I;
'1'." .V“
..r . yuan... 71
cm}!- L. D;
I
J. 3 ‘
'* rig... .5...
n .vlt 3:46

.1.
. its .. (as . :6
iv! 9. r

‘ 1...
5303.1.
.1153}? it.

.1 . 7.
titties»:

 

 

 

 

 

 

 

 

 

1.. .l

.q‘l.4.. ‘..

L. . :1 I . x . .
.rrfi....,sf. 5. x3 ... ‘ I ‘cw
. . Pa » -2“ .. I- . . .14.». fl... 533$. . .

. P n‘ p . .
1:. It?!» :fislvl ...l.|r

 

I HIGAN TATE UNIVERSITY LIBRAR

"lullrnlziiwlmlm“tumult ‘

3 1 93 00895 2990

 

 

 

 

 

 

This is to certify that the

dissertation entitled

An Oxidation State Study of Desulfovibrio
vulgaris Flavodoxin by X-ray Crystallography
presented by

William Watt

has been accepted towards fulfillment
of the requirements for

Ph . D . degree in Chemist rv

await

Major professo}

Date Mn] 2; [Sq 0-

MS U i: an Affirmative Action/Equal Opportunity Institution

0-12771

 

       
   

——'V

W

LIBRARY

“hm“ State
University

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or betore date due.

II DATE DUE DATE DUE DATE DUE

l:‘ l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU is An Affirmative Action/Equal Opportunity Institution
emails-e:

AN OXIDATION STATE ‘STUDY OF DESULFOVIBRIO VULGARIS
FLAVODOXIN BY X-RAY CRYSTALLOGRAPHY.

By

William Watt

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Chemistry

1990

V 7.4.“: 7

Ill/5' " " ___

ABSTRACT

AN OXIDATION STATE STUDY OF DESULFOVIBFIIO VULGAFIIS
FLAVODOXIN BY X-RAY CRYSTALLOGRAPHY

BY

William Watt

The focus of this study has been to determine the conformation of the
holoprotein - with the FMN in each of its three oxidation states - of recombinant
flavodoxin from Desulfovibrio vulgaris. Crystals, in the oxidized state, of
flavodoxin form yellow bipyramids from 3.2 M ammonium sulfate in 0.1 M
Tris-HCl buffer at pH 7.0 with protein concentrations ranging from 0.7-0.9% by
the standard hanging drop method. The reduced states of the crystals were
achieved through the addition of sodium dithionite at pH 7.0 for the
semiquinone (semi-reduced) and pH 9.0 for the hydroquinone (fully-reduced).
The experiments were conducted at low temperature (450°C) in order to
maintain the oxidation state of interest throughout the data collection. Data sets
consisting of one room temperature (oxidized form) and three low temperature
(each oxidation state) were collected on a Nicolet P3F/Xentronics area detector
x-ray diffractometer system.

Four structures were refined by Konnert & Hendrickson restrained
parameter least squares ranging from 2.25A (hydroquinone) to 1.9A (all others)
and crystallographic R values ranging from O.21(hydroquinone) to 0.17
(oxidized - room temperature).

A single-electron reduction of the oxidized flavodoxin is accompanied by
a change in the environment of the FMN. The conformation of residues
TRP60-ASP61-ASP62-GLY63-SER64 is different in the oxidized and
semiquinone forms; however, there doesn't appear to be any conformational
differences between the semiquinone and hydroquinone forms.

The orientation of N(5) of the isoalloxazine ring and the carbonyl oxygen of
ASP61 is in good proximity for a hydrogen bond in the semiquinone and
fully-reduced forms. The conformation of the flavin is nearly planar in all three
oxidation states, but in the hydroquinone form it has a slight propeller-type
twist.

In this thesis, some structural comparisons of the four are made, with
particular emphasis on the features that might be related to the low versus
room temperature data collections and differences in oxidation state.

For Nancy and Samuel, my family

ACKNOWLEDGMENTS

I would like to express my appreciation to Dr. Alexander Tulinsky for
allowing me to complete the majority of my research at my place of
employment, the Upjohn Company.

In addition, I would like to thank the Upjohn Company for allowing me to
use their facilities for completing my research and to Dr. Edward C. Olson and
Dr. David J. Duchamp for their assistance in obtaining permission from the
Upjohn Company for allowing me to conduct my research in their laboratories. I
am also grateful to the macromolecular laboratory personnel (Dr. Howard M.
Einspahr, Dr. Barry C. Finzel, Stephen W. Muchmore, Debbie Ft. Holland, and
Laura L. Clancy) for their assistance and helpful discussions on various
aspects of macromolecular crystallography; a special thanks to Dr. Barry C.
Finzel who seems to either know the answer to any question regarding
refinement and for letting me use his library of programs for plotting or
displaying crystallographic data in useful ways.

Finally, I want to express my appreciation to Dr. Keith D. Watenpaugh
whose interest and enthusiasm in flavodoxins has provided the incentive to
completing this project.

TABLE OF CONTENTS

CHAPTER
LIST OF TABLES.
LIST OF FIGURES.

I INTRODUCTION.

A. Electron Transfer Proteins.
B. Flavodoxlns.

II DATA COLLECTION.

A. Overview of Area Detection Methods.
1. Hardware and Principles.
a. Steps In Data Collection.
2. Software. . . . . .
a. Data Collection.
b. Data Processing.

III EXPERIMENTAL.

A. Overview of Cryocrystallographlc Methods.

8. Method of Sample Handling.
C. Crystallization. . . .
D. Oxidation State Change Experiments.
E. Data Collection. . .
1. Oxidized form. .
a. Room Temperature.
b. Low Temperature.
2. Semiqulnone form.
3. Hydroqulnone form.
F. Data Processing.

Iv REFINEMENT. .
A. Overview of Refinement Techniques.
1. Fourier Methods. .
2. Real Space Refinement.
3. Least Squares Methods.

vi

PAGE

. VIII

.17

. 17
. .17
. 19
.23

. 23
. 24

. .37
. 41

. 41

. .47
. 47
. 47

. .48
. 49
. 51

.51

62
62
.62
.63
.65

CHAPTER 1 PAGE

4. Model rebuilding through interactive computer
graphics:FRODO. . . . . . . . . . . . .76

V RESULTS. . . . . . . . . . . . . . . . . . 80
A. Description. . . . . . . . . . . . . 80
B. Flavodoxln Refinement. . . . . . . . . . . 85

1. Oxidized form. . . . . . . . . . . . . . 85
2. Semiqulnone form. . . . . . . . . . . . 93
3. Hydroqulnone form. . . . . . . . . . . . 96
C. Discussion. . . . 98
1. Comparison of the room-temperature and low
temperature oxidized state. . . . 1 O7
2. Comparison of the oxidized and semiquinone
states. . . . 118
3. Comparison of the oxidized and hydroquinone
states. . . . . . . .125
4. Comparison of the semiquinone and
hydroquinone states. . . . . . . . . . . 1 27
5. Electron transfer. . . . . . . . . . . . 1 32
LIST OF REFERENCES. . . . . . . . . . . . 1 34

vii

TABLE

5a
5b

10

11

12

13

14

15
16

LIST OF TABLES

Flavodoxin sequences.

Homologies and Minimum Base Changes/Codon.
Flavodoxin Internal Gene Duplication.

Data Collection Parameters File.

Data Collection Results for oxidized form.

Data Collection Results for semiquinone and hydroquinone form.

Summary of Data Collection Results.

Data Collection Statistics for room temperature-oxidized form.
Data Collection Statistics for low temperature-oxidized form.
Data Collection Statistics for semiquinone form.

Data Collection Statistics for hydroquinone form.

Torsion Angles for B Turns .................
Refinement Summary of Oxidized State (RT) .........
Summary of Least Squares Refinement Parameters.

Refinement Summary of Oxidized State (LT) .........
R-Statistics between Data Sets ...............

Refinement Summary of Semiqulnone State. . . ......

viii

PAGE

9

. 11
13

. 25
.50

.52

.53

. 57
. 59
60

.61

. .82
. .87
. 88

. .92
. .94
.95

TABLE
1 7

18

19
20

21

PAGE
Refinement Summary of Hydroqulnone State .......... 97
Hydrogen Bonds between the FMN and the apoprotein or solvent
for each oxidation state .................... 106
Crystal Data for D. vulgan's Flavodoxin ............. 1 12
Electropotentials of FMN ................... 124
Bending of lsoalloxazine Ring in Flavodoxin ......... . 131

LIST OF FIGURES

FIGURE

10
11
12
13
14
15
16

17

18

Flow Diagram of respiration in three stages.

PAGE

2

FMN (flavin mononucleotide) and FAD (flavin adenine dinucleotide). . 4

Evolutionary tree for flavodoxins.

Proposed pathway for sulfate reduction in Desulfovibn'o vulgaris.

Side view of detector assembly.

Diagram of Nicolet P3/F/Area Detector System.
Exploded view of detector assembly.

Flow Chart of XENGEN Software System.
Detector Face divided into eleven regions.
Schematic of Crystal Mounting Device.

Low Temperature Schematic for the diffractometer.

Crystal Mounting Method for Low Temperature Data Collection.

Photograph of the oxidized form of D. vulgan's flavodoxin.

Photograph of the semiquinone form of D. vulgan's flavodoxin.

Photograph of the hydroquinone form of D. vulgan's flavodoxin.

Parameter refinement File.

Line section through p(obs) - p(calc) synthesis showing error

in position ........................
Diagonal matrix ......................

X

12
.16
18
20
21
26
. 28
38
39
. 40
. 42
. 44
. 46

.55

. .64

..68

FIGURE

19

20

21

22
23

24

25
26
27
28
29

30

31

32

33

34

Block-diagonal matrix with 9 x 9 blocks for three positional and

six thermal parameters per atom ...............

Ribbon diagram representation of D. vulgaris flavodoxin
a-carbon backbone ....................

PAGE

. .70

..81

Stereo representation of Type l/Type ll [5 turn along residues 42-46. 83

Schematic showing packing of molecules ...........

Diagram showing hydrogen bonds between main chain atoms

in OXRT ..........................

Ball 8. stick representation showing polypeptide chain grouped

into segmented bodies ...................

Ramachandran plot of room temperature-oxidized form. .

Ramachandran plot of low temperature-oxidized form.

Ramachandran plot of semiquinone form ..... ' ......

Ramachandran plot of hydroquinone form ..........

Environment of the flavin in the oxidized state of the D. vulgaris

flavodoxin .........................

Three models of conformation of the semiquinone form. .

Average deviation in residue position between room and
low temperature models ..................

Stereo representation of the conformation of ARGZ4 in the
room temperature and low temperature oxidized form.

Stereo representation of the conformation of LYSB? in the
room temperature and low temperature oxidized form.

Stereo representation of the conformation of GLU42 in the
room temperature and low temperature oxidized form. .

xi

. .84

..86

..89

.99
. 100

. 101

. .102

. 103

. 104

. .108

.110

.111

.114

FIGURE PAGE

35

36

37

38

39

40

41

42

43

44

45

Stereo representation of the conformation of GLU99 in the room

temperature and low temperature oxidized form ......... 115
Difference in thermal parameter of the main chain between room and
low temperature oxidized models ................ 116
Difference in thermal parameter of the side chain between room and
low temperature oxidized models ................ 117
Average deviation in residue position between oxidized (LT) and
semiquinone models ..................... 119
Stereo view of electron density corresponding to residues 60 to 63 in
the FMN binding region in oxidized and semiquinone forms. . . . 120

Hydrogen bonding network around the FMN in the oxidized form at
room temperature ...................... 121

Hydrogen bonding scheme of the ribityl and phosphate groups of the
FMN............... .............. 122

Hydrogen bonding network around the FMN in the semiquinone
form ............................. 123

Stereo representation of the conformation of ARGZ4 in the room
temperature oxidized and semiquinone forms .......... 126

Average deviation in residue position between low temperature
oxidized and hydroquinone models ............... 128

Average deviation in residue position between semiquinone and
hydroquinone models ..................... 129

xii

I INTRODUCTION

Electron transport in mitochondria can be described as the flow of
electrons from intermediates of the tricarboxylic acid cycle and other
compounds down the respiratory chain to molecular oxygen, the final acceptor
in respiration (Figure 1) [1]. There are other membrane systems, such as
heterotrophic bacteria, that carry out electron transport. These organisms,
which have no mitochondria, contain electron carriers such as flavoproteins,
iron-sulfur proteins, and cytochromes. Mitochondrial, bacterial, and
photosynthetic electron transport processes are accompanied by
phosphorylation of ADP (adenosine diphosphate), as a means of conserving
the energy of oxidation. There are some other electron transport processes
associated with the reduction of oxygen without the accompaniment of
phosphorylation. For example, oxygenases use oxygen to insert into a
substrate, superoxide dismutase converts superoxide ions into hydrogen
peroxide and oxygen to help remove the extremely 'reactive superoxide ion
and prevent irreversible damage to various biomolecules, and catalase can
remove an equivalently damaging hydrogen peroxide by converting it to water
and oxygen.

The redox enzymes associated with the electron transport process are
more complex in structure and mechanism and less well understood than
other classes of enzymes. Indeed, many of these enzymes are buried in cell
membranes and are difficult to solubilize and purify. In addition, the
mechanism of free-energy release occurring during electron transfer that is
conserved and transformed into phosphate-bond energy during this process
is still not well understood. Hence, for such reasons this group of proteins is
still an exciting area of biochemical research.

 

 

   
 
   
 
 

F Amino Glucose Fatty Acids
acids ;
Mobilization
of acetyI-CoA< Pym“
2H + 002
K AcetyI-CoA
( Citrate
Oxaloacetate \
cis - Aconitate
Kreb's cycle 4 lsocitrate
Fumarate CO2
a-Keto utarate
Succinate
CO;
L SuccinyLCoA
NAD+
Flavoprotein } (ADP 1' Pi
Cwntme 0 ATP
ADP .L P,
Electron Transport Cytochrome b C
&oxidative < ATP
phosphorylation 0803“” c } ADP + Pi
Cytochrome as C ATP
x I

 

2H* +1/202—v» H20

Figure 1. Flow Diagram of respiration in three stages.

A. Electron Transfer Proteins

There are four types of oxidation-reduction enzymes or electron-transport
proteins involved in the transfer of electrons from organic substrates to
molecular oxygen. They are: (1) pyridine-linked dehydrogenases, which
require either NAD or NADP as a coenzyme, (2) flavin-linked dehydrogenases
or oxidases, which contain FAD (flavin adenine dinucleotide) and FMN (flavin
mononucleotide) as a prosthetic group, (3) iron-sulfur proteins, such as
ferredoxin, (4) cytochromes. which contain an iron-porphyrin prosthetic group.
Finally, there is a fat-soluble enzyme, ubiquinone, which also functions in
electron transport.

The pyridine-linked dehydrogenases, 200 or more, function in different
aspects of metabolism. They catalyze the following reaction:

 

Reduced substrate 4- NAD(P)+ Oxidized substrate 4- NAD(P)H + H“

 

depending on the particular pyridine nucleotide. The process requires the
direct transfer of a hydrogen from the reduced substrate to the NAD(P)+
moiety. The pyridine nucleotides are noncovalently bound to the
dehydrogenase protein and hence are substrates or coenzymes, as opposed
to fixed prosthetic groups, since they bind to, and dissociate from, the active
site during catalysis.

The flavin-linked enzymes contain a prosthetic group that is tightly
bound; these flavin mononucleotide (FMN) or flavin adenine dinucleotide
(FAD) groups use their isoalloxazine ring for the oxidation-reduction process
(Figure 2).

O
N
\ NH
\/K
T" °
R

whereR=

CH2
CHon
CH20H
a...
clezoraoaz-

FMN

 

FAD

Figure 2. FMN (flavin mononucleotide) and FAD (flavin adenine cfinucleotide).

Some of the most important flavin-linked dehydrogenases in respiration and
electron transport are: the NADH dehydrogenases, which catalyze the
electron transfer from NADH to the next member to the electron-transport
chain; succinate dehydrogenase, which is active in the Krebs cycle,

dihydrolipoyl dehydrogenase, which is in the pyruvate and a-ketoglutarate
systems, acyl-CoA dehydrogenase, which catalyzes the first hydrogenation
step during fatty acid oxidation, and other flavin-containing enzymes, not in
the mainstream of electron transfer, e.g. xanthine oxidase, aldehyde oxidase
and other oxidases. These dehydrogenases differ significantly from
pyridine-linked dehydrogenases in that the FMN is tightly bound to the
enzyme protein and thus functions as a prosthetic group rather than a
coenzyme; the flavin does not leave the enzyme during or after the catalytic
cycle. The flavin-linked redox enzymes can be classified as dehydrogenases
and oxidases, according to their ability to react with electron acceptors. In
dehydrogenases, there is a low tendency for the reduced flavin to be
reoxidized by oxygen; the oxidases, on the other hand, can have their
reduced flavin reoxidized by oxygen to yield hydrogen peroxide. Some
flavoproteins can shuttle between fully oxidized and fully reduced forms by
simultaneous two-electron transfers, while others transfer one electron at a
time. The transfer of only one hydrogen atom or electron to a molecule of FAD
or FMN leads to the semi-reduced, or semiquinone form of the flavin group; in
the case of the flavodoxins the electron shuttles between the semiquinone
and the hydroquinone forms.

The iron-sulfur proteins, such as ferredoxin, contain iron and sulfur
(usually donated by a cysteine) coordinated in tetrahedral arrangements.
These iron-sulfur centered proteins undergo electron transfer through
reversible Fe(ll)-Fe(lll) transitions. Based on electron spin resonance
experiments, there are at least seven different iron-sulfur centers in the
electron transport chain. Four are located in the NADH dehydrogenase

complex, two associated with cytochrome b, and one with cytochrome c1.

Although they are important in electron transport, their function is not well
understood.

Cytochromes are electron-transferring proteins containing iron-porphyrin
groups; some are found in aerobic cells. Some of the cytochromes are buried
deep in the inner mitochondrial membrane and appear to act in a concerted
manner and transfer electrons from various dehydrogenase systems toward
molecular oxygen. The cytochromes undergo Fe(II)-Fe(lll) changes during
catalysis. Their reduced forms usually do not oxidize in the presence of
molecular oxygen, except the terminal cytochrome of the respiratory chain,
cytochrome a3, which contains a tightly bound copper. In animals. where the

respiratory chain has been most thoroughly studied, five cytochromes have
been identified in the inner membrane of the mitochondrion. They are:
cytochromes b, c1, c, a, and a3.

In addition to the aforementioned electron-transport proteins, there is a
fat-soluble electron-transfer coenzyme known as ubiquinone, or coenzyme Q._
This coenzyme has a reducible quinone with a long isoprenoid side chain.
Coenzyme Q is the only electron carrier that is not tightly bound or covalently
attached to a protein.

B. Flavodoxins

The complex electron-transfer mechanism associated with
sulfate-reducing bacteria belonging to the genus Desulfovibrio has offered
many opportunities to conduct research in the area of structure/function
relationships that are part of the biochemical redox chains. Although many
redox carriers are known, the mechanism of electron transfer is still far from
being completely clarified. This is in part due to a still incomplete
understanding of the pathway for sulfate reduction.

Sulfate-reducing bacteria are probably present day representatives of
very ancient organisms. This renders the study of redox proteins extremely
attractive, such as flavodoxins, rubredoxins, ferredoxins, and cytochrome c,
that constitute families of homologous proteins.

Of all the flavoenzymes, the most structural information available is of the
flavodoxins [2]. From visible CD spectral studies [3], flavodoxins have been
divided into two classes: "pasteurianum" and ”ruanm" types, where the
positive CD band maxima reflect a red shift of the rubrum type relative to the
pasteurianum type. The ”pasteurianum” type includes the bacteria
Peptostreptococcus elsdenii, Clostridium pasteurianum, and Clostridium MP,
whereas the "rubrum“ type includes the bacteria Rhodospirillum rubrum,
Azotobacfer vine/andii, and Desulfo vibrio vulgaris. Flavodoxins contain one
equivalent of FMN (riboflavin 5’-phosphate) which is their only known
prosthetic group and they lack transition metals such as iron commonly found
in ferredoxins. They are functionally equivalent to the ferredoxins via their
redox potentials between the semiquinone and hydroquinone states (ca. -400
mV). It was first observed [3] that clostridial (nitrogen-fixing) flavodoxin is
synthesized only in iron deficient media. In sulfate-reducing bacteria the
picture is somewhat more complex, since the relative amounts of ferredoxin
versus flavodoxin is dependent on the species and also according to growth
conditions. As it turns out, D. vulgaris, strain Hildenborough, synthesizes large
amounts of flavodoxin and only very little ferredoxin, even in high
concentrations of iron [4].

Flavodoxins do not react directly with small molecules such as pyridine
nucleotides and their only known ”substrates" are other redox proteins. For
example, they are reduced by NADPH and the ability to couple NADPH
oxidation to cytochrome 0 reduction using NADP-ferredoxin reductase as a
catalyst. In addition, flavodoxin also takes part in the phosphoroclastic
cleavage of pyruvate:

2H“ cyt. c3 ' Flavodoxin Pyruvate
(red.) (ox1d.)
H2889
H2 cyt. c3 Flavodoxin Acetyl phosphate
(oxid.) (red.) 4- C02

Since these reactions are known to depend on ferredoxin, than flavodoxin, a
small redox protein, must function similarly to ferredoxin [5].

A comparison of the published flavodoxin sequences is listed in Table 1.
The alignment is based on sequence homology rather than structural
similarity. Overall, there is between 30 and 40% identity between the proteins.
Certain regions show greater homology than others. For example, the
N-terminii have the greatest homology between the four bacterial (top four)
types. The crystal structures of the oxidized form of flavoxdoxin from D.
vulgaris at 2.0A [37], at 1.85A from C. MP [92] and at 2.0A from the A. nidulans

[38] have shown that flavodoxins are a/[S proteins with a parallel )3 sheet at its
core. There are five regions of the bacterial type flavodoxins that form

fi-pleated sheet structure having 40-50% identity levels, depending whether
the region is defined from comparing D. vulgaris or the C. MP protein. The

p-pleated sheet stmcture for the D. vulgaris are residues 3-10, 32-38, 52-58,
87-93 and 118-126 and for the 0. MP are residues 1-6, 30-35, 48-55, 80-89,

and 108-119. A similar comparison on the a-helical regions shows a 20-30%

identity level. The a-helical structure for the D. vulgaris are residues 14-30,
71-83, 102-115, and 133-148 and for the 0. MP are residues 10-27, 66-74,

93-107, 124-138. It appears that the B—structure is the most invariant region
between the different flavodoxins.

om4

mN4

004

Mr

om

mm

44Q«OUo:m
4 :4 8:444 .

