Design, synthesis and assay of potential inhibitors of E. Coli asparagine synthetase B

MISSING IMAGE

Material Information

Title:
Design, synthesis and assay of potential inhibitors of E. Coli asparagine synthetase B
Physical Description:
Book
Creator:
Dribben, Anthony, 1970-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 28018283
oclc - 38268352
System ID:
AA00022847:00001


This item is only available as the following downloads:


Full Text











DESIGN, SYNTHESIS, AND ASSAY OF POTENTIAL INHIBITORS
OF E. COLI ASPARAGINE SYNTHETASE B
















By


ANTHONY BRIAN DRIBBEN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1997





























This work is dedicated to my parents, whom I respect,

admire, and love dearly. This dedication is extended to my

brother, sister, grandparents, aunts, uncles, cousins, and

others who have supported me during the course of my

education.














ACKNOWLEDGMENTS


I would like to thank my advisor, Nigel G. J. Richards,

for his guidance, patience, support, and valuable input

throughout my graduate studies. The members of my

supervisory committee, R. J. Bartlett, W. S. Brey, W. R.

Dolbier, and S. M. Schuster, also have my appreciation.

Thanks to John Stanton and John Watts for helping me get

started using the ACES II ab initio program, and to Mike

Bartberger for occasional advice with GAMESS. I thank Ian

Parr, Jose Rosa, and Margaret Deyrup for helpful advice

throughout the synthetic portion of this research. I thank

Susan Boehlein and Tom Soper for results from various enzyme

assays, and also for instructive conversations about the

enzyme activity. I could not have done it without the help

from these important people. They have my appreciation and

best wishes for the future.

I also want to thank the professors in the chemistry

department at UF, particularly Jon Stewart and Ben

Horenstein, as well as all of the chemistry professors at

Mississippi College, who have given me helpful advice

regarding my future career in chemistry.

Next, I would like to thank all my colleagues in the

Richards Group: Ian Parr, Jennifer Brown, Craig Porter, Jose


iii








Rosa, Susan Rasmussen, Srini Iyengar, Rick Smith, Nagraj

Bokinkere, Yannis Karayannis, Rajiv Moghe, Steven Nowicki,

Astra Dinculescu, Maria Capobianco-Perez, Saul Jacchieri,

Ramanan Thiurmoorthy, Mihai Ciustea, and Amy Boone for daily

productive and nonproductive conversations in the topics of

chemistry and life. I wish all of them the best in the

future and hope to see them again. Special thanks to Yannis

for his quick, but accurate, proofreading of this manuscript.

Sincere thanks go to my parents, Al and Mary Jo Dribben,

who have been my primary support system throughout my

studies. Thanks also to my sister, Traci, and my brother and

sister-in-law, Taylor and Melanie. You have all been there

to listen to my gripes and complaints, yet still, for some

reason, love me anyway. Thanks also go to my grandparents

and other extended family members who have kept me in their

thoughts and prayers all this time. I also want to thank my

pals Lucian, Melania, Nagraj, and Carlton for helping me

spend my free time away from the lab. I also thank all my

Mississippi pals for their support, and for at least feigning

interest in what I have been doing. I truly appreciate it.

Finally, I would like to thank the Florida Section of

the American Cancer Society, and the National Cancer

Institute of the National Institutes of Health for funding

for this research. Thanks also go to the Florida State

Supercomputing Center for generous allocations of

supercomputer time.
















TABLE OF CONTENTS


ACKNOWLEDGEMENTS .......................................... iii

ABSTRACT ..................................................vii

CHAPTERS

1 STRUCTURE AND FUNCTION OF THE AMIDOTRANSFERASES
AND E. COLI ASPARAGINE SYNTHETASE B ................. 1

The Amidotransferases ................................. 1


Asparagine Synthetase.............................
Structure and Function .......................
Reaction Mechanism...........................
E. Coli Asparagine Synthetase B...................
Structure and Function ........................
Reaction Mechanism...........................
Specific Aims....................................


... 10


...18
... 18
...20
...30


2 PROBING THE ASPARTIC ACID BINDING SITE OF E. COLI
ASPARAGINE SYNTHETASE B USING SUBSTRATE ANALOGS .... 32


Introduction .........................................
Synthesis of Constrained Aspartic Acid Analogs .......
Studies on the Alkylation of Aspartic Acid...........
Mapping the Aspartic Acid Binding Site of E. Coli
Asparagine Synthetase B Using Substrate Analogs ....
Crystallization Studies of the Amidotransferases.....
Conclusions ............ ..............................


32
32
36

44
49
52


3 MODELING THE REACTION MECHANISM OF E. COLI
ASPARAGINE SYNTHETASE B USING COMPUTATIONAL
APPROACHES ....................................... 54


Introduction ................................
Structural Comparison of Amides and Imides..


Computational Studies of Formamide
Computational Studies of Formimide
Modeling the Reaction of AS-B.........
Preliminary Studies ...............
Methylthiolate Attack Calculations
Tetrahedral Intermediate Breakdown


......... 54
......... 55
......... 55
......... 58
S. ........ 64
......... 64
......... 69


Calculations ................................... 75
Conclusions .......................................... 81








4 CHEMOENZYMATIC SYNTHESIS OF L-4,4-DIFLUOROGLUTAMIC
ACID ............................................... 83

Introduction ............................... 83
Historical Perspective ........................... 83
Biological Studies ............................... 84
Synthesis of L-4,4-Difluoroglutamic Acid............. 90
Synthesis of DL-4,4-Difluoroglutamic Acid ........ 91
Resolution of DL-4,4-difluoroglutamic Acid ....... 95
Conclusions ......................................... 101

5 STUDIES TOWARD THE SYNTHESIS OF TRANSITION-STATE
BASED INHIBITORS OF E. COLI ASPARAGINE
SYNTHETASE B ................................... 102

Introduction.................................... ... 102
Synthesis of Phosphorus-centered Transition State
Analogs .............................. ......... 104
Synthesis of Glutamine Phosphonamidate Analog ... 104
Synthesis of Asparagine Phosphonamidate
Analog ......................................... 115
Conclusions ......................................... 123

6 FUTURE WORK ....................................... 125

Chemoenzymatic Synthesis of L-4,4-Difluoroglutamic
Acid .............................................. 125
Synthesis of Transition-State Based Inhibitors of
E. Coli Asparagine Synthetase B................... 126

7 EXPERIMENTAL ........................................ 129

Materials, Instruments, and Methods ................. 129
Theoretical Calculations ............................ 131
Synthetic Procedures and Chemical Data of
Compounds ......................................... 132

APPENDICES

A ABBREVIATIONS FOR AMINO ACIDS ....................... 179

B CALCULATED PARAMETERS FOR FORMAMIDE AND TAUTOMERS...180

C CALCULATED PARAMETERS FOR FORMIMIDE AND TAUTOMERS... 182

D CALCULATED REACTION COORDINATE TABLES ............... 185

E CALCULATED REACTION COORDINATE FIGURES .............. 192

LIST OF REFERENCES ........................................ 198

BIOGRAPHICAL SKETCH ....................................... 213














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

DESIGN, SYNTHESIS, AND ASSAY OF POTENTIAL INHIBITORS
OF E. COLI ASPARAGINE SYNTHETASE B

By

Anthony Brian Dribben

August, 1997




Chairman: Dr. Nigel G. J. Richards
Major Department: Chemistry


P-Alkylation of protected aspartate derivatives has

allowed the preparation of a number of stereochemically

defined aspartate analogs suitable for probing the molecular

features of the aspartic acid binding site in E. coil

asparagine synthetase B (AS-B). Further investigation of the

alkylation reaction has determined that the addition appears

to proceed via a Z-lithium enolate geometry, with complete

stereochemical control in some cases. The use of substrate

analogs has shown that AS-B appears to be extremely selective

and that it can tolerate only relatively minor structural

alterations in the aspartic acid substrate.

The computational modeling of possible AS-B reaction

mechanisms and intermediates through semi-empirical and ab

initio methodology has shown the feasibility of several

vii








alternate mechanisms which do not involve nitrogen transfer

via ammonia. Moreover, AS-B appears to behave in a similar

manner as proteolytic enzymes which also utilize a catalytic

sulfur nucleophile.

A route for the chemoenzymatic synthesis of L-4,4-

difluoroglutamic acid has been developed which involves the

enzymatic resolution of a racemic precursor. This fluori-

nated amino acid is a valuable precursor in the synthesis of

other fluorinated substrate analogs. Potential applications

of these fluorinated analogs include use as mechanistic

probes of AS-B due to their enhanced reactivity, or use as

possible suicide inhibitors of AS-B.

A series of protected phosphorus-based transition state

analogs have been synthesized. These types of compounds,

which contain a phosphonamidate functionality, are useful

inhibitors of other hydrolytic enzymes, and may prove to be

effective inhibitors of AS-B.


viii














CHAPTER 1
STRUCTURE AND FUNCTION OF THE AMIDOTRANSFERASES AND E. COLI
ASPARAGINE SYNTHETASE B


The Amidotransferases


The amidotransferases are a family of enzymes that use

the amide functionality of glutamine 1 in the biosynthesis of

other biomolecules. These biomolecules include amino acids

(1), purine and pyrimidine bases (2), amino sugars (3),

coenzymes (4), and antibiotics (5). The fate of the amide

nitrogen atom in some of these biomolecules is illustrated in

Figure 1-1. The amidotransferase family of enzymes has been

thoroughly reviewed by Buchanan (6) and Zalkin (7).

Most of the amidotransferase enzymes catalyze at least

three different reactions (6,7). They can use ammonia as the

nitrogen source in the absence of glutamine (ammonia-

dependent activity). They also can use glutamine as the

nitrogen source, (glutamine-dependent activity), or hydrolyze

it to form glutamic acid 2 and ammonia (glutaminase

activity). The glutamine-dependent and glutaminase

activities can be inactivated by alkylating an active site

cysteine residue with a glutamine analog, or by replacing the

cysteine residue with an alternate amino acid.














I
H
amino acids
(Trp, Glu, Asn, His)
I





0
H3+ N
coH2
C 02


NH2





R
pyrimidines
(Cyt, Ura, Thy)


amino sugars
(Glucosamine 6-P)


NCH2R '


Cl

Cl
0


coenzymes
(Folic acid, Coenzyme B 12 )


02H


Antibiotics
(Chloramphenicol)


Figure 1-1. Fate of the amide nitrogen of L-glutamine 1 in
several biosynthetic pathways.



These properties indicate the presence of two

independent domains, or subunits, operating together for

enzyme activity. The synthetase domain utilizes ammonia as

the nitrogen source, and the glutamine amide transfer (GAT)

domain utilizes glutamine as the nitrogen source. The


NH2


R
purines
(Ade, Gua)








chemical reactions catalyzed by the synthetase domain (i) and

the GAT domain (ii) are illustrated in Figures 1-2 and 1-3,

respectively. These two domains are either fused together

and present on a single protein chain, or present as

independent subunits of the enzyme (6,7).

Based on the conservation of GAT domains throughout

these enzymes, the amidotransferase family has been divided

into two subfamilies. The first subfamily, known as Class I,

was originally designated TrpG based on the trpG gene

encoding the GAT domain of anthranilate synthase component

II. Members of this family include anthranilate synthase

(13), carbamoyl-phosphate synthase (14), cytidine

triphosphate (CTP) synthetase (3), FGAM synthetase (15),

guanosine monophosphate (GMP) synthetase (16), imidazole

glycerol-phosphate synthase (1), and aminodeoxychorismate

synthase (17). The second subfamily, known as Class II, was

originally designated PurF based on the purF gene encoding

the GAT domain of glutamine phosphoribosyl pyrophosphate

amidotransferase (glutamine PRPP amidotransferase) (8,9).

This family includes the enzymes asparagine synthetase (10),

glucosamine 6-phosphate synthase (2), glutamate synthase

(11), and glutamine PRPP amidotransferase (12). Other

amidotransferases which have not yet been classified as Class

I or Class II include arylamine synthetase (5), Gln-tRNAGln

amidotransferase (18), NAD synthetase (19), cobyrinic acid









F AS
R-L + (NH3)F AS R-NH2 + U (i)
-H+


R = nitrogen acceptor.
L = leaving group.
F
(NH3) = free ammonia derived from an ammonium salt source.



Figure 1-2. Reaction catalyzed by the synthetase domain of
Amidotransferases.



ONH2 AC02H
+ AS
H3+N + H20 +-- H3+N + (NH3) (ii)

C02 co2
1 2

G
(NH3) = unidentified form of ammonia derived from the amide nitrogen of 1



Figure 1-3. Reaction catalyzed by the glutamine
amidotransferase domain (GAT) of Amidotransferases.



a,c-diamide synthetase (20), and cobyric acid synthetase

(21).

The mechanism of amide nitrogen transfer from the GAT

domain to the synthetase domain was initially investigated by

Levitzki and Koshland (22) using the Class I enzyme CTP

synthetase from Escherichia coli (E. coli). Through the use

of kinetic assay experiments, it was found that ammonia

derived from the glutamine amide functionality (NH3)G remained

bound to the enzyme prior to incorporation into the

substrate. Ammonium sulfate was used in competition





5


experiments to demonstrate that free ammonia from solution

(NH3)F competed with (NH3)G for a single site in the enzyme.

Similar studies by Zalkin and Truitt (23) conducted using the

Class I enzyme GMP synthetase also found that (NH3)G remains

enzyme bound. However, experiments using ammonium chloride

as the (NH3)F source to determine whether a single NH3 site

was common to (NH3)F and (NH3)G were inconclusive. These

authors demonstrated that the results obtained by Levitzki

and Koshland showing competition of (NH3)F for (NH3)G were

probably due to S042- inhibition and not due to (NH3)F

inhibition derived from (NH4)2S04. Nevertheless, the specific

form of (NH3)G for these enzymes was not addressed in either

of these studies and remains an open question. Also, since

these experiments were based on Class I enzymes, the same

conclusions cannot necessarily be extended to Class II

enzymes.

The role of the ammonia-dependent activity has also been

studied in other amidotransferases such as anthranilate

synthase (24), glutamine (PRPP) amidotransferase (25), and

carbamyl-P synthetase (26). In general, it was found that

the ammonia-dependent activity of E. coli strains in which

the GAT domain function had been inactivated by modification

or deletion of the active site cysteine residue, remained

unaltered but required higher concentrations of ammonium

salts in the reaction medium. This observation is attributed

to the low concentrations of NH3 present in solution relative

to NH4+ at pH 7.0-7.5 and from the high KM of NH3 relative to








the KM of glutamine. From these studies, two ideas for the

ammonia-dependent activity of the amidotransferases were

proposed. One proposed idea is that the ammonia-dependent

enzymes existed before the evolution of the GAT domain. This

evolution resulted in the expansion of the specificity of

these enzymes. The other proposed idea is that the ammonia-

dependent activity resulted from utilization of an existing

enzyme-bound ammonia site by NH3G. However, no evidence has

been presented to support either of these two proposals.


