Download Glucose 6-Phosphate Dehydrogenase Mutations

From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
Glucose 6-Phosphate Dehydrogenase Mutations CausingEnzyme Deficiency
in a Model of the Tertiary Structure of the Human Enzyme
By
C.E.Naylor, P. Rowland, A.K. Basak, S. Gover, P.J. Mason, J.M. Bautista, T.J. Vulliamy, L. Luzzatto,
and M.J. Adams
Human glucose 6-phosphate dehydrogenase (G6PD) has a
particularly large number of variants resulting from
point
mutations; some 60
mutations have been sequenced
to date.
Many variants, some polymorphic, are associatedwith enzyme deficiency. Certain variants have severe clinical manifestations; for such variants, the mutant enzyme almost always displaysareduced
thermal stability. A homology
model of human GGPD has been built, based on the threedimensional structure
of the enzyme from Leuconostoc mes-
enteroides. The model has suggested structural reasons for
the diminished enzymes t a b i l i and hence for deficiency.It
has shown that a cluster of mutations in exon 10, resulting
in severeclinicalsymptoms,occurs at or near the dimer
interface of the enzyme, that the eight-residue deletion in
the variant Nara isat a surface loop, and
that thetwo mutations in the A- variant areclose together in the three-dimensional structure.
0 1996 by The American Societyof Hematology.
G
both polymorphic and sporadic variants there is always some
residual enzyme activity and this is invariably lower in RBCs
than in other cells. The lack of null variants and thelow
activity in RBCs (which cannot make up for enzyme breakdown through de novo protein synthesis) implies that instability of mutant G6PD molecules is probablythe commonest
cause of G6PD deficiency. This has been confirmed in several cases by appropriate biochemical analysis.
Following the determination of the primary structure of
human G6PD through cloning of the corresponding cDNA,’
the molecular lesions of more than 60 G6PD variants have
been dete~mined~.~;
nearly all are missense point mutations
causing single amino acid replacements. These mutations,
with differing clinical manifestations, vary in their effect on
residual enzyme activityand on substrate binding; almost
all cause a decrease in enzyme stability. Figure 1 shows the
distribution of severe (class I) mutations with respect to the
primary structure of the protein; notable is a cluster of severe
mutations between residues 380 and 450. This analysis has
proved most valuable for diagnosis and to define which mutations account for G6PD deficiency in various populations.
However, it has been less successful in defining the role of
different domains and of individual amino acid residues in
the stability of the protein, andin explaining why some
mutants are almost asymptomatic whereas others cause severe CNSHA.
Some of us have recently solved and given
a detailed description of the three-dimensional structure of G6PD from
Leuconostoc mesenteroides.‘ The protein from this species has significant homology to human G6PD, and residues shown to be
important for activity are conserved. Thus, we can use knowledge of the structure of the bacterial enzyme to model normal
human G6PD B.The dimer is shown tobe the smallest molecular form of G6PD that is active in the human enzyme’ as in
all species8 The crystal structure of the L mesenteroides apoenzyme hasa dimer in the asymmetric unit. The
two monomers
are related by an axis which is very close to
a dyad and it is
anticipated that the subunit interface is retained in the human
enzyme. Our homology model allows us to predict, with some
confidence, the position and role of both normal and mutated
residuesinthetertiary
structure ofhuman G6PD,andthus
answer some of these questions.
LUCOSE-6-PHOSPHATE dehydrogenase (G6PD) is a
housekeeping enzyme that catalyzes the first and ratelimiting step in the pentose phosphate pathway. Its key role
in metabolism is to provide reducing power in the form of
NADPH. This role is particularly important in redblood
cells (RBCs) where NADPH is required for detoxification
of hydrogen peroxide and other compounds, via glutathione.
In these cells the reaction catalyzed by G6PD (and by 6phosphogluconate dehydrogenase, which depends on G6PD)
is the only source of NADPH. Human G6PD, a homodimer
encoded by a gene which maps to the region Xq28 on the
X-chromosome, exhibits an extraordinary degree of genetic
variability: more than 100 deficient variants have been reported and characterized.’ These variants are either true polymorphisms, with relatively mild clinical manifestations and
variant alleles reaching frequencies of 1% to 50% in various
parts of the world, or sporadic variants, brought to light
because they cause chronic nonspherocytic hemolytic anemia (CNSHA) in affected males. The former group of G6PDdeficient variants have most likely reached polymorphic frequencies because deficient individuals are protected to some
extent against severe Plasmodium fakiparum malaria.’ In
From the Laboratory of Molecular Biophysics, Universityof Oxford; the Department of Haematology, Royal Postgraduate Medical
School, Hammersmith Hospital, London, UK; Departamento de Binquimica y Biologia Molecular IV, Universidad Complutense de Madrid, Facultad de Vetinaria, Madrid, Spain; and the Department of
Human Genetics,Memorial-Sloan Kettering CancerCenter, New
York, NY.
Submitted March 27, 1995; accepted November 9, 1995.
Supported in part by a programme grant from the MRC (UK) to
L. L. and P.J.M. C.E.N. has a Wellcome Prize Studentship, S.G. was
funded by the Oxford Centre for Molecular Sciences and P.R. was
supported by a Science and Engineering Research Council studentship. M.J.A. is Dorothy Hodgkin-E.P.
Abraham Fellow of Somerville
College and an associate member of OCMS.
Address reprint requests toM.J.Adams,DPhil,Laboratoryof
Molecular Biophysics, University of Oxford, S Parks Rd, Oxford,
OX1 3QU, UK.