4mm. ”M2M40=4> 40000400on¢
4

4mm.
40>
04¢
04H
04H

04H
:04
:04

Q49
Q49
Q49
4>9

0:0
0:0
0:0
0:0

0:0

040
04¢
>40
040
>40
:40

:04
«>4
:«¢
04¢

04¢
040

>40

>40
>40
>40
s40

:04

04¢
«>4
:40
>40

.m\¢.
«>4
:40
04¢
«>4
04¢

:40
04¢
04¢
>40

4:4
4:4
x«¢

4>9
4>9
4>9
4>9

408
040
040
040

:40
400
:«¢
:40

04¢
«>4
04¢
:40
04¢
04¢

04¢ >40 :04

04¢

>40
:40
x40

40m
40m
40m
40m

:40
:40
:40
044

«>4
04¢
:04
0:0

044
04H
04H
04H
04H
04H

044
40>
40>

40m

40>
0:0
0:0

040
40>
40>
:04
:0:

04¢
40m
:04
:40
:04
4:9

04¢
044
40>

40>
40>
04H

Q«¢

40>
:40
Q«¢

0:0
Q«¢
Q«¢
:40
>40

«>4
«>4
«>4
:«¢
:40
:40

“OH.“
”.>.‘

=«¢
04¢
«>4
:«¢

4:4
40>
440

>40

400
400
Q«¢

40¢
Q«¢
Q«¢
Q«¢
>40

04¢
04¢
04¢
04¢
04¢
04¢

«>4
:40
«>4
«4:

04¢
«>0
040

«>0

:40
Q«¢
:40
:40

:40
:40
40>
04H
04¢

40>
40>
so:
40:
40:
4:4

.«00:0:00« 508090.". 4

>40
>40
>40
04¢

>40
>40
>40

>40
>40
>40
>40

040

:04

04¢
«>4
:«¢
c«¢
:40

>40
«>4
04¢
04¢
«>4
4>9

w 5
”0M :40 «>4 m>0 Q«¢
0:0 :40 044 044 Q«¢
044 >40 40> 044
4:4 040 :40 4:4 MM“
4>9 >40 :«¢ 00: >4
:04 c«¢ «>4 :04 «

0 0 :04 04¢ 40> «MM
0M0 :0: >40 40> «>4
0:0 :04 04¢ 40> «>4
0:0 «>0 040 40> «
0 400
40« 40> :4
“4M :04 :40 :40 “Wm
40 :04 40> :40 ¢
”“0 044 400 Q«¢ Q«¢
4:9 4:9
40« 40> 04¢ x«¢ “M0
04¢ Q«¢ 400 40> 04¢
4:9 Q«¢ :40 0:0 «¢
0> Q«¢ 40m 40> “«¢
M0> 40m c«¢ 04¢
:40 4:4 4:4 >40 >4m
04¢ 4:9 «>4 >40 4:
dduiludHiqudlxdulunH
:HuIIHAHIquIHHuIHnH

:40
:40
Q«¢

:40
:40
:40

«>4
«>4
«>4
04¢

>40
>40
>40
>40

04¢
x«¢
:04
40>
044
04¢

Q«¢
:«¢

>40 400
>40 400
>40 400
4:4 400

040

:40
04¢

«>4
:40
:40

>40
>40
>40
>40

00:
402
40:
Q49

40>
:04
304
40m
4:9
40m

400
400

.0304

“000m
"cameo

04¢
04¢
04¢

Q49
0:0
044

4:9
«>4
40m

x40

40m
04¢
044
4:4

«>4
04¢
«>4
:40
:«¢
Q«¢

>40
>40
Q49
Q49
Q49
>40

c«¢
:40
04¢

04¢
Q«¢

04¢

40>
:04
044
04¢

040
040
40m
40m

044
Q«¢
40>
40>
40>
40>

4>9
0:0
4>9
4>9
4>9
4>9

Q«¢
Q«¢
04¢

Q«¢
04¢
Q«¢

044
«>4
«>4
>40

40m
«>0
«>0
«>0

x«¢
400
:40
Q«¢
Q«¢
:40

04H
0:0
044
40>
40>
40>

O40
O40
O40

40:
40:
40>

40m
040
4:9
4:4

>40
>40
>40
>40

40>
002
04¢
04¢
«>4
4>9

40>
:04
04H
04H
04H
044

.40. Q2 anbw4umo4o
:«0Q
0. «44004:: 04444>ou4
4m

00:
:40
Q«¢

Q49
Q49
04¢

40m
04¢
40m
x40

0:0
:04
so:
:04

40>
4:4
>40
>40
>40
>40

4:4
>40
c«¢
:40

:04

:40
:«¢
>40

:40
«>4
>40

44>
so:
044
440

04¢
:04
04H
:04

«>4
x40
«>4
04¢
400
04¢

4:9
044
40>
40>

04¢

c«¢

=40
Q«¢

>40
>40
«>0

40>
Q«¢
:40
:04

40>
044
:04
40>

04¢
Xm‘
:40
044
:40
:40

>40
«>4
«>4

«>4
«>4

40>
044

40m
Q«¢
0:0

Q«¢
4:9
:40
40m

40>
40>
044
:04

x40
x«¢

:40
«>4

044
04¢

04¢

040

40>
044
04¢

>40
>40

:04
0:0
044
c«¢

Q«¢
Q«¢
Q«¢
Q«¢

:04
0:0
04¢
40>
04H
:04

402
402
402
40!

"QZ.U
u.>.D
.Q.U
.0.m
m:.U
.>.Q

.Q.U
.0.m
m2.0
.>.Q

erU
.0.“
QE.U
.>.D

QQOU
.0.“
QI.U
.>.o

.H.m
.>.<
.Q.U
.O.m
m:.0
.>.Q

.4.m
.>.<
.Q.U
.0.m
m2.0
.>.0

10

Hence, one could say this is the nucleus of the molecule and must be
important for the stabilization of the conformation of the protein. Some other
observations indicate: (a) the most homologws regions correspond to the
FMN-binding region; and (b) since the nucleus of the molecule is the most
conserved, suggests that flavodoxin biosynthesis is dependent on FMN
binding.

Some evolutionary relationships of flavodoxins can be determined by
examining the minimum base changes per codon (MBC/C) or the number of
homologous positions (Table 2) [6]. From this table one can construct an
evolutionary tree for the flavodoxins (Figure 3) and along With several other
lines of evidence, it appears that D. vulgaris is of ancient age [7]. There is
unfortunately very little evidence of homology when comparing flavodoxin and
ferredoxin from the same organism.

An interesting study of bacterial-type ferredoxins apparently shows
evidence of internal gene duplication [8] from a primitive fragment of
apparently one-half of the length of the contemporary protein. The plant-type
ferredoxins show a second duplication to establish a 50% longer protein. An
examination of bacterial flavodoxins can be carried out to see if the same
duplication is present (Table 3). If one examines the FMN phosphate binding
site of the bacterial flavodoxins, there appears to be a statistically significant
homology for two regions of the protein. This holds true for three bacterial-type
proteins except 0. vulgaris. Therefore, this suggests that this is an extra
mutation lending to a longer period of evolution for the D. vulgaris [6].

The sulfate reduction system in which flavodoxin plays a role by shuttling
electrons to a complex sulfite reductase system from cytochrome c3 [9] is

given by:

11

 

00... «3 0.0 so... .440 0.0
40... 0.2.0 .45 0.2.0
.20 .8
0.2.0 .8 0.2.0 .3
0.002 8.0285:
£220 :48.
«a... «4... .23 8... 3... 2.0 .48 .43 .444 .440 .404. 2.0
8.. 8.. 8.0 8.. .>.< .444 .49 .440 .444 .>.<
«m... 2... «4... 0.0 .48 .48 .49. 0.0
8.0 3.... 0.0 .44.. .44... 0.0
co... 0.2.0 .49. 0.2.0
.5 .8
.>.< 0.0 .o. 0 02.0 5.0 .>.< 0.0 .0. 0 0.2.0 .20
0.00.2 8.0228:

«0:0.«0. mm 43

 

000000000020 0.000 84.8.52 0:0 «£02050... .0 0.000.

12

 

A.V. HJ‘. D-V- C.MP C.p. RB.
gene
doubfing
I
gene
doubfing

Figure 3. Evolutionary tree for flavodoxins (See Table 1 for abbrevations).

13

Eacuwuamuma 34.204.444.040 4.44.0
4.44.2443 42000004Q044404Q4m " 344.44
43 054444444040 4442.0

4.4440445 04.320444:me u.>.o

004 44 44 44 44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44 04

...444 440 444 Hum 44a «am 044 444 444 444 444
44: 040 040 040 040 444 44> 040 444 440 442 444 040 man 440 «am 444 44> 444 044 444 .0.0
44 44 44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44 04 44 44

404 404 404 404 404 404 404 404 004 44 44 44 44 44 44 44 44 44 04 44 44

444 444 044 44> 444 440 440 4:0 444 :40 444 444 444 444 440 440 440 4:0 440 444 44>

444 04¢ 040 044 040 444 444 44¢ 44¢ 440 444 4:4 444 440 440 044 044 44> 044 044 4:0 .>.0
04 44 44 44 44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44 04

004 44 44 44 44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44 04

040 4:0 04¢ 44¢ 442 044 444 440 man .440 444 440 444 444 44a 4:0 044 444 44> 444 444

4:0 040 444 040 :40 044 44> 040 and .440 44: 444 444 440 44a 044 444 044 444 04¢ 040 02.0
44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44 04 44 44 44 44

404 004 44 44 44 44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44

444 444 444 _q44 44: 444 :40 440 nun .44u 444 440 444 444 44a 4:4 :44 440 444 444 444

44> 44> 444 044 040 044 :40 040 Hum .44u 442 444 o40 440 44a 044 :44 444 444 444 444 .4.0
44 44 44 44 44 44 44 04 44 44 44 44 44 44 44 44 44 04 44 44 44

40.06.0024: 5044 0.5.3 400.500. 44.244. 4.44 4.49455 00:40:04.0 4040 _4:..4E_ 0480024.“. .4 4.000.

14

 

$032“
H2 cytochrome c3 Flavodoxin
(oxud.) (red.) Sulfite
Hzase Reductase
J System
2H” cytochrome c3 Flavodoxin V
(red.) (oxid.)
H28

The sulfite reductase system [10] was found to be composed of three
reductases: sulfite reductase. trithionate reductase, and thiosulfate reductase.
It was also found that trithionate was formed from extracts of D. vulgaris and
the following cyclic process was proposed with trithionate and thiosulfate as
intermediates linking the reduction of (bi)sulfite to sulfide:

H H H
3032. "'——2"> $3032. fi’ 32032. i» H23
1 3032' 3032‘

 

The sulfite reductase [11] and thiosulfate reductase [12] systems have been
studied. An experiment from the D. vulgaris shows that cytochrome c3 plays
the role of an intermediary electron carrier in thiosulfate reduction. It also
shows that direct electron transfer occurs via cytochrome c3 from hydrogenase
without any other factors intervening. Finally, the experiment indicated that
flavodoxin had no effect on the thiosulfate reduction; this is not surprising
since flavodoxin does not accept electrons from hydrogenase [13]. From the
use of natural electron carriers, the thiosulfate-forming pathway of D. vulgaris
has been constructed. In order to reduce bisulfite to thiosulfate, flavodoxin and
cytochrome c3 were necessarily present.

15

The reaction showed cytochrome c3 transferring electrons from hydrogenase
to flavodoxin. Trithionate accumulates but is immediately used by the
thiosulfate-forming reaction, to form thiosulfate. The thiosulfate reductase then
can complete the pathway by forming sulfide (Figure 4).

The foregoing summarizes the current scheme for the complete pathway
of sulfate reduction to sulfide which is still not completely satisfactory. Since
trithionate reductase activity has been observed with crude extracts, its role
still remains a puzzle. Further studies on thiosulfate reductase and
trithionate-reducing enzyme and on the reconstitution of the sulfite reducing
system should help to elucidate the mechanism of sulfite reduction in D.

vulgaris.

16

>4f<

cytochrome Ca cytochrome Ga

(OX1d.) (red) 2H+ H2

H2886

Flavodoxin Flavodoxln cytochrome c4. cytochrome (33
(red) (oxid-) (oxid.) (red.)

     
 

 

3HSOa' saoez' 5203' + 2H503'
Blsulfite reductase Thiosulfate-forming
HSQa' enzyme
2_ Thiosulfate
3203 + H2 > 9032' + H25
Reductase

Figure 4. Proposed pathway for sulfate reduction in Desulfovibn'o vulgaris.

ll DATA COLLECTION

A. Overvlew of Area Detection Methods

Since the amount of observational data for a stmcture determination of a
macromolecule by x-ray diffraction techniques is in the thousands to hundreds
of thousands of reflection intensities, the development of area detector
diffractometers that use an image proportional counter coupled with a video
display has been carried out in several laboratories in order to reduce data
collection times [15, 16, 17, 18, 19]. Collection of reflections from crystals with
large unit cells, either through conventional counting circuitry (diffractometer)
or with film, is inefficient. The diffractometer has high counting efficiency but it
only counts one reflection at a time; while film records many reflections
simultaneously, it has low counting efficiency. In addition, these data
collection methods require the crystals to undergo long exposure times
throughout the entire data set which are likely to deteriorate in the x-ray beam.

1. Hardware and Prlnclples

The multiwire area detector (MAD) diffractometer is capable of measuring
more than 1000 reflections per hour. A small curved-window high-resolution
multiwire detector, employing xenon at high pressure as the ionizable gas,
using a capacitive readout of the photon events, has been developed by Fl.
Burns and is available commercially (Nicolet-Xentronics (now Siemens),
Madison, WI). The (Figure 5) curved front beryllium window of the detector has
a diameter 11.5 cm and a radius of curvature of 24 cm. Data are received as a
series of 512 x 512-pixel 16-bit frame images. The data quality, as determined
from agreements among symmetry-related observations, is comparable to
collimated counter diffractometry.

17

 

 

 

. ~ ~
\ _\
.‘4‘\‘ \
.-"\'.
.\\ .\l
I EKV\\

 

 

 

 

 

 

.- «(K 25 “"‘:}\:|
‘.\\\.\\\4\§\

 

 

 

4b
0

 

 

‘4
‘ 3
\

Figure 5. Side view of detector assembly.

I
\‘ \\‘:;‘; ‘. ‘
t

 

19

The Nicolet Xentronics area detector/P3F set-up is shown in Figure 6.

The rotation of the crystal occurs about a vertical spindle with the detector

attached and moved through the 20 arm. Radiation (Cu Ka) is provided with a

conventional sealed tube source and passed through a graphite
monochromator; the choice of monochromatized radiation over Ni-filtered

arises from an improved signal-to-noise ratio. The detector central 20 value is
determined by the diffracting power of the crystal and the resolution of data the
investigator wants to collect. For example, for a crystal with a 70A axis, one

would place the detector distance at approximately 9 cm, and the central 20
value of 25° would place the beam slightly off the detector image. At this
setting, data from 50-1.8A could be collected. Higher resolution data would

obviously require the central 20 value to be moved beyond 25°.

The theory of operation of area detectors has existed for a number of
years in the scientific literature. When x-ray photons ionize the xenon gas the
positive ions produced drift to a multiwire anode and are read on a set of finely
spaced wires (cathodes), one set of wires in x and one set in y (Figure 7).
These wire planes are read out capacitively, with the charge division being
roughly one pixel at full width half maximum (ca. 0.2 mm). The spatial
resolution is high in both xand ydirections because the electrons initially
created diffuse in a drift space creating an ionization avalanche of thousands
to a million ion pairs (depending on the potential applied to the anode wires).
The distribution of the electrons is recorded on a number of wires from which
the 'centroid' (and hence the photon position) can be calculated to a fraction
of the actual wire spacing.

a. Steps in Data Collection

The first correction that is made before starting to collect data frames is a

20

Nicolet P3P Area Detector System

' Area
Detector |

 

 

 

 

 

Terminal
PCS -

Peripheral
Computer

 

 

 

Serial Une

 

 

Nova 4/C

 

 

 

 

 

 

 

2'3'43§%"R< ‘. .'-.<¢: 341'-
k” We *0 *4
,

2:1": 3,. ...:
34:45:4- 4444441i§t44§mh ..-...

 

 

 

 

 

Remote Computing System

Figure 6. Diagram of Nicolet P3P! Area Detector System.

6mm

2cm

2cm

24m

21

DETEC‘T”: FRONHI
HINDOH ‘1 mm, B.)

 

 

Dam sucraoo;

51M PINS ELECT‘R ODE

/ ANODC
/////// + . m xv.

CATHOD: Y putPut

/ ӣ55m: VESSEL

Methanerlo‘)

Figure 7. Exploded view of detector assembly.

22

flood-field correction. This correction is carried out by mounting an iron-55

source on the goniostat and exposing the detector to the iron y-rays. This
procedure is necessary because of the nonuniform sensitivity of the detector
face; different parts of the detector face have different quantum efficiencies
and hence scaling these different areas to a common mean value is required.

The second correction is the calibration of the x-y positions on the
detector face using a brass plate. The plate has many holes drilled at
accurately placed positions; it is mounted on the front of the detector face and
an iron-55 source or another diffracting source is used to lrradiate the detector
face for approximately thirty minutes. The data from this image are employed
to calibrate the detector addresses in pixel positions against the accurately
placed holes of the brass plate. This brass plate image is repeated every time
the crystal-to-detector distance is changed.

The final correction is the adjustment of the beam stop. After reducing
the current, the beam stop is carefully aligned by driving the detector out of the
beam path and irradiating a fluorescent paddle. If the beam stop is improperly
aligned, there will be a bright spot on the fluorescent paddle from the
unblocked x-ray beam. The position of the beam stop can be adjusted until it

appears centered. The detector can then be moved to 26 of 0° and a 30 sec
exposure can be taken. From the video display, box cursor, and the cursor
keys, one can make the necessary final adjustments of the beam stop.

Since there is no accurate method to align the x-ray tube and
monochromator with an area detector, the method followed at The Upjohn
Company is to replace the area detector with a standard collimated
scintillation detector. The standard alignment procedure provided by Nicolet is
used and when completed the area detector is replaced.

23

All data are collected as discrete ”frames” taken either with the crystal
stationary or with it rotating about a fixed axis. In a modified version of the
Electronic Stationary Method (ESP) method of data collection [19], one
obtains reflection intensities from a series of electronic pictures, or subframes,
each of which is exposed while the crystal is moving slightly. Each frame
consists of a set of subframes of equal time. The crystal is rotated

approximately 0.02° about a fixed axis (co) between subframes. Hence, if
samplings are taken over a 020° range, than ten pictures would comprise a
frame. This method is in contrast to standard film methods in which the crystal
is oscillated by a couple of degrees during the time the detector is recording
the diffraction pattern. One could compare this method with the step-scan
method [20] used in conventional diffractometry except many reflections are
measured simultaneously.

The acquisition of the data and the movement of the crystal and detector
are controlled by a PCS (Periphere Computer Systeme, GmbH) 9600
microcomputer with 1 MB of memory and many more MB of hard disk storage
space on a VAX 11/750 (PACVAX). At Upjohn, the data are transferred over
Ethernet from the PACVAX to THOR (VAX 8800) and processing takes place
there (Figure 6). The software system for processing data (XENGEN [21]) has
been written in C and runs under the UNIX operating system.

2. Software
a. Data Collection

The Harvard software package as modified by Nicolet was used to collect
all the data for this project. The data collection software performs a number of
complex, interrelated tasks: crystal alignment, crystal survey, data collection,
background correction and other related operations.

24

The first step is to manually optically center the crystal with the microscope

attached to the x circle. This is carried out with the aid of an optical alignment
program, part of the standard programs supplied by Nicolet, which drives to
the appropriate diffractometer angles to aid one in visually centering the
crystal. The next task is to examine the crystal integrity and diffracting power.
The programs ALlGN and CAMERA (subprogram of ALIGN) provide one the
ability to move instrument angles, set the angles to a desired value, open or
close the x-ray beam shutter, enable or disable the detector, make an image
on the detector face, check peak profiles, and view the image. From these
operations a single still 'photo' is taken, from which reciprocal space axis
assignments can be made as well as examination of crystal integrity. The
crystal alignment is only critical for knowing how to orient the crystal during the
data collection to insure that a complete data set is being taken. The final task
uses the program COLLECT. This routine accepts the necessary data
collection instructions into a file and carries them out when instructed (Table
4). '

b. Data processing

The XENGEN software package [21] was used for processing all data of
the experiments for this project.

A flowchart diagram (Figure 8) shows the steps for processing the frame
image sets. The steps involved in data reduction are: (1) W:
generate a "spline" file from a brass plate image; (2) EQBQEB : used to define
the active pixels that make up the usable part of the detector face; (3) $3015:
determine the centroids of a group of bright spots appearing in each frame; (4)
W: index the reference reflections and obtain an estimate of the crystal
orientation and refine the crystal and detector parameters; (5) W:
compute the integrated intensities and estimated standard deviations;

25

Table 4. Data Collection Parameters File

 

Harvard-Xentronics X-Ray Area Detector - Data Collection System PCS P3F 0.0

1. Date : Fri Jant 10200001990 2. Experimenter: WWatt
3.Nameofdataset: UX0123a 4. Crystal: Flavodoxin
5. Crystal Orientation: None

6. Directory (Data Frames): PACVAX::USER_DISK:[DATA}

7. X-Ray Set: 6. Power. 50kV 35mA
9.Phi : 30.00000 deg 10. Chi : 0.00000 deg
11.0mega(@ Frame 1) : -20.00000 deg 12. Two_theta: 20.00000 deg

13. Beam (@ 2theta - 0) : x 241 y 244 14. Camera length: 180m
15. Comments: Flavodoxin (LT) Fully-reduced form - Hydroquinone

16. Framesize (Omega Scan) : -0.20000 deg 17. Exposure: 120 s
18. First Frame: 1 19. Last Frame: 305

20. Options: (Off/On) tape disk view 20. View Parameters: 6 60

Reading save OOIIGCHOD parameters from UX0123.par

Copyright 1985, President and Fellows of Haward College.
All rights reserved. Reproduction without express permission is prohibited.

Make changes? [y/n]

CVTCAD
Callb. Frame

 

 

 

 

26

 

 

 

 

 

 

CVTCAD
D.ta Frames

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Caldata (-.brs) Frames(- .trm) Wmask(- .wms) Frames( frm)
CALIBRATE I SPOTS tI BORDER

l Centrolds(-.cen) l
<—
CENMERGE __.

Uspline(-.uca) I EDWCEN Amask(-.ams)
REFINE Uparams(-.upr)
Wmask(-.wms) ”“1“”,
fl INTEGRATE l:—

 

 

 

 

 

 

 

MERGESHIFT

 

 

 

 

 

 

 

 

 

 

 

MRMEFIGE i

 

 

—’H7 SCALEI l-D Scaleout(-.so)

‘v-—> Deletions(-.del)
—> Checkrefi(-.ckr)

 

 

 

Scalein(-.si)
l Mretout( .mrl)
I , 3.__
warm-m.) <—_—-I STATS REJECT
* A

 

 

 

 

.[ MAKEMU '—

 

 

» Mrefout(-.mrf)

—> mum-m)

Figure 6. Flow Chart of XENGEN Software System

27

(6) EEQLLQE: apply Lorentz and polarization corrections and calculate
structure factor amplitudes; (7) MBMEBQE: merge together data from various
orientations of one crystal; (8) SCALE]: determine a scale function to reduce
systematic error in the data; (9) BEJEQI: eliminate outliers from data; (10)
SIAISI calculate statistics on merged data including R...ym between frames

sets; (11) MAKEMLL: compute scaled merged mean intensities or structure
factor amplitudes for unique reflections in the data.

The calibration ”spline" file is generated from the brass plate image by a
program called Qalitzrate. It works by finding the center of the spots on the
image and indexes them relative to one another. It uses a sophisticated
Hermite spline to convert a pixels-to-centimeters map, and the same spline is
used to help generate the necessary information to generate a
centimeter-to-pixel map. Extrapolation to the plate edge is carried out from the
pixel-to-centimeter map.

flame; is used to define the active pixels that make up the usable part of
the detector face. An 'active' pixel can be used in measuring intensities,
whereas the 'inactive' pixel is outside the active area, or obscured by some
object between the detector and crystal (e.g. beamstop). The output Annals file
is written to a bit-mapped bitmask file.

$0.015 generates a list of centroids of bright spots from reading in a set of
frames. These spots are merged with spots from neighboring frames, since
they are part of the same reflection. The program also checks to insure the
centroid is well-behaved. In addition to finding centroids, $11915, generates a
weighted mask (flmask), or model profile from a weighted sum of all bright
areas found in the particular parts of the detector face. The detector is divided
into eleven regions (Figure 9), so eleven weighted masks are generated. This
is necessary since the [magma routine that determines intensities are
regional and weighted masks for each region are needed.

28

 

Figure 9. Detector Face divided into eleven regions.

29

Once Spots has generated a list of centroids, the program Beflm, can be
used to get an initial orientation. Eating, can describe the crystal and detector
position by refining a set of parameters. if one wants to take advantage of the
auto-indexing feature, the program prompts with two questions: fractional error
in unit cell lengths allowed and fractional error in unit cell angles allowed.
Once values are entered, a list of the first few hundred centroids is examined
for difference vectors among themselves. The 'best' three noncoplanar
difference vectors are removed from the list and used as a reference set. A set
of trial (hkl) indices are then calculated from this reference set. This process is
carried out for other reference sets and a list of these sets can be examined.
The other criteria used for a 'good' reference set are checks on the error limits,
integemess of the indexing, and reasonableness according to a least-squares
residual test. From this list, a solution can be chosen. Because auto-indexing
fails some of the time, different common troubleshooting techniques are
available. For example, if the auto-indexing produces a good residual but the
distribution of indices is poor when compared to actual reflections (poor
integemess), the result is commonly an error in crystal-to-detector distance
and other various detector parameters. Adjustments before refinement can
lead to getting better integemess of the indices.