Asparaqine Synthetase



Structure and Function


Asparagine Synthetase (AS) is a member of the Class II

subfamily of amidotransferases, which are characterized by

the location of the GAT domain in their polypeptide chain and

an essential N-terminal cysteine residue (8). Asparagine

synthetases which strictly use ammonia as the nitrogen source

have been found in bacteria (27), whereas asparagine

synthetases capable of utilizing both ammonia and glutamine

as nitrogen sources have been found in more complex higher

organisms. Bacterial AS has been isolated from Streptococcus

bovis (28), E. coli (29), Lactobacillus arabinosus (30), and

Klebsiella aerogenes (31). AS has also been isolated and

purified from fungi such as Neurospora crassa (32), yeast

such as Saccharomyces cerevisiae (33), plants (34), and

various mammals (10,35-37).








Coding sequences for AS were first obtained from the

Chinese hamster (38) and from human fibroblasts (10). They

were found to encode primary sequences of 561 amino acids

having 95% identity. Alignment of these sequences with E.

coli glucosamine 6-phosphate synthase, E. coli PRPP

amidotransferase, and yeast glutamine PRPP amidotransferase

sequences confirmed that AS is a member of the Class II

subfamily which has an N-terminal GAT domain followed by a

synthetase domain (38).

Primary sequence alignment for several members of the

Class II subfamily is shown in Figure 1-4 (39). From this

alignment, at least 11 residues seem to be conserved

throughout the GAT domains of all Class II amidotransferases.

These correspond to Cysl, Gly2, Ile3, Arg30o, Gly31, Asp33,

Arg49, Asn74, Gly75, Asn79, and Asp98 in E. coli asparagine

synthetase B (see APPENDIX A for amino acid abbreviations).

Although the crystal structure for any of the asparagine

synthetase enzymes remains to be solved, the crystal

structures for other Class II enzymes, namely B. subtillus

glutamine PRPP amidotransferase (40), E. coli glutamine PRPP

amidotransferase (41), and the GAT domain of glucosamine 6-

phosphate synthase (42) have been determined. The crystal

structure of the Class I amidotransferase GMP synthetase (43)

has also been determined.











Ham ASasea
Hum ASasea
Eco ASasea
Pea ASasea
Pea AS IIa
Hum GFata
Yst GFata
Rhz NodMa
PRPPa


Ham ASase
Hum ASase
Eco ASase
Pea ASase
Pea AS II
Hum GFat
Yst GFat
Rhz NodM
PRPP


Ham ASase
Hum ASase
Eco ASase
Pea ASase
Pea AS II
Hum GFat
Yst GFat
Rhz NodM
PRPP


Ham ASase
Hum ASase
Eco ASase
Pea ASase
Pea AS II
Hum GFat
Yst GFat
Rhz NodW
PRPP


MCGIWALFGS
MCGIWALFGS
MCSIFGVFDI


DDCLSVQCL.
DDCLSVQCL.
KTDAVELRK.


MCGILAVLGC SDDSQAKRV.
MCGILAVLGC SDPSRAKRV.
MCGIFAYLNY HVPRTRREIL
MCGIFGYCNY LVERSRGEIT


MCGIVGIV..


.... GHQPVS


MCGIVGIAGV M ..... PVN


EANACKTQLI
..EAOSTFIY
TMDAGTLQRR
.... NCFRLR

100
P..LFGMQPI
P..LFGMQPI
V..NAGAQPL
P..ASGDQPL
P..ASGDQPL
EPSPVNSHPQ
RPEQVNCHPQ
APTERNAHPH
SSSASEAQPF

150
EIILHLYDK.
EIILHLYDK.
EVILALYQE.
DVIAHLYEE.
DVIAHLYEE.
ETIAKLVKYM
ECIAKLYLHL
EVVAHLLAKY
EILLNIFASE


KKKGKVKALD
KQIGKVSALK
RAEGKLGNLR
KANGLVSDVF


RVKKYPYLWL
RVKKYPYLWL
YNQQKTHV.L
FNEDKSII.V
FNEDNPSI.V
RSDKNNEFIV
RSDPEDQFWVV
FTE...GVAV
YVNSPYGITL


YDN..RESQD
YNTNLQNGHD
...... RRDG
LDNFRHYPLE


.SAMKIA...
.SAMKIA...
*.KALELSRLM
.RILELSRRL


.HRGPDAFRF
.HRGPDAFRF
RHRGPD...W
KHRGPD...W


.RVLELSRRL KHRGPE...W


ETLIKGLQRL
DTLVDGLQRL
ERLVEALEPL
QSIYDALTVL


EEVHKQQDMD
EEITK.QNPN
EKL .......
EARHMQR...


CYNGEIYNHK
CYNGEIYNHK
AVNGEIYNHQ
TVNGEIYNHE
TVNGEIYNHE
IHNGIITNYK
VHNGIITNFR
VHNG.IENFA
AHNGNLTNAH


..... GGIEQ
..... GGIEQ
..... KG.PE
..... HG.EN
..... YG.ED
TS.FTTLVER
LD.FHELTKL
LG.RREAMHA
ADNIFAAIAA


EYRGYDSACV
EYRGYDSTGI
EYRGYDSAGV
QHRGQDAAGI


...... TNCC
...... TNCC
...... DNAI
...... GDNY
...... GDCY
LDIEFDVHLG
RDVTFVSHCG
KEAPLSGTIG
.... LQGNMG


ALQQRF....
KMQQHF....
ALRAEY..GD
ELRKQL..PN
DLRKQL..SN
DL.KKFLESK
EL.KTLLINK
EL.KDELAAG
ELRKKLFEEK


TICMLDGVFA
TICMLDGVFA
FLDDLQGMFA
FVDMLDGIFS
FVDMLDGIFS
VIQQLEGAFA
VLLELEGSYG
MLKRVKGAYA
TNRLIRGAYA


ENVNGY ....
ENVNGY....
SGIYAS....
SGLHQH ....
SGLHQH....
GFDGGNDKDW
AIDGD.....
A..........
ISIDAN ....

99
FGFHRLAVVWD
FGFHRLAVVD
LAHERLSIVD
LAHQRLAIVD
LAQQRLAIVD
IAHTRWATHG
IAHTRWATHG
IAHTRWATHG
IGHVRYPTAG

149
EFEYQTNVDG
EFEYQTKVDG
RYQFQTGSDC
H.KFFTQCDC
H.TFRTGSDC
GYDFESETDT
GYKFESDTDT
GAEFQTETDT
RRHINTTSDS

199
FILLDTANKK
FVLLDTANKK
FALYDSEKDA
FVLLDTRDNS
FVPLDTRDNS
LVFKSVHFPG
LLCKSCHYPN
LAVILFEDDPS
CVAMIIGHGM


Notes: (a) Ham ASase = Hamster asparaqine synthetase (EC 6 3 5 4)
Hum ASase = Human asparagine synthetase (EC 6 3 5 4)
Eeo ASase = E. coli asparagine synthetase B (EC 6 3 5 4)
Pea ASase = Pea AS1 nodule asparagine synthetase (EC 6 3 5 4)
Pea AS II = Pea AS2 root asparagine synthetase (EC 6 3 5 4)
Hum GFat = Human D-fructose-6-phosphate amidotransferase
Yst GFat = S. cerevisiai D-fructose-6-phosphate amidotransferase
Rhz NodM = NodM encoded D-glucosamine synthetase (R. leguminosarum)
PRPP = E. coli purF amidophosphoriboxyltransferase (EC 2.4.2.14)
(b) N-terminal methionine is not present in any of the mature proteins.
Residue numbers correspond to those of human D-fructose-6-phosphate
amidotransferase.

Figure 1-4. Primary sequence alignment of Class II
glutamine-dependent amidotransferases (Taken from reference
39).








Asparagine synthetase catalyzes the ATP-dependent

synthesis of L-asparagine 4 from L-aspartic acid 3 using

either ammonia (iii) or L-glutamine (iv) as the nitrogen

source (Figure 1-5). Glutaminase activity (v) has also been

found in various sources of AS.


'CO2H CONH2
S+ ) AS,ATP
|+ (NH3)F ---- (iii)
H3 -02- -AMP H3N CO2-
-PPi
3 4


CONH2 C2H CO2H CONH2
IrASATP K
H3 N + + -A WH3N r + + (iv)
H3+N C02- -AMP H3+N C02-
CO2- -PPi C02-
1 3 2 4

CONH2 CO02H
I ~AS
H3N + H20 AS1 H3 N + NH3 (V)
-H+
CO2 CO2
1 2


Figure 1-5. Reactions catalyzed by Asparagine Synthetase;
(iii) ammonia-dependent activity; (iv) glutamine-dependent
activity; (v) glutaminase activity.



Some asparagine synthetases, such as asparagine

synthetase A (AS-A), can only use ammonia as the source of

nitrogen to produce asparagine, while the glutamine-dependent

enzymes can use glutamine or ammonia as the nitrogen source

(7). Interestingly, the glutamine-dependent activity of the

latter enzymes can be inactivated without affecting the








ammonia-dependent activity. This fact was demonstrated in a

series of experiments by Van Heeke and Schuster using human

AS expressed in E. coli (44) and S. cerevisiae (45). In

these experiments, a mutant enzyme was expressed in which the

N-terminal cysteine residue (Cysl) was replaced with an

alanine residue (ClA mutant). This mutation resulted in the

loss of glutamine-dependent activity and a slight increase in

ammonia-dependent activity. These results verify the

presence of a GAT domain with an N-terminal cysteine residue

essential for glutamine-dependent activity, as well as a

synthetase domain which is responsible for the ammonia-

dependent activity. Further studies by Schuster and

coworkers demonstrated that these two domains are

topographically separated from each other (46). This was

shown through the use of monoclonal antibodies which

selectively and specifically inhibited each of the two

activities independently (46). It is the interaction between

these two domains which allow the nitrogen transfer reaction

to occur (Figure 1-6).


Reaction Mechanism


As a result of studies on mammalian AS and bacterial AS-

A, 180 isotopic labeling techniques have established that, in

asparagine synthesis, the synthetase domain appears to

catalyze formation of the 3-aspartyl-AMP intermediate 6

(37,47). Cedar and Schwartz demonstrated the presence of 6


















Figure 1-6. Schematic representation of the conformational
reorganization of the GAT- and synthetase domains in the
nitrogen transfer reaction of asparagine synthetase.



by incubating E. coli AS with 180-labeled aspartic acid 3 and

ATP 5 (Figure 1-7) (47). These react to form 180-labeled 6

and release pyrophosphate (PPi) with no 180-label. The

resultant nucleophilic attack of ammonia on 6 forms 180-

labeled L-asparagine 4 and releases 180-AMP 7 as a side

product. This scheme of reactions illustrates the events

taking place in the ammonia-dependent reaction (iii) which

occur in the C-terminal synthetase domain of AS.

Of more interest is the function of the GAT domain.

Sequence studies by Tso et al. (48) and Vollmer et al. (49)

on glutamine PRPP amidotransferase established the presence

of an N-terminal cysteine residue (Cysl) essential for

glutamine-dependent activity. This residue is conserved in

all the members of the Class II family of amidotransferases.

It was demonstrated that alkylation of Cysi with 6-diazo-5-

oxo-L-norleucine 8 (DON) (Figure 1-8), or mutagenesis of the

histidine (Hislo0) and aspartate (Asp29) residues in glutamine

PRPP amidotransferase each abolished the glutamine-dependent


GAT-domain






Synthetase-domain











H3N N 180-
C02 18Q


NH2

0 0- o-
p p P
II II II o.
11 11 110^~
0 0 0

5 -
HO OH


-PPi

NH2


p ~Nq

o2- 180 0

6 -
HO OH


+NH3


H3 N *, NH2

CO2- 180


NH2

N
18 0- I 0

11 0

7 -
HO OH


Figure 1-7. E. coli AS catalyzed biosynthesis of L-
asparagine 4 from 180-labeled L-aspartic acid 3 and NH3 via
180-p-aspartyl-AMP intermediate 6.








0 0

/2 Cy
H3+N + S ___Cyl -N2 0 H3+N

C02 C02

8 9


Figure 1-8. Alkylation of Cysl with 6-diazo-5-oxo-L-
norleucine (DON) 8 to form inactivated Cysl residue 9 (12).



activity without affecting the ammonia-dependent activity

(8). Schuster et al. replaced Cysi with alanine (ClA) on

human AS and obtained a mutant enzyme with no glutamine-

dependent activity and unaffected ammonia-dependent activity

(45,50).

Mei and Zalkin (8) have proposed a mechanism for the

Class II subfamily of amidotransferases, including AS, based

on studies upon glutamine PRPP amidotransferase (Figure 1-

9). In this enzyme, the active site is composed of a Cysj,

Hisjol, and Asp29 catalytic triad (glutamine PRPP

amidotransferase numbering) which have a similar function to

the Cys-His catalytic diad found in the thiol protease family

of enzymes (51). In the proposed mechanism of AS, the

nucleophilicity of the sulfur atom is increased by the

presence of Hisjol in the active site, through polarization or

deprotonation of Cysj. The thiolate anion formed attacks the

amide functionality of glutamine to form the tetrahedral

intermediate 10. The primary amine generated abstracts a

proton from the imidazolium ion of Hisioi to form intermediate








11 which subsequently breaks down to give ammonia and y-

glutamyl-thioester enzyme intermediate 12. The ammonia
generated at this stage reacts with P-aspartyl-AMP

intermediate 6 to generate asparagine and AMP (Figure 1-7).

Nucleophilic attack of water, assisted by Hisol01 as before,

attacks intermediate 12 to form tetrahedral intermediate 13.

Breakdown of 13 releases glutamate and regenerates the

initial state of the enzyme.
Recent evidence, however, has been presented which does

not support this proposed mechanism. The primary sequence

alignment of several Class II amidotransferases have failed

to show conserved histidine residues (Figure 1-4) (39).

Furthermore, primary sequence alignment of 20 asparagine

synthetases from different sources have revealed no histidine

residue cognate to Hisol01 in glutamine PRPP amidotransferase

(52). In addition, the crystal structures of B. subtillus

glutamine PRPP amidotransferase (40), E coli glutamine PRPP

amidotransferase (41), and the GAT domain of glucosamine 6-

phosphate synthase (42) failed to reveal amino acid residues

in close proximity to Cysl which can participate in acid-base

catalysis. Therefore, residues Hisjol1 and Asp29, which are

implicated as catalytically important residues in glutamine

PRPP amidotransferase, do not seem to play a role in

catalysis by AS.


















10 Cys i


0 H /

a+N
H3 N
I 1+
CO 2 N H3
12 Cys1





-0 OH +


H3 + N 11* J.t%>

CO2 N H3
13 Cys1


+ H20
H 20O


H His


NH"3
N11 H 3
11 Cy


NN
'N H
A His


H3 N

CO2



H3N N
Cysi 0


N %-0

H His











H
His


Figure 1-9. Proposed ammonia-mediated mechanism of
Asparagine Synthetase based on studies of glutamine PRPP
amidotransferase (8).