The publication costsof this article were defrayed in part by page
chargepayment. This article must therefore be hereby marked
“advertisement” in accordance with 18 U.S.C. section 1734 solely to
indicate this fact.
0 1996 by The American Society of Hematology.
0006-4971/96/8707-0028$3.00/0
2974
MATERIALS AND METHODS
The human model was generated
by homology modeling from
the known structure of L mesenteroides G6PD6; the 33% sequence
Blood, Vol 87,No 7 (April l),1996 pp 2974-2982
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
10
5
5
20
5
30
5
40
5
50
5
A
M A E Q V A L ~ R T H V C G I L R E E L F Q G D A F H Q S D T H I F I I ~ ' ~ ~ ~ AA SK K*? I Y p T T ' ' 9 W-L
V S E 1 K T L V T F F " G : G T
A K R K L Y . P S V F N"L
5
10
-beta
60
5
5
70
80
5
5
90
5
alpha a-------
20
"""_
A----
100
5
5
P
A T P E E K L K L E D F F A R N S
K. Y V
F R D G L % P E N T F I ! C G Y A W S R L T V A D I R K Q S E P F F K
Y K K G Y L Q K H F A I V G ~ T A ~ R ~ Q A L N D D E F K Q L V R D S I K D F T D D Q A Q A AE F I E H F S Y R
5
70
5
80
30
5
40
5
50
5
60
"-
B""
"-beta
110
5
""""-alpha
120
b-------------
130
5
5
alpha
b.---
"
"
140
150
5
-beta C
160
5
d
A G Q Y D D A A S - Y Q R L N S H M N A L H L G S Q
A N R L F m L A L P @ T V - Y E A V T K N I H E S C M
S
A H D V T D A A S Y A V L K E A I E E A A D K F D I D G N R I F ' Y i M S V A r P , R F F G T I A K Y L K S E G L L A
5
-
5
5
280
5
100
90
5
----------alpha c
""""-
110
290
5
120
5
130
"----alpha d-----"""-
5
-beta D--
300
5
5
320
A A A A
F
K
A
~ K E ~ S A
T N S D D V Y D E
~ K V
v L K c I S Ev Q A N N v
LAIM:EK:P~ESFTDKDIR;AA!~:~NAAFNALKIYDEAEVNKYFVRAQY
v
310
5
5
250
5
260
5
330
270
5
340
5
5
V L G Q Y V G N P D G E G E A T K G
G A G
D S A D F K P
280
5
5
350
290
360
5
5
A"
K
5
370
&y
E R K A
Q * y
H D
v
F Y V ' R S - G ' K . ' R L A A K Q T R V D I V F K A G
340
5
--beta I---
5
390
5
400
5
410
H
C
R E C
H* 4
K
I
D
A G D I F H
QQCKRNELV"I:;RV,Q&NEA?$YTKMMTmKPGMF
T F N
F G S E Q E A Q E A V L S I I I D 5 P i K G A I E L K L N A ? f S V E
380
5
360
430
5
,;.
440
5
370
"-beta
5
L"--
_""
5
450
P".E
alpha m
"
"
"
5
490
5
380
5
beta M------
5
460
5
350
beta K----
420
5
K
F N P E E S E ' : E , D L T Y G N R Y
D A F N T R T I D L G W T V S D E
390
"_ 5
400
beta N------
470
5
5
5
alpha
480
H
V R S D E L R E A ~ R I F T P - ' L L H Q I E L E K P P P I P ~ I
A D W N G V S I A : - W ~ K F V D A ~ I S A V Y T A D K A P L E T : ! ~ K
K N V K L P D A m E R ' * L " T L'.'D V F C* S Q M H
D K K N T P E PZr!E R M f H D T M N;&D G S N
410
5
420
5
430
1
"
5
"-"
--beta J---
-------
500
5
5
"""""""-alpha
440
n
""
-
5
450
5
460
beta0
510
V
Y B S R G R T E A D E L M K R V G F Q Y E G T Y K W
V N P H K L
S:G.S M G'P:E A S D K L L A A N
G
D A W V F K G
5
470
5
480
485
alpha 0"-
_""
*Fully conserved residues
.-Highlyconserved residues
A = deletion
A- mutations in lower case
Fig 1. Sequence alignment of human and L mesenteroides GGPD. The alignment was used for model building of human GGPD. Residues
fully or highly conserved m e r 14 sequencesare indicated in different shades of gray.
The secondary structural elements for the L mesenteroides
enzyme are shown. The locations of amino acid substitutions found in severe (class I) variants are indicated by capital letters above the
alignment. Note the cluster found around amino acid 400.
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
NAYLOR
2976
ET AL
p + a domain
Fig 2.
6
Fig 4.
Fig 5.
Fig 6.
identity is sufficient to justify the assumption that it has a similar
fold. A multiple sequence alignment of 14 currently known G6PD
sequences (human,' L mesenteroide~,~
Synechococcus PCC 7942,"
Zymomonas mobilis," Saccharomyces cerevisiae," Drosophila melanogasrer.I3 rat,I4 mouse," Erwinia chrysanrhemi m. Hugouvieux-
Cotte-Pattat, J. Robert-Baudouy, P R database,
entry
S370531,
Esherichia coli,16Kluyveromyces marxianus lactis [M. WesolowskiLouvel, C. Tanguy-Rougeau, H. Fukuhau,PIRdatabase,Ihaentry
S 3 13371, Plasmodium falciparum," Pichia jadinii," Bacillus SP
HT-3H. [H Sagai, K Hattori, M Takahashi, US Patent No. 5,137,821,
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
A 3D MODEL
HUMAN
OF DEFICIENT
IN
MUTANTS
GGPD
2977
Fig 2. Model of the human dimer.In this figure subunit P (equivalentto subunit A in published illustrations of the dimer of L mesenteroides
G6PD') is shown in red, green, and yellow, andQ is in pale blue, pink, andwhite. The monomer consistsof two domains-a smaller coenzyme
domain encompassing residues 1-198 and a larger
p + a domain comprising residues 199-515. The requenca GASGDLA (residues
38-44)is at
the coenzyme binding site (arrow). The G6P binding site includes residues from the perfectly conserved 9-amino acid sequence RIDHYLGKE
(198-206). Three adjacent strands
of the p-sheet of the p + a domain (residues 380-425) arein the area of the dimer interface. Theorientation
used for this and other figures is shown schematically at upper right as a guide for the reader. Figures 2,4, 5, and 6 were drawn using the
program BOBSCRIPT, an enhancedversion of MOLSCRIPT."