Once indices have been assigned to the spots, the unit cell parameters,
orientation of the crystal, the detector position, and the rocking-curve behavior
of the crystal can be refined with 331mg. There are several modes of
refinement available for these parameters. In addition to the modes of
refinement, several parameters can be examined during the refinement stage.
For example, there are detector parameters as well as crystal parameters
which should refine to acceptable values based on some previous
crystallographic knowledge. The most powerful method, albeit slowest, is the
nonlinear refinement method. This method refines both the crystal and
detector parameters, simultaneously; the residual contains both the
integemess of the indices and errors in the detector parameters.

30

Hence, a high residual usually corresponds to poor integemess of the indices
or incorrect detector/crystal parameters.

The program [magnate performs a three-dimensional profile fit to each
reflection, applying the previously calculated mask. The background
correction uses an updating technique [22] where the background values are
assigned to pixels in frames where the pixels do not correspond to reflections.
These background values are obtained by averaging over many frames. The
background value for each pixel is than updated after each frame (provided it
still is not contained in a reflection). The new value is updated by a weighted
average of the old and new values. Thus, the background is averaged over all
of the frames and the mask is updated.

All intensity profiles are corrected for exposure time and dead-time loss
before being summed. In addition to these corrections, the profile-fitted and
summed intensities are corrected for the Lorentz effect [23] and the
polarization effect [24] and the small part of the profile lying outside of the
summed area.

During integration, in addition to background updating. mm are
updated whenever bright spots occur. Since there are eleven areas of the
detector face with each having its corresponding wmask. each bright spot
contributes only to the mags]; in its area. Finally, at regular intervals
(approximately every forty frames) the crystal and detector parameters are
refined [25].

Integration compares the summed intensity to the profile-fitted intensity
and is repeated if they disagree. If they agree, the mugs]; is then overlaid on
the profile-fitted intensity. If they do not agree, the integration is repeated and
tested again. Once [magma has completed the last frame, an output file mafia
is made.

31

The next step in processing involves reformatting, sorting, and merging
the integrated reflections. This is easily performed with the program Reduce.
The routine is very flexible: different flags direct the program to do different
operations. For example, "-5" flag uses the summed intensity instead of the
profile-fitted intensity. In addition to the removal of space group forbidden
reflections, the reflections are sorted in hkl order (khl - monoclinic) with the
symmetry-related reflections grouped together and merging of multiple
observations of the same reflection. The output file has the multiref format

(0110-

After reducing several frame sets, the program Mm is available to
merge the data sets. The program reads the multiref file and the output file has
the same format.

The final steps of the data processing entail scaling, deleting outliers,
statistical analysis, and writing out merged data. The program finale]
examines the symmetry-related observations and scales the merged intensity
data. The data are broken up into a variable-degree range scanning angle
with each range split into two shifts corresponding to either the top or bottom
part of the detector. The groups of crystallographic observations are scaled by
a closed-form linear least-squares algorithm. The program alternates between
scaling parameters and computing a scale factor until convergence. The
scaling parameters per shift are either one, two, or three terms. That is, one
parameter per shift is a linear scale factor, two parameters and three
parameters follow the formulas:

|cor = lraw * (9i ' Bi " 52)
loor- law. (girAi*S+BpsZ)
where gi, A] and Bi are refinable parameters for shift i and s = sine/7L for the

reflection of interest.

32

After scaling the data, one can examine the results to determine which
reflections are in need of deletion. The program Belem provides a way to
delete errant observations. This menu-driven program reads a multiref file and
the scale input and flags statistical outliers. One can examine the observations
that are suspicious and delete the suspicious reflections if any are present, or
one can select interactively a different reflection to delete. After the entire
multiref file is read the program allows one to apply the deletions, print a list of
deletions, or exit the program (with or without deleting reflections).

The final analysis of the data is carried out with a program called Stats.
This general statistical analysis program counts the number of data,
determines the number of missing reflections, performs a Wilson-like statistic
analysis of the data, and finally calculates agreement statistics and goodness
of fit parameters based on resolution range. Stats can also be used to check
the quality of data before Reject.

Prior to examining the multiref files by 5.0.8191. Bejegtpr Stats, the Bijvoet
reflections can be merged. The effects of one choice over the other can be
pronounced; the scaling with Bijvoet reflections merged or unmerged makes a
difference if there is good redundancy in the data. If there is poor redundancy,
however, less than two observations for every Bijvoet-unmerged reflection, the
scaling with the Bijvoet reflections merged will be necessary. This arises from
the inability to find enough symmetry mates for computing scaling functions.

The final program Makemu allows one to write out either intensities or
stmcture factor amplitudes to an ascii file, entitled mulist.

Ill EXPERIMENTAL

A. Overview of Cryocrystallographic Methods

The use of cryocrystallographic techniques is starting to be investigated
for the collection of data for macromolecules. It is well-known that cooling
crystals to around 0°C reduces the rate of radiation damage during data
collection [26] and it can be assumed that cooling further would provide
increased protection.

Low et al. [27] have studied insulin crystals at low temperatures (below
-150°C). The crystals, which contained no organic solvent and wiped clean of
excess mother liquor, were mounted in a special cell with mylar windows (ca.
0.15 mil thickness). The cell contained a droplet of mother liquor some distance
from the crystal. The cell was then dipped into liquid nitrogen and
photographed while a stream of liquid nitrogen flowed over it. The diffraction
pattern showed an increase in mosaicity on cooling which increased on rapid
rewarming to room temperature, and increased more on rapid recooling. They
believed the mosaic character of these crystals precluded their use for accurate
data collection. This enhancement of mosaic character was a result of thermal
strain on rapid cooling, not the formation of ice in the crystal. The conclusion of
their study revealed that the insulin crystals could not be cooled below -16°C in
their mother liquor without irreversible disordering.

This problem was addressed by attempting to reduce the disorder
through cross-linking lysozyme crystals with glutaraldehyde followed by
transferring to a 50% glycerol solution [28]. A slightly different approach
involved the addition of 3 M sucrose to a 2 M ammonium sulfate solution of
lactate dehydrogenase to form a glass at low temperature [29].

33

34

Lastly, the freezing of myoglobin crystals at a hydrostatic pressure of 2500 atm
to form the ice lll phase, which has a smaller volume of water, was also
investigated [30]. In each of these cases temperatures of -70°C and less were
achieved. As a general rule, additional problems were encountered with the
approaches. In the first case, the cross-linking of the lysozyme crystals would
affect the activity of the enzyme by restricting substrate access to the active site
and the cross-linking itself created disorder in the crystal. The next case has
problems with sucrose causing an increased mosaicity to the crystals, and the
formation of the glass makes substrate diffusion experiments impossible.
Finally, the high pressure approach to the myoglobin crystals renders the same
diffusion problems as the previous case and necessitates the use of expensive
and complex equipment.

A more recent approach to the collection of data at low temperature was
developed by Petsko et al. [31] using salt-free aqueous/organic liquids of low
freezing point. The basic feature to this technique was to replace the normal
mother liquor with a cryo-protective salt-free aqueous/organic solvent mixture
of high organic solvent concentration and low freezing point. One of the
objectives of this approach was to determine a fluid medium that would not
denature the protein or change its mechanism of catalytic action. This same
philosophy would have to be extended to the stability of the enzyme at low
temperature, as well. Douzou ef al. [32] have developed cryo-protective
artificial mother liquors of mixed solvents where the organic solvent is high
molecular weight (MPD: 2-methyl-2,4-pentanediol, propanediol) and low
volatility. These properties enable one to keep the protein out of solution and
are easy to control. The crystals are then handled in the traditional manner by
mounting in glass capillaries, then sealed, except the plugs of mother liquor at
each end of the tube were avoided (due to problems with distillation from
thermal gradients). Some of the results of this technique showed the effects of
cooling on resolution and the effects of cooling on radiation damage.

35

The resolution did not appear to be improved but there was areductlon in
thermal disorder. The radiation damage appeared to be immeasurable on a
crystal of ribonuclease A at -104°C for over 350 hours. The use of organic
solvents can have a disastrous effect on the protein; however, if proper care is
taken, such as monitoring the pH of the solvent and avoidance of sudden
changes in dielectric constant, the use of mixed solvents can be very mild to
the protein. Some of the advantages include the lack of need for special
equipment, minimal disorder, and possibility of substrate diffusion if the
viscosity is not too high. In addition, significant unit cell changes do not seem to
occur at these low temperatures, probably because at these temperatures the
values of the dielectric constant and pH are similar for the two mother liquors.
One of the chief drawbacks of this technique is the necessity of finding the best
cryo-protective mother liquor for the protein of interest. This step can be crucial,
especially if there are only a small number of crystals available. Another
problem is the effect of the mixed solvent on the protein conformation. Finally,
the viscosity of the mixed solvent can be and is usually a huge obstacle to
substrate diffusion. For some solvent systems where the viscosity is high at
room temperature, the problem is impossible at low temperature.

An extension of this method has been proposed by Dewan er al. [33].
They have studied ribonuclease A crystals mounted on glass fibers at 220K
and below, by the use of a nitrogen-gas cold stream and a computer-controlled
diffractometer. For all of their low temperature data collections, the
ribonuclease crystals were gradually transferred to a 50% MPD:H20

cryo-solvent mixture. Their first data set was collected at 160K, using a
conventional quartz capillary tube mount, with the capillary being bathed in a
cold nitrogen stream. This crystal showed a 21% decay (probably due to the
frost buildup) of the intensity standard reflections after 14 days. At this
temperature, there were problems with a thin film of frost building up on the
outside of the capillary resulting in spurious peaks.

36

The frosting problem was overcome by using a method developed for handling
air/moisture sensitive compounds [34]. The crystals were in their cryo-solvent
mixture and held at ~273K with an ice bath. The crystal mounting device
consisted of a glass fiber glued into a brass pin which in turn was mounted on
a standard goniometer head. On the tip of the fiber was a small bead of
uncured epoxy (normal 1:1 ratio of resin and hardener not allowed to harden).
The whole device was used to lift the crystal from its mother liquor. Once in
place the entire crystal/goniometer head assembly was rapidly transferred to
the diffractometer where the crystal was bathed in the nitrogen-gas cold
stream. Data collections of 180, 160, 130 and 98K were collected;
approximately 3% decay was observed for each of the data sets. A study at a
higher temperature was also performed (220K); however, the crystal had to be
coated with a silicone oil prior to being placed in the cold nitrogen-gas stream.
This was carried out in order to prevent the crystal from drying out in the cold
stream at the higher temperature. There was also no appreciable decay in this
experiment over an 8 day period. The results of this study indicated that the
reduction of deterioration was associated with a combination of the low
temperature and the lack of a capillary tube enclosing the crystal. Some of the
problems with the method are: there was a tendency to have the crystal
mounted in an unknown orientation, and, should the cold nitrogen stream be
interrupted during the data collection, the crystal would dry out and stop
diffracting. Some of the positive aspects of this method are the elimination of
slippage of the crystal which occurs during the standard capillary mounting
methods, and the elimination of the frosting problem previously mentioned.

Most recently, Hope [35] has developed a useful method for collecting
data at cryogenic (near liquid N2) temperatures. In this method, the crystals are
first transferred from their mother liquor to a hydrocarbon environment, then
mounted on a glass fiber, and flash cooled in situ with a cold nitrogen stream
on the diffractometer. This approach seems to prevent solvent from freezing
and disrupting the crystalline lattice.

37

For hanging drop experiments, the coverslip holding the drop is placed
with the drop facing up. The drop containing the crystal is then completely
covered with a large drop (ca. 0.5 ml) of Paratone-N (Exxon) (mixed with
mineral oil to attain suitable viscosity (ca. 25-50% mineral oil)). The crystal, with
a small amount of mother liquor, ls moved from the mother liquor to the oil with
the use of a pin or razor blade. The small amount of mother liquor adhering to
the crystal is removed by use of a strip of dense filter paper out to a point. The
transfer to the mounting fiber is carried as rapidly as possible. A standard
small-molecule-type glass fiber of appropriate length and thickness, attached
to a metal mounting pin, locked into a standard XYZ goniometer head, ls used
to lift the crystal from the oil (Figure 10). The adhesive used is an ultra-pure
high-vacuum grease (e.g. Teflon High Vacuum Grease - Distrib. by American
Scientific Products, Division of American Hospital Supply Corporation, McGraw
Park, IL 60085). The transfer into the cold stream is then carried out as quickly
as possible. The best procedure is to temporarily deflect the cold stream while
the crystal is being brought to its final position on the diffractometer. Once in
place, the obstruction is removed in order to let the cold stream flow over the
crystal (figure 11).

8. Method of Sample Handling

The approach used in this work differs only in the transfer of the crystal to
the hydrocarbon environment. in this experiment, the crystal is pipetted along
with some mother liquor and transferred to a spot depression plate containing
approximately two milliliters of Type A Cargille lens immersion oil (R. P.
Cargille Laboratories, Inc, Cedar Grove, NJ 07009). For the reduced state
flavodoxins, the lens immersion oil was previously purged with nitrogen for two
days before use. The excess mother liquor is removed from the crystal by
careful suction through a smaller capillary (ca. 0.2 mm) or a pointed filter paper
strip. The remainder of the procedure is the same as previously mentioned
(figure 12).

38

High vacuum grease \
I /

0.30 mm glass capillary

/ Tapered Brass Pin

  
 

 

:5: .....

Standard Supper
<— XYZ Goniometer
Head

 

Figure 10. Schematic of Crystal Mounting Device.

39

$582856 9: .2 28820...“ 229882 33 .2 059...

Eur... we.

 

w

- . . C(BUO

_ «2.. oSoSJW

 

 

 

ruozaxoxu,

4
5.742

 

 

 

\IIIIIIII\II.|||II/

x00 JOCPZOU any

©o©
Uoo

 

 

CHJOOU a;

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

     

5.53:3: \ 2.2 e
. . x. a . .
«:3: \X
1:; # >53»
agnnv ”(Omzum
5520.23
(trudge-.. a C
h d:u\ ,
rlr : u>4<>
c.0238 ~24
anti:
C . 233m
ouoo
Jeane
32.2.
623.. z 53
r x 35.53:
)\] uzammmze
8:23
mun-4h ¢<3uo cuumz<¢h AN. 30;: 940 N2 >10

40

 

Mother liquor

 

 

 

Side view a?
of depression
spot plate
Small glass
capillary or
Depression filter paper
spot plate strip
containing
crystal and

 

lens immersion -
oil

 

\Crystal mounting
device with adhesive
at end of tip

9.11

 

 

Figure 12. Crystal Mounting Method for Low Temperature Data Collection.

41
C. Crystallization

The material used in this study was generously supplied by Professor R.
P. Swanson; his relation to the work was to clone and nucleotide sequence the
flavodoxin gene. Since the redox potential of the semiquinone/hydroquinone
couple is one of the lowest of any known flavoprotein (ca. -450 mV vs. -175 mV
for free FMN, pH 7) and the three-dimensional stmctures of flavodoxin from
three sources (Closfn‘dium MP [36], Desulfo vibn'o vulgaris [37], and Anacysfis
nidulans [38] are known, then with the aid of genetic engineering technology
(in vitro site-specific mutagenesis) it should be possible to gain insights to the
factors that effect the redox properties of these proteins.

Flavodoxin, from D. vulgaris, was obtained through recombinant DNA
methods [39] and it differs from the wild-type at the N-terminus (i.e. Wild type .
PRO, Recombinant a ALA). The crystals are tetragonal, a,b = 51.96, c a 139.86,
space group P43212 with one molecule per asymmetric unit.

The material was provided at 9.98 mg/ml in 0.055 M Tris-HCl buffer (pH
7.5), and 0.25 M NaCl. The crystals were grown by the standard hanging-drop
method in 3.2 M ammonium sulfate in 0.1 M Tris-HCl buffer at pH 7.0 with
protein concentrations ranging from 08-10%. After one to two weeks, small
beautifully formed yellow tetragonal bipyramid crystals appeared in several
wells. After about one month, the average size of the crystals was 1.0 x 0.7 x
0.5 mm (figure 13).

D. Oxidation State Change Experiments

The goal of this project was to examine each of the three oxidation states
of the D. vulgaris flavodoxin by x-ray crystallography.

42

 

Figure 13. Photograph of recombinant D. vulgaris flavodoxin - oxidized form.

43

The reduction'to the semiquinone of the oxidized crystals (figure 13) was
achieved as follows: a well holding some of the yellow crystals was selected,
one was transferred from the coverslip to the well solution; a small amount of
sodium dithionite (a few mg, perhaps) crystals were then carefully added to the
well. After 15-20 minutes, the yellow flavodoxin crystals take on a red/purple
color (Figure 14)(some of the smaller ones actually appear blue). It appears
that the amount of sodium dithionite added affects the diffraction quality of the
crystal. This observation was made when several reduction experiments on
crystals with similar size (ca. 0.7 x 0.8 x 0.8mm), showing the expected
red/purple color, gave poor diffraction spots during the crystal survey on the
area detector video display. It was difficult to determine the cause of the loss in
diffraction quality, one reason could be the addition of too much sodium
dithionite. An experiment on optimizing the amount of sodium dithionite could
be profitable, but it may not be necessary. During the first reduction
experiments, it seemed that it might be necessary to provide an oxygen-free
environment for the experiment because the amount of sodium dithionite
added was dependent on the amount of oxygen present (dithionite scavenges
oxygen before reacting with the oxidant) and because of the difficulty in
maintaining the crystals in their reduced state. As it turned out, the amount of
sodium dithionite needed was probably greater than in an oxygen-free
environment, but it was not enough to damage the crystals during the
reduction. It was proved possible to add just enough sodium dithionite crystals
to reduce the protein and maintain quality diffraction.

The reduction to the hydroquinone of the oxidized crystals was carried in
a similar manner to the semiquinone experiment except the pH of the well
solution was changed from 7.0 to 9.0 before the addition of sodium dithionite.
This was necessary because the reducing power of sodium dithionite is
pH-dependent and the crystals will not achieve the hydroquinone state at pH
7.0.

44

 

Figure 14. Photograph of recombinant D. vulgaris flavodoxin - semiquinone
form.

45

The oxidized crystals were transferred from a coverslip to the well solution, and
the pH of the well solution was then carefully adjusted to 9.0 by the addition of
a few drops of 1.0 M NaOH. Once the pH was adjusted, a few mg of sodium
dithionite was added to the well solution containing the oxidized protein
crystals. The pH was monitored with pH paper as a rough estimate and the
final pH was measured with a pH electrode. The final pH of the well solution
was 9.0 for the fully reduced crystals. After 30-40 minutes, the yellow oxidized
crystals take on the paler (straw) yellow color (Figure 15) of the hydroquinone
state (some of the small ones appear colorless). Crystals at the higher pH
never showed the characteristic red/purple semiquinone form color. The order
of the procedure is somewhat critical because the reduction from the
semiquinone to the hydroquinone, through the adjustment of pH after the
addition of the sodium dithionite, seems to requires a longer time (a couple of
hours). In addition, the longer time also seems to affect the diffraction quality of
the hydroquinone crystals. Perhaps the simultaneous change in the pH and
ionic strength of the crystals dramatically affects the protein conformation in its
crystalline state while in the well. The crystals seem to be stable for a long
period of time (several days) in the well solution in their oxidized state, if the pH
has not been changed. In the semiquinone state, if the crystals are allowed to
sit for several hours or more (9.9. overnight), they reoxidize. In the
hydroquinone state, if the crystals are allowed to sit for just a few hours (three
to four), they start reoxidizing to the semiquinone state and finally to the
oxidized state in several hours. Furthermore, when reduced crystals sit for
several hours, they deteriorate by taking on rough edges compared to the
oxidized crystals that were originally reduced with the sodium dithionite. The

same observation also occurs in the protein crystals at higher pH. I

46

 

 

Figure 15. Photograph of recombinant D. vulgaris flavodoxin - hydroquinone
form.

47
E. Data Collection

All data collections were obtained on a Nicolet P3F/Xentronics area

detector x-ray diffractometer system using monochromatized Cu Kor radiation
from a sealed tube source (35 mA, 50 W).

The crystal-to-detector distance (18 cm for all data collections in this
project) chosen is dependent on the shortest reciprocal axis and is thus set as
close to the crystal as possible without producing overlapping reflections. The

data collections consisted of a consecutive series of electronic pictures with co
advancing by a pre-specified amount and each picture in the series (frame)

taken at the same 1 and 4: setting and exposure time. For example, at x = 0°,
data can be easily collected without obstruction of the detector by the x circle

and thus an entire run could be obtained by setting (0 at 0° and stepping 4:.
However, the data near the rotation axis would be poor and other runs at

different 2; settings are needed for a complete data set.
1. Oxidized form
a. Room temperature

A clear, yellow, bipyramidal crystal of dimensions 1.2 x 0.8 x 0.6 mm,
mounted in a quartz capillary, was used for the room temperature data
collection (23°C). The entire data set consisted of eight frame sets or 2410
frames. The strategy for data collection consisted of recording 0.20° frames for

various psettings at x = 0°, then adjusting x and collecting at another 4: setting.

48

Since the z circle would occasionally block the outer edge of the area detector
causing errant observations. all of the co scan ranges were from 45 to 340° (i.e.

ca. 60%» scan) or as close to those values as possible depending on the 1
setting. Each resulting frame set would consist of approximately 300 frames.

The first three frame sets were collected at p =- 0, 60, and 120° at x :- 0°. The x
and rpcircles were then each positioned at 90° for another frame set. These
four frame sets were collected at 20 - 17° with an exposure time of 60 seconds
per frame. The next four frame sets were collected ,7, :- 0° and at three 4: settings
of 0, 55, and 115° and one setting of x and p at 90°. These four frame sets were

collected at 26 = 30° with an exposure time of 120 seconds per frame (Table
5a).

b. Low temperature

A clear, yellow, pyramidal crystal of dimensions 0.6 x 0.6 x 0.7 mm, was
selected for diffraction measurements and transferred to a depression spot
plate containing approximately two milliliters of Type A Cargille lens immersion
oil. The excess mother liquor adhering to the crystal was removed by careful
suction into a 0.2 mm capillary tip. The crystal mounting device consists of a
standard XYZ goniometer head holding a tapered brass pin with a standard 0.2
mm glass capillary attached to it. The glass capillary, meanwhile, holds a small
amount of ultrapure high-vacuum grease at its tip. The whole device was used
to lift the crystal from the oil. The microscope was set up with all of the
necessary tools on the tabletop of the diffractometer in order to minimize
transfer time.

49

As soon as the crystal was free from the immersion oil, it was rapidly
transferred to the diffractometer. The cold stream was deflected by holding a
stiff piece of plastic across the nozzle. The crystal was then mounted on the
diffractometer and the plastic obstruction was removed to allow the cold stream
to flow over the crystal. The crystal was cooled to -144°C and entire data set to
1.9A resolution consisted of eleven frame sets or 2630 frames. The strategy for
data collection was similar to the room temperature set except the exposure

times ranged from 60 to 180 seconds and settings for 4: were at 180, 225, and
0° and x settings were 45, 315 and 270°. The a) scan values ranged from 45 to

330° depending on the position of 1. Table 5a summarizes the settings for
each frame set.

2. Semiqulnone form

A large, red/purple crystal of dimensions 1.1 x 0.8 x 0.8mm was selected
for diffraction measurements and transferred to a depression spot plate
containing approximately two milliliters of Type A Cargille lens immersion oil by
the same procedure as in the low temperature study of the oxidized form. The
lens immersion oil was purged with nitrogen gas for several hours before use.
Since the kinetics of reoxidation to the oxidized form is moderate. once the
crystal was mounted on the end of the glass capillary, it was necessary to
rapidly transfer the crystal to the diffractometer. There was a small amount of
time (less than 10 seconds) to allow this due to the thin coating of oil on the
crystal after removing it from the hydrocarbon environment. The entire data set
was obtained from eight frame sets or 1963 frames. The frames sets were

collected at x settings of 0 and 300°, 4: settings of 20, 50, 80, 200 and 260°, 0) ‘

scan ranges from 45 to 339° and 26 settings of 20 and 30°. All exposure times
were at 120 seconds (Table 5b).