Milman et al. (53) kinetically characterized mouse

pancreas AS and proposed a uni-uni-bi-ter ping-pong Theorell-

Chance mechanism (54) for the glutamine-dependent activity

(Figure 1-10A). In this mechanism, glutamine binds first

followed by release of glutamate. Sequential binding of








aspartate and ATP follow, at which point nitrogen is

transferred. The ordered release of PPi, AMP, and Asn

regenerate the original state of the enzyme. Markin et al.

(55) kinetically characterized beef pancreas AS and proposed

a bi-uni-uni-ter ping-pong mechanism (54) for the glutamine-

dependent activity (Figure 1-10B). In this mechanism,

glutamine and ATP bind sequentially, followed by release of

glutamate and addition of aspartate. At this stage, nitrogen

is transferred, followed by sequential release of PPi, AMP,

and Asn to regenerate the original state of the enzyme. In

the ammonia-dependent activity of this enzyme, ammonia binds

first followed by random binding of ATP and Asp. Nitrogen

transfer then occurs, followed by sequential release of PPi,

AMP, and Asn (Figure 1-10C) (55).

It is difficult to reconcile the binding of glutamine
and release of glutamate prior to the formation of 3-

aspartyl-AMP 6. Since aspartate binds after hydrolysis of

the putative y-glutamyl enzyme intermediate 12, it is not

clear how the enzyme can sequester ammonia throughout the

hydrolysis step, as well as maintain the unprotonated

nucleophilic character needed for the subsequent reaction

with P-aspartyl-AMP 6. Recent studies by Habibzadegah-Tari

on E. coli AS (56) have demonstrated that glutamine is the

last reactant to bind to the enzyme before the nitrogen

transfer step occurs, in contrast with previous results. If

this is the case, 3-aspartyl-AMP formation would precede

glutamine hydrolysis.








Gin Glu Asp ATP


1 V \ ...


PPi AMP Asn
A A


nitrogen
transfer
-----W_


Gin ATP Glu Asp


if


nitrogen
transfer
----A-


PPi AMP Asn

I I I


(ATP)(Asp)
NH3 Asp ATP


if if if


nitrogen
transfer
----- >*30.


PPi AMP Asn
I I


Figure 1-10. Kinetic mechanism of AS depicting order of
substrate binding, nitrogen transfer, and product release; A.
glutamine-dependent activity in mouse pancreas AS; B.
glutamine-dependent activity in beef pancreas AS; C. ammonia-
dependent activity in beef pancreas AS.


H H .. ........










E. coli Asparaqine Synthetase



Structure and Function


There are two different AS enzymes that have been

isolated from E. coli. Asparagine synthetase A (AS-A) is

encoded by the asnA gene and strictly catalyzes the ammonia-

dependent reaction (iii) (Figure 1-5) (27). In contrast,

asparagine synthetase B (AS-B) is encoded by the asnB gene

and catalyzes all reactions (iii)-(v) independently (57).

However, multiple asparagine synthetases contained in a

single organism other than E. coli have not been reported

(7).

The primary sequence of E. coli AS-A has been determined

from the cloned asnA gene and found to be composed of 330

amino acids with an Mr of 36,700 (58). Similarly, the

primary sequence of E. coli AS-B has been determined from the

cloned asnB gene and found to be composed of 554 amino acids

with an Mr of 62,700 (59). Inspection of the N-terminal

amino acid sequence of AS-B identified it as being a member

of the Class II subfamily. Also, the primary sequence of AS-

B was found to have a high degree of homology (48%) with

human AS and little or no homology with AS-A. Because of the

dissimilarity between them, it is accepted that the asnA and

asnB genes evolved independently. Thus, the ability of AS-B

to possess glutamine-dependent activity probably did not








arise from fusion of an ancestral GAT gene with the asnA gene

(59).

Site-directed mutagenesis on the GAT domain of AS-B has

revealed no histidine residues which can participate in

glutamine-dependent nitrogen transfer (39). Furthermore,

crystal structures of other members of the Class II subfamily

lack amino acid residues in close proximity to Cysi which can

participate in acid-base catalysis. Unfortunately, no

crystal structure is available for E. coli AS-B or any

asparagine synthetase. Nevertheless, alternative pathways

for the nitrogen transfer step in the mechanism of Class II

enzymes have recently been made.


Reaction Mechanism


Richards and Schuster (60) have proposed a mechanism for

glutamine-dependent nitrogen transfer in E. coli AS-B which

operates without participation of histidine or aspartate

residues in catalysis (Figure 1-11). In this mechanism,

nucleophilic attack of the amide nitrogen on P-aspartyl-AMP 6

forms the unsymmetric acyclic imide 14 after breakdown of a

presumed tetrahedral intermediate. Subsequent attack of Cys1

thiolate anion upon imide 14 produces y-glutamylthioester 12

and the hydroxyimine tautomer of asparagine 15 after

protonation. Finally, water attacks intermediate 12 to

ultimately form glutamate via intermediate 13 (Figure 1-9)

and regenerates the initial state of the enzyme.















SH

Cys 1


H3 +N) 0O
cO 0

6


HO OH


-AMP


H3+N,


CO2- SH

CysI


CO2


SH

CYS1


Figure 1-11. Proposed mechanism for glutamine-dependent
nitrogen transfer in E. coil AS-B involving acyclic
unsymmetric imide 14 as a reaction intermediate.



In this proposed imide-mediated mechanism, the need for

an active site histidine to perform general acid protonation

of the ammonia leaving group from tetrahedral intermediate 10

is eliminated (Figure 1-9). Furthermore, the transfer of


H3+N.


-Asn
+H20
V


+ H20








nitrogen via the imide intermediate 14 eliminates the

possibility of ammonia diffusing from the active site to form

unproductive NH4+. Additionally, there is no need for a

separate ammonia binding pocket or an NH3-enzyme complex

intermediate.

There is some precedence for imide formation from

primary amides in other biological systems as well as from

synthetic organic methodology (61-63). Gibbs et al. have

raised catalytic antibodies which catalyze the rearrangement

of an asparaginyl-glycyl peptide sequence to an aspartyl-

glycyl and an isoaspartyl-glycyl peptide sequence through the

formation of a cyclic succinimide intermediate (61). Solid

phase peptide synthesis using certain glutamine derivatives

form peptide mixtures arising from coupling at the Cc-

carboxyl group and the Cy-carboxyl group via a cyclic

glutarimide intermediate (62,63).

Recent mutagenesis studies on E. coli AS-B have

investigated the glutamine-dependent nitrogen transfer

reaction in more detail (39). Sequence studies have revealed

that His29 and His80 are conserved in the Class II subfamily

of amidotransferases. The His29 to alanine (H29A) and the

His8o to alanine (H80A) mutants each demonstrated essentially

no effect on the ammonia-dependent activity of AS-B,

indicating that the mutant protein had folded in a similar

manner to wild type AS-B. In the case of the glutamine-

dependent reaction, for H29A and H80A, although kcat was

essentially unchanged by the amino acid replacements, the Km








for glutamine was increased by a factor of approximately 4.5

relative to the wild type enzyme. These results suggest that

neither of these histidine residues appear to be involved in

mediating the nitrogen transfer reaction.

The Cysl to alanine (C1A) and serine (CIS) mutants of

AS-B were constructed and determined to have no glutaminase

or glutamine-dependent activity (60). Although the ammonia-

dependent activity was unaffected, it was inhibited by

glutamine. This observation was also made in the CIA and CIS

mutants of human AS (52). Furthermore, the CIA mutant

inhibition pattern was consistent with the formation of an

abortive complex in a similar fashion as the human AS CiA

mutant. These results are consistent with the proposed

mechanism, in which the imide intermediate 14 (Figure 1-11)

could represent the abortive complex.

Primary sequence alignment studies have also shown that

Asp33 (D33) in AS-B is cognate to Asp29 in glutamine PRPP

amidotransferase (Figure 1-4). Moreover, Asp33 is highly

conserved in the Class II subfamily including asparagine

synthetases from various sources (39). The Asp33 to

asparagine (D33N) and Asp33 to glutamic acid (D33E) mutants of

AS-B had little or no effect on the glutamine and ammonia-

dependent activities, however.

In recent studies conducted by Habibzadegah-Tari (56),

E. coli AS was kinetically characterized and a bi-uni-uni-ter

ping-pong Theorell-Chance mechanism (54) was proposed for the

glutamine-dependent activity (Figure 1-12). These results




















Figure 1-12. Kinetic mechanism of E. coli AS-B.



differ from the kinetic mechanisms reported for mouse

pancreas (54) (Figure 1-10A) and beef pancreas AS (56)

(Figure 1-10B). In this mechanism, ATP and aspartate bind

sequentially, followed by release of PPi and addition of

glutamine. At this stage, nitrogen is transferred, followed

by random release of asparagine and AMP, then glutamate

release to regenerate the original state of the enzyme.

Contrasting with the mechanisms proposed by Milman et al. and

Markin et al. (Figure 1-10A,B), this mechanism does not

generate free ammonia while simultaneously interacting with

Asp and ATP. Furthermore, these results are consistent with

the mechanism for glutamine-dependent nitrogen transfer

proposed by Richards and Schuster (60) (Figure 1-11). Since

the binding of glutamine and subsequent nitrogen transfer
take place after formation of P-aspartyl-AMP 6, this

mechanism also allows for the possibility of forming imide

intermediate 14.

In related studies on human AS, Sheng et al. (50)

pursued mutagensis studies to investigate the glutamine-


(AMP) (Asn)
ATP Asp PPi Gin Asn AMP Glu





nitrogen
transfer








dependent nitrogen transfer reaction in more detail. The

Cysl to alanine (ClA) and serine (CIS) mutants were

constructed and determined to have no glutaminase or

glutamine-dependent activity. The ammonia-dependent

activity, although unaffected, was inhibited by glutamine,

presumably through the formation of an abortive complex.

Sequence studies on human AS carried out by the Boehlein, et

al. (39) demonstrated that His102 of human AS is not cognate

to Hisol01 of glutamine PRPP amidotransferase as was previously

believed. These results infer the possibility that human AS

may also be operating by a mechanism analogous to the

proposed mechanism for E. coli AS-B (Figure 1-11).

Sequence studies have revealed that the residues Arg30o,

Asn74, and Asn79 appear to be conserved throughout the Class

II subfamily of amidotransferases (39). The roles of these

highly conserved residues on the glutamine-dependent activity

of E. coli AS-B were investigated through mutagenesis studies

by Boehlein, et al (64). The Arg3o to alanine (R30A) mutant

of AS-B resulted in a 182-fold increase of KM for glutamine

in the glutamine-dependent activity compared to wild-type AS-

B. In addition, the ATP-dependent stimulation of the

glutaminase activity was modified or completely lost when

Arg30o is replaced by other amino acids. This observation

suggests that Arg30 may mediate communication between the

synthetase and GAT domains of AS-B, in an analogous fashion

to that observed for Glu841 in carbamoyl-phosphate synthetase

(65). The kinetic parameters associated with the ammonia-








dependent and the glutamine-dependent activity of the Asn79

to alanine (N79A) mutant of AS-B were essentially unchanged

compared to wild-type AS-B, indicating that this residue

appears to possess no catalytic function in AS-B.

Three mutants of AS-B were constructed in which Asn74

was replaced by alanine (N74A), glutamine (N74Q), and

aspartic acid (N74D). All three of these mutants displayed

kinetic parameters associated with ammonia-dependent activity

which are similar to those of wild-type AS-B. However, the

N74A and N74Q mutants displayed a substantial difference in

glutamine-dependent activity compared to wild-type AS-B,

indicating that Asn74 plays a significant role in glutamine-

dependent nitrogen transfer. Glutaminase activity of the

N74A mutant was similar to that of wild-type AS-B; however,

the activity of the N74Q mutant was only 10% relative to

wild-type AS-B. Significantly, the glutaminase activities of

both N74A and N74Q were stimulated by ATP. Interestingly, it

has been determined that the N74D mutant confers nitrile

hydratase activity upon the mutant enzyme (66).

These results obtained from these studies on Asn74 allow

for further refinement of the proposed mechanism of E. coli

AS-B shown in Figure 1-11. The Asn74 residue can play a

catalytic role through polarization of the amide bond of

glutamine, possibly by formation of low-barrier hydrogen

bonding (67,68), to form the hydroxyimine tautomer 16 which

now has a more nucleophilic nitrogen atom (Figure 1-13).

This nitrogen atom can attack P-aspartyl-AMP 6 to form imide








intermediate 14, followed by the subsequent attack of Cysi

thiolate anion on imide 14 to form the tetrahedral

intermediate 17. The resultant breakdown of 17 to form the

y-glutamythioester enzyme intermediate 12 and asparagine

tautomer 15 ultimately generates glutamate and asparagine

(Figure 1-11).

A modified version of this mechanism which is also

consistent with recent results on Asn74 has been proposed (64)

(Figures 1-14). In this mechanism, Asn74 plays a catalytic

role by stabilizing the oxyanion formed in intermediate 10 by

the nucleophilic attack of Cys, on bound glutamine. The

resulting primary amine nitrogen of 10 is nucleophilic, and

can attack 3-aspartyl-AMP 6 to form the intermediate 17,

which ultimately generates asparagine and glutamate as before

(Figure 1-11).

Further studies on E. coli AS-B using alternate

substrates have provided further evidence for the involvement

of Asn74 in catalysis. Boehlein et al. (69) have investigated

the ability of the enzyme to use glutamic acid y-hydroxamate

(LGH) and hydroxylamine as alternate substrates for glutamine

and ammonia, respectively. Using wild-type AS-B and LGH as

the substrate, the catalytic efficiency (kcat/KM) increased 3-

fold in the glutaminase activity relative to glutamine, while

it decreased 3-fold for the glutamine-dependent activity

compared to glutamine. The catalytic efficiency for








H H

.H V Asn74 Y -Asnr74
N"

H3+N H H3+N H
1 l 16
C02- C02

y N-aspartyl-AMP

Cysi
I + +
-o 0 H3 0 0 NH3

(202N 2
H3N2 H H3+N H
S+NCysl Si
Co2- 17 C- 14

-Asn



OH
0H 0

H3N H20 H3+N J
-IT Cys1 -Cys1 _1

C02 12 CO2



Figure 1-13. Proposed mechanism for nitrogen transfer in E.
coli AS-B involving imide intermediate 14.



hydroxylamine in the ammonia-dependent activity of wild-type

AS-B increased 2-fold relative to ammonia. Using the N74A

mutant, using both LGH and hydroxylamine in all three

reactions catalyzed by AS-B (Figure 1-5) gave essentially no

difference in catalytic efficiency as compared to using









0s N s i
O -O S -0 S 0 NH3+


+ 2 Cysi + 2P-Aep-AMP 1. 2
H3+N H3+N H3 N
T T i0 T 1
C02 C02 C2 17





0 0 OH

OH +
+ -+n +
H3 N H3 N
H20 Cys H3 C 2-
2 i-Cys12
12 15



Figure 1-14. Proposed mechanism for nitrogen transfer in E.
coli AS-B involving tetrahedral intermediate 10.



glutamine and ammonia as substrates. However, because the

nitrogen transfer with LGH proceeded with a turnover number

(kcat) that was 10-fold lower than that observed in glutamine-

dependent asparagine synthesis, the results were interpreted

to be further evidence against ammonia-mediated nitrogen

transfer in E. coli AS-B.