Fig 4. Serious mutations at the dimer interface. Amagnification of the dimer interface region of the model. Subunit P is inred, green, and
yellow, Q in pale blue, pink, and
white. The C a carbons of mutations causing seriously deficient (class
I)variants areindicated by black spheres
in subunit P and grey spheresin 0. All variants are labeled andthe region of the dimer enlarged isshown at upper right. The class I variant
residues form two symmetry-relatedclusters at the dimer interface.The cluster at the lower right hand cornerof the diagram involves rasidues
from subunit P 213Val --t Leu (Minnesota=), 216Phe
Leu (HarilaouZ7)(not visible), 274GIu
Lys (Corum"), 278Ser
Phe YWexham'"),
393Arg + His (NashvillP), 394Val- Lys (Alhambra"), 396Pro -,Leu (Bari") and 398Glu
Lys (Puerto Limonu); andfrom subunit 0: 385Cys
+ Arg (Tomaha), 386Lys + Glu (Iowa=), 387Arg --t Cys (Guadalajara3') 387Arg
His (Beverly Hills?, 405Met -t Ile (Clinic?, 410Gly --t Cys
(Riversidea); 41OGly + Asp (Japan3') and 416Glu
Lys (Tokyo3*). The other cluster contains the same residues in the alternative subunit.
Further from the cluster are 363Asn + Lys (Loma Lindaz6), and 439Arg Pro (Pawnee3'). These residues interact across the dimer interfaca.
generating the major interaction energy between the two subunits. Mutations disrupt and weaken the dimer-dimer interactions, leading to
destabilization. Highlighted in the diagram is Lys 386, originally predicted as a residueinteracting with NADP+.
-
--
+
+
+
+
Fig 5. The aight-residue deletion situated in a flexible loop. A magnificationof the region of the eight-residue deletion in the class I variant,
Nara?5which is predicted to be missing residues 318-325. Thisis the only variant identified so far that has more than two residues deleted.
The first and last deletad residues are indicated by black spheres at their C a positions. The chain trace highlighted in red in (A) shows the
modeled conformation of the normal enzyme a t this point. The blue chain trace in (B) shows a suggested alternative conformation after
deletion of residues318-325. This alternative conformationwas createdby removingresidues 318to 325 and then modelinga new conformation
using 0; surrounding residues were altered as little as possible. No energy minimization was performed. lt is possible to model the deletion
without affecting any secondary structural elements or the active site of the protein.
Fig 6. The region including 68Val- Met and 126Asn --t Asp of the A- deficient variant. A magnification of helix ac, and the @sheet of the
coenzyme binding domain. The residues mutated in the variant A- are shown in ball and stick (grey, carbon; blue, nitrogen; red, oxygen;
yellow, sulphur) to indicate their proximity. The left-hand diagram showsthe residues in the common G6PD B (Val 68 and Asn 126). The
righthand diagram shows those in G6PD A- (Met 68 and Ay, 126). Val 68, on pB, is also a valine in L mesenteroides G6PD; the sheet strand is
immediately followed by the conserved Arg 72 and the position of the sidechain of Val68 can thus be modeled with confidence. The common
African variant A (126Asn Asp) is notdeficient nor is the engineered mutant 68Val -t Met. The normal behavior of the mutant 68Val- Met
allows us to assume that pB remains unchanged andthat the sidechain of Met 68 will be on the same side of the sheet as is Val 68. In the
model, the sidechains of Val 68 and Met 125 are in direct contact; direct contact between 68 and 126 may be achieved by a small rotation of
a c on itsaxis.
<
+
19921)wasgenerated,usingAlignment
of MultipleProtein Sequences (AMPS).I9The resulting alignmentis shown for the human
and L mesenteroides enzymes (Fig I). The secondary structural elements for the bacterial enzyme, also given in Fig 1, were identified
from its refinedthree-dimensionalstructureusingtheprogram
DSSP'" as is described in Rowland et al.6 This alignment was used
withtheprogramMUTATE
(R. Esnouf, D. Phil.thesis,Oxford
University, 1992) toexchangethesidechains
of L mesenteroides
G6PD with those in the human enzyme and place each altered sidechain as its most common rotomer; at this stage insertions were not
modeled and residues in deletions were replaced with dummy residues (Ala).
Eight insertions totaling 16 residues were modeled into the structure, and six deletions totaling 14 amino acids were removed, using
the molecular graphics package 0." For each insertion or deletion,
a plausible main chain conformation was chosen that avoided strain
andbad contacts with the surrounding protein. Any severe clashes
between sidechains were remedied on inspectionof the protein. The
first 26 residues of the human protein were not modeled; they form
an amino-terminal tail of unknown structure present in all the mammalianG6PDssequenced so far,butnotpresentinG6PDsfrom
other genera, including L mesenteroides. The final three residues of
human G6PD were also not modeled.
