of Formation of Major Metabolites,
Arch Biochem Biophys. 305, 288-297.

59. Needleman, P., Whitaker, M. O., Wyche, A., Watters, K., Sprecher, H. & Raz, A.
(1980) Manipulation of Platelet-Aggregation by Prostaglandins and Their F atty-Acid

Precursors - Pharmacological Basis for a Therapeutic Approach, Prostaglandins. I 9, 165-
181.

109

60. Knapp, H. R. (1990) Prostaglandins in Human Semen during Fish Oil Ingestion -
Evidence for Invivo Cyclooxygenase Inhibition and Appearance of Novel Trienoic
Compounds, Prostaglandins. 39, 407-423.

61. Kaduce, T. L., Figard, P. H., Leifur, R & Spector, A. A. (1989) Formation of 9-
Hydroxyoctadecadienoic Acid from Linoleic-Acid in Endothelial-Cells, J Biol Chem.
264, 6823-6830.

62. Lands, W. E. M., LeTellier, P. R., Rome, L. H. & Vanderhoek, J. Y. (1973)
Inhibition of prostaglandin biosynthesis, Adv Biosci. 9, 15-28.

63. Kozak, K. R., Rowlinson, S. W. & Marnett, L. J. (2000) Oxygenation of the
endocannabinoid, 2-arachidonylglycerol, to glyceryl prostaglandins by cyclooxygenase-
2, J Biol Chem. 275, 33744-33749.

64. Kozak, K. R., Crews, B. C., Ray, J. L., Tai, H., Morrow, J. D. & Marnett, L. J.
(2001) Metabolism of Prostaglandin Gycerol Esters and Prostaglandin Ethanolarnides in
Vitro and in Vivo, J Biol Chem. 276, 36993-36998.

65. Kozak, K. R., Crews, B. C., Morrow, J. D., Wang, L. H., Ma, Y. H., Weinander, R.,
Jakobsson, P. J. & Marnett, L. J. (2002) Metabolism of the Endocannabinoids, 2-
Arachidonlyglycerol and Anandamide, into Prostaglandin, Thromboxane, and
Prostacyclin Glycerol Esters and Ethanolamides, J Biol Chem. 277, 44877-44885.

66. Rome, L. H. & Lands, W. E. M. (1975) Structural requirements for time-dependent

inhibition of prostaglandin biosynthesis by anti-inflammatory drugs, Proc Natl Acad Sci
USA. 72, 4863-4865.

67. Bhattacharyya, D. K., Lecomte, M., Rieke, C. J., Garavito, R. M. & Smith, W. L.
(1996) Involvement of arginine 120, glutamate 524, and tyrosine 355 in the binding of
arachidonate and 2-phenylpropionic acid inhibitors to the cyclooxygenase active site of
ovine prostaglandin endoperoxide H synthase-1, J. Biol. Chem. 271, 2179-2184.

68. Rieke, C. J ., Mulichak, A. M., Garavito, R. M. & Smith, W. L. (1999) The role of
arginine 120 of human prostaglandin endoperoxide H synthase-2 in the interaction with
fatty acid substrates and inhibitors, J Biol Chem. 274, 17109-17114.

110

69. O'Neill, G. P., Mancini, J. A., Kargman, S., Yergey, J., Kwan, M. Y., Falgueyret, J.-
P., Abramovitz, M., Kennedy, B. P., Ouellet, M., Cromlish, W., Culp, S., Evans, J. F.,
Ford-Hutchinson, A. W. & Vickers, P. J. (1993) Overexpression of Human Prostaglandin
G/H Synthase-1 and -2 by Recombinant Vaccinia Virus: Inhibition by Nonsteroidal
Anit-Inflammatory Drugs and Biosythesis of 15-Hydroxyeicosatraenoic acid, Mol
Pharm. 45, 245-254.

70. Lecomte, M., Laneuville, O., Ji, C., DeWitt, D. L. & Smith, W. L. (1994)
Acetylation of human prostaglandin endoperoxide synthase-2 (cyclooxygenase-2) by
aspirin, J Biol Chem. 269, 13207-13215.