Indirect support for the proposed polarization and

tautomerization of the glutamine amide group has been

obtained in model studies upon the glycosylation of

asparagine residues in Asn-X-Ser/Thr sequences (70-72).

Imperial et al. evaluated several synthesized

conformationally constrained peptides containing asparagine.

The primary amide nitrogen of the Asn residue becomes N-

glycosylated through nucleophilic attack on an electrophilic








oligosaccharide. This process is catalyzed by the enzyme

oligosaccharyltransferase (70). Studies on this enzyme

system indicated that the Ser and Thr residues in the active

site presumably assist the tautomerization of the Asn primary

amide group.

In experiments aimed at determining the mechanism of

nitrogen transfer of E. coli AS-B, 13C and 15N kinetic isotope

effects have been measured in the glutaminase and glutamine-

dependent synthesis reactions (73). For the glutaminase

reaction, substitution of heavy atom labels in the sidechain

amide of the glutamine gave observed values of 1.0245 and

1.0095 for the amide carbon and amide nitrogen, respectively.

In the glutamine-dependent synthesis of asparagine, the amide

carbon and amide nitrogen isotope effects were 1.2031 and

1.0222, respectively. Based on comparison of these values

with those obtained with papain and other thiol proteases,

these results were interpreted to mean that nitrogen transfer

does not proceed by free ammonia formation, rather, it

probably involves a series of intermediates in which

glutamine becomes covalently attached to aspartate. These

results are consistent with the proposed mechanism for

nitrogen transfer through the tetrahedral intermediate 17

(Figure 1-14) (64). However, these results are also

consistent with the proposed mechanism for nitrogen transfer

via the imide intermediate 14 (Figure 1-13) (60).

Further experiments aimed at determining the mechanism

of E. coli AS-B were recently carried out by Rosa-Rodriguez,








et al. (74). A sample of the imide intermediate 14 was

synthesized using synthetic organic methodology and screened

for activity against AS-B. Incubation of 14 with wild-type

AS-B did not form products L-Glu and L-Asn. Furthermore,

incubation of L-GIn, L-Asp, and ATP with the CIA and CIS

mutants of AS-B did not generate the imide 14. These studies

indicated that the lack of imide 14 was not dependent on the

energetic of the enzyme-substrate complex. While these

results are inconsistent with the imide-mediated nitrogen

transfer mechanism (Figure 1-11), they do not necessarily

rule it out as a possibility.


Specific Aims


The enzyme L-asparaginase (75,76) is often employed in

the treatment of acute lymphoblastic leukemia (ALL). The

mechanism of action for L-asparaginase is to deplete the

level of circulating asparagine synthesized in the liver by

asparagine synthetase (77,78). Resistance of human neoplasms

to protocols involving L-asparaginase arises from the

synthesis of endogenous asparagine in malignant cells

(79,80), and there is also evidence to suggest that blood

asparagine concentration may also play a role in T-cell

response (81). Therefore, inhibitors of asparagine

synthetase represent targets with potential application in

treating leukemia, and in exploring cellular mechanisms of

immunosuppression (82,83).








The first part of the present study involved the

synthesis and use of constrained substrate analogs to probe

the active site of E. coli AS-B. The lack of a crystal

structure of the enzyme necessitates the use of probes to

analyze the structural properties of the active site.

Further studies were done to clarify several stereochemical

features observed in the alkylations of aspartic acid.

The second part of the present study involved the

computational modeling of possible reaction mechanisms and

intermediates through semi-empirical and ab initio

methodology. From the results of these calculations, a

comparison of reaction energetic for the proposed mechanisms

versus the currently accepted ammonia-mediated mechanism was

performed.

The third part of the present study involved the

chemoenzymatic synthesis of fluorinated glutamate substrate

analogs. Potential applications of these fluorinated analogs

include use as possible suicide inhibitors, or use as

mechanistic probes of E. coli AS-B due to their enhanced

reactivity.

Finally, the fourth part of the present study involved

the development of synthetic routes toward the synthesis of

phosphorus-based transition state analogs. Regardless of the

mechanism of action, compounds that mimic a tetrahedral

intermediate or transition state may prove to be effective

inhibitors of the enzyme.














CHAPTER 2
PROBING THE ASPARTIC ACID BINDING SITE OF E. COLI ASPARAGINE
SYNTHETASE B USING SUBSTRATE ANALOGS


Introduction


Functionalized aspartic acid analogs have proven useful

tools in the elucidation of structure-function relationships

associated with the N-methyl-D-aspartate (84) and glutamate

receptors (85), and in the preparation of chiral

intermediates in the synthesis of a number of complex natural

products (86). This study involves the preparation and assay

of a series of P-functionalized aspartates to probe the

specificity and stereochemical preferences of the E. coil

asparagine synthetase B active site. Additional studies have

also probed the stereoselectivity of the aspartate alkylation

reaction.


Synthesis of Constrained Aspartic Acid Analogs


The synthesis of the constrained aspartate analogs 18

and 19 begins with the commercially available N-

benzyloxycarbonyl-L-aspartic acid 20 (Figure 2-1). Using

the method of Wang et al. (87), formation of the N-protected

diester 21 was achieved by reacting the dicarboxylate cesium

salt of 20 with benzyl bromide (48% yield). Alkylation of

the dianion of 21 has been shown to proceed without









1 CO2H CO2Bn R. 2

Hilloi
fi ii ^ "
f, %%
ZHN CO2H ZHN C02Bn ZHN C02Bn
H
20 21 2 2: RI=CO2Bn, R2=H
23: RI=H, R2=CO2Bn



Figure 2-1. Synthesis of dibenzyl [2S,3R]-2-benzyl-
oxycarbonylamino-3-propen-2'-yl-succinate 22 and dibenzyl
[2S,3S]-2-benzyloxycarbonylamino-3-propen-2'-yl-succinate 23.
Z=C6H5CH20CO-; Bn=C6H5CH2-. Reagents and conditions: i.) a.
Cs2CO3, H20, b. BnBr, DMF, 48%; ii.) LiHMDS, allyl bromide,
THF, -78 oC, 60%.


racemization at the chiral carbon using a wide variety of

electrophiles (88-90). Treatment of 21 with lithium

hexamethyldisilazide (LiHMDS) (2.0 eq.) in THF at -78 C

gave a 1:6 mixture of the diastereomers 22 and 23 after

column chromatography. No N-alkylated product was observed,

and the recovered starting material had an unchanged optical

rotation, suggesting that the deprotonation at C-2 did not

occur under these conditions. Two equivalents of base are

needed in this reaction due to the relatively acidic

carbamate N-H present. As the mixture of diesters 22 and 23

could only be separated through careful chromatography and

fractional crystallization, it was decided that the

separation of the diastereoisomeric products would be done at

a later stage in the synthesis, if possible.








R 2 RI

22+23 HO AN
^C02Bn X CO2Bn
Z H Z H

24: RI=CO2Bn, R2=H 26: Rl=CO2Bn, R2=H
25: Ri=H, R2=CO2Bn 27: RI=H, R2=CO2Bn


iii


RI


H2 H

18: R1=CO2H, R2=H
19: Ri=H, R2=CO2H



Figure 2-2. Synthesis of pyrrolidine-2(S),3(R)-dicarboxylic
acid 18 and pyrrolidine-2(S),3(S)-dicarboxylic acid 19.
Z=C6H5CH20CO-; Bn=C6H5CH2-. Reagents and conditions: i.)
OS04, NaI04, MeOH, H20, RT, 85%; ii.) Et3SiH, CHCI3, TFA, RT,
26 27%, 27 40%; iii.) NH4OCHO, Pd/C, EtOH, A, 18 85%, 19
25%.


Cleavage of the double bond and subsequent cyclization
(Figure 2-2) was accomplished using sodium periodate and
catalytic Os04 (91,92) in an aqueous methanol solvent, giving
the mixture of substituted pyrrolidines 24 and 25 in a good
yield. Dehydration and reduction of the resulting imine was
done by treatment of 24 and 25 with triethylsilane and
trifluoroacetic acid (93) in chloroform to give the








pyrrolidines 26 and 27 in moderate yields. At this time, it

was possible to separate the diastereoisomers without too

much difficulty by column chromatography. This purification

afforded 26 and 27 as a 1:2.5 ratio. Removal of the N-

benzyloxycarbonyl protecting group, as well as the benzyl

esters, was performed independently on 26 and 27 with

catalytic hydrogenation using palladium and ammonium format

in ethanol (94,95). Purification of the crude diacids using

Dowex-1 strong anion exchange resin (97) gave purified 18 in

a 85% yield, and 19 in a 25% yield.

During our experiments, a study by Humphrey et al. (97)

appeared describing a synthesis of the enantiomers of 18 and

19 (synthesis began with D-aspartic acid) and their potential

use as neurotransmitter conformer mimics. The alternate

synthetic route is conceptually identical to the one
described herein, employing a stereocontrolled P-alkylation

of a protected aspartic acid derivative, except that the N-

phenylfluorenyl protecting group was used. In this study,

the relative stereochemistry of the alkylated material was

proven by X-ray crystallography.

While the synthesis and use of constrained aspartic acid

analogs to probe the aspartic acid binding site of E. Coli

AS-B is the goal of this part of the project, the observation

of diastereoselectivity in the alkylation reaction warrants

further investigation.









Studies on the Alkylation of Aspartic Acid


The first reported alkylation of a protected aspartate

derivative was done in a study by Seebach and Wasmuth (98) in

which the N-formyl di-tert-butyl ester of L-aspartic acid was

alkylated with a series of alkyl halides (Table 1, entry 1).

Using lithium diethylamide as the base, 3-alkylated products

were produced with some stereoselectivity, however the

product mixture also contained a-alkylated product,

indicating that little regioselectivity existed for the

reaction. Further work by Baldwin et al. (90) on P-methyl a-

tert-butyl diester of aspartic acid found that increasing the

steric bulk of the nitrogen protecting group and using

sterically bulky bases would give the 3-alkylated products as

the sole regioisomer (Table 1, entry 2-3). However, little

or no stereoselectivity was seen at the 3-carbon, probably

due to the smaller P-methyl ester that was used.



Y Y
|H |H
C02Rp C02Rp iljC02Rp
base, E-X EE + E
X=Br,I,CHO .f J^
RNHN CO2Ra X=BrICHO RNHN CO2R RNHN CO2Ra
H H
a b
Y=H,OH


Figure 2-3. Alkylation of protected aspartic acid with
various electrophiles (E-X) to produce diastereomers a and b.










Table 1. Examples of Alkylations of Protected
L-Aspartic Acid (Figure 2-3)

a:b yield
RN Ra Rp basea electophile ratiob (%) Ref.


1
2
3
4
5
6
7c
8C
9c

10C
11C
12
13
14
15
16
17
18
19
20
21
22
23
24
25


HCO
Cbz
Cbz
Cbz
PhFl
PhFl
PhFl
PhFl
PhFl
PhFl
PhFl
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz
Cbz


tBu
tBu
tBu
tBu
Me
Me
Me
Me
Me
Me
Me
tBu
tBu
tBu
tBu
tBu
tBu
tBu
tBu
Me
tBu
CPh3
tBu
tBu
Bn


tBu LiEt2NH
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me KHMDS
Me KHMDS
Me KHMDS
Me LiHMDS
Me KHMDS
Me LiHMDS
Me KHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Me LiHMDS
Bn LiHMDS


CH3I
PhCHO
PhCH2Br
CH2=CHCH2Br
PhCH2I
CH3CH2OTf
CH2=CHCH2I
CH2=CHCH2I
CH3I
CH31
PhCH2Br
ClCH2CO2tBu
CH2=CHCH2Br
CH2=CHCH2I
CH3CHO
PhCHO
propargyl Br
NCCH2Br
CH3I
BrCH2CO2tBu
BrCH2CO2tBu
BrCH2CO2tBu
ClCH2CO2tBu
PhCH2Br


a All reactions done in THF. b The products a and b
correspond with diastereomers a and b in Figure 2-3. c
Alkylations performed on D-aspartic acid.


>50:1
1:1
5:1
3:1
7:1
>50:1
1:10
23:1
2:1
>50:1
>1:50
>50:1
>50:1
8:1
>50:1
>50:1
4:1
2:1
1:1
5:2
5:4
4:5
>50:1
2:1


98
90
90
88
100
99
97
97
97
97
97
102
102
102
102
102
102
102
102
102
102
102


CH2=CHCH2Br 6:1


w w









M+ ++ R+
RN 0 RN N 0
-N" O-" I -Nt


RPO2C RpO02C V "-Q
i ii



E-X E-X

H0


RPO2C OR
E


Figure 2-4. Proposed nitrogen-chelated enolate geometry for
the alkylation of protected L-aspartic acid derivatives with
various electrophiles; i (Z)-enolate; ii (E)-enolate.


Rapoport et al. (99,100) reported the stereoselective

alkylation of 9-phenylfluoren-9-yl-N-protected L-aspartic

acid derivatives with various electrophiles giving higher

diastereomeric excess of the 2(S),3(R) products (Table 1,

entries 5,6). The proposed explanation of this observation

is alkylation through a chelation-controlled (Z)-potassium

enolate in which the nitrogen group is involved in chelation

with the cation to form a cyclic-chelated enolate (Figure 2-

4). Precedence for this type of nitrogen-chelated enolate

has been reported by Rapoport and Lubell (103) and Yamamoto

et al. (104) in related systems.








Further work by Humphrey et al. (97) on the same system

investigated the role of the cation of the base used. It was

found that the potassium and lithium bases gave opposite

diastereoisomer ratios (Table 1, entries 7-11). In this

study, the respective potassium and lithium enolates were

trapped with TMS-Cl, and their configurations were determined

by 1H NMR. The trapped potassium enolates were found to

exist exclusively in the Z-configuration, while the trapped

lithium enolates were exclusively in the E-configuration.

Although the reasons for this difference were not given, the

assigned geometries are consistent with reports that (Z)-

lithium enolates are significantly more reactive than the

corresponding (E)-lithium enolates (101). From the assigned

enolate configurations (Figure 2-5), the facial selectivity

of alkylation for the (Z)-potassium enolate was explained by

a proposed a-ester chelation which gave a cyclic chelate.