Energyminimizationwasperformedonthemodelusing
XPLOR." The protocol constrained all main chain atoms not in insertions or deletions, and consisted of two stages: an initial minimization with a monomer model, followed by a second using the dimer.
Totalenergydecreasedfrom
mol".
265,116 kcalmol"to
-7,649 kcal
RESULTS
The use of 14 sequences and a multiple alignment
algoof the
rithm such as that in AMPS enhances the reliability
alignment over that from a pairwise comparison of the two
sequences of interest. The method uses a matrixof pairwise
identities and optimizes the conservation between all pairs
of sequences; thus, it can take advantage of subgroups with
higher conservation in comparing two more distantly related
sequences. Although the human sequence has
33.1% identity
with that fromL mesenteroides, uncertainties can be resolved
by using, for instance, the 48.1% identity of human G6PD
withthat of S cerevisiae, which itself has 35.7% identity
with L mesenteroides G6PD. ThealignmentinFig
1 was
achievedwithoutdisruption of secondarystructure except
for a shortening of helices a b and of a k by one turn and a
lengthening of ab' to the same extent. All three helices are
attheextreme edges of thedimer.The multiple sequence
alignmenthasresultedinseveralregionswherethereare
one or two residue insertions and deletions, usually loops.
in
The position of these discontinuities results from optimizing
sequencesimilaritiesover the completeset of sequences;
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
2978
NAYLOR ET AL
I
coenzyme
I
H HH H H H HHH
H H HH
A a B b b ’ C c D d E e F Y ’ f g C h i
jkH
H
H
H
H
H
H
H
H
B+*
I J K
H H H
HH H H
LMNlm n O o
HHH
H
HELIX
SHEEI
20.0
%
17.5
15.0
3
.-
e
12.5-
v)
10.0U
a
7.55.0-
2.50.Q
Sequence Number
one-to-one correspondence of residues cannot be assumed
in these regions. Most insertions and deletions arise from
preserving the alignment of the L mesenteroides sequence
with those of other prokaryotes and the human sequence
with the eukaryotic sequences. Typical examples are the loop
between PK and @L, whose length varies from 14 to 19
residues, and the insertioddeletion at ab-ab‘ where the exact
position of the comer between the two external helices may
vary between the two groups of structures. After residue 497
there is considerable variation between sequences and the
alignment is rather arbitrary. Although the PM-PN loop is
lengthened by one residue in two of the bacterial sequences
andby three residues in that of Z mobilis, the structural
elements which make up the dimer surface, af, ag, PMand
PN, are not subject to insertions or deletions.
The model shows the extended nature of the dimer (Fig
2) with “P + a” domains forming the dimer interface and
“coenzyme” domains distant from the dyad axis. The following description of the human enzyme is based on the
detailed description given of L rnesenteroides G6PD.6 The
domain structure as defined for the bacterial enzyme is retained; the domain boundary is at the end of sheet strand
@F.This is at the terminus of a standard dinucleotide binding
foldz3and is in a region of high sequence conservation. The
N-terminal dinucleotide binding domain comprises residues
27-200 in the human enzyme (1-177 of the L mesenteroides
enzyme). The conserved NAD(P) fingerprint forms a tight
turn that begins at Gly 38; its position is indicated in the
figure. We have shown this to be the NADP+ binding site
in L mesenteroides G6PD (C.E.N., unpublished results, November 1994). The coenzyme specificity is ensured by the
Fig 3. Schematic
showing
residuesinvolvedin the dimer
interface.ThedHerencein
the
mean accessible surface area of
the atoms in each residuein the
dimer from their accessible surface area in the putative monomer. The dimer surface consists
of the residues with changed
solvent accessibility. Secondary
structure elements are shown.
presence of the conserved Arg 72, which binds the 2”phosphate of NADP. The active site can be identified at the
domain boundary and includes residues from a nonapeptide
(198-206), which is conserved in the 14 G6PD species considered, as well as residues distant in the primary structure.
His 201, whichis predicted to interact with the substrate
(G6P), is 17 A from Gly 38, the neighbor of the adenine
ribose of NADP+.
The dimer interface of human G6PD is formed by association of the sheets and two helices in the second (p + a )
domain of each subunit to form a barrel. The residues involved in the dimer surface (defined as those which contain
atoms which are less accessible to solvent in the dimer than
in the putative monomer) are shown schematically in Fig 3;
they are primarily from af,ag, PM, PN, and loops connecting these elements (see Fig 2). The two active sites (His 201
of the G6P binding sites) in the dimer are more than 50 A
apart, separated by the large, predominantly antiparallel
sheets of the dimer interface which together form a halfbarrel.