71. Dewitt, D. L. (1999) Cox-2-selective inhibitors: The new super aspirins, Mol Pharm.
55, 625-631.

72. Gierse, J. K., McDonald, J. J., Hauser, S. D., Rangwala, S. H., Koboldt, C. M. &
Seibert, K. (1996) A single amino acid difference between cyclooxygenase-1 (COX-1)
and -2 (COX-2) reverses the selectivitiy of COX-2 specific inhibitors, J Biol Chem. 271,
15810-15814.

73. Marnett, L. J. & DuBois, R. N. (2002) COX-2: a target for colon cancer prevention,
Annu Rev Pharmacol T oxicol. 42, 55-80.

74. van der Donk, W. A., Tsai, A.-L. & Kulmacz, R. J. (2002) The cyclooxygenase
reaction mechanism, Biochemistry. 41 , 15451-15458.

75. Garavito, R. M. & Mulichak, A. M. (2003) The structure of mammalian
cyclooxygenases, Ann Rev Biophys Biomol Struct. 32, 183-206.

76. Schneider, C., Boeglin, W. E., Prusakiewicz, J. J., Rowlinson, S. W., Marnett, L. J.,
Samel, N. & Brash, A. R. (2002) Control of prostaglandin stereochemistry at the 15-

carbon by cyclooxygenases-l and -2. A critical role for serine 530 and valine 349, J Biol
Chem. 277, 13354-13360.

77. Spencer, A. G., Thuresson, E. A., Otto, J. C., Song, 1., Smith, T., DeWitt, D. L.,
Garavito, R. M. & Smith, W. L. (1999) The membrane binding domains of prostaglandin

endoperoxide H synthase-1 and -2: Peptide mapping and mutational analysis, J Biol
Chem. 274, 32936-32942.

lll

78. Malkowski, M. G., Theisen, M. J., Scharmen, A. & Garavito, R. M. (2000) The
formation of stable fatty acid substrate complexes in prostaglandin H-2 synthase-1, Arch
Biochem Biophys. 380, 39-45.

79. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffraction data collected in
oscillation mode, Macromol Crystallogr, Pt A. 276, 307-326.

80. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse
Kunstleve, R. W., Jiang, J. S., Kuszewski, J., Nilges, M., Pannu, N. S., Read, R. J., Rice,
L. M., Simonson, T. & Warren, G. L. (1998) Crystallography & NMR system: A new
software suite for macromolecular structure determination, Acta Crystallogr D. 54, 905-
921.

81. Sack, J. S. (1988) Chain — A Crystallographic Modeling Program, J Mol Graphics. 6,
224-225.

82. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993)
PROCHECK: a program to check the stereochemical quality of protein structures, J A pp]
Cryst. 26, 283-291.

83. Evans, S. V. (1993) SETOR: hardware-lighted three-dimensional solid model
representations of macromolecules, J Mol Graphics. I I , 134-138.

84. Mead, J. F. & Slaton, W. H. (1956) Metabolism of Essential Fatty Acids.3. Isolation
of 5,8,11-Eicosatrienoic Acid from F at-Deficient Rats, J Biol Chem. 219, 705-709.

85. Lefltowith, J. B., Flippo, V., Sprecher, H. & Needleman, P. (1985) Paradoxical
Conservation of Cardiac and Renal Arachidonate Content in Essential Fatty-Acid
Deficiency, J Biol Chem. 260, 5736-5744.

86. Cox, J. W., Rutecki, G. W., Francisco, L. L. & Ferris, T. F. (1982) Studies of the
Effects of Essential F atty-Acid Deficiency in the Rat, Cir Res. 51 , 694-702.

87. Bonta, I. L., Chrispij, H., Noordhoe, J. & Vincent, J. E. (1974) Reduction of
Prostaglandin-Phase in Hind-Paw Inflammation and Partial Failure of Indomethacin to
Exert Antiinflammatory Effect in Rats on Essential F atty-Acid Deficient Diet,
Prostaglandins. 5, 495-503.

112

88. Burr, G. O. & Burr, M. M. (1930) On the nature and role of the fatty acids essential
in nutrition, J Biol Chem. 86, 587-621.

89. Elliott, W. J., Morrison, A. R., Sprecher, H. & Needleman, P. (1986) Calcium-
Dependent Oxidation of 5,8,11-Icosatrienoic Acid by the Cyclooxygenase Enzyme-
System, J Biol Chem. 261, 6719-6724.