The electrophile presumably attacks preferentially opposite

to the bulky protected nitrogen group. The (Z)-lithium

enolate, in which the metal ion cannot form a cyclic chelate

because of its (E)-geometry, assumes a hydrogen in the plane

conformation (Figure 2-5) that is attacked opposite the

bulky nitrogen group, giving rise to the opposite

diastereomer product (Table 1, entries 7-11).
















R02C'


Z-potassium
enolate


Z-lithium
enolate


Figure 2-5. Proposed potassium and lithium enolate
geometries for the alkylation of protected D-aspartic acid
derivatives with various electrophiles (Taken from reference
97).



The results obtained by Parr (102) using the benzyloxy-

carbonyl-N-protected L-aspartic acid derivatives with LiHMDS

and various nucleophiles (Table 1, entries 12-19) are in

agreement with these previously reported studies.

Exceptional diastereoselectivity was seen with several

electrophiles, leading to an initially proposed enolate

geometry similar to that seen by Seebach et al (98, 103),

Rapoport and Lubell (104), and Yamamoto et al (105) on

related systems. This proposed enolate involves chelation

control by the nitrogen anion formed to give the

corresponding (E) or (Z)-lithium enolate (Figure 2-4).

Presumably, electrophilic attack on the least hindered face

opposite the bulky a-ester group gives the product in

diastereomeric excess.


H

R2N








To investigate this idea, further studies by Parr (102)

were done to ascertain the effect of steric bulk at the a-

ester position of N-protected L-aspartic acid derivatives

with LiHMDS and tert-butyl bromoacetate (Table 1, entries 20-

22). It was demonstrated that increasing the size of the a-

ester substituent did little in determining the diastereo-

selectivity of the reaction. The relative stereochemistry of

the products formed were established using NOE difference

spectroscopy (106,107) and by examination of scalar coupling

constants (108,109). These results appear to rule out the

proposed nitrogen-chelated enolate, and favor a similar (Z)-

lithium enolate geometry as proposed by Humphrey et al

(Figure 2-5) (97).

While the studies by Parr (102) clearly demonstrated the

relative stereoselective and regiospecific preference for the

alkylation reactions, the absolute stereochemical assignment

for the alkylated products remained ambiguous. It was

determined that the enantiomeric purity of an alkylation

product could be determined by a chemical resolution and

subsequent 1H NMR study. From the results of this resolution

and stereochemical assignment, a tentative assignment of

stereochemistry could then be made to the other alkylated

products.

The synthesis of tert-butyl-2(S),3(R)-2-benzyloxy-

carbonylamino-3-carbomethoxy-glutarate 28 begins with the

protection of the commercially available L-aspartic acid, 3-

methyl ester 29 (Figure 2-6). Nitrogen protection with a









C02tBu
2COMe ,CO2Me [ |H
iii I iiij ZHN,,T
I -- I-> H" C02Me
HI
H3+N CO2- ZHN C02tBu CH2tBu

29 30 28


Figure 2-6. Synthesis of 1,5-di-tert-butyl-2(S)-(N-
benzyloxycarbonyl)amino-3(R)-carbomethoxypentanoate 28.
Z=C6H5CH20CO-. Reagents and conditions: i.) Benzyl
chloroformate, dioxane, H20, Na2CO3, RT, 70%; ii.)
isobutylene, conc. H2S04, RT, 45%; iii.) LiHMDS, tert-butyl
chloroacetate, THF, -78 C, 39%.


benzyloxycarbonyl group was done using literature methods
(110) to give the Cbz-N-protected acid, which was
subsequently subjected to isobutylene gas with catalytic H2S04

in methylene chloride (111) to give the N-protected diester

30. Alkylation using LiHMDS (2 eq.) and tert-butyl

chloroacetate in THF gave 28 as the single diastereomer, as

determined by 1H NMR.
Semi-empirical calculations were made using the AM1
Hamiltonian (112) as implemented in MOPAC 6.0 (113) on the

possible diastereomers of 28. This study indicated that the

2(S), 3(R) diastereomer is more stable than its C3 epimer by
2.5 kcal/mol. While the alkylation of aspartate is an

irreversible reaction which occurs under kinetic control,
these calculations suggest that the reaction also produces

the thermodynamically favored product.









CO2tBu C2tBu
.,H H H
H2N IP ii I N
28 11 H1" CO2Me ii1 P yHI, C02Me
C02tBu Me 0 CO2tBu
33 31

Siii


C02tBu
H H
Ph N N 0,
H1 III C02Me
Me 0 C02tBu
32


Figure 2-7. Synthesis of the (R)- and (S)-N-(a-methyl-
benzyl)urea triesters 31 and 32. Reagents and conditions:
i.) H2, Pd/C, MeOH, RT, 66%; ii.) (R)-a-methylbenzyl
isocyanate, THF, A; iii.) (S)-a-methylbenzyl isocyanate, THF,
A.


The enantiomeric purity of 28 was ascertained after
conversion to the (R)- and (S)-N-(a-methylbenzyl)urea tri-

esters 31 and 32 (Figure 2-7). Hydrogenation of 28 was
performed using hydrogen gas over catalytic palladium in
methanol (94,95) to form the free amine 33 (Figure 2-7).
The amine 33 was then acylated with either (R)- or (S)-a-

methylbenzyl isocyanate in THF (109,114) to form 31 and 32,
respectively. Observation of the diastereomeric methyl ester
singlets by 500 MHz 1H NMR spectroscopy in CDCI3 during

incremental additions of the opposite isomer demonstrated 31








and 32 to be of >95% purity (Figure 2-7). Hence 28 is

presumed to be of >95% purity.


Mapping the Aspartic Acid Binding Site of E. Coli Asparaqine
Synthetase B Using Substrate Analogs


The inhibitory effects of a series of aspartic acid

analogs on glutamine-dependent asparagine synthesis were

determined using recombinant, wild-type AS-B1 (115).

Standard reaction solutions in these experiments contained 1

mM glutamine, 1 mM ATP, 1 mM aspartic acid and 10 mM MgCI2 in

TrisHCl buffer at pH 8. Asparagine production was assayed by

measuring the amount of pyrophosphate formed under steady

state conditions (39,64).

Initial studies focused on the commercially available

compounds 34-39 (Figure 2-8). All of these compounds

exhibited only low levels of inhibition even at 10 mM

concentration. The inability of malic acid 36 to inhibit

asparagine synthesis demonstrates that the hydroxyl cannot

act as an isostere of the charged amino group, consistent

with results reported for the asparagine synthetase present

in RADA1, a murine leukemia (116). It therefore appears that

AS-B requires the presence of all of the ionizable groups in

aspartate for substrate binding. The significance of the 3-

carboxylate in recognition is underscored by the failure of

L-proline 39 to inhibit AS-B, within the limits of the assay,

at concentrations up to 50 mM.

1 Enzyme assay experiments performed by Dr. Susan K. Boehlein
(115).











CO2H R R R
C23 --O2H < R6


R 2 H3N rC02N- _CO2H
HH
H
34: RI=CO2H, R2=H 37: R3=OH, R4=H 18: R5=CO2H, R6=H
35: R1=H3N+, R2=H 38: R3=H, R4=OH 19: R5=H, R6=CO2H
36: RI=OH, R2=CO2H 40: R3=nPr, R4=H 39: R5=R6=H
41: R3=H, R4=Me
42: R3=Me,R4=H



Figure 2-8. Aspartic acid analogs 18, 19, 34-42 used in
inhibition studies of AS-B.



Having established that all three ionized functional

groups were essential to the recognition of aspartate by AS-

B, the P-functionalized aspartic acid analogs 18, 19, 40,

41, and 42 were then assayed2. Previous experiments by

Mokotoff et al. on asparagine synthetase isolated from

asparaginase-resistant Novikoff hepatomas (117) demonstrated

27% AS inhibition in synthetase activity in the presence of

the racemate of erythro-p-hydroxyaspartate 37 (Figure 2-8).

However, no reduction in AS-B activity was observed for 3-n-

propylaspartate 40, and the remaining analogs were weak AS-B

inhibitors (Table 2). Furthermore, there was a significant

effect of stereochemistry at C-3 upon the level of AS-B

inhibition. Again, this observation is consistent with

reports that threo-B-hydroxyaspartate 38 (Figure 2-8) cannot


2 Compounds 40-42 synthesized by Dr. Ian B. Parr (102).









TABLE 2 Inhibition constants for selected aspartic acid
analogs in the glutamine-dependent synthetase
activity of E. coli AS-B

Substrate Kisa Kiia
Inhibitor varied Inhibition pattern (MM) (mM)

41 Asp Non-competitive 18.0 >50
41 Gln Non-competitive 93.0 16.5
41 ATP Non-competitive 8.0 15.0
42 Asp Competitive 0.25 n.a.b
42 Gin Non-competitive 1.62 1.71
42 ATP Non-competitive 0.45 3.28
19 Asp Competitive 2.65 n.a.
19 Gin Uncompetitive n.a. 2.57
19 ATP Non-competitive 12.65 4.80
a Kis and Kii are the inhibition constants computed from
the intercepts and slopes, respectively, of the double
reciprocal plots of 1/Vo versus 1/[aspartate] obtained at
various concentrations of the inhibitor (118). b n.a. = not
applicable.



inhibit AS isolated from Novikoff hepatomas (117). The

constrained aspartate analog 18 gave an apparent KI of 60 mM.

In contrast, the constrained analog 19, differing only

in its C-3 configuration, was competitive with respect to

aspartate, having a KIs of 2.65 mM. A similar dependence of

inhibitory properties upon C-3 stereochemistry was also

observed for the methylated analogs 41 and 42 (Table 2).

Detailed kinetic analysis confirmed that 42 and 19 were

competitive only with respect to aspartate, suggesting that

these compounds did indeed bind to the same form of the

enzyme (probably the E.ATP complex) (56) as aspartic acid,








preventing the natural substrate from binding and taking part

in subsequent chemical transformations (118).

Given the conformationally restricted nature of 18 and

19, and the observation that these compounds are only

competitive with respect to aspartate, it is reasonable to

assume that 19 defines the bound conformation of aspartic

acid. In addition, the identical correlation between C-3

stereochemistry and the ability to inhibit AS-B observed for

the diastereoisomers of P-methylaspartate suggests that 42

adopts a bound conformation on the enzyme similar to that of

19. Both 42 and 19 possess a carboxyl group capable of

undergoing reaction with ATP to form reactive intermediates

cognate to 3-aspartyl-AMP 6 (Figure 1-7).

In the absence of detailed structural data upon the

complexes between AS-B and these aspartate analogs, we also

propose that 19 binds within the same pocket as 42,

especially given the correlation between C-3 stereochemistry

and inhibition for the aspartate analogs. It is reasonable

to suggest that aspartic acid binds, at least initially, to

AS-B in a conformation that is identical to that of the rigid

analog 19 (Figure 2-9A). In this shape, all of the polar

functionality is placed upon one face of the molecule, with

the hydrogen atoms defining a hydrophobic surface that makes

contact with the enzyme (Figure 2-9B). Given that the

methylated analog 42 also interacts with this site, there is

some flexibility in the protein that can be used to










(A) H C02- (B) H Arg

N C0-2' H- -
A+ C"N --

H0C H N-H

0 H
H C
H0 C02-
19 H3+N GC02

GO2-

H3+N C02- C2H H
H

H *H
H
A Hydrophobic pocket
Asp
Figure 2-9. (A) Newman projection of the bound conformation
of L-aspartate in the AS-B active site based on the
constrained analog 19. (B) Schematic model for the aspartic
acid binding site in AS-B. (Taken from reference 117).



accommodate the larger substitutent. On the other hand, the

inability of 40 to inhibit AS-B synthetase activity argues

that this pocket in the enzyme cannot be distorted

significantly.

There are now three examples of aspartate binding sites

for which the X-ray structures have been reported (119-121).

In all cases, arginine residues interact with the carboxylate

groups of aspartate, although the number employed is

variable. Given that both carboxylates are placed in close

proximity in our model of the bound aspartate conformation,

it is possible that only a single arginine residue will be

present in the binding site of AS-B. Interactions that








stabilize the protonated amine of L-aspartate, however, seem

less predictable. For example, the amino group is placed

over the face of a tryptophan ring in adenylosuccinate

synthetase, while in the bacterial aspartate receptor,

backbone carbonyl groups are used to bind the amine. Our

data suggest that while two of the three protons in aspartate

are placed within a well-defined pocket on the surface of the

enzyme, there is room in the site to accommodate a small

hydrophobic substituent in place of the pro-(R) hydrogen at

C-3 (Figure 2-9B).


Crystallization Studies of the Amidotransferases


To date, no X-ray crystal structure for E. coli

asparagine synthetase B is available. The X-ray crystal

structures are available for the Class II amidotransferases

E. coli glutamine PRPP amidotransferase (41), B. subtilis

glutamine PRPP amidotransferase (40), and the GAT domain of

glucosamine 6-phosphate synthase (42), as well as the Class I

amidotransferase GMP synthetase (43). However, neither of

the Class II amidotransferases which have been crystallized

have a high degree of homology with AS-B (39), making a

structural prediction by comparison difficult. Therefore,

obtaining crystals of sufficient quality for X-ray analysis

has become a priority for future studies on AS-B.

Recent approaches in the crystallization of

amidotransferase enzymes has been through the method of

isomorphous replacement (122). Proteins and nucleic acids








crystallize with a large amount (30-90%) of water in the unit

cell, so that there are aqueous channels in the crystals

(123). These channels provide routes along which solutions

of compounds containing heavy atoms may diffuse and interact

with side chains on the surface of the protein. If the heavy

atoms attach to the macromolecule in well defined positions

that are the same from unit cell to unit cell, intensity

differences due to the addition of the heavy atom will be

observed in the X-ray diffraction pattern (124). This method

allows one to derive phases for the "native" macromolecule

(no heavy-atom binding). The isomorphism between the

"native" protein structure and crystals soaked with the

appropriate heavy-atom salt or reactant is the basis of the

method of relative phase determination (123).

The method of isomorphous replacement with mercury ions

(Hg++) was used successfully in the structure determination of

the Class I amidotransferase GMP synthetase (125). Salts

containing mercury ions are common isomorphous replacement

compounds because they readily bind to the sulfur atoms of

cysteine residues (126). However, use of mercury salts has

not produced viable crystal structures of AS-B.

Studies by Tesmer et al. have successfully determined

the structure of the Class I amidotransferase GMP synthetase

by forming a product complex of AMP and pyrophosphate (PPi)









NH2 NH2

N

N N I II N N I

HO 4 HO/ 0
U ~ HO U

HO OH HO OH
43 44
Figure 2-10. Synthesis of 2-Iodoadenosine monophosphate (2-
I-AMP) 44. Reagents and conditions: i.) POCI3, H20, C6H5N,
CH3CN, 0 C, 34%.


with the enzyme (43). In this study, the adenine nucleotide

analog 2-iodoadenosine triphophate (2-I-ATP) was probed as a

possible heavy-atom derivative. While 2-I-ATP did bind to

the protein, it gave a substantially different structure than

the structure produced with AMP/PPi, excluding the purine

ring from the adenine-binding pocket.