Some 29 class I variants have been considered; they comprise two deletions, of 2 (StonybrookZ4)and 8 (Nara”) residues, and 26 point mutations. The mutated positions are
shown in Fig l. Both deletions occur in surface loops distant
from the dimer axis; they are close to each other. Seventeen
of the remaining point mutations (20 variants) are in or close
to the dimer surface (Table 1). The variant residue is defined
as being a part of the surface if one of the atoms in the
corresponding residue in G6PD B has a changed solvent
accessibility as discussed above. The class I residues involved in the dimer surface are 2 13 (Minnesotaz6),2 16 (Hari-
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
MODEL
A 3D
HUMAN
OF DEFICIENT
IN
MUTANTS
G6PD
2979
Table 1. Class I Mutations Close to the Dimer Surface
Variant
Minnesota
Harilaou
Corum
278 “Wexham”
Tomah
386
Iowa
Guadalajara
Beverley Hills
Nashville
394 Alhambra
Bari
Puerto
398
Limon
Clinic
Riverside
Japan
Tokyo
Pawnee
Kobe
Santiago de Cuba
--
Mutation
Site
213 Val Leu
216Phe Leu
274 Glu Lys
Ser Phe
385 Cys Arg
Lys Glu In
387 Arg Cys
387 Arg His
393 Arg -t His
Val Leu
BL
396 Pro Leu
Glu Lys
405 Met Ile
410 Gly Cys In
410 Gly -t Asp
416 Glu Lys
439 Arg + Pro
440 Leu Phe
447 Gly Arg
-+
-
-+
-+
Secondary
Structure
Element Active
Distance From
Dimer
Distance From
Interface
Reference
Subunit Involved
in Patch
In surface
In surface
6.7 A
In surface
1.4 A
surface
2.2 A
>l0 A
>l0 A
>l0 A
>l0 A
>l0 A
>l0 A
>l0 A
P
P
P
P
>l0
>l0
>l0
>l0
A
A
A
A
>IO A
P
P
P
P
PM--PN
4.2 A
5.2 A
8.6 8,
4.1 A
In surface
surface
>l0 A
Q
Q
BN
am
am
End a m
In surface
In surface
4.8 A
3.2 A
>l0 A
>l0 A
5.9 A
>l0 A
Q
P
P
P
af
af
ai-aj
ai--aj
PK--PL
BK--BL
PK--PL
Q
Q
Q
-+
PL
-+
-
-+
-+
-+
-+
-
-+
BL--BM
DL--PM
BM
laou2’), 278 (“Wexham”28),386 (Iowa29),405 (Clinic3’),
410 (Riverside29 and Japan3’), 416 (Tokyo32), and 439
(Pawnee3’); of those close to the surface, residues 385 (Tomah29),387 (Guadalajara3’),393 (Nashville26),394 (Alhambra3’), 398 (Puerto Limon33),440 (Kobe”) and 447 (Santiago
de Cuba3’) have an atom less than 5 A from one with changed
accessibility.
DISCUSSION
Based on this model of the human G6PD dimer, we have
attempted to rationalize the previously reported pattern of
naturally occurring mutations and to analyze some specific
examples. The model of the human enzyme has allowed
us to locate the different variants on particular secondary
structural elements and to see their spatial relationships to
the coenzyme binding site, the active site and the dimer
interface. Almost all residues seen to be important in binding
NADP in L mesenteroides G6PD, and those predicted to
bind substrate and to promote catalysis, have been conserved
in the human enzyme. None of these residues has been mutated in any of the variants described so far. Coenzyme and
substrate binding have been modeled by Rowland et a16; the
three-dimensional structure of a binary complex of the L
mesenteroides enzyme confirms the features of NADP binding (C.E.N., unpublished results, November 1994) and this
will not be elaborated further here.
Regions of the surface of the dimer that may be of importance for enzyme stability may be recognized in that class I
variants cluster in such an area. It is observations of this
kind which we will now address in this report; it is not our
intention to use our homology model to answer detailed
questions concerning the particular contacts made between
variant residues and their neighbors in the three-dimensional
structure. It is clear that a detailed analysis of the different
26
27
28
28
29
29
31
29
26
31
40
33
30
29
31
32
31
34
35
sidechain-sidechain interactions that arise in each variant
and their consequences in terms of the stability and activity
of the human G6PD molecule must await the three-dimensional structure determination of both human G6PD B and
of the different variants. We will confine our detailed discussion to three particular examples.
Class I variants: Subunit contacts. A striking previous
observation has been that a set of severely deficient variants
associated with CNSHA is clustered in exons 10 and 1l,29
and it has been suggested that they correspond to the NADP+
binding site. It is now clear that the binding site is elsewhere,
and this cluster corresponds in fact to the subunit interface.
These variants are located close to the ends of the sheet
strands of the P + a domain; they form two surfaces, each
shared by both subunits (Fig 4). The surfaces also include
some severely deficient mutations mapping outside exons 10
and 11. The fold brings residues from different regions of
primary structure in the same monomer close together in
space: residues 213 and 216 on a f are a part of the same
surface patch as 393, 394, 396, and 398 on PL and in the
PL-PM turn. Similarly, the dyad axis brings residue 405, at
the end of BM, and 410, in the PM-PN turn, of the second
subunit into the same surface patch. It can be seen that
residues at the two ends of PL contribute to different patches;
it is the residues that precede @Lof the second subunit, 385,
386, and 387, which are in the same cluster as 393, 394,
396, and 398 of the first subunit (Fig 4). It should be pointed
out that, of the class I residues close to or in the dimer
surface, only residue 440 is within 10 A of the modeled
substrate site of its own subunit. None of the residues is
within 10 A of either coenzyme site or of the substrate site
in the second subunit. The large distance between the two
active sites of the dimer has already been noted.
Mutations affect both hydrophobic and charge-charge in-
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
2980
NAYLOR ET AL
teractions. Salt bridges may be broken: for instance in the
variants G6PD Corum,28274Glu ”+ Lys, and G6PD Iowa,29
386Lys Glu. Glu 274 makes a salt bridge to Arg 348 and
Lys 386 is a neighbor of Glu 417; G6PD Iowa is described
as having greatly decreased thermal stability (E. Beutler, L.