90. Oliw, E. H., Homsten, L. & Sprecher, H. (1997) Oxygenation of 5,8,1 1-
eicosatrienoic acid by prostaglandin H synthase-2 of ovine placental cotyledons: Isolation
of 13-hydroxy-5,8,1l-eicosatrienoic and l1-hydroxy-5,8,12-eicosatrienoic acids, J
Chromatogr B. 690, 332-337.

91. Adkisson, H. D., Risener, F. S., Zarrinkar, P. P., Walla, M. D., Christie, W. W. &
Wuthier, R. E. (1991) Unique Fatty-Acid Composition of Normal Cartilage - Discovery
of High-Levels of N-9 Eicosatrienoic Acid and Low-Levels of N—6 Polyunsaturated

Fatty-Acids, FASEB J. 5, 344-353.

92. Nugteren, D. H. & Christhazelhof, E. (1987) Naturally-Occurring Conjugated
Octadecatrienoic Acids Are Strong Inhibitors of Prostaglandin Biosynthesis,
Prostaglandins. 33, 403-417.

93. Harnmarstrom, S. (1981) Conversion of 5,8,1 1-Eicosatrienoic Acid to Leukotriene-
C3 and Leukotriene-D3, J Biol Chem. 256, 2275-2279.

94. Lagarde, M., Burtin, M., Rigaud, M., Sprecher, H., Dechavanne, M. & Renaud, S.
(1985) Prostaglandin-E2-Like Activity of 20-3N-9 Platelet Lipoxygenase End-Product,
FEBS Lett. 181, 53-56.

95. Porter, N. A., Caldwell, S. E. & Mills, K. A. (1995) Mechanisms of F rec-Radical
Oxidation of Unsaturated Lipids, Lipids. 30, 277-290.

96. Kapust, R. B., Tozser, J., Fox, D., Anderson, D. E., Cherry, S., Copeland, T. D. &
Waugh, D. S. (2001) Tobacco etch virus protease: mechanism of autolysis and rational
design of stable mutants with wild-type catalytic proficiency, Protein Eng. 14, 993-1000.

97. Kozak, K. R., Prusakiewicz, J. J., Rowlinson, S. W. & Marnett, L. J. (2002)
Enantiospecific, selective cyclooxygenase-2 inhibitors, Bioorg Med Chem Lett. 12, 1315-
1318.

113

98. Garavito, R. M. & DeWitt, D. L. (1999) The cyclooxygenase isoforms: structural
insights into the conversion of arachidonic acid to prostaglandins, Biochim Biophys Acta.
1441, 278-287.

99. Walker, M. C., Kurumbail, R. G., Kiefer, J. R, Moreland, K. T., Koboldt, C. M.,
lsakson, P. C., Seibert, K. & Gierse, J. K. (2001) A three-step kinetic mechanism for

selective inhibition of cyclo-oxygenase-2 by diarylheterocyclic inhibitors, Biochem J.
35 7, 709-718.

100. Gierse, J. K., Koboldt, C. M., Walker, M. C., Seibert, K. & lsakson, P. C. (1999)
Kinetic basis for selective inhibition of cyclo-oxygenases, Biochem J. 339, 607-614.

101. L011, P. J., Picot, D., Ekabo, O. & Garavito, R. M. (1996) Synthesis and use of
iodinated nonsteroidal antiinflammatory drug analogs as crystallographic probes of the
prostaglandin H-2 synthase cyclooxygenase active site, Biochemistry. 35, 7330-7340.

102. Kalgutkar, A. S., Crews, B. C., Rowlinson, S. W., Marnett, A. B., Kozak, K. R.,
Remmel, R. P. & Marnett, L. J. (2000) Biochemically based design of cyclooxygenase-2

(COX-2) inhibitors: Facile conversion of nonsteroidal antiinflammatory drugs to potent
and highly selective COX-2 inhibitors, Proc Natl Acad Sci USA. 97, 925-930.

103. 4, C. C. P. N. (1994) The CCP4 Suite: Programs for Protein Crystallography, Acta
Crystallogr D. 50, 760-763.

104. Winn, M. D., Isupov, M. N. & Murshudov, G. N. (2001) Use of TLS parameters to
make ansiotropic displacments in macromolecular refinement, A cta Crystallogr. D. 5 7,
122-133.