In order to attempt to determine the crystal structure

of E. coli AS-B, the heavy-atom nucleotide analog 2-

iodoadenosine monophosphate (2-I-AMP) 44 has been synthesized

and crystallization experiments attempted. It is hoped that

2-I-AMP will give a product complex as seen with GMP

synthetase.

The synthesis of 44 begins with 2-iodoadenosine 433

(Figure 2-10). The monophosphorylation of the unprotected

nucleoside 43 was done with phosphoryl chloride in the

3 2-Iodoadenosine was a generous gift from Dr. V. J.
Davisson, University of Purdue.








presence of water and pyridine in acetonitrile, following the

procedure of Yoshikawa et al. (127,128). The reaction was

purified by anion exchange high performance liquid

chromatography (HPLC) using a water:methanol gradient to

obtain pure 44 in a 34% yield.

Experiments to crystallize AS-B as a product complex

with 2-I-AMP, to date, have not produced crystals of

sufficient quality for structural analysis4.


Conclusions

3-Alkylation of protected aspartate derivatives has

allowed the preparation of a number of stereochemically

defined aspartate analogs suitable for probing the molecular

features of the aspartic acid binding site in E. coli AS-B.

Further investigation of the alkylation reaction has

determined that the addition appears to proceed via a Z-

lithium enolate geometry, with complete stereochemical

control in some cases.

The use of substrate analogs has shown that AS-B appears

to be extremely selective and it is able to discriminate

between metabolites that have similar structures to aspartic

acid, which is clearly important in terms of cellular

metabolism. However, the protein residues that are

responsible for mediating this selectivity remain to be

defined by site-directed mutagenesis or X-ray

4 Crystallization experiments of 2-I-AMP 44 with E. coli AS-B
performed by Dr. Todd Larsen, University of Wisconsin-
Madison.





53

crystallography. Additional studies using heavy-atom labeled

substrates and products, such as 2-I-AMP, have failed to give

the structural information needed to aid in this assignment.

While it appears that AS-B can tolerate only relatively

minor structural alterations in the aspartic acid substrate,

these results suggest that functionalized aspartate analogs

can be developed that are selective, tight-binding AS

inhibitors.














CHAPTER 3
MODELING THE REACTION MECHANISM OF E. COLI ASPARAGINE
SYNTHETASE B USING COMPUTATIONAL APPROACHES


Introduction


There are currently two proposed nitrogen transfer

mechanisms for E. coli AS-B which do not involve free

ammonia. One mechanism involves nitrogen transfer through

formation of an imide intermediate (Figure 1-13), while the

other involves nitrogen transfer through prior formation of a

tetrahedral intermediate, followed by the nitrogen transfer

step (Figure 1-14). In order to acquire additional

information on the possible nitrogen transfer mechanisms,

detailed, high-level ab initio calculations have been done in

order to provide insight into the relative activation

energies involved in the possible reaction mechanisms. These

calculations include the effects of electron correlation

(129), which is important in modeling processes involving

bond formation and bond breaking. Because of the

impracticality of modeling these large systems using current

computational techniques, smaller model systems have been

used in this analysis. This chapter describes the

computational modeling of these two proposed mechanisms, as

well as the accepted ammonia-mediated mechanism. A detailed








analysis of the structural properties of simple amides and

imides is also described.


Structural Comparison of Amides and Imides



Computational Studies of Formamide


The formamide 45 and hydroxyimine tautomer 46 system

(Figure 3-1) and similar species have been studied

previously (130-133) as models for tautomerization in nucleic

acid bases. Many theoretical studies employing a variety of

methods have focused attention on the relative stabilities of

tautomers 46-49, and have examined the computed stabilities

as functions of choice of basis set (134) and treatment of

electron correlation (132). Several other workers (135,136)

have examined the predicted geometry of 45 as a function of

atomic basis set, in particular, the planarity (or lack

thereof) at the H2N center.

Studies by Wang et al. (137) on formamide 45 and the

lowest energy hydroxyimine tautomer 46 employed higher levels

of theory, using Dunning's "correlation consistent" polarized

valence double-zeta basis set (PVDZ) (138). Vibrational

frequency analysis and electron correlation techniques using

single and double excitations from the single configuration

reference function (CISD) method (139) and the second-order

M0ller-Plesset perturbation theory (MP2) method (140,141)

were also utilized in this study. Using these methods, the








0 0/H HO0

HNHb H-"N H-kN
I I I
Ha H H

45 46 47

HI-10 0/.,H-

H-"N ,,H 0--"'
H-^N- HA^NA

48 49


Figure 3-1. Formamide 45 and tautomers 46-49.


energy difference between 45 and 46 was calculated to be 12.1

kcal/mol.
In the present study on 45 and the hydroxyimine
tautomers 46-49, geometry optimizations were done at the SCF
level using Dunning double-zeta basis set (142) with

polarization functions (DZP) (143) for all the atoms in the
study. Vibrational analysis was used to verify that
optimized structures were potential energy minima, and to

compute the zero point energy correction. Single point
correlation energies were calculated using CCSD(T) methods
(144). These calculations were done using the ACES II ab

initio software package (145,146).

The SCF-optimized bond lengths and angles of formamide
and it's tautomers are given in Appendix B, Tables 7 and 8,
respectively. For 45, experimental structural data (147) are








included for comparison; analogous data for the hydroxyimine

tautomers 46-49 is not available. Total and relative

energies for 45 and tautomers are given in Table 3. The SCF-

level energy difference between 45 and the lowest energy

hydroxyimine tautomer 46 is calculated to be 12.6 kcal/mol
using the DZP basis. The 45--46 correlation energy

difference, which was obtained at the singles and doubles

coupled cluster (CCSD(T)) level of correlation, is 10.8

kcal/mol. Consideration of the zero-point vibrational
energies give a 45- 46 correlated-electronic-plus-zero-point

energy difference of 11.8 kcal/mol. This data indicates that

the gas-phase electronic endothermicity for the

tautomerization of 45 to the most likely formed tautomer, 46,

is 11.8 kcal/mol. This is the gas phase thermodynamic energy

required to tautomerize the amide bond of formamide, and it

gives an idea of the thermodynamic energy that will be

required of the AS enzyme. Relative energies for the other

tautomers were calculated in a likewise manner, and were also

corrected for zero-point energies using the SCF harmonic

vibrational frequencies (Appendix B).

The local harmonic vibrational frequencies of formamide,

45, as well as comparison to experimental data (148,149) are

given in Appendix B, Table 9. Included in this table are the

calculated vibrational frequencies of a series of deuterated

formamide structures, which were done to aid in the

assignment of the vibrational frequencies which were in

question. The SCF approximation often overestimates









Table 3. Total Energies (atomic units) for SCF and
Correlated Wave Functions and Relative Energies (kcal/mol)
for Formamide 45 and Tautomers 46-49

species SCFa CCSD(T)b Relative

Energiesc
45 -168.975 99 -169.534 60 0
46 -168.955 87 -169.517 34 11.796
47 -168.950 02 -169.512 34 17.917
48 -168.944 87 -169.506 99 15.064
49 1-168.949 14 -169.511 62 14.749
a Hartree-Fock SCF-level calculation using DZP basis set. b
With all electrons treated at the singles and doubles coupled
cluster level. c Relative energy calculated by subtracting
zero-point corrected CCSD(T) calculated energy of 45 from
each zero-point corrected CCSD(T) calculated energy and
converting to kcal/mol.



vibrational frequencies by 10%, especially for low-frequency

vibrations like the H2N out-of-plane bending motion (137).

Although anharmonicities were likely to be important, their

treatment was not included in this study. Therefore, caution

should be used when using these results. All the data in

Table 9 (Appendix B) pertain to the Cl symmetry case (i.e. -

all structures were calculated as planar).


Computational Studies of Formimide


In the proposed mechanism of E. coli AS-B which involves
nucleophilic attack of glutamine on P-aspartyl AMP 6 (Figure

1-13), the subsequent loss of AMP produces an imide

intermediate. This intermediate can exist as several

conformational isomers, with the imide proton (N-H) and the

















Figure 3-2. Formimide conformational isomers 50-52.


carbonyl oxygen(s) existing in either the cis or trans

orientation.

A search of the literature revealed that little work has

been done in calculating the expected form of formimide, or

N-formylformamide, and the relative energy between the "keto"

and "enol" forms. Radom, et al. (150), using a STO-3G basis

set SCF optimization (151), calculated the E,E (Figure 3-2)
conformational isomer, or conformer, 52 to be lower in

energy than either the E,Z (51) or Z,Z (50) conformers by

0.05 kcal/mol and 2.63 kcal/mol, respectively. Kupfer, et

al. (152), using a 3-21G basis set (155) SCF optimization
calculated the relative energy of 52 to be higher in energy

than 51 by 1.37 kcal/mol and lower than 50 by 5.16 kcal/mol

(Table 4).

Experimental results have not been in agreement either.

Steinmetz (156) used microwave spectroscopy to assign the

major conformer of formimide in the gas phase as the

asymmetric conformer 51. Noe, et al. (157) assigned the E,E

configuration 52 as the predominant conformer in acetone


o 0 0 Hb H H

H' N)H HANAO 0'N O
I aI I
H H H
50 51 52










o0H 0

Ha "'N Hb

53
OH Hb
O-H Hb

Ha N'-O


0 Hb

a I
H
57
Ha Hb



H


HAN
H0 0


HaHb

54


0 Hb

HA NA)NO',H
a


Ha Hb

OAN'--O


Figure 3-3. Formimide tautomers 53-60.


solution using NMR. A ratio of 85:15 (E,E : E,Z) was found

in the study.

Because of the conflict of data, both computational and

experimental, it was decided to determine the

relativeenergies of formimide and tautomers (Figure 3-3)

using high level ab initio calculations. The calculations on

the formimide system were done using the same methods as the

formamide system. Geometry optimizations were done at the








SCF level using the DZP basis set. Vibrational analysis was

used to verify that optimized structures were potential

energy minima, and to compute the zero point energy

correction. Electron correlation energies were obtained

using CCSD(T) methods. All structures were optimized as

planar, but upon scrutiny of vibrational analysis output

(Appendix C, Table 3), it was seen that the tautomers of 51

contained an imaginary vibrational frequency, indicating that

the optimized structures were not at a minimum.

The relative energy differences between the formimide

conformers 50-52 can be best seen in Table 4. Previous

calculations done on this system by Radom et al. (150) and

Kupfer et al. (152) are given for comparison. Data from a

second set of calculations on the three conformers of

formimide which uses the MBPT(2) gradient (153) while using a

TZ2P basis set (154) are also included in this table. The

calculations done in this work, which utilize the higher

degrees of theory, along with the larger basis sets, are the

most comprehensive calculations done on this system to date.

Note that in both sets of calculations on the formimide

conformers done in this study, the E,E conformer 52 was

calculated as the lowest energy structure in the gas phase.

The SCF-optimized bond lengths and angles of the

formimide conformers and tautomers are given in Appendix C,

Tables 10 and 11, respectively. Total and relative energies

for 50-60 are given in Table 5. As before, this relative

energy is the difference between the SCF optimized structures









Table 4. Calculated Relative Energies of Formimide
Conformers (kcal/mol)

0 0 0 H, H H

HANA H HAN'O' O NKO
H A H
50 51 52

STO-3G 2.63 0.05 0
3-21G 5.16 0 1.37
DZP CCSD[T] 6.40 0.51 0
TZP MBPT(2) 6.25 0.67 0


which were treated with all the electrons at the CCSD(T)

level done with the DZP basis set less the calculated energy

of the lowest energy formimide conformer, 52. Zero-point

corrections using the SCF harmonic vibrational frequencies

(Appendix C, Table 12) have also been made.

The local harmonic vibrational frequencies of the

formimide conformers are given in Appendix C: Table 12. As

before, to aid in the assignment of vibrational frequencies

in question, the calculated vibrational frequencies of a

series of deuterated formimide structures are also given.

All the data in Appendix C: Table 12 pertains to either the

C2v (50 and 52) or Cl (51) case.

Also of interest is the higher energy of the

hydroxyimine tautomer 53 (Figure 3-4). The higher energy of

53 is in accord with the experimental findings of Allenstein

et al. (158), who deduced from the gas-phase infrared

spectrum of formimide that the enol form is not present in









Table 5. Total Energies (atomic units) for SCF and
Correlated Wave Functions and Relative Energies (kcal/mol)
for Formimide Conformers and Tautomers 50-60

species SCFa CCSD(T)b MBPT(2)C Relative
~____________ ~____________Energiesd
50 -281.724 60 -282.625 80 -282.758 77 6.40


53 -281.714 75 -282.616 85 12.74
54 -281.695 32 -282.596 86 24.51
51 -281.736 58 -282.635 57 -282.767 65 0.51
55 -281.694 79 -282.597 38 24.02
56 -281.697 48 -282.599 01 23.02
57 -281.712 60 -282.611 62 15.41
58 -281.703 08 -282.602 77 20.74
52 -281.736 42 -282.636 29 -282.768 72 0
59 -281.708 98 -282.610 04 _____16.39
60 -281.697 67 -282.599 75 _____22.54


a Hartree-FocK SCF-level calculation using DZP basis set. D
With all electrons treated at the singles and doubles coupled
cluster level. c Second order many body perturbation level
calculation using TZ2P basis set. d Relative energy
calculated by subtracting zero-point corrected CCSD(T)
calculated energy of 52 from each zero-point corrected
CCSD(T) calculated energy and converting to kcal/mol.



significant amounts, as well as the work of Noe et al. (157),

who obtained no evidence for the presence of 53 from NMR

experiments with formimide (in acetone solvent). The

decreased tendency of imides to tautomerize, as compared with

P-ketones, is not surprising in view of the fact that

conjugation is significant even in the "keto" form of imides.


















Figure 3-4. Formimide tautomer 53.



Modeling the Reaction of AS-B


In the proposed imide-mediated nitrogen transfer

mechanism of asparagine synthetase (Figure 1-13), the

breakdown of the imide intermediate occurs with nucleophilic

attack of a thiolate anion to form the tetrahedral

intermediate 17. The resultant loss of asparagine anion

gives a thioester intermediate 12, which is subsequently

hydrolyzed by water to release glutamate. In the alternate

proposed mechanism (Figure 1-14), the primary amine of the

tetrahedral intermediate 10 formed from thiolate attack on
glutamine undergoes nucleophilic attack on 3-aspartyl AMP 6

to form intermediate 17. Because both of these proposed

mechanisms involve the nucleophilic attack of the thiolate

anion of Cysj, it was determined that this reaction should be

analyzed in more detail using computational methodology.


Preliminary Studies


Previous studies by Kollman et al. (159,160) involved

modeling the attack of hydroxide (OH-) and thiolate (SH-)


0,H-. 0 O"
or
HAN H H N

53a 53b








anions on formamide as a model for proteolytic enzymes. In

these studies, no local minimum for a tetrahedral adduct

structure involving thiolate was found, in contrast to

similar calculations with hydroxide. Instead, only an ion-

dipole complex minimum was found. Further studies were done

by the same authors (161) to examine the possibility for

concerted proton transfer in the sulfhydryl protease, but no

definite mechanism was determined.