Forman, P.A. Alarcon, unpublished results, 1986). The effect
of NADP+ and NADPH in changing the equilibrium between
monomers, dimers, and tetramed6 together with the possibility of reactivating mutants such as G6PD Iowa in high
concentrations of NADP+ has led to the proposal of Lys 386
as an NADP+ binding re~idue.’~
However, it is 36 W from
His 201, thought to be the site of substrate binding and is
even more distant from the known coenzyme site of the L
rnesenteroides enzyme, and so is unlikely to bind catalytic
NADP+. The rate-limiting dimerization stage in the reactivation of urea denatured L rnesenteroides G6PD has been
shown to be enhanced by addition of coen~yme.’~
Stimulation of these processes by ligands is likely to arise by their
stabilizing an important intermediate (or the end product) in
the folding pathway by forming a binary complex; for L
mesenteroides G6PD an intermediate is indicated. The position of Lys 386 in the model for the human enzyme, in the
dimer interface and distant from the proposed NADP+ site,
would suggest a similar mechanism.
Hydrophobic contact surfaces at the interface are decreased in some variants: for instance, in G6PD Harila~u,’~
216Phe + Leu. Two variants, namely G6PD Clinic,’”
405Met ”* Ile, and G6PD
213Val +Leu, disrupt
the same hydrophobic contact, which involves residues from
exon 7 as well as exon 10. The sidechains of these residues
on different monomers make contact across the dimer interface. The large number of class I variants in this part of the
dimer contact area suggests that the contribution of these
residues to inter-subunit contacts is important for the integrity of the dimer andthat this region of the interface is
particularly sensitive to change and is crucial to stability of
the enzyme.
Class I variant with an eight-residue deletion. Only one
deletion of more than two residues has been described thus
far. In G6PD Nara,” associated with CNSHA, residues 3 18325 are deleted. Some of these residues are part of a surface
loop that was found to be highly mobile in L rnesenteroides
G6PD. There are insertions and deletions in the sequence
alignment in this region (Fig l). The deletion in the human
enzyme at 304 is secure and is a prokaryote/eukaryote difference; the alignment of pH, which is well conserved, generally preserves the amphiphilic character and has equivalent
small and large residues. The residues of the loop between
3 1 1 and 3 17 need notbe equivalent in the three-dimensional
structure, but equivalence is certainly achieved by Tyr 322.
In the crystal structure of the bacterial enzyme, the residues
that are aligned with those in the deletion immediately follow
the only loop for which sidechain conformations are not
clearly identifiable. The loop has the highest temperature
factors seen in the L mesenteroides G6PD structure6; the
average main-chain temperature factor f?r the residues 292297 (312-317 in human G6PD) is 64.5 A? whereas that for
the whole protein is 33.7 A.’ The great flexibility of this
+
loop in G6PD from L mesenteroides suggests that a loop of
different length and conformation as proposed for the normal
human enzyme is readily accommodated (Fig 5A). Removal
of the eight residues 3 18-325, including the totally conserved
Tyr 322, can be accommodated by a local rearrangement of
the main chain (Fig 5B). A direct connection between residues 317 and326canbe
made without disruption of the
structure butat the expense of exposing otherwise buried
residues. The major loss of activity in the human enzyme is
likely to arise from deletion of Tyr 322, which shields these
hydrophobic residues, Val 309 and Gly 310, from solvent.
This would explain why G6PD Nara still has residual activity
but is highly unstable.
Less deleterious variants: G6PD A-. Not all sequenced
variants have been classified in terms of activity. As well as
the class I variants, considered in detail above, we have
considered the locations of those 15 of the least deleterious
variants that have known single-residue mutations, retain at
least 10% residual activity and are not associated with
CNSHA. Withonly a model of the human enzyme, we
should not seek explanations for the ways in which thestructure may alter to accommodate each of these changes. Nonetheless we havenotedthat there is a strong tendency for
these variant residues to be accessible, on external helices
or loops. Only one variant, 285Arg -+ His, involves mutation
ofan inaccessible residue that is well conserved between
species, and our modelwill accommodate a histidine. In
contrast, of the class I variant residues, 12 have verylow
accessibility and 2 of these and 5 others are totally conserved.
Among the less deleterious variants, G6PD A-, the common African deficient variant contains two mutations,
68Val+ Met and 126Asn + Asp relative to G6PD B.?*The
126Asn Asp mutation is found on its own in the nondeficient variant G6PD A.By contrast, the mutation 68Val
Met, which is the commonest in G6PD A-, has not yet been
encountered on its own and when artificiallyengineered into
normal G6PD B it does not cause deficiency.” It has
been concluded that the two mutations 126Asn + Asp and
68Val+ Met act synergistically to cause enzyme deficiency.
This so far unique findingisnow justified by our threedimensional model: indeed, we find that these twy residues,
on OB and (YC, are spatially very close (some 8 A apart, as
illustrated inFig 6). In the absence of the experimentally
determined structure of the human enzyme, it is not reasonable to speculate further on the detailed differences in the
sidechain interactions of the residues intheA-variant.
However, we may surmise that the two substitutions interact
specifically with each other causing disruption in the protein
structure, possibly in the ,&sheet of the coenzyme domain.
To understand infiner detail what amino acid changes
may affect the dimer-tetramer equilibrium, and what weight
must be assigned to different changes that may contribute to
enzyme instability in the many known variants with differing
severity of G6PD deficiency, it will be necessary to solve
the three-dimensional structure of human G6PD. However,
the structure modeled here on the basis of the L rnesenteroides G6PD has already shown that residues in the active site
+
+
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
A 3 0 MODEL OF DEFICIENT MUTANTS IN HUMAN GGPD
have not been modified in known variants, that a weakened
dimer interaction is responsible for a large class of variants
which give rise to CNSHA, and that residues distant in the
primary structure and from different subunits are involved
in the same sensitive region of the dimer interface.