50

com
com
com
com
omN
0mm
0mm
omN
omw
com
com

mom
mom
new
com
omm
mom
mom
mom

a:

canon vendcwuo non uuaaoem aoauoednou oven

ONH
ONH
om
om
cm
on
ONH
ONH
cud
om
ow

ONH
oNH
can
ONH
om
ow
om
ow

chm
chm
chm
chm
mam
mam
mam
mam

mv

mc

mv

0°
OOOGGOOO

oma

omH
mNN
omH
mNN
oma
omH
mNN
oma

mHH
mm

om
om
ONH
om

mmmrmm
mmmrmm
ommrom
ommrom
memrom
ommro¢
mmmtmv
mmmrmv
wmmer
ommrcm
ommrom

muzuouoaeeu

vvmtmm
v¢mtm¢
mcmro¢
««mtov
vmm1ov
amniow
mmmtc¢
annrom

om
om
pH
ha
ha
ha
cm
on
om

o

o

GGJUQIIICI-ALDmI-lbx

30H 1 Snow uwuwcwxo

om
cm
cm
on
ha
ha
EH
EH

GIDUDEIIIHOI:

ousuouomswu Eoou r Show kufiufixo

udddulda

nuasnom cowuooaaoo puma mo mumEEdm

. an OHAQB

am

uuwldqdd

51
3. Hydroqulnone form

A clear, pale yellow pyramidal crystal of dimensions 0.5 x 0.4 x 0.4 mm,
mounted on a glass capillary - by the same method as the semiquinone form -
was used for a low temperature data collection. After many painful attempts at
trying to obtain a crystal suitable for x-ray analysis some comments are
necessary: (1) since the kinetics of reoxidation to the semiquinone is fast, it was
crucial to use freshly purged lens immersion oil and to work rapidly since the
time allowed to transfer the mounted crystal from the oil to the diffractometer
was 1-2 seconds; (2) the quality of the diffraction spots were strongly
dependent on the amount of base added for pH change and the amount of
sodium dithionite added to complete the reduction.

The entire data set comprised of ten frame sets or 2816 frames. The

frame sets were collected at x a 0, 30, 330, and 345°, 28 = 0, 20, and 30° and p
settings of 0, 45, 50, 60, 100, 130, 175, 245 and 300°. The exposure times

were either 90 or 120 seconds (depending on the position of the 28 arm).
Table 5b summarizes the settings for each frame set.

A summary of the data collections for each oxidation state is shown in
Table 6.

F. Data Processing

The XENGEN data processing software [21] was used to evaluate and
reduce the frame sets for each oxidation state of flavodoxin. The processing of
the data for each oxidation state was carried out in a similar manner. Before
using any of the frame data the format was changed from Harvard to XENGEN
with the aid of the program mg; each frame occupies approximately 512
blocks of disk space in the Harvard format and after conversion occupies

52

Ohm
0mm
mhm
mam
mam
oHN
com
com
mom
mom

MNN
mha
aha
mom
com
mom
mom
mam

dufidHNlNdlfdz ufiflqlidfifl

.nfiuom 23825035: 65— encode—Ugo Mom ova—Sex aowuooaaoo even

owa
ONH
oma
ONH
ONH

ca

ca

cm
ONH
ONH

ONH
ONH
ONH
ONH
ONH
ONH
ONH
ONH

com
com
com
com

COO

com
m¢N
mha
ooa
m¢
cm
oma
om
cm

on
ow
om
om
omN
com
omN
com

Hmmlmo
mmmta¢
mmmrov
mmmlN¢
mmmtmv
mmmrhm
ovm10¢
o¢m10¢
H¢mlmv
avMINv

ommrmm

olmm

olmm
mmmrmm
mmmtmv
¢mmlmv
mmmloe
mmmto«

nuaonom coauooaaoo puma mo aumEEdm

mm
cm
on
om
om

mN
ow

om
ON
on
on
on
an
ow
ow

(mUQthrtDZl-fb

macawsvou0>m

GmUDh‘lf-nwfl':

macaw:VwEom

.nn 03!...

uudldqud

Table 6. Summary of Data Collection Results

Space Group

53

P43212 (tetragonal)

Oxidation State W W Semis.
Unit Cell Parameters
a - b (A) 51.96 50.72 51.44
c (A) 139.86 139.04 139.62
Resolution Range oo-1.9A ”-1.9A ”—1.91;
No. Possible Refl. 15,055 14,745 15,439
No. Refl. Measured 61,291 41,222 60,332
No. Unique Refl. 13,588 11,288 13,507
Collected
Raym* 0.066 0.093 0.106
.N
2 2 |<Ihkl> - Ihklil
hkl 1-1
*Rsym =

 

.N
2 2 Ihkli

hkl 1-1

51.36
139.38

oo-Z .2511

9,614
53,148

9,313

0.068

where N is the number of symmetry related reflections.

54

approximately 520 blocks of disk space. Hence, in order to collect a single
frame set of 300 frames, one needs at least 156,000 blocks of disk space. In
order to begin processing the data a plate image determination was carried out
with the aid of the program Qanntata. Once the calibration database (uspline)
had been constructed, before processing the data, the program Basia; was
used on the first thirty frames of the initial room temperature frame set to define
the active pixels on the detector face. The program, Spats, was then used on
the first forty frames of the same frame set to obtain a list of the bright spots
(reflections) for refining the crystal and detector parameters. There were 225
well-behaved spots placed as the Centroids output on the disk. These centroids
along with the spline file were used as input parameters for the program
Eafina; the initial detector and crystal parameters were read from the uspline
and centroids files. This information was then edited using the menu (Figure
16) which appeared on the screen during the refinement of the crystal and
detector parameters. For example, the swing angle was changed to -17° which

corresponds to the 17° position of the 28 arm. The cell parameters were
auto-indexed with the fractional error on the unit cell lengths and angles set at
0.1. The parameters were refined with two cycles of closed-form refinement
and several hundred cycles of nonlinear least squares (choices are 0-9: 9 was
chosen) on the crystal and detector parameters, simultaneously. The output
file, uparams, was generated containing the refined parameters. The
refinement reached convergence when most (greater than 90%) of the
reflections showed reasonable integemess (i.e. when most of the reflections
had index errors less than 0.1). After refinement, the first sixty frames were
integrated for intensities and their standard deviations with the program
intagtata. The program read the two mask files, amask and wmask from flame:
and Santa. respectively, as well as the uspline, uparams, and frame data; at
regular intervals the crystal parameters were automatically refined and a new
uparams file was constructed. After integration, the data were reduced with the

program Baguaa using the flag -c;

55

Parameter refinement performed Mon Aug 21 16:09:39 1989

 

Refining? a:Y b:N c:Y alpha:N beta:N gamma:N omega:Y chi:Y phi:¥
Value: 51.712 51.712 140.144 90.000 90.000 90.000 92.534 -63.409 118.141
Shift: 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
p1ateD:Y XcenzY YcenzY tilt:Y swing:Y gam0:Y gam1:Y gam2:N
Value: 18.2316 0.0384 -0.0533 -0.2224 -17.053 2.1961 -0.5491 0.0026
Shift: 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Max Errors Allowed in: Resolution Wave- Crystal Space Step
index leix) Yipix) phi(fm) Min Max length System Group .size
0.2000 5.0000 5.0000 5.0000 20.0000 2.1477 1.5418 4 96 12
Alignment:a along X , b along Y c along .Mainbeam-58.378 265.331
Command(R,L,P,H,A,W,E,Q,S,C,U,0-9): U ( TYPE F1 TO DO IT, F2 FOR HELP }
----- Root-Mean-Squared Errors In: <#frms # of 9 of
Index X Y phi Gamma Profile moved) refls indices
0.0453 0.8914 0.7120 0.6109 0.0000 0.0000 0.0000 165 495 all
0.0449 0.8749 0.7141 0.4706 164 492 refined
Shells:D> 7.62 D) 5.35 D> 4.54 D> 4.10 D> 3.78 D> 3.53 D> 3.31 D) 2.68
errlh) 0.0379 0.0472 0.0359 0.0414 0.0337 0.0481' 0.0433 0.0627
ind/ref 60/ 20 60/ 20 60/ 20 60/ 20 60/ 20 60/ 20 60/ 20 75/ 25
Index errlhl> 0.000 0.050 0.100 0.150 0.200 0.250 0.300 0.350 0.400 0.450
h 159 5 1 0 0 0 0 O 0 0
k 149 15 l 0 0 0 0 0 0 0
l 84 56 22 3 0 0 0 0 0 0

Commanle,L,P,H,A,W,E,Q,S,C,U,0-9):
Meanings of the

<RIGHT>: next value <CR>: next line

<DOWN>: down 1 line 2: go to top
(LEFT): previous val $:
<UP>: prev line “L: redraw menu
?: define param i:

Commands
R: refine rocking curve; L: linear lsq
H: nonlinear ref on H; A:
E: write params & exit; S:
U: update statistics; 0-9:

Figure 16.

go to bottom f3: undo delta

nonlinear ref,all vbls; W:
specify hkls in file
nonlinear ref: 0-P,9-H:

R { TYPE F1 TO DO IT, F2 FOR HELP }

Control Keys:
f1: perform command @: restore orig-
f2: print help inal value
#: restore all

f4: undo all deltas orig values

redraw stats <BS>,<DEL>: erase previous character

refinement: P: nonlinear ref on phi;
write params: Q: quit:
closed-form refinement;

auto-index: M: remap

C:
I:

Parameter Refinement rile

56

this created a new centroids file which was then used as input for another cycle
of refinement with Bafina. This was carried out so that new, improved uparams ‘
and wmask files could be used for integrating the entire frame set. After
integrating the full set of frames, the output file, urefls, was reduced with the
program am. The flag -z was chosen so that misindexed reflections were
deleted before generating the multiref file. Since the eleven different areas of
the detector face must be scaled together, a group size of 5° scanning angle
(default) was used. The program SaaLaj, was then used to perform least
squares scaling of the different groups. The scaling was carried out with two
parameters (-p2) and in twenty cycles (£20) or until convergence.

The programs Qatjtzrata and 8.01031. were executed before and on the
initial frame set because all of the frame sets were collected at 18 cm and
because the active portion of the detector face only needed to be defined once;
this same procedure was carried out for each of the eight frame sets of the
room temperature-oxidized form data. After reducing all of the frame sets, the
program Manama was used to merge the eight multiref files into one large
multiref file. This multiref file was then scaled (with Ssalaj) and at this point, the
final rms R-factor on intensity was about 0.20. The program Bajagt was used to
examine outlying or suspicious data; approximately 480 reflections were
deleted from the multiref file (about 3%). This left about 90% of the data
between infinity and 1.9A which was examined with the program Stats (Table
7). There are 1074 reflections missing from the data set; 13558 out of 15055
possible reflections were collected. The overall R3,,“ is 6.7%.

The same recipe was used for processing all of the data sets for the three
low temperature flavodoxin structures. The program BQmaL was used at the
beginning of each initial frame set for each structure. The remaining programs
from Santa to Stats were executed as stated previously.

£37

mm.m

mm.mm
mw.mm
ho.mH
om.mH
mp.HH
vw.h

mm

mv.v
mm.NH
cm.m
vm.h
wH.w
mm.v
H¢.N

em

HmNHw hamm
Nth o
mmmm o
nmam o
mmnh ooH
vmmmH moo
mvomw oon

«no sauce 5 A

mew
o

o

o
Hm
mmm
mmN

h

mcoHuo>udmno no

ww.w
vv.hH
mm.MH
vm.oH
Nm.m
wo.m
m~.m

mm

mN.>
oo.wH
mo.~H
HQ.OH
0H.OH
am.m
Ho.m

mm

N~.h
mh.hH
hm.vH
co.HH
mo.0H
mm.m
mv.m

H“

0mm
0

h
0N
mHH
owv
mom

m

«UHumHumum ucoeuduOC

.umuHque ecu cH unmade mcoHum>uomno cc 30Hns uOu mcoHuomHumm

wom
N
mm
mm
wNN
mNm
mNN

m

on.wmm
mH.v0H
mh.mNH
mm.wvH
mo.mmH
ow.NmN
hm.mmn

.h.o.U

mmmH
onH
OMv
vow
awn
emu
vow

v

HQNHw
whmn
amen
nmam
mane
vmmmH
mwonN

nno ‘

uncoHuooHuwm cchmHz

.coHuooHuou o no mEOHum>uombo ooHoom HHm ecu auHmcmucH wcmuc>< x»
.ooH « >uHmcoucH co uouomurm ozHo>rouaH0mnm mmquHmm "mm

.OOH a m co Houomurm maHo>roboH0mnm councHoch "em

.ooH a auHmcoucH co nouoourm oaHo>rdu=H0mbo counoHoch "mm

.ooH a aanceucH co uouoourm pmumsvmrcouanwsca "mm

.oOH a auHmcoucH co uouoourm couosvmrcouonoz "Hm

omHN
own
NHw
ovm
ave
05H
«m

m

oncw
Now
ebb
Now
on
onH
NM

N

NmoH
vmm
chm
wmv
vmv
Hm
oH

H

.02 co>Hu :uHs chHuooHucm no .02

who
avm
mmH
mmH
wuH
mm
m

H

.ooH . goueeauem usuruormmmcoooo “.a.o.o

mmmMH
«HmH
mnNN
hmww
thN
mmvN
comm

mmomH me.N
mva ooo.~
anew HNH.~
00cm oom.~
«new hmm.m
hmnw vho.m
mmww mmr.v

OcHnqu oouueHHoo oHanmom Honda

acoHuOdHumm oomum no amassz

”aHHonm an noHocopcapom coHuo>uombo uo >un5§=m

ommMH
vaH
mnww
thN
hmww
mmvw
wmow

numx a

Honn mmo
o H

m ov
mN «NH
NmH Ham
vooH MHv
«mow me
omA owv

H QMHN
Na
nmN
mwm
how
Hmv
omH

ovv
omcom

vaN whoH HovH wma
vmw mvm mmv -m
mHm Hem Non vmm
mmm «av pom th
com hon HmH o~H
arm vnH hoH Hm

noH me mm mm

oNv oHv mv Nv
ce>HU :H .xvmeonxx

mmm
hhH
va
00H
mm
on
pH

0

and: «coauomauom no bones:
umHHucm an muouuourm one meHuHmceucH coHuoeHuom uo aucee=m

.auou bewHuonrewaueuemleu noon won nuHueHueum dowuoeHHoo oven

uszHBHZHhma

umHMuOB
Nth.H
oweo.~
HmON.N
NQN¢.N
wmhh.N
owom.m

unoc<.uaeHa

.m>< umsoq HHonm
www.mv mn.~HOH "mHmuos
-m.m m~.>m~ mm.H
pen.oH oo.mmn mo.~
www.mH HH.vwv H~.~
paw.v~ wo.mmc mv.~
mow.am am.thH oh.~
mom.va ~m.~mv~ om.m

>uHm .unoc<
rcducH .uHEHA
.xemam\» .com uozoq
oomuo>< domud>< Hszm
..b .waaunr

58

Table 8 shows the distribution of data from Stats for the low temperature -
oxidized form of flavodoxin. The data set consists of 11288 out of 14745
possible reflections from infinity to 1.9A without any data being deleted; 3240
reflections are missing (ca. 22%). The overall Ray,“ is 9.3%. Table 9 shows the
distribution of data from Stats for the semiquinone form of flavodoxin. The data
set consists of 13507 out of 15439 possible reflections; 1707 reflections are
missing (about 11%). The overall R3,," is 10.6%. finally, Table 10 contains the
data distribution from Stats for the hydroquinone form from infinity to 2.2511. The
data set consists of 9313 out of 9614 possible reflections with 235 reflections
being deleted from the data set by Bajast.; 301 reflections are missing (3%).
The overall Rsym is 8.0%

56)

wm.~H
mN.Nm
w¢.m~
om.v~
cw.mH
0N.mH
hm.HH

mm

mv.m
mm.HH
mm.oH
mm.m
mo.w
oo.m
vv.v

vm

NNNHv
mme
mmNm
vmwm
mvmm
NNNm
ommow

m
o
o
o
H
m
h

mNH

NN
moH

mno Houoa b A

m~.m

n~.wH
mw.mH
so.MH
mm.HH
No.oH
H~.m

mm

mcoHuo>uonbo uo

hm.oH
Nw.oH
wa.mH
wNuvH
om.NH
n~.HH
mo.0H

Nm

5

mm.oH
nm.oH
wN.hH
mv.mH
nm.nH
o~.~H
mm.m

Hm

hmm
o

H

w
mm
moH
CNN

w

moHumHumuw uceeuwnoc

«Hm

H

«H
vw
em

NeH
mhm

m

o>.mhm
mm.vhH
wm.wm~
mo.Hm~
ww.hmn
mw.vmm
No.waw

nwm
mm

HOH
anH
vNH
awn
vow

v

Nwva
mmoH
nmwm
vnwn
ovum
Nwwm
ommow

nbo o

vohH

mm

Non
hmm
hnn
mam
mmw

n

Nmmm
nvm
ohm
vvm
new
Nnm
NmN

w

.02 co>Ho nuH: acoHuooHuom uo

mmmm
van
Hmp
mwh
mam
va
#5

H

.02

ovm
ONH
mvo
mom
mow
0mm
we

ochmHz pououHHou poHmmom Homom

n
H

mmNH
NooH
hva
nmmH
vaH
vmmw
ommN

H

.a manna mew .

mvwv
mmnn
ommm
mmvm
NNVN
vaN
vvww

monthHmmo

H mmw.~ "menace
amm.H seam.fl
HHH.~ wvmo.~
nm~.~ nHmH.m
chm.~ m~H4.~
owo.m «Hop.~
nn~.v ~a~..n

unoc<.uasaa

acoHuooHuom vveum no amassz

"nHHezm >9 moHuceccaoom coHum>uonno

ooNHH
NmOH
fivhH
nan
vHoH
vnww
ammw

«wax u

mwmm onH
N mH

ow on

on owH
FMN Haw
«VOH pom
woow mmH
ooA owv

H ohmH
boH
HNM
woe
OHm
Hoe
HPH

ovv
occom

nHmH
Hnm
va
va
mom
HNN
NoH

owv
co>HU

HwNH
hmw
Hon
wmw
cmH
ooH
mm

OHV

HOOH
nww
on
oHN
HnH
mv
on

mv

no auoesam

oav
wnH
mvH
NOH
Hm
mm
mH

NV

:5 lxemeoan\»
£UH3 MCOHUUOHuflm HO H0963:
umHHozm an nuouoourm 0cm «OHancoucH EOHDUoHuom uo auuee=m

.o>< uozoq HHozm
«mm mvp.ow mo.onH “mHmuoa
we non.m -.n~m H¢.H
ow th.vH mH.~vv mo.~
mm Hmo.o~ mn.-m mH.~
mw nmo.on mH.mmm Hv.~
mH coo.o> om.v~oH op.~
EH «no.va 0H.omo~ ov.n
aan .umocc
o rcOucH .UHEHA
.».o«»\» .uom pesos
ooouo>¢ coouo>¢ HHmnm

.afiou UequHHOreusueuemaeu IOH non nuduedueum aoHuoeHHou even

.0 .eHAfiUnr

6C)

mm.hH
0H.oh
mv.mv
rm.Hm
Nm.HN
vh.vH
mN.oH

mm

hp.h
No.>N
vm.hH
Hm.NH
ow.m
wm.m
mm.m

vm

Nmnoo
moon
mhmw
vao
thHH
NvQVH
ommmH

mno Hmuoa

vm.oH
om.nm
mm.mN
wo.mH
mH.vH
o>.m

vm.o

mm

meH

vm

awn
are
raw

FA

acoHum>uoano uo

oo.mH
mn.wm
om.wN
hw.ON
no.mH
cw.NH
Hm.oH

Na

ohw

Nm

HmH
voN
HMN

h

mH.mH
vh.>n
ov.hN
wm.HN
oo.mH
mv.NH
wm.mH

Hm

one

on

va
GON
th
omn

w

«UHunHumum ucoeuouot

H mohH
bN
oNN
th
omn
va
ovm

m
.02

mn.Nmm
o>.oMH
ov.omH
mN.oHN
wm.¢mN
Nw.mHv

mo.mn0H

.m.O.U

Hth

PNH
Haw
mow
mHm
oov
cnm

v

Nnnom
omen
whom
vam
thHH
NvovH
ommmH

«no 4

vaN

mac
mHo
won
wen
va
mNN

n

mmoN
can
New
ans
mmN
and
now

N

ce>Ho :uHI acoHuooHuom no

mmNH
Hev
HoN
ahH
oNH
co
mMH

H
.oz

For
who
HmN
mmH
NmH
DOH
NoH

ocHunHz cuuoeHHoo anHmmom Homom

H

homm
mmmH
vaN
MQMN
nowN
movN
ammN

H

h mHnea wwm r WZOHHHthmo

acoHuouHuom ooeum uo guess:

unHchm an moHococcauom :oHuu>uemno uo >umEE=m

homnH
nomH
vaN
mmnN
nevN
movN
mmmN

mama Q

hNon mno
o o
N mH
NH mm
mmH GEN
ONoH mHv
bomH ovN
owA omv

H nmaH owON mooH vva moHH

wH

OPH
va
NHo
omv
vNN

ovv
mocom

mm

MOv
wHw
hmm
omN
hnH

ONv
co>Ho

hmH
Nmm
hon
0mm
omH
Nb

oHv
cH

can 0H4
mam mom
wen «am
hem mwa
no on
me pm
mv Nv
lxeeeoam\»

mmva mNo.~ unHouoa
vovN omm.H Heom.H
momN heo.N vnNo.N
Nva omN.N pth.N
mmmN vmm.N omam.N
NmmN ono.n vap.N
HopN own.e oowc.m
umo:<.uHEHA
.o>¢ Hosea HHunm
has NHH.Nq NH.mmo "mHouoe
anN HoH.n Hv.vw om.H
mNN v0v.n 0N.MMH No.N
NvH who.nH Hp.oHN mH.N
op Hmo.nN mp.ann o«.~
mm coo.Hu vh.ope mp.N
hN HNn.hHH mH.mmmH ov.m
>uHm .uaoc<
o rcuucH .uHEHA
.».eqa\» .uom pesos
emouo>¢ omouo>¢ HHezm

:uHa acOHuooHuom uo Hogan:
"aHHunm >n auouomMrm can moHuHmceucH coHuooHuom uo aum255m

sfluom enoadnvdade wow eoaueaueum aOHuoeHHou seen

.m__eHxaunv

.h OHQMB 03m l mZOHBHszmo

61

mvam moMN moo thH eNOH NNvH equ MeNH mNN Hon mHmm «Hem Hvo.m umHouos
mNNm NOH or NmN omH va owN NmN rm Hm swvH mva th.N oomN.N
mHHH omH ooH mmN pmH mmN HnN omH an on mth vomH opv.N onn.N
mHmo mow mOH mNN oHN NmN MNH NmH on on hHmH HhmH wmo.N wmnm.N
Nmmm Nnm th MNN NHH HHN em mmH HN HN ommH oonH mHo.m ovno.N
HmooH «em mvH mHH va NvN vNH mON HN m onH meH nom.n HmvN.m
HmoHH ohm mNH th an mnH NON mmN Hm vm NomH mmpH ONm.m memo.v
moo Hobos n A p w m e n N H Ochon couooHHoo oHnHmmom Henom unoc<.uHEHq
mcoHuo>uombo we .02 co>HO :uHs nEOHuooHuom we .02 ecoHuocHuom madam uo Hanan: .o>¢ “are; HHocm
"nHHonm >3 moHoccccspem coHum>uumno uo >ucEE=m
m>.HH en.m No.m mv.m on.p ow.an ovam anm pmvH oooH mHHH mva oovH HNmH men oo~ oHN.Hn vo.an "mHouoe
mv.vv HN.mH mp.mN mm.vN mn.nN mm.mHH mNNw pwvH o m we omH own Non owN Hm wnp.m pp.om mN.N
mm.mm HH.mH on.HN no.0N om.HN pm.vNH nHHh mva H oH bHH «on an pHv NvH he mo~.o mm.noH mm.N
Ho.vN ON.HH mm.mH No.mH om.wH cm.mMH meo pHmH mH mm mom nHv men va Ho «N ONN.vH mo.HmH om.N
Nm.vH Hv.o vm.m Hv.m ep.oH N>.mvH Nmom ommH HMH MMN mom pmn omH Nm oN ON ONn.pN mn.mmN no.N
om.m wm.m om.m om.m om.w wo.pwH HmooH onH mac van NNv me an on mH e me.mv ov.mwm mN.n
mN.> mo.m pm.v om.v NN.w ov.omN HmoHH NomH mvo pom HHm mHH Hm FN «H v onH.op nv.ooo oo.v
>an .unocc
mm vm mm Na Hm .m.o.o mno e mama e omA owv cvv ONv oHv mv NV 0 rcoucH .uHaHH
cocoa ce>HU cH AxooeoHaxw .xooana .uea Hosea

«UHumHuoum ucoeoeuoc :uHa meoHuooHuox uo uonaaz coouo>< ooauo>¢ HHozm

umHHonm >3 «Housewrm oco noHancoucH cOHuooHuom no huoeanm

.33 303.63an now 836383 33338 33 .3 .333.