Before carrying out similar studies on the formimide/

thiolate system, it was necessary to duplicate the previous

work in order to "calibrate" the methods being used.

Preliminary semi-empirical studies were made using MOPAC 6.0

(113) as implemented on the CACHE Scientific worksystem.

Using the geometry optimized structures obtained from this

AM1 (112) calculation as an input structure for the ab initio

calculation, a SCF geometry optimization was done using a 4-

31G basis set (162), as implemented in the ACES II ab initio

software package.

All degrees of freedom were relaxed during the

optimization calculations except the C-S distance. This

value was constrained to several values to simulate

nucleophilic thiolate attack on the carbonyl carbon of

formamide. The selected C-S distances were: 1.5, 2.0, 2.63,

2.68, 3.0, 3.87, 4.0, and 6.0A. As in the case of Kollman

(160), it was necessary to include further constraints on the

6.0k C-S distance calculation, in order to insure

nucleophilic attack was under investigation, and not simply a





















Figure 3-5. Thiolate attack on Formamide.



proton abstraction. These constraints included the CSH angle
(990), the NCS angle (90 ), as well as the NCSH dihedral

angle (0 ) (Figure 3-5).

Once it was established with the formamide system that

the calculation method being used was valid, and that

previous results could be duplicated, the investigation of

nucleophilic attack of thiolate on formimide was done using

the same procedure. The first formimide isomer chosen for

the calculation was the Z,Z conformer 50, for a couple of

reasons. First, 50 was calculated as the highest energy

conformational isomer, which may be the most likely choice

for an enzymatic-type reaction. Also, with larger groups on

the imide, namely the Asp and Gln residues, the Z,Z structure

50 should be the favored isomer (74). However, the

nucleophilic attack of thiolate on the E,Z conformer 51 was

also done, for comparative purposes.

Semi-empirical calculations, as well as a few trial ab

initio calculations, showed that, as in the case of


0
d = C-S distance
H)/H e = C-S-R angle
H N = N-C-S angle
dR
d:,^R [= H, CH3
E) -!
iR I-*









0 0 0 H

a d = C-S distance
I I
H N H H N = C-S-R angle
di I
d d I = N-C-S angle
; R = H, CH3
i I





Figure 3-6. Thiolate attack on Formimide.



formamide, further constraints would be needed to insure

nucleophilic attack of the thiolate anion on the carbonyl

carbon of formimide was taking place, not proton abstraction

of the relatively acidic N-H. The CSH bond angle and the NCS
0 0
bond angle were constrained to 99 and 90 respectively

(Figure 3-6). It was determined that the NCSH dihedral

angle constraint used in the formamide system was not

necessary in the formimide system.

Further studies were done to investigate the nature of

the attacking nucleophile. In AS-B, the attacking

nucleophile is the N-terminal cysteine residue. For this

reason, it was felt that perhaps the thiolate (SH-)

nucleophile used in these calculations is not the best

choice. Therefore, the calculations were repeated using the

same procedure with methylthiolate (CH3S-) as the attacking

nucleophile. Methylthiolate was chosen because it

incorporates an alkyl group so as to be more similar to the









Table 6. Ab initio (4-31G) Calculated Relative
Energies (kcal/mol) for the Nucleophilic Attack of
(Methyl)Thiolate on Formimide (gas phase)a

C-s 50 51 50 51
distance with with with with
(A) -SH -SH CH3S- CH3S-
1.50 _____ +45.61
2.00 _____+1.13 -3.65 -3.63
2.63 -18.14 -17.86 -17.19 -18.21
2.68 -19.24 -19.32 -19.26 -18.99
3.00 -22.42 -26.15 -25.61 -25.63
3.45 -29.10 -30.52 -30.63 -29.75
3.87 -26.45 -26.61 -27.02 -25.96
4.00 -25.06 -24.94 -25.45 -24.35
6.00 -9.40 -8.21 _____-8.06
a Calculated energies are not corrected for zero-point
energies.



natural cysteine residue, while still being small enough to

be a feasible ab initio calculation.

The ab initio energies calculated for the attack of

thiolate and methylthiolate on the two conformers (50,51) of

formimide are shown in Table 6. The corresponding semi-

empirical calculations are shown in Appendix D: Table 13.

As can be seen, the energies for the thiolate and

methylthiolate systems are comparable, indicating that the

thiolate and methylthiolate nucleophiles behave similarly.

The shift of calculated minima from 3.0 A for the semi-

empirical calculation (Appendix D: Table 13) to 3.45 A

(Table 6) for the ab initio calculated structure is

presumably a function of the more complete electronic








contributions calculated using ab initio methods. From these

studies, it was determined that replacing the hydrogen of

thiolate with the methyl group of methylthiolate gave little

difference in the energetic calculated for each system.


Methylthiolate Attack Calculations


Once the preliminary studies were completed, a

systematic computational study of the nucleophilic attack of

methylthiolate on formamide and formimide was undertaken.

These studies were done using the angle constraints outlined

above. A comparative study on the effect of protonating the

carbonyl oxygen of formamide and formimide with the

corresponding nonprotonated species was also done (Figure 3-

7). For comparison, the protonated studies were done with

the NCOH dihedral angle initially set at 0 or 180, but this

angle was not constrained.

Semi-empirical studies were made using MOPAC 6.0 as

implemented on the CACHE Scientific worksystem. Using the

geometry optimized structure obtained from the AM1

calculation as the input structure for the ab initio

calculation, a SCF geometry optimization was done using a 6-

31G basis set (163), as implemented in the GAMESS ab initio

software package (164,165). Upon locating the stationary

points, the local harmonic vibrational frequencies were

calculated and zero-point energy corrections were obtained.

The second-order Moller-Plesset (MP2) perturbation theory

method was used to evaluate the single point electron























Figure 3-7. Methylthiolate attack on Protonated Formamide
45 and Formimide 50,51.



correlated energies of the species using a 6-31G** basis set

(166,167).

The overall reaction profile of the nucleophilic attack

of methylthiolate on formamide 45 is shown graphically in

Figure 3-8, and the total and relative energies for the

complexes are given in Appendix D, Table 14. The corre-

sponding semi-empirical calculated reaction profile can be

found in Appendix E. A gradual decrease in energy occurs as

the sulfur-carbon distance is reduced from infinity to -4 A.

The energy subsequently steadily increases as the sulfur-

carbon distance is further reduced. As before, this minimum

at 4.0 A corresponds to an ion-dipole complex. No local

minimum for a tetrahedral adduct structure involving

methylthiolate was found for nonprotonated formamide.

In the case of protonated formamide 45, a local minimum

for a tetrahedral adduct structure was found at 1.75 A, and a

higher energy structure was found at 2.5 A. While this


( H ) 0
0
,d = C-S distance
H Z 'R 0 = C-S-CH3 angle
H N = N-C-S angle
d I = N-C-O-H+ dihedral
R = H, CHO


CH3







































Figure 3-8. Ab initio calculated gas phase reaction
coordinate for methylthiolate ion + formamide 45.



higher energy structure is not a true transition state, as no

saddle-point analysis was done to find the true transition

state, presumably the complex at 2.5 A closely resembles the

transition state. Therefore, the difference in relative

energies can be calculated and used to predict the gas phase

energy of activation (Eact) for the nucleophilic addition of

methylthiolate to protonated 45. From this calculation, Eact

for the nucleophilic addition of methylthiolate to protonated








45 is 21.7 kcal/mol. The energy difference between the

higher energy structure at 2.5 A and the tetrahedral adduct

local minimum structure is 23.9 kcal/mol.

As seen in Figure 3-8, the relative energies were

calculated as the energy of the respective reaction

coordinate structure less the energies of the isolated

reactants. The shift between the nonprotonated and

protonated reaction coordinate paths occurs because the

energies of the protonated species have not been corrected

for gas-phase proton affinities (161,168).

The overall reaction profile of the nucleophilic attack

of methylthiolate on Z,Z formimide conformer 50 is shown

graphically in Figure 3-9, and the total and relative

energies for the complexes are given in Appendix D, Table 15.

The corresponding semi-empirical calculated reaction profile

can be found in Appendix E. As in the case of formamide, no

local minimum for a tetrahedral adduct structure involving

methylthiolate was found for nonprotonated formimide 50,

rather, an ion-dipole complex minimum was found at 4.0 A.

For protonated formimide 50, the structures along the

reaction coordinate were initially protonated with the HOCN

dihedral angle at 0 and 180. For most points along the

reaction coordinate, these structures optimized to identical

or nearly identical structures, however, the local minimum

for a tetrahedral adduct structure was found at 1.75 A and

2.0 A for the 0 and 180 initially-protonated species,

respectively. A higher energy "transition-state-like"











i 100 -----
-0- tt imide nonprot
---- tt imide protO
50
^ 50 -E-C tt imide protl80



C0,

CU
4>
10


ID -50
4-I

0
1-100
0

(U

150



U -200
a 1 2 3 4 5 6 7
I C S distance (A)


Figure 3-9. Ab initio calculated gas phase reaction
coordinate for methylthiolate ion + formimide 50.



structure was found at 3.0 A for both species. From these

calculations, the gas phase Eact for the nucleophilic addition

of methylthiolate to protonated 50 is 5.7 and 6.5 kcal/mol

for the 0 and 180 initially-protonated species,

respectively. The energy difference between the higher

energy structure at 3.0 A and the tetrahedral adduct local

minima structures is 36.8 and 39.7 kcal/mol, respectively.

The overall reaction profile of the nucleophilic attack

of methylthiolate on formimide conformer 51 is shown





74




S100-
0
--C)0- tc imide nonprot
-""- tc imide protO
50
S5-0--- tc imide protlO80


0



C -50
o


"a-I00
0
r4-100-
(D
0





-150


--200 .. .......... .......
S1 2 3 4 5 6 7
(D
C S distance (A)


Figure 3-10. Ab initio calculated gas phase reaction
coordinate for methylthiolate ion + formimide 51.



graphically in Figure 3-10, and the total and relative

energies for the complexes are given in Appendix D, Table 16.

The corresponding semi-empirical calculated reaction profile

can be found in Appendix E. Again, no local minimum for a

tetrahedral adduct structure involving methylthiolate was

found for nonprotonated Z,E formimide conformer 51, just the

ion-dipole complex minimum at 4.0 A. One would not expect

the orientation of the nonreacting-carbonyl group of

formimide to cause a substantial difference in the energies








of nucleophilic addition of methylthiolate. This is true in

the case of the nonprotonated species, as both 50 and 51 gave

nearly identical results. However, while a local minimum for

a tetrahedral adduct structure was found at 2.0 A, as in the

Z,Z conformer 50, the higher energy "transition-state-like"

structure was found at 2.5 A. The reason for this shift can

be attributed to the lack of intramolecular hydrogen bonding

which can exist between the carbonyl oxygens in 50, but

cannot happen in 51. This is similar to the intramolecular

hydrogen shift seen in formimide tautomer 53 in the initial

structural studies done on formimide (Figure 3-4). From

these calculations of methylthiolate + protonated 51, the gas

phase Eact for the nucleophilic addition is 16.1 kcal/mol.

The energy difference between the higher energy structure at

2.5 A and the tetrahedral adduct local minimum structure is

52.7 and 39.3 kcal/mol for the 0 and 180 initially-

protonated species, respectively.


Tetrahedral Intermediate Breakdown Calculations


Once the thiolysis studies were completed, a systematic

computational study of the breakdown of the tetrahedral

intermediates was done. In this case, the carbon to nitrogen

bond was constrained at different values to simulate the

lengthening and eventual bond breakage to form the thioester

intermediate. Again, semi-empirical studies were made using

MOPAC 6.0, and the geometry optimized structure obtained from

the AM1 calculation was used as an input structure for the ab








initio calculation. As before, a SCF geometry optimization

was done using a 6-31G basis set, as implemented in the

GAMESS ab initio software package. Upon locating the

stationary points, the local harmonic vibrational frequencies

were calculated and zero-point energy corrections were

obtained and MP2 calculations were done using a 6-31G** basis

set to evaluate the single point electron correlated energies

of the species.

The overall reaction profile of the breakdown of the

tetrahedral intermediate formed by the nucleophilic attack of

methylthiolate on formamide 45 is shown graphically in

Figure 3-11, and the total and relative energies for the

complexes are given in Appendix D, Table 17. The

corresponding semi-empirical calculated reaction profile can

be found in Appendix E. In the series of breakdown

calculations on formamide 45, the nitrogen was protonated to
-+NH3. For the nonprotonated carbonyl species, a gradual

decrease in energy occurs as the nitrogen-carbon distance is

increased from 1.35 A to 3.0 A. The energy subsequently

increases slightly as the nitrogen-carbon distance is further

lengthened. However, no local minimum for a tetrahedral

adduct structure was found for nonprotonated formamide, and

the minimum at 3.0 A corresponds to an ion-ion complex.

For the breakdown of the tetrahedral intermediate formed

from methylthiolate attack on protonated formamide 45, two

local minima were found on the reaction coordinate (Figure

3-11). The minimum found at 1.5 A corresponds to the





77




H -300
o

'-4 _________________________
o -0--- amide neut
S-320 amide protO
-D--- amide protl80

4-)
d -340

ID

0 -360

CO

0
4..)

-380


(D

N -400 .... ....... ........ "'.
1 2 3 4 5 6 7
i^C N distance (A)



Figure 3-11. Ab initio calculated gas phase reaction
coordinate for breakdown of methylthiolate ion + formamide 45
tetrahedral intermediate.



tetrahedral intermediate. Lengthening the nitrogen-carbon

bond gives a higher energy structure at 2.5 A, which

resembles the transition state for the breakdown of this

complex. A second minimum is seen at 3.0 A and 3.5 A, for

the protonated species at 180 and 0, respectively. While

both of these species correspond to an intermolecular

complex, the 00 protonated species (Figure 3-6) has involved

a proton exchange from the carbonyl to the leaving ammonia








group, and results in a shift of the minimum. From these

calculations, Eact for the breakdown of the tetrahedral

intermediate formed from the nucleophilic attack of

methylthiolate on protonated 45 is 10.7 kcal/mol. The energy

difference between the higher energy structure at 2.5 A and

the second intermolecular complex minima at 3.0 A and 3.5 A

for the 180 and 0 protonated species is 23.4 and 28.9

kcal/mol, respectively.