105. Winn, M. D., Murshudov, G. N. & Papiz, M. Z. (2003) Macromolecular TLS
Refinement in REFMAC at Moderate Resolutions, Methods in Enzymol. 3 74, 300-321.

106. Garavito, R. M. (2001) Prostaglandin endoperoxide H synthase-1 and -2 in
Handbook of Metalloproteins (K. Wieghardt, R. H., T. L. Poulos, and A. Messerschmidt,
ed) pp. 245-63, John Wiley, Sussex.

107. Seibold, S. A., Smith, W. L. & Cukier, R. I. (2004) Peroxidase site of prostaglandin
endoperoxide H synthase-1: Docking and molecular dynamics studies with a
prostaglandin endoperoxide analog, J Phys Chem B. 108, 9297-9305.

114

108. Seibold, S. A., Ball, T., Hsi, L. C., Mills, D. A., Abeysinghe, R. D., Micielli, R.,
Rieke, C. J., Cukier, R. I. & Smith, W. L. (2003) I-Iistidine 386 and its role in
cyclooxygenase and peroxidase catalysis by prostaglandin-endoperoxide H synthases, J
Biol Chem. 278, 46163-46170.

109. Kulmacz, R. J. (1989) Topography of Prostaglandin-H Synthase -
Antiinflammatory Agents and the Protease-Sensitive Arginine-253 Region, J Biol Chem.
264, 14136-14144.

110. Chen, Y. N., Bienkowski, M. J. & Marnett, L. J. (1987) Controlled Tryptic
Digestion of Prostaglandin H-Synthase - Characterization of Protein-Fragments and
Enhanced Rate of Proteolysis of Oxidatively Inactivated Enzyme, J Biol Chem. 262,
16892-16899.

111. Kalgutkar, A. S., Crews, B. C. & Marnett, L. J. (1996) Kinetics of the interaction of
nonsteroidal antiinflammatory drugs with prostaglandin endoperoxide synthase-l studied
by limited proteolysis, Biochemistry. 35, 9076-9082.

112. Moth, C. W., Prusakiewicz, J. J., Marnett, L. J. & Lybrand, T. P. (2005)
Stereoselective Binding of Indomethacin Ethanolamides Derivatives to Cyclooxygenase-
l, J Med Chem. 48, 3613-3620.

113. Xiao, G., Tsai, A. L., Palmer, G., Boyar, W. C., Marshall, P. J. & Kulmacz, R J.
(1997) Analysis of Hydroperoxide-Induced Tyrosyl Radicals and Lipoxygenase Activity
in Aspirin-Treated Human Prostaglandin H Synthase-2, Biochemistry. 36, 1836-1846.

114. Mancini, J. A., Oneill, G. P., Bayly, C. & Vickers, P. J. (1994) Mutation of Serine-
516 in Human Prostaglandin G/H Synthase-2 to Methionine or Aspirin Acetylation of
- This Residue Stimulates 15-R-HETE Synthesis, FEBS Lett. 342, 33-37.

115. Rowlinson, S. W., Crews, B. C., Lanzo, C. A. & Marnett, L. J. (1999) The binding
of arachidonic acid in the cyclooxygenase active site of mouse prostaglandin
endoperoxide synthase-2 (COX-2) - A putative L-shaped binding conformation utilizing
the t0p channel region, J Biol Chem. 2 74, 23305-23310.

115

116. Kiefer, J. R, Pawlitz, J. L., Moreland, K. T., Stegeman, R. A., Hood, W. F., Gierse,
J. K., Stevens, A. M., Goodwin, D. C., Rowlinson, S. W., Marnett, L. J ., Stallings, W. C.
& Kurumbail, R. G. (2000) Structural insights into the stereochemistry of the
cyclooxygenase reaction, Nature. 405, 97-101.

117. Kurumbail, R. G., Kiefer, J. R. & Marnett, L. J. (2001) Cyclooxygenase enzyme:
catalysis and inhibition, Curr Opin Struct Biol. 11, 752-760.

118. Marco, S. D., Priestle, J. P., Grutter, M. G., Wennogle, L. P. & Boyar, W. (1997)
Characterization, crystallization and preliminary crystallographic analysis of human
recombinant cyclooxygenase-2, Acta Crystallogr D. 53, 224-226.

116

   

l1111111111111llllll1