IV REFINEMENT

A. Overview of Refinement Techniques

Once a three-dimensional model of a macromolecule has been constructed
from its electron density map, refinement is carried out because the model is
approximate. This is due in part to the relatively low resolution and inaccurate
initial phases of average quality initial maps. The initial models are useful for
showing gross features, such as folding, and outline of the molecule. In order
to understand the chemistry of the molecules, it is necessary to refine the
model parameters to the fullest extent. In the early 70's, Watenpaugh et al.
[101] demonstrated that the quality of protein models could be improved by
refinement methods similar to those commonly used to refine small molecule
models. The term refinement refers to the adjustment of the positional
parameters (x, y, 2}) and the thermal parameters (31') of the j atoms in the unit

cell of the structure in order to improve the agreement between the observed
structure factor amplitudes and calculated structure factor amplitudes from the
model. The process is iterative, with calculated phases changing as the cycles
progress. Convergence is obtained when the parameter changes are no
longer significant. One problem that can arise is: if the parameters are too far
away from the true solution convergence can be reached prematurely at a
false minimum which can lead to an inaccurate solution.

1. Fourier Methods
The easiest approach to refining a structure involves using the observed
structure factors and the calculated phases with the use of a Fourier series.

The resulting electron density map can be used to determine the peak
maxima by means of graphical interpolation.

62

63

Some of the problems with the method are the difficulty to refine the scale and
thermal parameters easily.

The difference density, Ap, makes use of the coefficients AF, which are the
differences between the observed and calculated structure factors, AF - Fo -
Fc. If the observed and calculated electron densities are expressed in terms of

F0 and Fc,

po(x.y.z) = 1N 2 E 2 F0...” exp (~2rri(hx + ky + lz))
h k r

and pc(x,y,z) = 1N 2‘, 2‘, X chk] exp (-2:ri(hx + ky + lz))
h k I

then

po - pc = UV 2 2 2 AFth exp (-21tl(hx + ky + '2»
h k I

Difference syntheses have the ability to show clearly the errors in the structure
and can be used as a basis of refinement in improving parameters of the

model. For instance, the errors in Ap resulting from the positional parameters

is shown in Figure 17. The result is a gradient in the AF synthesis and a

correction to a more positive region is in order.
2. Real Space Refinement

The formidable size of the computing task for refinement of protein
structures was the impetus for conceiving the Real Space Refinement method.
The two principal characteristics of this method are: first, the minimization of
“pobs - pmode,)2 dv for the observed and gaussian model densities and

second, the manner of dealing with the stereochemical aspects of the problem

64

do
:58 :_ .98 9.56% 2855...“ 8.83 - $83 .6325 8.83 6:... .2 65mm

8.83 - Amnooa /

 

 

 

 

 

65

through flexible chain techniques [45]. The first case considers refining p0.

The method, hence, does not refine the phases and so it is appropriate at the
early stages when the model is approximate and is being adjusted to an
multiple isomorphous replacement map. After the early stages, processes
which refine the phases are preferable. The second property of the

minimization makes use of the weighting residual 1N 2 IAFIZ, so all the
density is used with uniform weighting of the reflections. Another property

associated with such a refinement is that, once p0 has been determined, the

size of the computational task rises linearly (instead of quadratically as with
reciprocal space data) with the number of atoms. Finally, it is easy to refine
selected portions of the structure against data relating to that portion
independent of the other regions.

The stereochemical aspects of the problem consider treating the protein as
a flexible chain. The method treats only the internal rotations of the chain as
independent variables. Some advantages of this method deal with the
handling of planar groups such that the out-of-plane displacements are small
and large rotations about a bond; this method is discussed by Mandel et al.
[46].

3. Least Squares Methods

The principle of least squares states that the best values for the parameters
are those which minimize the sums of the squares of the differences between
observed and calculated values of the function for all observational points.
Thus, the following function is minimized:

m
2 Wr (For ' Fc,r)2

r-1

66

where wr is the statistical weight of observation r, F0, is the observed value at

r and Fa,r is the corresponding calculated value. In x-ray diffraction, the
functional form of the structure factor is transcendental and so must be
approximated by a truncated Taylor series. Thus,

D = X thi (lFol - lkFcl)2
m

where thr is the weight of the observation and Z is the summation over all

observations. The scale factor k is applied to Fc because if it is applied to F0,
least-squares will minimize by decreasing k and increasing the thermal
parameters, thus minimizing D and arriving at an incorrect solution. Since the
solution of D involves solving a system of equations, the use of matrix notation
is convenient. The normal equations may be written as follows:

a11x1+ 812 X2 + a13x3 + + 81")!" 3V1
821X1 + 322 X2 + 323X3 + + aZan = V2

an1x1+ 8.12 X2 4' an3X3 + ... + 8.1an 'Vn

m
where a: = prachrl/api achfl/Bpj x,=Ap,

I-1

m

"j = z Wr (AFr) a":crl/apj

I-1

67

The matrix form would be written as follows:

a11a12.........a1n X1 V1
821 822......"Qn X2 V2
3
a1“ 32".........ann Xn vn
or, Ax = v

Hence, if the normal equations have a solution, the following is true:

A'1AX = A'1V
x = A'1 v

where A'1 is the Inverse matrix. The matrix of the normal equations always
results in a square, that is, the number of rows is equal to the number of
columns. The line from the upper left to the lower right in a square matrix is
called the principal diagonal. Since the matrix is symmetric, that is a]; = aim the
elements of this principal diagonal are the sums of the squares. The
diagonal-least-squares approximation in which only the diagonal elements
are nonzero is illustrated in Figure 18. The actual magnitude of the ijth
coefficient depends on joint variation of |Fc| and the parameters off the

diagonal. If the parameters are correlated, however, there will not be a
random cancellation of these terms. One approach to circumventing this
problem is the block-diagonal least-squares approximation. The
block-diagonal approximation uses correlations between scale, thermal
parameters and positional parameters for each atom.

68

 

oo o

 

 

 

 

Figure 18. Diagonal Matrix

 

69

Thus, all atom blocks are a 4 x 4 matrix - each consisting of three positional
and one (over-all isotropic) thermal parameter. In figure 19, a variation on the
block-diagonal matrix least-squares approximation makes use of six thermal
parameters per atom (each atom is individually anisotropic) in which a 9 x 9
matrix replaces the 4 x 4 matrix [47].

One technique for refining the structures of macromolecules is through
increasing the number of observations by making use of restraining variances
of the interatomic parameters (mean atomic positions and the thermal
parameters). This is necessary since there is usually a paucity of diffraction
data and the problem is severely underdetermined. Next, the sheer size of the
computational problem is daunting. Finally, initial models of macromolecular
structures are inaccurate. Despite these problems, the wealth of knowledge
about the stereochemistry of macromolecular structures supplements the
limited diffraction information and hence makes the refinement problem
tractable. At the same time, these models that conform with known
stereochemical features have inherent chemical reasonableness. These
conditions on refinement against diffraction data serve to restrict models to a
realistic range of possibilities. Thus, they are termed restraints as opposed to
constraints, which confine to certain values.

The ratio of observations can be increased by increasing the number of
observations through restraints on bond lengths, bond angles, and torsion
angles. The number of parameters can be reduced by imposing constraints on
the same quantities. Whereas restraints introduce more observations by
providing ideal values for the relevant quantities, constraints specify strict ’
relationships between the quantities, thus reducing the number of
independent variables. Restraints act like springs pulling real atoms toward
each other; constraints are like rigid rods between atoms, reducing their
degrees of freedom.

70

 

Figure 19. Block-diagonal matrix with 9 x 9 blocks for
three positional and six thermal parameters
per atom.

71

Sussman etal. [50] have developed a program CORELS (Constrained -
Restrained Least Squares) in order to increase the ratio of observations to
parameters by constraining certain groups of atoms to a certain
stereochemistry and then refining them as rigid groups.

The principle of least squares states that the 'best' set of parameters are
those which minimize the weighted sum, over all observations, of the squared
residuals:

40‘) = 2 Wu [9°b3 - 9°“le

where the appropriate weights wh are the inverses of the variances of the
observations and g is an observed or calculated parameter. In this case there

are several qualitatively different classes of observations. The grand function
for minimization is

<D=X¢i

The observational function corresponding to structure factors is

reflections
41 = 21/0201) (|F°°°l -l|=°fi"°li2
h

where the weighting factor is determined by c (h substitutes for hkl). The class
of stereochemical information comprises three types of distances: actual bond
distances, next-to-nearest-neighbor distances from atom groups that define
bond angles, and first-to-fourth-atom distances that relate across planar
groups.

72

The resulting function is

dances
4,2 . 2 1 @20) (dideal - dmodel)2
l

where o is the standard deviation for the distribution of values expected from
distances of the particular type. The planar groups are another class of
importance. The method for dealing with planar groups, uses restraints of the
deviations of atoms from a least-squares plane through the group. The
corresponding function is

mplana
pla'iesabms

$3 a 2 z 1/0’2(1,k) (mk ‘ 11k - dk)2
k l

where m, and dk are parameters of the least-squares plane and o is the
standard deviation to be permitted [48]. The next group considers the
stereoconfiguration at chiral centers. This restraint treats the central atom as a
chiral volume and this volume is obtained from the triple scalar product of the
vectors from the central atom to the attached atoms that determine the

chirality. i.e. V = (rN - tea) . [(rc 40,1) x (r03 'VCall- The corresponding function is

Chiraloenhrs
.34 = Z 1/02(i) (videal . vmodel)2
i

The contacts between nonbonded atoms are characterized by potential
energy functions U(d). These functions feature a steep repulsive barrier and a
shallow attractive well.

73

The repulsive barrier is frequently represented as a Lennard-Jones or
Buckingham potential and can be approximated by

U(d) ' U(dm‘") - 1/02"(d - dmln)2n, where d < dmin

The best fittings in the range between - 2A and have n = 2 for contacts among
H,C,N,O, and S. The following observational function exemplifies this,

Nonbonded
oor'tacts

95 = 2 1/oN4(m) (dmmin - dmmodel)4
m

when taken only over ”repulsive” contacts, i.e., dmode. < dmin. The value of dmin
depends on the atomic elements in contact and on the type of contact:
single-torsion separated atom pairs, multiple-torsion separated pairs, or
possibly hydrogen-bonded pairs. The last group considers the conformational
torsion angles. These are among the least restricted of the stereochemical
features. The observational function is *

7.819108
4’6 g 2 1,62“) (xideal - xmod6|)2
i

which 3099*" is the angle of the torsional potential minimum to which x"‘°d°' is

currently nearest. o(t) is the standard deviation of the expected distribution. The

same kinds of restraints can be applied to thermal parameters as well as

positional parameters. Since the rms deviations of the atomic positions are

relatively high in macromolecules, the thermal parameters that describe these

deviations - whether they are static variations in the crystal lattice, dynamic

disorder, or actual thermal vibration - should be in good agreement with known

74

stereochemistry [49]. For example, the variance is in the magnitude of the
differences in the mean square displacements u2 of the two atoms joined through

a bond. Since B = 81ru2, the observational function is

Distances
¢7 - 2 1,620) (Barium . Btarget)2
I

where Borioin and Blame! are the thermal parameters of the pair of atoms related
through the bond.

A relatively new package of programs (TNT) has been developed to be as
flexible and general purpose as possible [51]. The package uses the principle of
restrained least-squares refinement. Stereochemistry is defined in a general way
that can be applied to proteins, nucleic acids, coenzymes, solvent atoms, etc..
The package also implements the Fast Fourier Transform algorithm to carry out
all crystallographic transformations. The two terms for minimization are the
crystallographic term and the stereochemical term. The crystallographic term
uses the space-group-specifio FFT [52] to calculate structure factors. The
gradients are calculated by a modified version of the procedure outlined by
Agarwal [53, 54]. The major goal of the stereochemical term is to make it as easy
as possible for the user to specify ideal bond lengths and angles. The
stereochemical restraints can be introduced as energy terms [42] or by
expressing all types of stereochemical restraints as distances [49]. There are
drawbacks to both approaches. For example, it may be difficult to obtain reliable
energy terms for unusual chemical groups or an inappropriate energy term might
disguise an unusual feature of the structure. Hence, proper weighting would be
imperative. in the case of interatomic distances, such distances need to be
determined from examples with ideal geometry. This kind of information might not
be available.

75

in addition, restraints on interatomic distances by means of specifying bond
angles can lead to distorted planarity of trigonal atoms and aromatic ring systems.

Conventional refinement involves a series of steps, each of which consists of a
few cycles of least-squares refinement with stereochemical restraints or
constraints followed by model rebuilding through interactive computer graphics
[55]. Finally, solvent molecules are included and alternative conformations for
some atoms in the protein are considered. This process is time-consuming,
because of the radius of convergence of least-squares algorithms. The radius of
convergence is theoretically d/4, where d is the smallest interplanar spacing
available from the experimental data. In addition, least-squares refinement is
easily trapped in local minima so manual intervention is generally required.

Molecular dynamics simulations are performed by solving Newton equations
of motion [56] with forces on the atoms derived from an empirical potential energy
that describe stereochemical and nonbonding interactions [57]. Molecular
dynamics is added into the crystallographic refinement by adding the effective
potential energy term:

Est = S X [ iFol - Well2

to the empirical potential energy used in the CHARMM program [58] and other
dynamics programs. The strategy of the approach is to calculate the Es, and it
derivatives from an atomic model using a harmonic approximation (to reduce
computational time); if an atom moves more than 0.4A then the Es, and its
derivatives are recalculated. This refinement with the effective potential energy
E5, is known as "MD refinement". A program package called X-PLOR, [59] is
currently gaining use throughout the crystallographic community for early to
midrange stages of refinement. One of the chief advantages of this method over
existing traditional methods is the ability to move atoms over a large distance - by

76

sampling space - maintaining stereochemical reasonableness, without manual
intervention.

Finally, a program, CEDAR [60], which uses a combination of energy
minimization, dynamics (by Jan Hermans and Michael Carson) and
crystallographic least squares (by Keith Watenpaugh) refinement, is currently
undergoing development at the Upjohn Company. The force terms that are used
in its calculations include those from bond lengths, bond angles, dihedral angles,
interatomic distances, van der Waal contacts and partial ionic charges. The
crystallographic least squares considers crystallographic data; the
symmetry-related contacts are used in the energy minimization, dynamics and
crystallographic least squares. Solvent molecules can also be included in the
crystallographic or energy minimization calculations. This program employs a
somewhat different approach from the stereochemically-restrained least squares
programs, such as PROLSO. Its principal difference is the use of energy terms in
lieu of stereochemical restraints, and the modifiable weighting of these energy
terms. A target value is calculated from the blockodiagonal part of the matrix and
the weighted energy terms are minimized in a least squares sense towards this
target value. In essence, the program employs crystallographic least-squares and
energy minimization, simultaneously.

4. Model rebuilding through Interactive computer graphics: FRODO

There are classes of errors in protein models that cannot be corrected by least
squares refinement. These errors lead the refinement process to converge at a
local minimum R value. Some of the errors might be due to (1) misorientation of
the peptide chain or side chain (i.e. incorrect torsion angles); (2) incompleteness
of the model, such as lack of ordered solvent molecules; (3) incorrect tracing of
the polypeptide chain; (4) uncertainty in the primary sequence.

77

An extensive study of maps plotted on stacked plastic - Plexiglas - sheets was
the method for determining the MIR structure of D. vulgaris flavodoxin at 2.5A
resolution [41]. A slower method, somewhat time-consuming and tedious, makes
use of a Richards optical comparator [96] to optically simulate the superposition of
the model and the electron density. This method was shown to be effective in
determining the structure of KDPG aldolase by Mavridis et al. [43].

Computer graphics is the most powerful way to allow the crystallographer to
interact with the protein model. The display unit, such as an Evans & Sutherland
P8390, allows the investigator to fit the protein model to the electron density map
and hence determine the 'best fit'. Efforts to develop the necessary software have
been made available through several versions of a molecule-building program
entitled FRODO [44.65.9899]. The program allows the crystallographer to: (1)
display and examine the model; (2) label atoms and residues; (3) regularize bond
lengths and angles; (3) manipulate atoms or molecules; and (4) measure
distances, angles, torsion angles and specific distances. The program is divided
into five different routines: (1) SAM - to create, examine, and manipulate the
atomic coordinate file; (2) CHAT - to set data file names and program constants;
(3) INTER - interactive display mode: the mode in which the crystallographer uses
the data tablet, dials, and terminal function keys; (4) REFINE - model
regularization routine which has a predefined dictionary containing the ideal
bond lengths and angles to be applied to the atomic coordinates; (5) MOLCUL -
static molecular model routine which is used for comparison with the working
model and to display for photographic or demonstration purposes.

Once the program has been downloaded to the host computer (e.g VAX8800
at Upjohn) a constants file - which contains FRODO parameters, file names,
display parameters, contour and map parameters and regularization parameters -
is requested as input. The program then enters the CHAT mode and commands
can be entered from the keyboard to either enter the SAM, MOL, or INTER mode
or to select which atoms or range of atoms and electron density maps to display.

78

The routine SAM is used to manipulate the atomic coordinate file'or to return to
the CHAT mode. If one is creating a new coordinate file or if there is a need to
reinitialize the current one the command EMPTY is invoked and the program
prompts the user to enter the number of sequence records plus any extra space
for insertion or deletions such as solvent atoms. A coordinate file in either PDB
(Protein Data Bank), WH (Wayne Hendrickson - PROLSQ) or DI (Robert Diamond
- BUILDR) format can be read into the program for display and manipulation. After
reading in the coordinate file the symmetry operations and cell parameters can
be added in through the use of SYMMETRY (the Cambridge version of FRODO
will allow the user to enter the symmetry by giving the space group number).
Finally, an output file can be generated through the routines MAKEPDB,
MAKEWH or MAKEDIAM for the PDB, WH or DlAM formats, respectively.

In the CHAT mode, the routine NOW allows one to examine the current list of
FRODO parameters. Once commands are changed, their new values are stored
- in the constants file. For example, in fiavodoxin, to display the range of residues
near the FMN, one could enter the command ZONE 60 65 to display residues 60
to 65. Symmetry can be turned on or off through SYMM or NOSYMM,
respectively. There are a number of CHAT commands that allow the user to
display electron density: BACK and NOBACK turn the electron density on and off;
MAP allows one to select which maps to display (a typical example would be to
display 2i-‘o-Fc and Fo-Fc maps, simultaneously). Another feature of MAP is to

display the same map at a different contour level and/or color. For example, when
one of the models of flavodoxin was examined, the 2Fo-Fc map was displayed at
two different contour levels (10 and 0.80) and the Fo-Fc map was displayed at 36
and -30. All four of these maps were displayed simultaneously with four different
colors to facilitate interpretation. The command GO allows one to enter the INTER
(interactive display) mode.

In the INTER mode, there are twelve function key bars at the terminal to allow
the user to turn different parts of the display menu on and off.

79

The seven keys are: (1) MASTER - allows the user to select any of the seven
keys; (2) DISPLAY - allows the user to display the menu, axes, atoms, etc. on or
off; (3) VIEW - allows the user to look down a crystallographic axis or rotate
accordingly; (4) MAP - to turn electron density maps on and off; (5) MOLCUL - to
turn any one of eight MOL static images on and off; (6) DIALS - used to select
which of the ten dial sets are to be used; (7) MISC - to set focus cross on or off; to
blank dial labels. Also, In the interactive mode the menu and dials are active; the
eight control dials allow the user to rotate, translate, move atom fragments or
individual atoms, rotate torsion angles, change colors of MOL Images, electron
density maps, or atoms, and change to stereo. The graphics tablet - which Is
mouse driven - provides a means to select menu items such as FBRT
(Foreground Background Rotate Translate), MOVE, and TOR. FBRT is an
Interesting choice that lets the user select a molecular fragment and move (rotate
or translate) it in any direction. Once a fragment Is moved to the desired location,
the SAVE command Is selected and the new coordinates are written to disk.

The REFINE mode allows the user to regularize the new changes of the
fragment to the rest of the model. It uses the method of Hermans and McQueen
[97] which minimizes the weighted sum of terms representing the shift from its
starting positions, errors In bond lengths, bond angles, and fixed torsion angles. If
there is difficulty keeping atoms in the density during the regularization
procedure, the FIX option can be selected. The atom to be fixed Is then selected,
the REFI option is selected to allow the user to enter the range of residues to
regularize.

The MOLCUL mode (accessed through CHAT or the MOL menu item) Is to
generate static background images containing stick models of molecules or
selected'regions of a molecule. These images are useful for comparing with the
working model or for demonstration or photographic purposes. Its limitations are
the atoms displayed cannot be manipulated.

V RESULTS

A. Description

A ribbon representation [61] of the folding of flavodoxin is shown in figure
20. The overall shape can described as an oblate spheroid with approximate
dimensions of 25.4 x 40A x 40A. The core of the molecule consists of a

five-strand parallel pleated sheet flanked on each side by two sets of or helices.
There are 147 residues in the stmcture with about one-third of the residues

contributing to the 8 sheet, one third contributing to helices, and the final third
In turns with other conformations. The regions of the molecule that are involved

In the five [3 strands are: residues 2-10, 31-38, 51-60, 85-95, and 125-129. The
four or helical regions are: residues 13-28, 70-81, 103-115, and 132-148. The

remainder of the molecule consists mostly of ,8 turns: residues 10-13 and 27-30
are Type I, 37-40, 97-100 and 128-131 are Type III, 83-86 is Type II and 61-64
is Type II' (Table 11). One interesting turn region (Figure 21), which lies

between residues 42-48 (AGGLFE), shows a Type II followed by a Type I [3
turn. This secondary structure motif Is not completely unique since it Is also
present in thermolysin region 227-232 (NGGVHI) [62], a completely unrelated
structure. The packing of the molecule Is In a side by side, zig-zag, rod-like
fashion with adjacent parallel rods separated by large solvent channels. The
sheets of rods are stacked with successive layers alternating parallel to x and y
axes (figure 22).

The FMN lies mostly below the surface of the molecule. Some of the
hydrogen bonding gives rise to low thermal motion and minimal disorder.

80

82

HermNH

 

HHH HNI moi. NNH mot mmaw .mmm
HHH hMl mot bmH oml mem OOHIhm .mmm
HH . v on oNH mml xand omlmm .mmm
.HH oHI mml moi on Hmoaw molHo .mwm

H ml owl mm Not mme owlmv .mmm
HH mN mm ONH vml mHOU bvlvv .mmm
HHH le oml mNI mml >m¢a Hvlhm .mmm

H ml Hml oml ool M0<Q< HMIHN .mwm

H OH moHl Nvl mar GBBm MHloH .mwm

52 case no sea. m5 me «a: we coconcom

mouse a you moncc conuoa .HH oHnnB

83

 

Figure 21. Stereo representation of Type I/Type II [3 turn along residues 42-46.