The overall reaction profile of the breakdown of the

tetrahedral intermediate formed by the nucleophilic attack of

methylthiolate on formimide 50 is shown graphically in

Figure 3-12, and the total and relative energies for the

complexes are given in Appendix D, Table 18. The

corresponding semi-empirical calculated reaction profile can

be found in Appendix E. For this species, a tetrahedral

intermediate structure is seen at 1.35 A, with a higher

energy species at 2.0 A. A second minimum is found at 3.0 A,

which corresponds to the intermolecular complex. The

activation barrier calculated for the breakdown of the

nonprotonated formimide 50 is 25.5 kcal/mol. The energy

difference between the higher energy structure at 2.0 A and

the second intermolecular complex minima at 3.0 A is 2.2

kcal/mol.

For the breakdown of the tetrahedral intermediate formed

from methylthiolate attack on protonated formimide 50, the

higher energy structure corresponding to the transition state





79




100

0
---O--- tt imide nonprot
o 50 -4- tt imide protO


o-:
L1)
4 0



Id
C -50
4-
0

-100
0



-150

I4-

S-200 .... ... . .. ..

S1 2 3 4 5 6 7
3 C N distance (A)



Figure 3-12. Ab initio calculated gas phase reaction
coordinate for breakdown of methylthiolate ion + formimide 50
tetrahedral intermediate.



was shifted to 1.75 A, as compared to 2.0 A for the

unprotonated calculation (Figure 3-12). From these

calculations, Eact for the breakdown of the tetrahedral

intermediate formed from the nucleophilic attack of

methylthiolate on protonated 50 is 17.7 kcal/mol, which is

significantly lower than that seen for the nonprotonated

species. A second minimum is found at 3.0 A, which

corresponds to a intermolecular complex. The energy











100-

I--I
0
&)


0-1
50


4-)

fl
'd


(D -50-
0P 0






(D
H

-rj

2 -50
+4


(t?
r-15




-
'rb -200 -

rz


0 tc imide nonprot
tc imide protO
- tc imide protl80


2 3 4 5 6 7
C N distance (A)


Figure 3-13. Ab initio calculated gas phase reaction
coordinate for breakdown of methylthiolate ion + formimide 51
tetrahedral intermediate.



difference between the higher energy structure at 1.75 A and

the second intermolecular complex minimum at 3.0 A is 18.2

kcal/mol. The stability of this second complex, as compared

to the nonprotonated species, is due to an intermolecular

proton transfer which has occurred between the protonated

carbonyl of the forming thioester, and the amide anion

leaving group.








The overall reaction profile of the breakdown of the

tetrahedral intermediate formed by the nucleophilic attack of

methylthiolate on nonprotonated and protonated formimide 51

is shown graphically in Figure 3-13, and the total and

relative energies for the complexes are given in Appendix D,

Table 19. The corresponding semi-empirical calculated

reaction profile can be found in Appendix E. No stable

tetrahedral species was found for either the nonprotonated or

protonated species. However, a stable intermolecular complex

minimum was found at 2.5 A and 3.0 A for the nonprotonated

and protonated species, respectively.


Conclusions


The point of these calculations is to assess the

feasibility of the two proposed mechanisms of E. coli AS-B

which preclude the use of ammonia. The calculations on the

thiolysis of formamide and formimide suggest that a proton

transfer from an active site residue may be necessary prior

to or concerted with the nucleophilic attack of CysI on the

reaction substrates or formed imide intermediate. This

result is in agreement with that seen in similar studies

performed on the thiol protease papain (161). Moreover, the

activation barriers calculated for the thiolysis reaction

were significantly lower for the Z,Z conformational isomer of

formimide than for formamide.

The activation barriers calculated for the breakdown of

the resultant tetrahedral intermediates are similar for both








formamide and the Z,Z conformational isomer of formimide.

This would seem to indicate that the energetic involved in

the loss of the amide anion is very similar to that needed

for the loss of ammonia. Moreover, the activation barriers

calculated for the resultant breakdown of the tetrahedral

intermediates formed was greater than the activation barriers

calculated to reform the starting materials. This

observation is consistent with the mechanisms of other

protease enzymes that utilize sulfur as the nucleophile

(161,169).














CHAPTER 4
CHEMOENZYMATIC SYNTHESIS OF L-4,4-DIFLUOROGLUTAMIC ACID


Introduction



Historical Perspective


The synthesis of 9a-fluorohydrocortisone acetate 61

(Figure 4-1) by Fried and Lobo (170) in 1954 represents the

first significant report of a successful application of

selective fluorination to modify biological activity. The

publication served to ignite a new era for medicinal and

biological chemists to use fluorine as a substituent to

modify reactivity in a variety of biochemically interesting

molecules.

The attractiveness and utility of fluorine as a

substituent in these molecules results from the pronounced

electronic effects that can occur, without any substantial

steric penalty. With its small Van der Waals radius (1.35

A), fluorine closely resembles hydrogen (Van der Waals radius

1.20 A), while the carbon-fluorine bond length, 1.39 A, is

comparable to that of the carbon-oxygen bond, 1.43 A (171).

The Pauling electronegativity of fluorine (4.0 vs. 3.5 for

oxygen) can have a pronounced effect on the electron

distribution in the molecule, affecting the acidity or
























Figure 4-1. 9a-Fluorohydrocortisone acetate 61.



basicity of the neighboring groups, dipole moments within the

molecule, and overall reactivity and stability (172). Also,

fluorine can function as a hydrogen bond acceptor because of

its available electron density (173,174).


Biological Studies


Fluorinated analogs of naturally occurring biologically

active compounds often exhibit unique physiological

activities. For example, the introduction of fluorine into a

pharmacologically active substance often leads to the

development of more potent agonists or antagonists (175).

Fluorine-containing amino acids have been synthesized

and studied as potential enzyme inhibitors and therapeutic

agents. Specifically, P-fluorinated amino acids have been

found to be irreversible inhibitors of certain pyridoxal

phosphate-dependent enzymes (176,177). a-Monofluoromethyl

and difluoromethyl amino acids have been recognized as potent








enzyme-activated irreversible inhibitors of parent a-amino

acid decarboxylases (178,179). Other fluorinated amino acids

have been found to exhibit antibacterial and cytotoxic

activity (180). Some of these fluorinated amino acids may

also be useful in the treatment of central nervous system

disorders (172,181).
Fluorinated glutamic acid analogs have been used to

probe several glutamic acid metabolic pathways. L-3-

Fluoroglutamic acid has seen extensive use as an inhibitor of

pyridoxal phosphate dependent enzymes that use glutamic acid

as a substrate (177). These enzymes include transaminases,

glutamate racemase, and glutamate decarboxylase. Racemic 4-

fluoroglutamic acid has been studied as a potential

antimetabolite in certain strains of Mycobacterium

tuberculosis (182). While 4-fluoroglutamate alone proved to

be inefficient as a cancerostatic agent (183), its derivative

fluoromethotrexate 63 (Figure 4-2) has shown promising

cancerostatic activity (184,185). 4-Fluoroglutamic acid has

also become a focus of interest in studies of its effects on

central neuronal mechanism (181,186,187).

4-Fluoroglutamic acid, 3,3-difluoroglutamic acid, and

4,4-difluoroglutamic acid have been used to probe the role of

glutamic acid in folate-dependent one carbon biochemistry

(188-191). The predominant forms of intracellular folate

cofactors exist as pteroyl y-glutamyl polypeptides (192),

which contain a long y-glutamic acid polypeptide chain

(peptide formation involves the y-carbonyl of glutamic acid,









0 C02H


Figure 4-2. Methotrexate 62 and fluorinated methotrexate
analogs 63,64.


rather than the normal a-carbonyl-peptide linkage). One of

the more important folate cofactors is N5,N10-methylene-

tetrahydrofolate (N5,N10-methylene-THF), which is critical for

the biosynthesis of deoxythymidine nucleotides by the enzyme

thymidylate synthase. It is also essential for the

biosynthesis of inosine monophosphate, the precursor for both

adenosine and guanosine nucleotides, as well as the amino

acids histidine and methionine (193).

In the course of deoxythymidine monophosphate synthesis,

N5,N10-methylene-THF is oxidized to dihydrofolate (DHF). DHF

is then recycled back to N5,N10-methylene-THF by reduction of

DHF to THF by the enzyme dihydrofolate reductase, then

hydroxymethylation of THF by the enzyme serine hydroxymethyl

transferase. Inhibition of dihydrofolate reductase quickly

results in all of a cell's limited supply of THF being

converted to DHF by the thymidylate synthase reaction.

Therefore, inhibition of dihydrofolate reductase not only


CH3 R2
62: RI=R2=H
63: Ri=F, R2=H
6 4: Ri=R2=F








prevents synthesis of the nucleotide deoxythymidine

monophosphate, but also blocks all other THF-dependent

biological processes. Thus, dihydrofolate reductase offers

an attractive target for chemotherapeutic agents.

Methotrexate 62 (Figure 4-2) is an excellent inhibitor

of dihydrofolate reductase, and it has proven to have wide

potential utility as an antitumor agent (194). However,

extended use can give problems because of its relatively high

toxicity (190). Introduction of fluorine into the amino acid

moiety (Figure 4-2) enhances the acidity of the y-carboxylic

acid group, and thereby diminishes the toxicity of

methotrexate (180). This enables the fluorinated analogs to

be used in high-dose treatment regimens (195).

This enhancement of activity caused by the introduction

of fluorine offers several potential uses in elucidating

mechanistic information about E. coli AS-B (196). The

hydrolysis of y-fluorinated L-glutamine analogs, for instance,

should occur very quickly with AS-B, due to the enhanced

electrophilicity of the neighboring carbonyl carbon (195).

If this chemistry step is sufficiently rapid, one can obtain

mechanistic information for AS-B based solely on

conformational changes of the enzyme that occur during the

course of the reaction.

Formation of the y-fluorinated L-glutamine analogs

needed for this type of study require the development of a

viable synthetic route for the formation of necessary

precursors. This study describes efforts made in the












0

ICF2 F OMe

66


0 0
AX
F OMe 6 8Ph6

F OSiMe3i

67


Siii


F F


MeO'


F F


V


H3+N


Figure 4-3. Stereospecific synthesis of L-4,4-
difluoroglutamic acid 65. Taken from reference 197.
Ph=C6H5-. Reagents and conditions: i.) Zn, TMS-Cl, CH3CN;
ii) acryloyloxazolidinone 68, RT, 15hr; iii.) Bu2BOTf,
iPr2NEt, CH2C12, NBS; iv.) NaN3, DMF; V.) a. LiOH, THF; b. H2,
Pd(C), EtOH.



development of synthetic routes for the stereospecific


synthesis of L-4,4-difluoroglutamic acid.










Synthesis of L-4,4-Difluoroglutamic Acid


A stereospecific synthesis of L-4,4-difluoroglutamic

acid 65 has been reported by Kitagawa and coworkers (197)

(Figure 4-3). Their method, however, requires a

commercially unavailable starting material, methyl

difluoroiodoacetate 66, which is costly to prepare (198,199).

The key step in this asymmetric synthesis involved a Michael

addition of the 2,2-difluoroketene silyl acetal 67 to the

chiral oxazolidinone 68. The fluorinated oxazolidinone 69

was then stereoselectively brominated through a boron enolate

and NBS (200). The bromide 70 was subsequently converted to

the azide 71 with inversion of stereochemistry. The

oxazolidinone group of 71 was then removed by hydrolysis, and

the azide group was reduced to the amine through catalytic

hydrogenation with palladium to give 65.

Because of the unavailability of the iodoacetate 66, an

alternate route for the synthesis of 65 must be taken. At

first glance, the synthesis of 65 seems to be analogous to

the stereospecific synthesis of L-4-fluoroglutamic acid 72

reported by Hudlicky and coworkers (Figure 4-4) (201,202).

This synthesis takes advantage of a relatively cheap and

readily available starting material, 4-hydroxyproline. Using

the appropriate starting material, the authors were able to

make all four stereoisomers of 4-fluoroglutamic acid. The











N:
/ C02H
H


HO_


N~/ co2Me


ii


F



/ CO2Me
Ac


F

0 / co2H
N/ C02H


F

i v O N=
- C02Me
N/ 'CO2Me


Figure 4-4. Stereospecific synthesis of (+)-threo-L-4-
fluoroglutamic acid 72. Taken from reference 201. Ac=CH3CO-
. Reagents and conditions: i.) a. Ac20O, AcOH; b. CH2N2; ii)
DAST; iii.) Ru02, Na0IO4, H20; iv.) conc. HC1.


key step in this sequence involved the oxidation of the
fluoroproline 75 (Figure 4-4) to give the fluorinated
pyrrolidine-5-one 76. Subsequent hydrolysis in concentrated
hydrochloric acid afforded the desired diastereomer 72.
Tsukamoto and coworkers (191) attempted to synthesize L-
4,4-difluoroglutamic acid 65 using similar reaction
conditions as in the reported stereospecific synthesis of
monofluorinated glutamate 72. In this study, the 4-
oxoproline derivative 78 (Figure 4-5) was successfully
converted to the difluoroproline derivative 79 by
(diethylamido)sulfur trifluoride (DAST)-mediated fluorination









0 F F
F [0] F
DAST
< 1 ~ D --- N N
/ CO2Me/ C02Me / CO2Me
tBoc tBoc tBoc
78 79 80


Figure 4-5. Attempted synthesis of L-4,4-difluoroglutamic
acid 72 through oxidation of N-protected difluoropyrrolidone
79. Taken from reference 191.



(203). Oxidation of 79 with ruthenium tetraoxide for

extended reaction periods failed to yield the oxidized

pyrrolidinone 80, presumably as a result of the strong

electron-withdrawing effect of the two adjacent fluorine

atoms (191).

An alternate route for the synthesis of L-4,4-

difluoroglutamic acid 65 involves the resolution of the

racemic DL-4,4-difluroglutamic acid 81. While this route may

not be as elegant as an asymmetric synthesis, it should

produce the target enantiomer desired.


Synthesis of DL-4,4-Difluoroqlutamic Acid


The proposed synthetic route for the preparation of L-

4,4-difluoroglutamic acid 65 is illustrated in Figure 4-6.

The synthesis of 81 has been reported by Tsukamoto and

coworkers (191). The key step in this synthesis is the








F F F
F CO2H F CO2H F CO2H
H,+N N v H3+N

C02 CO2H C02
65 86 81


\1
F F
O2N CO2Et F 'oCONEt2 F %CONEt2
+ 02N o C H2N
F FV T
H CONEt2 CO2Et CO2Et

0 82 84 85


Figure 4-6. Retrosynthetic analysis for the formation of L-
difluoroglutamic acid 65 starting from the difluorinated
malonaldehydic acid derivative 82 and ethyl nitroacetate.


Knoevenagel condensation of the ethyl nitroacetate anion to
the difluorinated malonaldehydic acid derivative 82 (Figure
4-7) to give the fluorinated nitroalcohol 83. Subsequent
transformations of 83 gave racemic 81 in a 15% overall yield.
While a leading method for the synthesis of difluoro-
methylene-containing compounds has been the fluorination of
ketones with DAST, in contrast to the reaction with alcohols,
DAST-mediated fluorination of ketones generally requires
higher temperatures and reaction times, and often gives low
yields of the desired product (191). This synthetic approach