ACKNOWLEDGMENT
We are grateful to Prof L.N. Johnson for facilities and support
and to Robert Esnouf for use of his program BOBSCRIPT.
REFERENCES
1. Luzzatto L, Mehta A: Glucose-6-phosphate deficency, in
Scriver CR, Beaudet AL, Sly WS, Valle D (eds): The Metabolic
Basis of Inherited Disease (ed 7). NewYork, NY, McGraw-Hill,
1995, p 3367
2. Greene LS: G6PD deficiency as protection against falciparummalaria- An epidemiologic critique of population and experimental
studies. Yearbook Physi Anthropol 36:153, 1993
3. Persico MG, Viglietto G, Martini G, Toniolo D, Paonessa G,
Moscatelli C, Dono R, Vulliamy T, Luzzatto L, D’Urso M: Isolation
of human glucose 6-phosphate (G6PD) cDNA clones: Primary structure of the protein and unusual 5’ non-coding region. Nucleic Acids
Res 14:2511, 1986
4. Vulliamy TJ, Beutler E, Luzzatto L: Variants of glucose 6phosphate dehydrogenase are dueto mis-sense mutations spread
throughout the coding region of the gene. Hum Mutat 2:159, 1993
5. Hirono A, Miwa S: Human glucose-6-phosphate dehydrogenase: Structure and function of normal and variant enzymes. Haematologia (Budap) 25:85, 1993
6. Rowland P, Basak AK, Gover S, Levy HR, Adams MJ: The
3-dimensional structure of glucose 6-phosphate dehydrogenase from
Leuconostoc mesenreroides refined at 2.0 A resolution. Structure
2:1073, 1994
7. Cancedda R, Ogunmola G, Luzzatto L: Genetic variants of
human erythrocyte glucose 6-phosphate dehydrogenase. Discrete
conformational states stabilised by NADP+ and NADPH. Eur J
Biochem 34:199, 1973
8. Levy HR: Glucose 6-phosphate dehydrogenases. Adv Enzymol
Relat Areas Mol Biol 48:97, 1979
9. Lee W, Flynn TG, Lyons C, Levy HR: Cloning of the gene
and amino acid sequence for glucose 6-phosphate dehydrogenase
from Leuconostoc mesenteroides. J Biol Chem 266:13028, 1991
10. Scanlan DJ, Newman J, Sebaihia M, MannNH, Carr NG:
Cloning and sequence analysis of the glucose-6-phosphate dehydrogenase gene from the cyanobacterium Synechococcus PCC 7942.
Plant Mol Biol 19:877, 1992
11. Bamell W, Yi KC, Conway T: Sequence and genetic organisation of a Zymomonas mobilis gene cluster that encodes several
enzymes of glucose metabolism. J Bacteriol 172:7227, 1990
12. Nogae I, Johnston M: Isolation and characterization of the
ZWFI gene of Saccharomyces cerevisiae, encoding glucose 6-phosphate dehydrogenase. Gene 96:161, 1990
13. Fouts D, Ganguly R, Gutierrez AG, Lucchesi JC, Manning
JE: Nucleotide sequence of the Drosophila glucose 6-phosphate dehydrogenase gene and comparison withthe homologous human
gene. Gene 63:261, 1988
14. Ho Y-S, Howard AJ, Crap0 JD: Cloning and sequence of a
cDNA encoding rat glucose 6-phosphate dehydrogenase. Nucleic
Acids Res 16:7746, 1988
15. Toniolo D, Filippi M, Dono R, Lettieri T, Martini G: The
CpG island in the 5’ region of the G6PD gene of man and mouse.
Gene 102:197, 1991
2981
16. Rowley DL, Wolf RE Jr: Molecular characterization of the
Escherichia coli K-l2 zwf gene encoding glucose 6-phosphate dehydrogenase. J Bacteriol 173:968, 1991
16a. Sidman KE, George DE, Barker WC, Hunt L: The Protein
Identification Resource (PIR). Nucleic Acids Res 16(S):1869, 1988
17.O’Brien E, Kurdi-Haidar B, Wanachiwanawin W, Carvajal
J-L, Vulliamy TJ, Cappadoro M, Mason PJ, Luzzatto L: Cloning
of the glucose 6-phosphate dehydrogenase gene from Plasmodium
falciparum. Mol Biochem Parasitol 64:313, 1994
18. Jeffery J, Persson B, Wood I, Bergman T, Jeffery R, Jornvall
H: Glucose-6-phosphate dehydrogenase. Structure-function relationships and the Pichia jadinii enzyme Eur J Biochem 212:41, 1993
19. Barton GJ: Protein multiple sequence alignment and flexible
pattern matching. Methods Enzymol 183:403, 1990
20. Kabsch W, Sander C: Dictionary of protein secondary structure: Pattern recognition of hydrogen bonded and geometrical features. Biopolymers 22:2577, 1983
21. Jones TA, Kjeldgaard M: 0-The manual (Version 5.9). Uppsala University, Sweden, 1993
22. Briinger AT: X-PLOR. Version 3.1. A system for X-ray crystallography and NMR. New Haven, CT, Yale University Press, 1992
23. Rossmann MG, Liljas A, Brandtn CI, Banaszak W:Evolutionary and structural relationships among dehydrogenases, in Boyer
PD (ed): The Enzymes Vol XI (ed 3). New York, NY, Academic,
1975, p 61
24.Xu WM, Westwood B, Bartsocas CS, Malcorraazpiazu JJ,
Indrak K, Beutler E: Glucose-6-phosphate-dehydrogenasemutations
and haplotypes in various ethnic-groups. Blood 85257, 1995
25. Hirono A, Fujii H, Shima M, Miwa S: G6PD Nara-A new
class-l glucose-6-phosphate-dehydrogenasevariant with an 8 aminoacid deletion. Blood 82:3250, 1993