 

85

The hydrogen bonding also stabilizes the or helices and 8 structure and so the
side chains and carbonyls are easily seen despite the high solvent content. For
example, there Is a hydrogen bond between ASP95 and M97) which occurs in
D. vulgaris as well as In A. nidulans, and Clostridium MP so it appears that Asp
may be conserved at this site not because of its interaction with the FMN group
but because it stabilizes the backbone conformation by forming the hydrogen
bond [63]. The isoalloxazine ring appears to be planar and is buried between
two segments of polypeptide chain and two side groups TRP60 and TYR98.
The TYR98 side-chain Is almost coplanar with the flavin group - approximately
8° between the plane normals - whereas TRP60 is not coplanar -
approximately 46° between plane normals. Depending on the bacterial source,
TYR98 is replaced by PHE or TRP in the sequence, so the presence of an
aromatic group does not appear to be coincidental. The hydrogen bonds
between the main chain atoms is shown in figure 23. Distances less than 3.2A
(arbitrarily chosen) between peptide nitrogens and carbonyl oxygens have
been considered hydrogen bonds.

B. Flavodoxin refinement

1 . Oxidized form

The protocol and parameters of the refinement of the room temperature
flavodoxin stmcture is summarized in Tables 12 and 13. The starting MIR
model [37] was refined for 12 cycles by segmented-body least squares using
CRYLSO, a small-molecule least squares program from the XT AL [64] system
of crystallographic programs. A segmented body (Figure 24) is defined as
either the side-chain (from CB to the end of the side-chain) or the planar part of

the main-chain (i.e. Ca-C(O)-N). The segmenting was carried out throughout

the whole structure and 16 refinement cycles decreased the R-value from
0.42-0.33 using data ranging from 10-2.0A and an overall thermal parameter of
20.0.

86

.829. a .o 33 92 .3 Exam:

“$5 a page 9: 9: 32m 2 :66 was 538 on... .AEofi 3.85:5
mesa casein—e c323 maven $58.06 863 .298 29808

0 as 2 £3828 o5 a .88 1sz seem as a 828. 65 as. .58
«323 9:. .5532 8528. 35888 $858 as. has: 2:. ...me
5 mean 520 can. c323 «econ c323: 8.25% 8835 .3 2:9...

 

 

 

 

 

 

 

 

 

 

GrahaJku. inmolutuut Sokmun:

0

33

20

55

83

111

123

Table 12.

10.0-2.00A

10.0-2.00A

8.0-2.80A

8.0-2.50A

8.0-2.25A
8.0-2.00A

8.0-1.90A

87

0

45

32

52

87

130

0.

0

42

.27

.24

.20

.19

.18

.17

Refinement Summary of Oxidized State (RT)

B:uflms Commuma

MIR model,Wild-type

CRYLSO,
Groups,

Seg. Body
CEDAR, 4

Frodo sessions

PROLSQ,
2 Frodo

PROLSQ,
3 Frodo

1 Frodo
2 Frodo

2 Frodo

B-20.0,
sessions

Ind. B's,
sessions

session
sessions

sessions

88

Table 13. Summary of least-squares refinement parameters‘l

 

OXRT OXLT SO HO
Observed reflections 13,588 11,288 13,507 12,471
Reflections used In refinement 11,964 10,459 10,568 8,694
R 0.170 0.199 0.203 0.213
f(s) 2.46 3.05 2.42 1.82
Geometrical conformity
Targets Final model

Distances (A)

1-2 0.030 0.024 0.022 0.034 0.030

1-3 0.040 0.054 0.056 0.060 0.059

1-4 0.050 0.050 0.057 0.065 0.057
Planes (A)

Peptides 0.040 0.018 0.019 0.034 0.025

Other 0.040 0.018 0.018 0.044 0.025
Chiral volumes (A3) 0.300 0.241 0.245 0.337 0.341
Non-bonded contacts (A) .

Single Torsion 0.500 0.188 0.199 0.210 0.240

Possible I-I-bonds 0.500 0.320 0.373 0.353 0.356

Other 0.500 0.193 0.235 0.219 0.304
Thermal parameters (A2)

1-2 (main-chain atoms) 1.500 1.811 1.650 1.300 1.263

1-3 3.000 2.667 2.355 1.988 1 .987

1-2 (side-chain atoms) 2.000 3.638 2.951 2.630 2.030

1-3 4.000 5.601 4.000 3.981 3.059

 

tFinal model values are the root mean square deviation from the expected
geometry for the restraint class. Thermal parameter values are the mean
difference in isotropic temperature factors between pairs of atoms. The notation
1-2, 1-3, 1-4 refers to atoms related by a bond, bond angle, or dihedral angle,
respectively. The target a Is the Inverse square root of the weight applied to this
type of geometry restraint throughout the refinement. Reflection residuals are
weighted by 1/f(s)2, where f(s)~1/2 <|Fo - Fc|>.

OXRT = Oxidized form - Room Temperature (Resolution used: 8.0-1.9A)
OXLT = Oxidized form - Low Temperature ( (Resolution used: 8.0-1.9A)
SQ = Semiqulnone form (Resolution used: 8.0-1.9A)

HO a Hydroquinone form (Resolution used: 8.0-2.25A)

89

3536258930. 25 8956 520 3268on @5265 5295852 x26 w :3 .vN 35mm

 

 

 

:35 85

 

 

520 59: co :3 BEE

....

...........

90

Two electron density maps were calculated (2Fo - Fc and F0 - F6), and the
model was rebuilt as needed using FRODO, a computer-graphics molecule
building program [55, 65].

The model was then refined with 8 cycles of CEDAR [60] with each cycle
of CEDAR calculating ten cycles of energy minimization. Two electron density
maps (2Fo - Fc and F0 - Fc) were'computed and the model was rebuilt as
needed. Solvent water molecules were added to the model during the
rebuilding session with much care taken to Insure good selection. In summary,
a total of 45 waters were added to the model using good hydrogen bonding
distances and discrete globular sections of positive difference density as
criteria for solvent selection. After 32 cycles of CEDAR refinement, the R-value
converged to 0.27 and appeared to reach a local minimum. Since CEDAR was
undergoing development and thus not completely tested, it was decided to
switch to PROLSQ [49], a restrained least squares program which minimizes
differences between observed and calculated structure factors as well as a
residual based on differences between idealized and observed geometry.

The refinement began with this final CEDAR model minus the solvent
atoms. This time, a smaller amount of data (8-2.8A) was used with an overall
temperature factor of 20.0. After 20 cycles of least squares with PROLSO and
two graphics rebuilding sessions, the R-value decreased to 0.24. The data
were then extended to 2.5A and individual temperature factors were assigned.
After 35 more cycles with PROLSO. three graphics rebuilding sessions and
addition of 35 waters, the R-value decreased to 0.20. The remainder of the
refinement was carried out in a similar manner. The final R-value Is 0.17, with
data from 8-1.9A, and 130 waters.

After several cycles of least squares refinement and the construction of the
two aforementioned electron density maps, the first model rebuilding pass
through the molecule was mostly to check for gross errors - such as misplaced

91

main or side chain atoms - that could 'easIly be corrected by the simple
adjustment of torsion angles. The next pass was for checking the geometrical
integrity of the molecular model such as proper main chain folding through
verification by a Ramachandran plot (menu option - RAMA). While fitting the
side chains to their corresponding density, low energy conformations were

initially considered; 3.9. x, values should be -60°. 60°, i180°. This type of

density fitting was achieved by Initially adjusting the torsion angles to low
energy values; then if most of the side chain was lying out of the density, the
following was carried out: first, one of the bonds was broken between the
fragment lying outside the density and the rest of the molecule; second, FBRT
was chosen from the menu and the fragment was selected; third, the fragment
was rotated Into the density. This technique maintains the low energy torsion
angle geometry and still places the fragment in the electron density.

The refinement of the low temperature flavodoxin (oxidized) is
summarized in Tables 13 and 14. The final room temperature model without
the solvent atoms was used as a starting model for the low temperature data.

The process began by refining the model as a rigid-body with data from
s-3.5A for four cycles with CRYLSQ. The R-value decreased from 0.559 to
0.497. The next eight refinement cycles, using the same range of data was on

the model partitioned into secondary structure groups. The or helices were

each partitioned into groups, as were the [3 strands, and so forth. The R-value
decreased from 0.497 to 0.400. The data were then extended to 2.8A and four
cycles with CRYLSQ decreased the R-value to 0.361. An overall thermal
parameter of 20A2 was held constant during the refinements with CRYLSQ.
This model was then used as a starting model for PROLSO. The procedure for
refinement is the same as that which was carried out for the room temperature
model. The final model is based on 67 cycles of PROLSO. data ranging from 8
- 1.9A, 253 waters, and an R-value of 0.195.

WWW

4

12

14

28

33

67

92

Table 14. Refinement Summary of Oxidized State (LT)

8.0-3.50A

8.0-2.80A

8.0-2.50A

8.0-2.25A

8.0-1.90A

0

89

257

R:mflms Commute

0.49

0.36

0.32

0.29

0.25

0.19

CRYLSQ, Rigid Body

CRYLSQ, Sec. Struc.
Groups

PROLSQ, B-15.0,
1 Frodo session

Ind. 8's, 2 Frodo
sessions

1 Frodo session

3 Frodo sessions

93
2. Semiqulnone form

From a similar structural study with Clostn'dium MP [36], It appeared that
the molecular conformations must be the same for both oxidized and
semi-reduced states, despite the large differences in diffraction intensifies
(T able 15). The R-factor differences - for the low temperature studies - are
probably due to differences in relative positions of the molecules In the unit
cell rather than conformational changes in the molecules themselves.

The coordinates from the low temperature oxidized model were employed
to Interpret the electron density for the semiquinone form. The refinement of the
semiquinone form of flavodoxin Is summarized in Table 13 and 16. The initial
model was refined for four cycles with CRYLSQ as a rigid-body in the same
manner as previously described for the low temperature oxidized state. The
R-value decreased from 0.51 to 0.41. The next four cycles using CRYLSQ, with
the model divided into groups of secondary structure, decreased the R-value t0 '
0.331. This model was used for refinement with PROLSO for 5 cycles with data
from 8-2.8A and an overall temperature factor of 20.0; the R-value decreased to
0.325. Electron density maps (2Fo - Fc and F0 - Fc) were calculated and the

model was rebuilt according to the density. During the Interactive graphics
session the region around the FMN, residues 61, 62, and 63, was in need of
reconstruction. It appeared that a conformational change of the apoprotein had
occurred around N(5) of the FMN, but the type of change was difficult to
determine. It was apparent that this region of the protein should be deleted and
then replaced later to prevent its Influence in biasing the phases for the new
maps. Hence, thirty atoms (Residues 62-65) were deleted from the model and it
was refined for four cycles using PROLSO, this time with an overall
temperature factor of 15.0 - a temperature factor in better agreement with the
low temperature oxidized model; the R-value decreased to 0.314.

94

Table 15. R-Statistics between Data Sets

Data Sets* R

 

Oxid. form (RT) vs. Oxid. form (LT) 0.35
Oxid. form (RT) vs. Semiquinone 0.39
Oxid. form (RT) vs. Hydroquinone 0.58
Oxid. form (LT) vs. Semiquinone 0.42
Oxid. form (LT) vs. Hydroquinone 0.57
Semiquinone vs. Hydroquinone 0.45

 

*Since the range of data for the hydroquinone form is only
2.25A, the R-value was only calculated to that resolution.

R is a standard crystallographic R-value comparison between
data sets based on their observed F's.

95

Table 16. Refinement Summary of Semiquinone State

Shmanku. Rmuumtnul Salaam. Rzmuus Cmmmmta

4 8.0-3.5A 0 0.41 CRYLSQ, Rigid-body

8 8.0-2.8A " 0.33 CRYLSQ, Sec. Struc.
. Groups

5 8.0-2.5A " 0.32 PROLSO, B-20.0,

1133 atoms, 1
Frodo session

8 " " 0.31 B-15.0, 1103 atoms,
1 Frodo session

12 " " 0.31 1114 atoms, 1 Frodo
session
15 " " 0.31 1125 atoms, 1 Frodo
session
26 " 1 0.28 1129 atoms, 2 Frodo
sessions, Ind. 3'3
32 " " 0.27 1126 atoms, 1 Frodo
session
47 " " 0.28 1128 atoms, 2 Frodo
sessions
58 " 51 0.24 1179 atoms, 2 Frodo
sessions
61 8.0-2.25A 129 0.24 1209 atoms, 1 Frodo
session
69 " " 0.24 1232 atoms, 1 Frodo
session
81 8.0-2.00A 140 0.23 1243 atoms, 1 Frodo
session
128 ” 236 0.20 1369 atoms, 4 Frodo

sessions

96

The two electron density maps were calculated and eleven atoms were
restored at better positions, is. ASP62, 64C, 640, and 650A(I.e. Ca). Three

refinement cycles of this newer model allowed better Interpretation of ILE65
and SER64. The only residue missing at this point was ASP63, due to 'poor
phasing' for this region. In actuality, this final residue was built with help of the
hydroquinone model. During the refinement and model rebuilding stages, It
was difficult to determine the position of Ca despite all of the other atoms fitting

the density well. Every time the atom was placed in its correct position, after
further refinement, negative density would appear at its position suggesting
either a higher temperature factor or an incorrect placement of the atom.
Neither of these cases actually applied. After examining this region In the
refined hydroquinone model, the conformation and placement of ASP63
showed good agreement with the semiquinone model. One possible
explanation for this refinement problem could be the presence of both oxidized
and semiquinone oxidation states in the crystal during the data collection (i.e.
the reduction to the semiquinone was not totally complete). The refinement with
PROLSO was completed in 128 cycles, data from 8-2.0A, 236 positions for
water, and an R-value of 0.20.

2. Hydroqulnone form

The coordinates from the semiquinone model were used to Interpret the
electron density for the hydroquinone form. The refinement of the
hydroquinone form of flavodoxin is summarized in Tables 13 and 17. The initial
model was divided Into groups of secondary structure in the same manner as
previously described for the semiquinone species. It was refined for eight
cycles with CRYLSQ and the R-value decreased to 0.34. This model was then
refined with PROLSO for 5 cycles with data from 8-2.8A and an overall
temperature factor of 15.0; the R-value decreased to 0.325. Electron density
maps (2Fo - F6 and F0 - Fc) were calculated and the model was rebuilt

according to the density.

97

Table 17. Refinement Summary of Hydroquinone State

QxQULnnl Basdhudan Sdhmmt R:mflms Cmmmmms

8 8.0-2.8A " 0.34 CRYLSQ, Sec. Struc.
Groups
5 " " 0.32 PROLSO, 8-15.0,
1 Frodo session
13 8.0-2.50A 24 0.28 Ind. B's, 1 Frodo
session
19 " 49 0.26 1 Frodo session
26 " 68 0.25 1 Frodo session
35 8.0-2.25A 78 0.27 1 Frodo session
40 " 118 0.26 1 Frodo session
47 " 167 0.24 1 Frodo session
53 " 182 0.23 1 Frodo session
59 " 215 0.22 1 Frodo session
64 " 236 0.22 1 Frodo session

72 " 244 0.21 1 Frodo session

98

The data were extended to 2.5A, individual temperature factors were assigned,
26 cycles of refinement, and a couple of graphics rebuilding sessions (mostly
for adding 68 water molecules) decreased the R-value to 0.25. The remainder
of the refinement was carried out in a similar manner with a final R-value of
0.21 and 244 sites for water.

C. Discussion

In assessing the final models of the four aforementioned structures, some
general comments are needed. One interesting point to note is the
Ramachandran plot [100] of the two oxidized stmctures. In each case, the main
chain dihedral angles conform to the 'allowed' regions of the plot except the

values for ASP62 (¢=80°, wan-70°)(Figures 25 and 26) despite the good fit of the
side chain to the electron density. The same plots of the semiquinone and

hydroquinone shows ASP62 having moved into the 'allowed' region (can-110°,

w=-63°)(figures 27 and 28).

Since there were significant differences in the diffraction data between the
oxidized and semiquinone states, as Indicated from the calculated R-value
between the two data sets, three models (Figure 30) [66] of conformational
changes between these states were proposed a few years ago. One of the
goals of this study was to confirm which of these models is correct.

The environment around the flavin In the oxidized state is shown in figure
29. The flavin lies between the two aromatic residues TRP60 and TYR98 and Is
nearly coplanar with TYR98. The distances between 0(2) of the flavin and
N(95) and N(102) are 3.14 and 2.95, respectively; 0(2) hydrogen bonds to
N(95) and N(102). The flavin hydrogen bonds from M3) to O(100) with a
distance of 3.13.4.

180

120

60

-60

-120

-180

99

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

vs“... .7 ‘ '1 ,
---.... \\ ‘4. - ~ : .- -
I- "O—e-" ‘ " \ ' . . .
+ \\ 1‘ ‘n . .0 . : '
i- 1’ «l- + \ l ‘-l '2‘. : - ~ i '-~
~“ ----- T + + 4 Q " l "'1 = : '1 ‘-
K ‘ \ 1. .r ‘ J+ ‘. “\ '-. ‘t. I. I .‘ E E
P +- 1‘ + + “ ‘1!” 1 v '- ' ‘. 'r
. - -. ... I s = = '-
b + .‘. + ’9' a," + + E ‘1 ‘7 2 .’ '\ - '\
. ' ‘P ' '\ ' .’ i '. l
fl‘un-‘nn 1 r \L al- 3; 1 \ 'x {\+ fit 1 L
I" ++ \‘ ) \\ I S I e‘
I- - ... + | 1+ l” 'v ‘. /\ 1 ‘ 1
I" ‘\ /---"""":.-J ' 3 ': I I
h , . .- ......... 2 ‘ = ,
.-/ ‘3 ’1” ' ' / \. I 5’
L ./ ' s ‘ 1 : ‘l :
/" ," a” -/ .N ' ‘ ' ;'
' ’.—- -' '_ I .
l- /.’ + I."/../ I :- ,..-.‘ “ a, ‘| \. I a .'
. .- . I ’ ' : . 1
—- " ,',’z/ 5' i ‘: : . r
v ..u . 1 - . . .
t m; 1 . : :
' m L. A. 0‘ . .
1 ”If ‘ 7‘. .v ;' . '. 1
s I‘ 7’ 541' \'\‘ i 1 5 "-. ‘- ‘-
. ~‘. I'M" “‘I s". s \ i. 's
t. \ - .- 'f m i r l
s‘ _/ - - ,u 4‘ J 1 . . ~,
'- ' a. - \, I I " \\- 1 g \ \

\ x ‘7’ ' ': b \ ' ' . 4
’- .\ "5.,H1 .0 .'; \‘*.’ : I
{\.\ \\‘ f :1 ... \Q 5 2

' ~ -‘ :: . 7 s

\.‘ at! ..\ + + g ,. \nl-s ,

r \ u. 8‘ . e 1 . a
-.‘ x ‘ _ ¢ _. 3 1 ‘l I
b “- \ ‘1 . 4. I ~. 1. \. .J
\_ N ‘\ r s
_"x_ 'L \ y \A a 9
\ \ V" \.f-.. ———.—r
K -. .
)- \.‘ K, + ;! fix“,
; |' I." + ’\ "\‘
' 1 . + + f‘.--'---.

- /' I " ’ "en.‘
........... .— , . .5.
.I --------------------- ' 3; fi 1 "3+- ".x‘ -° /

I ‘. ' .‘ r

-....-‘ 1' 9 __l 3‘“ : 2 P.

r .y , + . \ : : . -
"l . \
k .\ 0‘ ++ / . \- ._ . f
s s. .-., - \‘ L -
V. ...... f k .‘i ’. ‘ _. \'
,. o“..“. ' : ‘ ‘\.
r. ..... 4]", ,........x + him i,

~~ ...................... ‘1 """"""""" ‘r”. '5

r _, ...... . q
\‘s- """"""""""""""""
p .......................
I.
h
I-
L ................ . ................. .. .
'- ‘ .~.“~_~-/ ‘ I’.\ ’ 1
1"“.1 ..... ‘ """"" I. “1 " " -/
,_ ‘ ................ ' v - I I
I 4 ...... J \ g \ e .1 )
,..-- ......... ...... ,- \\. - . , - .
" """f ------------- -a‘ "‘\ \ " :_\\ \ '1 I /
...... + 1 4“ | \. . \ I ’-
‘l 4 1' ~44 n 41 1 r r 1 1.1 1 1 1 1 L 1 r J ‘ .4 L 1 1 A

 

 

Figure 25. Ramachandran plot of room temperature - oxidized form
(The boxes correspond to GLY residues).

180

120

60

-120

'180

100

 

 

 

 

 

 

 

 

 

 

 

 

 

 

we-..- “..‘M‘ \\ \‘.\ ‘ v :v ' f.
" -..""“”¥"m \ V‘ \_ 1 r : ;
+ \ \ '\\. '. ‘1 | 1
st- ‘- n t : B
' ‘ - l .
4, + 4'“ \ \‘ ‘1“ 1‘. r I ‘. 1‘ \
:. ‘ --. ... r 1' ‘i K: 1“ I: .1 It
‘.\ + ‘3". J ‘- \i. t. .' .‘ : '4
f- + ' I I, ‘.‘l l v ... 3 K
7' ‘ I o- ' ‘ ‘\ \ . . .
'.. + + 4? .- W .. t 1‘1‘ ,1 ,1 g ‘.
r- ‘. f 4' 33’ .‘r ... '- \\ ‘ I, E ‘.
'h ....... +* A I LA 2 'l l L
+ +' 4- + *\ IT- ) \_\ 'v : g
I + ’o‘ \ ' \ ‘4
F ' ‘ -’ ' a f.
.I + + t, .- ...... ,1 , . ,‘
\ / ... .......... .r . .' ' I
1. ‘1- 1 ’4'" / '. l
_ /" : ;‘,-’/ -. 3
. + I! I! l ..0 ‘4» .- l :1
" I /' 7 .f' / "_ t. .3
/ + l’.’/ (I. ’0- .‘. v ‘. "
)- ': 3r": ;' i ‘ .
i f". ‘ . 0 '
2 ”9f \‘ a r" \\ . '
"x. 3;". \4‘ g .1 5 3 D . fl 1.
'- . ,A‘ 'if,‘ “\1 ’1 ; I 'l ‘5 '
'5 ‘1‘ ; " / , r! ‘- -\ ‘ 1' . f \ .
--. / \..- :4 ‘: x .1 = s “-
i. \ ~~‘. “A ...! \ ‘u . : : \
R"‘\ ‘.‘ ', .l. \‘u. 5 : N
« \- 5:: § :
.' ‘ O 0'
p .~‘\ “‘§ ‘.‘:5 + ,. :3 "I.”3 I!
’~'\ “\_ ‘ + s. .0 .‘ I l
b \ -\ 51 l ._ + t, 0“" K”!
\ \I ’u, \\ I
' . :r A L. I
K $5" \f.
,. \\ ‘-. 2‘4“”.
‘ ‘- ‘1? «aunt‘-
’ 2 . \ a
-- I ’ " 3'4 * +1b;\\
P """""""""" r z/ " + + ‘\H .I "s.
P ........ ~ ........ fi’ ." + ‘P ‘,\' -' I, r.
' """ ‘ " 1' *‘“\ 5 l .
h “‘ ; 4" : " i. '\ l'
'\ \ ‘ ... + '4' 2‘ 3 ‘
4 1.4‘ .4°* ‘L -‘ fl \'
i ssssss 0" ‘ ‘
.-*~~.._) '. ‘ \
'4 ..--4” .......... J +- we \5 - j
..................................... """ 0" .-
- .......... ‘IV”
I ..... ‘.‘
A‘\\ ...........................
b ‘‘‘‘‘‘‘‘‘‘‘‘‘‘‘‘
h
P
)-
)-
"oo .............................. I ‘
p ‘ .... ‘-ee-'/. ‘ ,.\ ,-
I: .............. \‘ ' "x a l
o ........... h ......... ,e. . I .
1 ....... a \ I. 1 K I (I
"" ............. e - ' .e .t I I
...... _. ”"‘~--~-. ./ \_ \. ... . .'
F I """" ' "4R\ ‘ \ WK \ I, ,‘
'L.-1-‘.3 ‘I—1 l l 1 1 1 1 L 1 1 4 1 L 1 ‘4 ‘ at ‘L 1 L r t 1 I ‘ . ‘

 

 

Figure 26. Ramachandran plot of low tem

rature - oxidized form

(The boxes correspond to GL residues).