26. Beutler E, Kuhl W, Gelbart T, Forman W: DNA sequence
abnormalities ofhuman glucose-6-phosphate dehydrogenase variants. J Biol Chem 266:4145, 1991
27. Poggi V, Town M, Foulkes NS, Luzzatto L: Identification a
single base change in a newhuman mutant glucose-6-phosphate
dehydrogenase gene by polymerase-chain-amplificationof the entire
coding region from genomic DNA. Biochem J 271:157, 1990
28. Mason PJ, Sonati MF, MacDonald D, Lanza C, Busutil D,
Town M, Corcoran CM, Kaeda JS, Stevens D, Al-Ismail S, Altay
C, Hatton C, Lewis DS, McMullin MF, Meloni T, Paul B, Pippard
M, Prentice AG, Vulliamy TJ, Luzzatto L: New glucose-6-phosphate
dehydrogenase mutations associated with chronic anaemia. Blood
85:1377, 1995
29. Hirono A,Kuhl W, Gelbart T, Forman L, Fairbanks VF,
Beutler E: Identification of the binding domain for NADP of human
glucose-6-phosphate dehydrogenase. Proc NatlAcad
Sci USA
86:10015, 1989
30. Corrons JLV, Rovira A, Pujades A, Vulliamy T, Luzzatto
L: Molecular heterogeneity of glucose-6-phosphate-dehydrogenase
(G6PD) in Spain and identification of 2 new base substitutions in
the G6PD gene. Blood 84:551a, 1994 (abstr, suppl 1)
31. Beutler E, Westwood B, Prchal JT, Vaca CS, Bartocas CS,
Baronciani L: New glucose-6-phosphate dehydrogenase mutations
from various ethnic groups. Blood 80:255, 1992
32. Hirono A, Fujii H, Hirono K, Kanno H, Miwa S: Molecular
abnormality of a Japanese glucose-6-phosphate-dehydrogenasevariant (G6PD Tokyo) associated with chronic non-spherocytic hemolyticanemia. Hum Genet 88:347, 1992
33. Beutler E, Kuhl W, Saenz GF, Rodriguez W Mutation analysis of glucose-6-phosphate dehydrogenase (G6PD) variants in Costa
Rica. Hum Genet 87:462, 1991
34. Hirono A, Nakayama S, Fujii H, Miwa S: Molecular abnormality of a Japanese glucose-6-phosphate dehydrogenase variant
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
2982
NAYLOR ET AL
(G6PD Kobe) associated with a greatly increased affinity for galac38. Hirono A, Beutler E: Molecular cloning and cDNA sequence
tose-6-phosphate. Am J Haematol 45:185, 1994
for human glucose-6-phosphate dehydrogenase. Proc Natl Acad Sci
35. Vulliamy TJ, D’Urso M, Battistuzzi G, Estrada M, Foulks NS,
USA 85:3951, 1988
MartiniG,Calabro V, Poggi V, Giodono R, Town M, Luzzatto L,
39. Town M, Bautista JM, Mason PJ, Luzzatto L: Both mutations
Persico MG: Diverse point mutations in the human glucose 6-phosphatein G6PD A- are necessary to produce the G6PD deficient phenodehydrogenase gene cause enzyme defieciency and mild or severe he- type. Hum Mol Genet l :17l , 1992
molytic anaemia. Proc Natl Acad Sci USA 85:5171, 1988
40. Filosa S, Cai WW, Galanello R, Cao A, Demattia D, Schettini
36. Bonsignore A, Cancedda R, Nicolini A, Damiani G, De Flora
F, Martini G: A novel single-base mutation in the glucose-6-phosA: Metabolism of human erythrocyte glucose-6-phosphate dehydrophate-dehydrogenase gene is associated with chronic non-spherogenase-VI. Interconversion of multiple molecular forms. Arch Biocytic hemolytic-anemia. Hum Genet 94:560, 1994
chem Biophys 147:493, 1971
41. Kraulis PJ: MOLSCRIPT: A program toproduceboth de37. Haghighi B, Levy HR: Glucose-6-phosphate dehydrogenase
tailed and schematic plots of protein structures. J Appl Crystallogr
from Leuconostoc mesenteroides. Kinetics of reassociation and reac24:946, 1991
tivation from active subunits. Biochemistry 21:6429, 1982
From www.bloodjournal.org by guest on April 30, 2017. For personal use only.
1996 87: 2974-2982
Glucose 6-phosphate dehydrogenase mutations causing enzyme
deficiency in a model of the tertiary structure of the human enzyme
CE Naylor, P Rowland, AK Basak, S Gover, PJ Mason, JM Bautista, TJ Vulliamy, L Luzzatto and
MJ Adams
Updated information and services can be found at:
http://www.bloodjournal.org/content/87/7/2974.full.html
Articles on similar topics can be found in the following Blood collections
Information about reproducing this article in parts or in its entirety may be found online at:
http://www.bloodjournal.org/site/misc/rights.xhtml#repub_requests
Information about ordering reprints may be found online at:
http://www.bloodjournal.org/site/misc/rights.xhtml#reprints
Information about subscriptions and ASH membership may be found online at:
http://www.bloodjournal.org/site/subscriptions/index.xhtml
Blood (print ISSN 0006-4971, online ISSN 1528-0020), is published weekly by the American
Society of Hematology, 2021 L St, NW, Suite 900, Washington DC 20036.
Copyright 2011 by The American Society of Hematology; all rights reserved.