101

 

 

 

. § :
_ ; '. =. '.
\ ' I ' '._
‘. e, '- ‘ s
I " . 1
. I. A 3
.' : \\
IN '
r' \' j
‘l
I \ I
\
... ‘
/ 1 ' e
s
..--""" 1‘4". i '
.f’ ,.“ ' ..Inh' '- a
' " ‘0 I l
\ A”. :
t ‘7. XV : ,- t .
\\ E 1 : 5 ._
\\'- fl 3 ' ‘
\"\ I.) : : \ ~
\\‘J' E 5
'\ i I

 

 

 

 

U
K I I
‘1
I
I‘\
I".-‘
0"
.1
CI
.3 fl
.1
32‘.
-\
is
\
\\
\‘i
I
D

-120

 

-. A
......... “h“... /’

 

.....C... ‘o- ~

3
(«fix 1\
.. ...... 0-..... \\
. 18° "'1 1 "TWA—l 1 L 4 1 N

I
x , .~
' A. I _I
'2 {"~4K\
1 \h 1 L 1 1 \ I
-120 -60 O

 

 

 

 

 

 

 

Figure 27. Ramachandran plot oi semiquinone form
(The boxes correspond to GLY residues).

'0

P.

-180

102

 

 

 

 

 

 

 

 

+ $.. '1 ........ .v-OQ \Q‘ ‘ ‘A\ '1‘ 2‘5 ‘ z 1'. ' I..

b ........ f ‘. ‘ \ ‘ . : : 2
\ .‘ *‘x \‘ f ; g. z‘

1- + 1 “. I f | e ‘ n

+ 'P + \ s \ ". ; z : '..
‘ ... * I ‘.“ 1 ‘. .' ' a .\ ‘- ‘ '1
,_ no—~“ + ’ m + , \ .. ‘ 1 ,. 3 ‘- "

p \ # 4.“ 7” e ‘. \“\‘ l ..v; . . ‘1 ’,
‘- \ + 1 "+ ' -. ‘ 3 3 X
+ . + + / 4 ‘2 \\ ‘ 9' ; z I‘
u . b .‘ g Q
“ + ¢ + \ + ‘ + + \\". ‘ I '- ‘:
““ on - \L \ '1 1‘ Z I
120 I 1!- +\ a \\ 1' '. 1
I! t ‘ ,r \ '\ : x
)- / + \ 4, .‘° 1. I\ : Ii
.' I.- + + '\ / _.---"';“:.o4’ A" \\ E E
P + ‘5 f '1 .- / ‘ E I
L ./. u: ! ’/ ,‘I ‘b 5 a:
/' xw’ -.." ..u= e a
'- h. ‘. a" ' '. .' :
/ ’14:" o"! ------- ' I. \X : 3
h f. 1? III [I i a, ‘g E
I 5:” .\'\ : g
.A 1 '0 . 'n 1 '
6° ‘\ i5" Ii \ ‘ "i :I 1 I:
. ‘.. ‘ Mr’ ‘1 ‘- '. ‘.
\\ / I'f’l: \‘0‘\‘ €\ : \‘ \ .
‘-. ' .‘ \ .' 2 .
h - \s ' as.

\ ._ . .‘_‘ / + \ r‘ ' U a \é.‘ 3 1’ a \. \
Na \_ \‘ng' + !;: *$~..r : '
,_ -. \\ ‘ :' . = Z .-
\> x.‘ ‘\5_ g‘ g. “\‘s j :

’- \\ ‘\ ~ \\ : .‘g N“. t

‘N " \ .\\ ’ '1 I.

\ \, .' 1‘ ‘ B
b ‘5 \. r h‘s + f ‘. \ ‘1‘
0 L‘\ t. 1, ‘ "x
\ V» + .. \‘N
P ‘1‘ \‘ '\
: 1 '

r ‘i' I E + f: 4" ' \
"‘" ------- ’ . /i '9, + J.- .\\ /"
“.....-“ """ ,/ [a + *5 . \\+ I .- ,"m
. -. ........... .. . ++ _, a z:

....... f 4’ L). ‘. z ' "a
v.- \‘,‘- + (‘ + 4* 4* + E-k‘s‘ E g : n
\ ‘ \t . * €L“+’ \‘ z, ‘1 W

- 60 1.. -..r _— ‘1‘
+ t"""--- '. - J
1 .. i)- ------------ M? ’ , “' """" d \ --
- -------------- t' I \.\
C m ........ . ------ 0'0“". N
h-
'120
h
1-
’ ac- '''''''''' ..._— 0" _‘
i- a w‘- .”—/ \ ’P'\. /
t'.‘.’ .................. .-. ‘ V I I.
end-oooun..‘-.- '1 \ N. I '
I" """""""""""" ‘ \ \ . I I
L .......... V . """"" "' \JK‘ \ \t' . :0];
‘1' 1 -‘T‘m1 1 L 1 1 1 u 1 4 1 1 1 1 1 1 l 1 '\ 1 k 1 1 1 11 L

 

 

 

 

 

 

 

 

Figure 28. Ramachandran plot of hydroquinone form

(The boxes correspond to GLY residues).

103

       

al." £10

0'

Figure 29. Environment of the flavin in the oxidized state of D. vulgaris
flavodoxin.

 

6:05:98?“ 2: so cozmctecoo 8 £89: 8.5. .8 2:9“.

EUOE EB... . .muoE ccoomw .238 “9.1..

O
.... :2 OJ

N.. .2. 0.2

:25 , . 0 I
0 05032.: n O O

:9...

..u.

  

105

The distances of 0(4) of the flavin to N(62) and to WAT(155) are 3.36 and
2.58A, respectively and it also forms a hydrogen bond to N(62) and to
WAT(155). Finally, WAT(155) hydrogen bonding distances to C(62) and N(100)
are 2.64 and 3.02A (Table 18). Now, if a hydrogen were added to N(5) the
distance from N(5) of the flavin to N(62) would decrease from 3.39A to 2.39A
causing an unfavorable contact. Thus, a conformational change of the
polypeptide chain is necessary to relieve the unfavorable contact as well as
electrostatic interactions.

After examining difference maps between the oxidized and semiquinone
states, there appeared to be significant changes around the region of residue
62. Since reduction of the flavin to the semiquinone form places a hydrogen at
N(5), the first model, which actually is the correct one, proposed that peptide 62
rotates along the Ca(61)- Ca(62) vector to form a hydrogen bond between

C(61) and N(5). This model removes the unfavorable hydrogen-hydrogen
contact through the conformational change. The second model proposed the
same peptide rotated to form a hydrogen bond between C(61) and M64), but
the hydrogen bond to N(5) is not formed. Finally, a third model has the peptide
bond rotated to form a hydrogen bond between N(62) and C(64) again the
absence of the hydrogen bond to N(5).

The final part of the study was to examine the fully-reduced
(hydroquinone) state of the FMN. In this case, if a hydrogen is placed on the
N(1) position of the semi-reduced flavin, a similar conformational change
would be expected to occur along the peptide chain of residues 94-96;
however Ludwig et al. [67] have not observed any conformational change
between the semiquinone and hydroquinone states of the Clostridium MP
flavodoxin. There has been some speculation that a single electron transfer
occurs between these two states and so there is no hydrogen present to cause
packing rearrangements.

106

Ucon Umumousmflms

 

 

mo.~ Sim mm.~ S.~ Sflso 8
SA $.N med mh.m 3:2 8
:6 $4” mim $.N 8:3 8
86 Sim .. . 3:2 8
2.~ $.N mm.~ 22m as; 8
I 36 $4 «fin 3:2 8
$.N mm.~ $4.. om.~ 85:0 mo
«fin mm.~ mm.~ mad 8:5 so
-.m .. I .. 332 so
mod $.N mod Ed 822 no
i $.N mm.~ Sam Siaz «so
om.m Sum mo . m S.N 33m: «so
u .. om.~ 36 83¢: mac
n .. and $.m «3.23 meo
«flu $.N $.m $.N 830 who
36 56 8d m~.m .832 Hz
e~.~ S.~ $.m om.~ 83.2 No
eH.m mm.m HH.m mH.m .Ammvz No
36 86 was... .26 835 m2
1 i and i 832 so
mm.~ om.~ Sum mm.~ 83$. 3
l aim .. 1 22.5.. «o
86 mm.~ .. u :80 m2

am cm 5.58 958 ucmfioflfieuoumomm 22m

Afivoocmumfla

.ououo coaudUduo none new
unabdou no odououmomo 0:3 and ZZH can doosaon boson aooouuhn .au canes

107

Another reason for studying the hydroquinone form was to’ examine the
planarity of the isoalloxazine ring. From Dreiding model studies, the
isoalloxazine ring has a butterfly bend along the N(5)/N(10) line. This bend
was the topic of much discussion [68, 69] several years ago. Unfortunately, the
crystallographic model at 2.25A resolution does not provide any conclusive
evidence regarding this bend. The overall shape appears to be planar,
however, the atoms deviate slightly out of the ring plane to such an extent that
its conformation appears to be a slightly twisted chair.

1. Comparison of the room-temperature and low-temperature
oxidized state

The principal reasons for examining the oxidized form at low temperature
were to determine if the protein crystals can survive the sudden change in
temperature and still maintain quality diffraction and to decide if it is beneficial
to collect data at cryogenic temperatures. In order to assess the merit of
collecting data at low temperature, one needs to compare room and low
temperature structures and examine places where low temperature data
collection has improved the quality or clarity of the electron density maps. In the
room temperature model some of the side chain conformations were difficult to
determine; however, in the low temperature model, these same side chains
appear to be better ordered. Some examples of this observation are discussed
below. A plot [70] of the average deviation (main chain is 0.3A; side chain is
0.5A) in position between the room temperature and low temperature models
as a function of residue for both the main and side chains is shown in Figure
31. The number of waters nearly doubled as the temperature was lowered.

The positions of the side chain of HiSf42 appears to be better ordered in
the low temperature model over the room temperature model. Similarly, many
of the long floppy side chains, such as LYS3, GLU79, GLUBO, ARGZ4, and
others, appear to be in a more rigid conformation in the low temperature model.

108

.389: 229qu2 26. EB Eco. 5223 5:58 3ng E conflict 39m}. .8 2:9”.

m:c_mmm

Egzaazsaaaégs31:3§§§§E§§§§aEafisasééutéémsasaigfi5.532324%
a: am“ aw. s: 83 ac am 3. am am 8... am am E

c_o;o c_mz

c_m£o os_m

 

S
u011180d U!

.—0

N

.s.w.a

eaueqaggio

109

The greatest deviation between the two models occurs at ARGZ4. in the
room temperature model, the side chain extends outward and then curls
slightly back toward the main chain. The conformation of the side chain is

stabilized through a hydrogen bond network with NH1 and N; with 051 of
ASP28, and NH1 with WAT162. The low temperature model has the side chain

extending straight out with hydrogen bonds stabilizing it through Ne with 051

and 052 of ASP28, NH1 with WAT374, and NH2 with 052 of ASP62 (related
through symmetry).

Another example of different conformation for the two models occurs in
LYSB7. The room temperature model has the side chain extending straight out
into the solvent channel with the formation of a salt bridge between "C and 021

and 082 of GLU118, whereas in the low temperature model the side chain curls

slightly back toward the main chain with the formation of a salt bridge between
N; and the C terminus. The conformational differences are shown'in Figures

32-33.

As the temperature was decreased, an adjustment in packing of the
molecules resulted in a change in the cell parameters as well as changes in
the diffraction intensities (Table 15). in addition, as the temperature was
lowered, the crystal volume decreased by approximately 5% (Table 19). The
corresponding decrease in the molecular volume was about 3%. The
molecular volume calculation was based on the volume of the molecular
envelope from a Connolly dot surface calculation [71]. Since the crystals
contain approximately 60% solvent, the change in solvent volume from cooling
is then almost 7%.

110

4 ~‘Wati62 ‘éwmtsz

 

Figure 32. Stereo representation of the conformation of ARG24 in the room
temperature (top) and low temperature (bottom) oxidized form.

111

 

Figure 33. Stereo representation of the conformation of LYSB? in the room
temperature (top) and low temperature (bottom) oxidized form.

112

£28528

oomtam on 2.850 m E9. oocfiao ooo_o>=o 8.38.9: 9.. So: 88.3.8 m. 95.2, 8.39....22.~

 

 

.5385
«.85 8. SE P5988. €2.85 2:832
58.888
«.888 8.388 5.8.885 .&.oe=.o> Euro
5.88.2
v8.» :32 5.38.: 22.8.3". .0 .oz
<88 <3 <3 8.5.83.
8.8. 8.8. 3.8. o
8. B 3.6 8.8 .5. one
- - 8.8. o
- - 8.5 pm. gum
acoficoEE .80
28.3693... o:o:.:o_Eom 859.5

 

5.3.83: «.39.: .o .2 58 Euro .2 each

113

The other aspect of this study was to determine the utility of cryogenic data
collections. Since there was no increase in diffraction as a result of freezing the
crystals and the deterioration was minimized as discussed previously, perhaps
some of the disordered side chains in the room temperature model might
become ordered as a result of the cooling. This appears to be the case in the
following examples: in the room temperature model, GLU42 extends out into

the solvent channel with no apparent density (at to contour level) beyond Ca

and there are no hydrogen bonds to this side chain. Since this side chain is
disordered in the room temperature model and hence the conformation is
difficult to determine, perhaps at low temperature, the thermal parameters
would be a little lower and allow easier interpretation of the density. Indeed, in
the low temperature model, the side chain becomes slightly more ordered with
density covering the carboxyl group and a hydrogen bond between 081 and
WAT389 and WAT406 to help stabilize the conformation. In a similar case to
GLU42, the side chain conformation is better defined (albeit different) for
GLU99 in the low temperature model versus the room temperature model. That
is, there is density in both models but the low temperature model is improved
with a hydrogen bond between 032 and WAT221 and WAT308 and 021 with

WAT279 for support. The conformation of the side chains and the improved fit
to the electron density are shown in Figures 34-35. These are just two
examples of an improvement in density fitting as a result of cooling the crystals
to cryogenic temperatures. Most of the structure is well ordered at room
temperature and so the cooling did not cause a significant improvement in
fitting the model to density. A plot (Figures 36 and 37) of AB as a function of
residue [86] between the two different models shows the expected trend of
decreasing thermal parameter value as a result of the lowering the

temperature, i.e. the average AB of the main chain between the low and room

temperature model is 2.09; the average AB of the side chain between the low

and room temperature is 2.02.

114

03-” 0‘1 0; fi 021
h. 5.,
- 451‘? {y ‘ s“.
- ' : ’9" ¢ “: s : ‘ ’ ;§: 1-
J" “" 5'? 3““:7’??? “ 1‘- ‘ - ‘“ . .1 :31"? 4'
I 7- ‘ ’l’ ‘W’ “2. .‘.‘ ‘, " q . - 4‘ . r.“_ ‘ I
C ~ 6*... - g -"

No! 406 Wof 406
{‘1‘\ l2 ‘
‘5‘, - VI
[‘3' .43; war 339 E‘- .41“ Viol 389
‘o L 0 . .
if ‘ 322'“) f, 'pcz
GP 9’

"fi'r'fflfr\ t .
J ”5‘5!” ”’35 ”1‘ , “ I:
l ’ —- ’ ‘~ -
' 1 11ut-:/." U.“ 1“ ’, ‘ I I," x'.. 5"" .I' &$. -1 1
- 5 ‘ ”:‘L’ 05‘th ‘* " “£\ "QJ. .8 Q ~ I
.- 3; 43 *3; . .22- .~ t-wl 43 mi?“ '2:
r- .. II" * 1“ 1- + 2* \
'1“: ' ‘ I’_H? - “ g..." ' 5"..- ‘3
w" yam: =’ ' “5:“
fl.~.9 0":
u, N,

Figure 34. Stereo representation of the conformation of GLU42 in the room
temperature (top) and low temperature (bottom) oxidized form.

115

f éwlcl A l’; )-.c1
'1 1‘: ~ 4“ & r1 '1‘?” ~
'f.)“ d * " "ch 'gk‘ ‘W ’P‘. t‘ogz
hf. f.‘ J: t @‘ __ ’ .
’Q ‘I I " a": e.”
. a“;

r 305 ”Ml 305
v r
‘-I

. .- .f 2
{3127 fig, ‘3 of 221

‘3'; J
' I
l ‘ a:
7
. 7 f, a,
I

f
‘*’ .2 "u
my}... [{‘._ r,

-'.«5p/S’ '0’

*‘V‘. \ S ’7. 9
o r '

”In" 9 8

 

Figure 35. Stereo representation of the conformation of GLU99 in the room
temperature (top) and low temperature (bottom) oxidized form.

116

£80.:

852.8 2292.89 26. van Eco. 5922.. Sago 59: 9.. .o .29:qu RES... c. 85550 .8 2:9“.

 

 

 

 

 

 

m:u_mmm

o3 cm; 3. o: 2: cm on as cm on 3 on cm a.

_ — p r _ _ k _ — p _ — p — ”Pl
1:-
.ml

. > > 71} > c

I .m
.2
m—

=_e.o c.02

<8>-<8>

117

on— om— am—

e.—

n

6&—

.mo:o.mo. >40 2 9.83:8 8:55.885 8.89:
852.8 23.9382 26. new Eco. 52%.: 59.0 38 on. .o 38:53 .959: c. 85850 Km 239“.

cm

oaommom

on ms

cm on OV

p h n

on

ON 0—

n -

 

I‘ll

nus

.our

. “Fl

r OPI

 

 

 

 

 

I I I ll

 

 

.o—

.m—

.ON

 

 

c.o;o ou_m

nu

<8>-<8>

118
2. Comparlson of the oxldlzed and semiquinone states

The most obvious differences between the oxidized and semiquinone
structures occur at the peptide linking GLY61 and ASP62. The orientation of

this peptide, the central unit in a 6 turn, is changed significantly as a
consequence of reduction of the FMN by sodium dithionite (Figure 38). The
shape of the density of TRP60-GLY61-ASP62-ASP63 in the oxidized and
semiquinone forms is shown in Figure 39; the carbonyl group of GLY61 points
away from N(5) in the oxidized flavodoxin whereas the carbonyl appears to
point in the opposite direction (toward) with respect to N(5) in the semiquinone
form. The geometry at C(61) and the isoalloxazine N(5) favors a hydrogen
bond involving the proton at N(5) of the semiquinone form; the O to N distance
is 2.99A and the O...H-N atoms form an angle of 166°. The hydrogen bonding
network along this peptide in the two states is different. Figures 40 and 41
show the hydrogen bonding schemes to the isoalloxazine ring and ribityl and
phosphate sections of the FMN for the room temperature model, respectively.
This hydrogen bonding network is very similar to that which was proposed by
Watenpaugh [37]. This same network seems to remain present in each of the
low temperature structures; the only difference is the presence of some
additional waters near the ribityl oxygens. This is a result of the water structure
ordering as a result of cooling. The hydrogen bonding scheme to the
isoalloxazine ring in the semiquinone form is shown in Figure 42. Smith et al.
[36] have carried out a similar study of the CI. MP flavodoxin and they have
observed a similar conformational change in the polypeptide chain near N(5)
with the carbonyl group flipping to make the hydrogen bond. The folding
around the FMN of the D. vulgaris and CI. MP flavodoxins is different for each
case. The structural variability in this region, and the dependence of the redox
potential on the particular species raises the possibility that flavodoxins have
developed slightly different mechanisms for modulating redox potential (Table
20) [101].

119

.952 385.283 Em A... “33.2.8 5223 5:88 2689 c. 5.83% @0992 .8 2:9“.

m:v.wmm

.¢§§3,.a§§_«gygaéfiééfigaggmgg53.552533858322222:2:§:a:=§
3: am. am. a: as. aw so E. 8 sm av am am a.

c.m;o c.mz

c.0go oe.m

 

'— S
uolqlsod u;

N

'S'w'a

aoueuegggg

120

.z- ‘1
”'- ‘ 53-“ -_.
h K- $3"- - 's‘n T211
5? ‘ .1, «=2 . «an.
' (‘a‘r’R' pl v J 3:31N‘f”?
$191121“ 5.: ' “
. § . . _
, + 1w..- tfl . ~
. ‘5' E 'zw'. 1; -. .‘
11‘ ~" 1.
o ‘ v ’ - “
u‘ 3 - 13m
1 . , ‘5
..x £52: may .
“a .
Trp6o"- '~1;.‘:‘(,
\ '_ ‘i x . 3
1? 4.,
‘5‘» ma ‘5’
~‘ ' ' ~- ‘3. '3
‘ ' ' . . vu-
K”‘33- -
. . \
‘ I
‘t- 95"?"3
1335:: ix 4’3};
V12: *3."
at,’
,1",-
I‘ -
.' " . .M"... :, (F41
‘J'W - w, 5 .
1" " ‘91"? _v, (_ ..~-‘
. 21:11:} “3‘33“ .341. “f'
: :31; . ..-
" ' '11 45

‘ ‘
. Jig:- »
. . V 'f’ 4
' 1:, '5 -.
1 I I

I #3
’ - : Il‘ _.
\‘.§ "‘5 \M.":
I

 

J.
(3"

50,,

. fly
”on
1

hi9

m.

...-g.

”N
at
‘ .

””3

~ ~ D ‘
‘ '31 '33-. $31 $3..
1.. * 818152.35

.\
l.".

l.
.2

I.

. ’ 3 _

‘ 3'3““. ‘ I.
-' 5;“ ‘» .
'93:? 51335356 $13.33:
-. évzrg . ‘ - 1': :
is“? yi‘ ~ ‘7 "15'
{231‘ ‘Efii'i’w ‘it- ,t

n a. 3 afi‘}
.‘ . '2

Figure 39. Stereo view of electron density corresponding to residues 60 to 63 in
the FMN binding region of the oxidized (top) and semiquinone

(bottom) forms.

121

 

Figure 40. Hydrogen bonding network around the FMN in the oxidized form at
room temperature. Open circles indicate oxygen atoms, darkened
circles are nitrogens. Hydrogen bonds are indicated as dashed lines.

122

.85. banana 3 85.065 2m 3:3 536»: .239? 2m 8.2.0 8:956 .285 .8958
285:. 8.2.0 :30 .22... m5 .0 8:20 2938.3 cam 25: on. .o 2:28 5528 59.2: .3 9:9“.

 

:3:
2:6
. N
. s
. s
u ..
“—5.: \\\ . u
“S 2:: :2... ‘ Sufi. .
.iiit ' lllll s s.
...:5/ .... _ .3
A. ...0. \(p \~
00 .M% \\ or. \
:3. 3.3

«at:

 

123

. 8...: 2m 8.2.0 8:228 .mEofi
.85 58 mm 8.855 2m 8:3 895»: m8 . .
828 286:. waffle :80 .88. 2.2.5388 2: E 2.2“. 2: 2396 £228 0528 c8231 9. 2:9”.

3.... 2.:

 

124

Table 20. Electropotentials of FMN

 

Species éox/sq fisq/hq Conditions
Free FMNa -238 -l72 pH 6.95/20”:
D. vulgarisb -149 -480 pH 7.78/25”:
A. nidulansc -221 -447 pH 7.0/30%:
c1. MPd -92 '-399 pH 7.0/20°C

 

a[102] b[103]

c[104] d[105]

125

Some other differences in the structure are a result of different
conformations in the side- chains. One very interesting example is ARGZ4: the
semiquinone model is conformationally similar to the room temperature model
and if one examines the cell parameters (Table 19) between these two models,
‘they are very similar. The interpretation of the side chain for each model was
carried out independently of one another and is shown in Figure 43. The result
associated with this side chain raises an interesting question regarding
interpretation of the density during the refinement stages. According to Figure
43, the conformation of the side chain in the room temperature model could

easily be fitted to the semiquinone density; however the torsion angle, x3, is

different for each case. In the room temperature model, the torsion angle was
adjusted to a value of -70° in order to fit the density. In the semiquinone model,
the torsion angle was -20° with a hydrogen bond to WAT162 (the solvent atom
was used to account for the difference density). Which conformation is correct?
One would believe that at high resolution (e.g 1.9A), misinterpretation of
density at the latter stages of refinement would be minimal, however, this
example shows that interpretation of the density was different for each model.
This again raises the question as to whether different refinement paths lead to
different interpretation of atomic positions in poorly determined regions of a
structure.

3. Comparlson of the oxidized and hydroquinone states

The comparison between these two oxidation states is a little more difficult
because of the lower resolution data set obtained for the hydroquinone
refinement. The majority of differences between the oxidized and
hydroquinone states are not much different than the differences between the
oxidized and semiquinone states. For example, the large conformational
change observed at GLY61 in the semiquinone was also observed in the
hydroquinone and so the differences with respect to the oxidized form are

126~

 

a . "i
‘35 - ‘ a 5:1; ’ . 3
“$9.:Ey ‘ " .:‘\
r
\ H1 H1
fr. 5 0 fl“
«3 " ,AWArisz . ‘ ~ .WAT'GZ
\ “‘$(12 3” v" ‘ 's 3 7
‘\ ' ,Q
&‘ ‘ “gt“
“.6.“ ’j' *4
st. : $ " ’ . (”1%
pie ' 3 ;'.
for ..., ~' ”5:
. v f ‘- -
A’s.» flfi‘“ A“ ... u T} ”4”" -|

Figure 43. Stereo representation of the conformation of A8624 in the room
temperature oxidized (top) and semiquinone (bottom) forms.

127

small (See Figure 44). There are some main chain differences observed
throughout the structure as well; however, it is difficult to determine if these
differences are a result of differences in resolution of the data sets.

4. Comparison of the semiquinone and hydroquinone states

The differences between the semiquinone and hydroquinone structures
appear to be too subtle to observe crystallographically (at least at the 2.25A
resolution of the hydroquinone structure). The majority of differences between
these two structures are slight main chain differences that propagate
themselves into side chain differences which could be related to differences in
resolution of data used in the refinement. A plot of the average deviation ln
position as a function of residue for both the main and side chains is shown in
Figure 45.

The two principal reasons for examining the hydroquinone state were to
determine if a hydrogen was present at N(1) and to examine the planarity of the
isoalloxazine ring and to provide an accurate model to do theoretical
calculations on those factors that effect the redox potentials. The results of
crystal structure analyses of several N(1) and N(5)-substituted 1,5
dihydroisoalloxazines demonstrate that the isoalloxazine ring adopts a butterfly
conformation with the outer rings having 30° normals [68]. Even spectroscopic
measurements [69] suggest that the ring is nonplanar at room temperature. A
comparative ‘30, 15N, and 31P NMR study on flavodoxins from Clostn'dium MP,
Megasphaera elsdenii, and Azotobacfer vine/andii [72] and a study with
flavodoxin from Desulfovibn‘o vulgaris have been repeated [73,74]. The study
was in part to try and determine the interaction between the FMN and
apoflavodoxin and the different contributions to redox potential. For example, it
is believed [75] that the bend along the N(5)-N(10) axis of the flavin molecule is
associated with regulation of redox potential.

.2an Sosa—86>: new 3538 2283.23 So. 5223 Samoa 338. E 5:926 399% .3. 2:9“.

mzv_mmm

giggsgmaéqgsafiéfigfisaagiggsaE3§§§2és§e§§2§§§§§E.§
a: sf em. s: 35 sc am am so am aw am am E

c_mco c_mz

u01aisOd U!

aoueuegggo

'S'ul'a

128

N

c_mco ¢a_m

 

129

£59 80:50.83 new 2052593; .8223 Samoa 323. E :2333 3992. .9 2:9“.

m:n_mmm

_s§a§_a§a§as:gasgésegfizgfigfiégEaggéafiggggg39.2525
83 am.— sm_ 8: SE ac so 3. am am av am am E

c_mgo c_mz

c_m;o me_m

 

S
uolqlsod u!

—I

N

'S'w'a

eaueuegggg

130

That is, the energy barrier associated with the transition from the bent
conformation to the planar conformation is related to the redox potential of the
flavin bound to the protein. This is supported from crystallographic studies in
the Closfn'dium MP [36.76], A. nidulans [38] and D. vulgaris flavodoxin by
observing a reduced isoalloxazine ring with a bending angle ranging from a
few degrees to 8.8°. Ludwig et al. [67] have examined the CI. MP flavodoxin in
the hydroquinone state with a focus to determine the shape of the
dihydroisoalloxazine ring. In their study, they found the ring to be bent along
the N(5)/N(10) line, with the atoms on each side of the line being in the same
plane, by as much as 8.6°. This is much less than the 30-35° found for most
substituted dihydroflavins [68,77] and is in good agreement with the 9.1 ° found
in 5-diethyl-3,7,8,10-tetramethyl-1,5-dihydroisoalloxazine [78].. In the D.
vulgaris structure, the overall shape appears to be planar (Table 21); a
bending angle (angle between the plane normals) of 16.7° was obtained from
the restrained least squares refinement in thehydroquinone form, however,
some of the pyrimidine moiety atoms moved out of the plane. An experiment to
determine the planarity of the flavin was carried out subsequent to the last
cycle of least squares. The planar moieties of the flavin were kept planar and
refined separately as rigid bodies. The resulting angle (twist or bend) was 3.4°.
Finally, the ring seems to remain nearly planar in all three oxidation states and
the bending or twisting that occurs appears to be within the experimental error
of the atomic positions.

13C and 15N NMR spectra of fully-reduced flavodoxin has led to an upfield
shift of the 13C and 1‘5N resonances except C(10a). which is shifted downfield
as compared to the oxidized form. A comparison of the chemical shifts from the
C(10a) and 0(2) in fully-reduced tetraacetylriboflavin, FMNH', and flavodoxin
shows that the protein-bound FMN is ionized. Since there appears to be a
hydrogen bond in the oxidized and semiquinone states between N(95) and
N(1), it appears that there is no hydrogen present at N(1) in the hydroquinone
state.

131

Table 21. Bending of lsoalloxazine Ring in Flavodoxin

Oxidation State R Resolution Bending
(A) angle(°)

 

Oxidized - RT 0.17 8.0-1.90 4.02
Oxidized - LT 0.19 8.0-1.90 1.72
Semiquinone 0.20 8.0-1.90 6.33
Hydroquinone 0.21 8.0-2.25 16.68
Hydroquinone 0.21 8.0-2.25 3.36

(after rigid-body
refinement)

 

132

This is supported by the N(95)-N(1) distance of 3.2A in the hydroquinone state;
a hydrogen on N(1) would cause a unfavorable contact with the peptide
hydrogen.

5. Electron Transfer

Since flavodoxin transfers electrons'to other flavoproteins, such as heme
proteins like cytochrome c3 and because the redox potential for the
semiquinone/hydroquinone couple is approximately that of ferrodoxins, it is
believed that flavodoxins alternate between their two lower oxidation states in
vivo [79].

According to Moonen er al. [80] a strong hydrogen bond on O(2a)

influences the n electron density on C(8), C(6). N(5), C(9a), and C(10a)
through conjugative effects. in order to stabilize this mesomeric structure. a
solvent of high permittivity is needed. This is observed in the 130 NMR
spectrum from observed shifts of C(8), C(8a). and C(6) and the crystallographic
results of the three oxidation states. From these results, the FMN is most

polarized in the direction from C(8), C(6). N(5), C(9a), C(10a), to C(Za). The
stabilization of the mesomeric stmcture is a combined effect of hydrogen

bonding with. C(Za) and a high permittivity at the dimethylbenzene edge of the
protein-bound flavin. The chemical shifts of N(1) of FMN bound to all of the
apoflavodoxins are very similar to that of FMNH'. This already indicates that the
prosthetic group in reduced flavodoxins is ionized; this has already been
mentioned for A. vinelandii [81] and M. elsdenii [69] by other methods. Access
to the flavin ring with solvent or electron donors is somewhat restricted by the
surrounding protein atoms in all three oxidation states. in order to make contact
with the isoalloxazine ring, the incoming ”substrate“ would need to interact with
the FMN methyl groups at C(7) and 0(8).

133

The remainder of the FMN is buried below the protein surface and access to
the flavin ring is not significantly different in any of the three oxidation states. It
seems apparent that these crystallographic results suggest that a different
electron transfer mechanism in which the electrons are transferred through the
dimethylbenzene portion of the FMN [82] rather than the "direct" electron
transfer mechanism at N(5) or 0(4) found in model flavin radical chelates as
discussed by Hemmerich et al. [83]. Computer modelling studies [84, 85] have
shown that interprosthetic group distances, favoring complementary salt
linkages, play an important role in stabilizing orientation and position for
electron transfer over the protein-protein interface. These calculations are
based on electrostatic fields and NMR experiments.

In order to gain a better understanding of the mechanism of electron
transfer in this class of proteins a few different experiments might be useful. For
example, point mutagenesis along the reverse turn (residues 60-63) could be
helpful in examining the role of hydrogen bonds in the mechanism. Another
experiment would be to purify and crystallize the flavodoxin-”substrate"
complex. This problem is being examined indirectly by Karplus et al.[87] who
have data on the spinach ferrodoxin-reductase enzyme complex and are
currently examining medium resolution maps.

 

' REFERENCES

10.

11.

12.

13.

14.

LIST OF REFERENCES

Lehninger, A. L. (1970). Biochemistry, Worth Publishers, lnc., New York.
Fox, J. L., Smith, SS. and Brown, J.Fi. (1972). Z' Naturfarsch 27b,
1096-1100.

D'Anna, J. A. Jr. and Tollin, G. (1972). Biochem. 11, 1073-1080.

LeGall, J., DerVartanian, D. V. and Peck, H. D. Jr. (1979). Curr. Topics in
Bioener. 9, 238-265. ‘

Akagi, J. M. (1967). J. Biol. Chem. 242, 2478-2483.

Fax, J. L. (1976). in: Fiavins and Flavoprateins (T .P. Singer, ed.) Elsevier
Scientific Publishing Co. pps. 439-446, Elsevier, Netherlands,

Peck. H. 0. Jr. (1974). Symp. Sac. Gen. Microbiol. (London) Evolution in
the Microbiological World 24, 241-262.

Eck, R. V. and Dayhoff, M. O. (1966). Science 152, 363-366.

Irie, K., Kobayashi, K., Kobayashi, M. and lshimoto, M. (1973). J. Biochem.
73, 353-366.

Kobayashi, K., Tachibana, S. and lshimoto, M. (1969). J. Biochem. 65,
155-157.

Kobayashi, K., Tachibana, E. and lshimoto, M. (1972). J. Biochem. 72,
379. '

lshimoto, M., Koyama, J., Yagi, T. and Shiraki, J. (1957). J. Biochem. 44,
41 3.

Aketagawa, J., Kobayashi, K. and lshimoto, M. (1985). J. Biochem. 97,
1025-1032.

Drake, H. L. and Akagi, J. M. (1977). J. Bacteriol. 132, 132-138.

134

 

15.
16.

17.
18.

19.

20.

21.

23.

24.
25.

26.

27.

28.

30.

31.

135

Amdt, U. and Ambrose, B. K. (1968). IEEE Trans. Nucl. Sci. NS15:N63,
92-94.

Minor, T. C., Milch, J. R. and Reynolds, G. T. (1974). J. Appl. Cryst. 7,
323-333.

Abrahamsson, S. (1972). Acta Cryst. A28, 8248.

Hashizuma, H., Kohza, K. and Kinoshita, K. (1972). Acta Cryst. A28,
S249. _

Xuong, N. H., Freer, S. T., Hamlin, Fi., Nielsen, C. and Vernon, W. (1978).
Acm Cryst. A34, 289-296.

Amdt, U. W., and Willis, 8. T. M. (1966). Single Crystal Diffractometer,
Cambridge University Press, London and New York.

Howard, A. J., Gilliland, G. L., Finzel, B. C. and Poulos, T. L (1985). J.
Appl. Cryst. 20, 383-387.

Howard, A. J. (1985). A Guide to Macromolecular X-ray Reduction for the
Nicolet Area Detector System: The XENGEN Manual.

Blundell, T. L and Johnson, L. N. (1976). Protein Crystallography,
Academic Press, New York.

Azaroff, L. V. (1955). Acta Cryst. 8, 701-704.

Howard, A. J., Nielsen, C. and Xuong, N. H. (1985). Meth. Enzymal. 114,
464-465.

Lipscomb, W. N., Hartsuch, J. A., Fieeke, Jr., G. M., Quiocho, F. A.,
Bethge, P. H., Ludwig, M. L., Steitz, T. A., Muirhead, H. and Coppola, J. C.

(1968). Brook. Symp. Biol. 24-90.

Low, B. W., Chen, C. C. H., Berger, J. E., Singman, L. and Fletcher, J. F.
(1966). Biochem. 56, 1746-1750.

Haas, D. J. (1968). Acta Cryst. 824, 604.

Haas, D. J., and Hossmann, M. G. (1970). Acta Cryst. B26, 998-1004.
Tomanek, U. F., Parak, F. Messbauer, R. L., Formanek, H., Schwager, P.
and Hoppe, W. (1973). Acta Cryst. A29, 263-265.

Petsko, G. A. (1975). J. Mol. Biol. 96, 381-392.

32.

33.
34.

35.
36.

37.

38.

39.

40.

41.

42.

$1.3

47.

48.

136

Douzou, P., Hui Bon Hoa, G. and Petsko, G. A. (1975). J. Mol. Biol. 96,
367-380.

Dewan, J. C. and Tilton, Ft. F. (1987). J. Appl. Cryst. 20, 130-132.
Hope, H. (1985). Proc. Am. Crystallogr. Assoc. Meet, Stanford Abst. PA3,
p. 24.

Hope, H. (1988). Acta Cryst. B44, 22-26.

Smith, W. W., Burnett, R. M., Darling, G. D. and Ludwig, M. L. (1977). J.
Mol. Biol. 117, 195-225.

Watenpaugh, K. D., Sieker, LC. and Jensen, L.H. (1973). Proc. Natl.
Acad. Sci. USA 70, 3857-3860.

Laudenbach, D. E., Straus, N. A., Pattridge, K. A. and Ludwig, M. L.
(1987). In: Flavins and Flavoproteins (D. E. Edmondson and

D. B. McCormick. eds.), de Gruyter 8: Co., 250-260.

Krey, G. D., Vanin, E. F. and Swanson, R. P. (1988). J.Bial. Chem. 263,
15436-15443.

Booth, A. D. (1948). In: Fourier Technique in X-ray Organic Structure
Analysis 62-65, Cambridge University Press, New York.

Watenpaugh, K. D., Sieker, L.C., Jensen, L.H., LeGaIl, J. and
Dubourdieu, M. (1972). Proc. Natl. Acad. Sci. USA 69, 3158-3188.
Jack, A. and Levitt, M. (1978). Acta Cryst. A34, 931-935.

Mavridis, l., Hatada, M. L., Tulinsky, A. and Lebioda, L. (1982).

J. Mol. Biol, 162, 419-444.

Jones, T. A. (1978). J. Appl. Cryst. 11, 268.

Diamond, Ft. (1976). In: Crystallographic Computing Techniques (K. Huml,
F. R. Ahmed and B. Sedlacek), 291, Munksgaard, Copenhagen.
Mandel, N., Mandel, G., Trus, B. L., Rosenberg, J., Carlson, G. and
Dickerson, R. E. (1977). J. Biol. Chem. 252, 4619-.

Stout, G. H. and. Jensen, L H. (1968). In: X-ray Structure Determination,
369-379, MacMillan Publishing Co. Inc., New York.

Schomaker, V., Waser, J., Marsh, Ft. E. and Bergman, G. (1959). Acta
Cryst. 12, 600.

 

49.

50.

51.

52.

61.
62.
63.

64.

65.

137

Konnert, J. H., and Hendrickson, W. A. (1980). In: Biomalecular Structure,
Function, and Evolution (R. Srinivasan, ed.), 43-57, Pergamon Press,
Oxford.

Sussman, J. L., Holbrook, S. R., Church, G. M. and Kim, S. H. (1977).
Acta Cryst. A33, 800.

Tronmd, D. E., Ten Eyck. L F., and Matthews, 8. W. (1987). Acta Cryst.
A43, 489-501.

Ten Eyck, L F. (1977). Acta Cryst. A33, 486-492.

Agarwal, R., Llfchitz, A., and Dodson, E. (1981). In: Refinement of Protein
Structures(P. A. Machin Campbell, J. W. 8. Elder, M., eds.), Science and
Engineering Research Council, Daresbury Laboratory, Warrington,
England.

Agarwal, R. C. (1975). Acta Cryst. A34, 791-809.

Jones, T. A. (1985). Meth. Enzymal. 115, 157-171.

Veret, L (1967). Phys. Rev. 159, 98.

Karplus, M. and McCammon, A. (1983). Ann. Rev. Biochem. 52, 263.
Brooks, 8. R. (1983). J. Comput. Chem. 4, 187.

Bn‘inger, A. T., Kuriyan, J. and Karplus. M. (1987). Science 235, 458-460.
Watenpaugh, K. D. (1985). Conformational Energy as a Restraint in
Refinement (J. Hermans), 77-80, University of North Carolina, Chapel Hill.
Priestle, J. P. (1988). J. Appl. Cryst. 21, 572-576.

Holmes, M. A., and Matthews, B. A. (1982). J. Mol. Biol. 160, 623.

Baker, E. N., and Hubbard, R. E. (1984). Prog. Biophys. Mal. Biol. 44,
97-179.

OIthof-Hazenkamp, R. (1987). In: CRYLSO, Least-Squares Refinement of
Atomic Parameters (S. R. Hall and J. M. Stewart, eds.), 137-148,
Universities of Western Australia and Maryland.

Pflugrath, J. P., Saper, M. A. and Ouiocho, F. A. (1985). In: Methods and
Applications in Crystallography (S. Hall and Ashida, T., eds.), 404-407,
Oxford Univ. Press, Oxford.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.
76.

78.

79.

80.
81.

138

Watenpaugh, K. D., Sieker, LC. and Jensen, L.H. (1976). In: Flavins and
Flavoproteins (T. P. Singer, ed.), 405-410, Elsevier Scientific Publishing
Co., Amsterdam.

Ludwig, M. L, Burnett, R. M., Darling, G. 0., Jordan, S. R., Kendall, D. S.
and Smith, W. W. (1976). In: Flavins and Flavaproteins (T. P. Singer, ed.),
393-404, Elsevier Scientific Publishing 00., Amsterdam.

Norrestan, R., Kierkegaard, P., Stensland, B. and Torbjomsson, L. (1969).
Chem. Commun. 1227, 1250-1251.

Ghisla, S., Massey, 8., Lhoste, J. M. and Mayhew, S. G. (1974). Biochem.
13, 589-597. .

Honzatko, R. B. (1986). Acta Cryst. A42, 172-178.

Connolly, M. L. (1983). Science 221, 709.

Vervoort, J., Miller, F, Mayhew, S. G., van den Berg, W. A. M., Moonen, C.
T. W. and Bacher, A. (1986). Biochem. 25, 6889-6799.

Favaudon, V., LeGaII, J. and Lhoste, J. M. (1980). In: Flavins and
Flavaproteins K. Yagi and Yamano, T., eds.), 373-386, Japan Scientific
Societies Press, Tokyo.

Vervoort, J., Miller, F., LeGaII, J., Bacher, A. and Sedlmaier, H. (1985).
Eur. J. Biochem. 151, 49-57. ‘

Massey, V., and Hemmerich, P. (1980). Biochem. Soc. Trans. 8, 246-257.
Burnett, R. M., Darling, G. D., Kendall, D. S., LeQuesne, M. E., Mayhew, S.
G., Smith, W. W. and Ludwig, M. L. (1974). J. Biol. Chem. 249.
4383-4392.

Werner, P. E. and Ronnquist, O. (1970). Acta Chem. Scand. 24,
997-1009.

Werner, P. E., Linnros, B. and Leijonmarck, P. (1971). Acta Chem. Scand.
25, 1297-1312.

Mayhew, S. G., Faust, G. P. and Massey, V. (1969). J. Biol. Chem. 244,
803-810.

Moonen, C. T. W., and Miiller, F. (1984). Eur. J. Biochem. 140, 311-318.
Edmondson, D. E., and Tollin, G. (1971 ). Biochem. 10, 133-145.

 

82.

85.

87.

89.

90.

91.

92.

93.

94.

95.

96.

97.
98.

139

Favaudon, V., and Lhoste, J. M. (1975). Biochem. 14, 4731-4738.
Hemmerich, P., Miiller, F. and Ehrenberg, A. (1965). In: Oxidases and
Related Redox Systems (King, T. E., Mason, H. S. and Morrison, M., eds.),
157-178, Wiley, New York.

Weber, P. C., and Tollin, G. (1985). J. Biol. Chem. 260, 5568-5573.
Stewart, D. E., LeGaIl, J., Maura, I., Moura, J. J. G., Peck, H. D. Jr.,

Xavier, A. V., Weiner, P. K., and Wampler, J. E. (1988). Biochem. 27,
2444-2450. _ ’

Finzel, B. 0., private communication.

Karplus, P. A., private communication.

Richardson, J. S. (1981). Adv. Prat. Chem. 34, 167-339.

Duboudieu, M., LeGall, J. and Fox, J. L. (1973) Biochem. Biophys. Res.
Commun. 52, 1418-1425. '

Tanaka, M., Haniu, M., Yasunobu, K. T., Mayhew, S. G. and Massey, V.
(1973). J. Biol. Chem. 248, 4354-4366.

Tanaka, M., Haniu, M., Yasunoba, K. T. and Mayhew, S. G. (1974a).

J. Biol. Chem. 249, 4393-4396.

Tanaka, M., Haniu, M., Yasunobu, K. T., Mayhew, S. G. and Massey, V.
(1974b). J. Biol. Chem. 249, 4397.

MacKnight, M. L, Gray, W. L. and Tollin, G. (1974). Biochem. Biophys.
Res. Commun. 59, 630-637.

Tanaka, M., Haniu, M., Matsueda, G., Yasunoba, K. T. and Mayhew, S. G.
(1971 ). Biochem. 10, 3041-3046.

Burnett, R. M., Darling, c. o., Kendall, o. s., LeQuesne, M. E.,

Mayhew, S.G., Smith, W. W. and Ludwig, M. L (1974). J. Biol. Chem. 249,
4383-4392.

Richards, F. M. (1968). J. Mol. Biol. 37, 225.

Hermans, J. and McQueen, J. E. (1974). Acta Cryst. A30, 703-739.
Jones, T. A. (1982). In: Computational Crystallography (D. Sayre, ed.),

p. 303. Oxford Univ. Press, London and New York.

99.

100.

101.

102.

103.

1 O4.

1 05.

140

Bush, 8. L. (1984). In: Computers and Chemistry Vol. 8, p. 1. Pergamon
Press, Oxford.

Watenpaugh, K. D., Sieker, L. C. and Jensen, L. H. (1979) J. Mol. Biol.
131, 509-522.

Ludwig, M. L, Pattridge, K. A., Smith, W. W., Jensen, L. H., and
Watenpaugh, K. D. (1982). In: Flavins and Flavoprateins (Massey, V. and
Williams, C. H., eds). pps. 19-27, Elsevier, North Holland.

Draper, R. D., and Ingram, L L. (1968). Arch. Biochem. Biophys. 125,
802-803.

Dubourdieu, M. LeGaII, J. and Favaudon, V. (1975) . Biachim. Biophys.
Acta 376, 519-532.

Entsch, B. and Smillie, R. M. (1972). Arch. Biochem. Biophys. 151,
378-386.

Mayhew, S. A. (1971). Biochem. Biophys. Acta 235, 276-288.

GRN STATE UNIV

IIIIIII 9|II9|9 9I||I9 |||I|I|||9IIIII9I|I|I|IIIII