Full PDF

Document technical information

Format pdf
Size 9.7 MB
First found May 22, 2018

Document content analysis

Category Also themed
Language
English
Type
not defined
Concepts
no text concepts found

Persons

Robert Huber
Robert Huber

wikipedia, lookup

Organizations

Places

Transcript

J . Mol. Biol. (1992) 225, 739- 753
Refined Crystal Structure of a Recombinant Immunoglobulin
Domain and a Complementarity-determining
Region 1-grafted Mutant
Boris Steipe 1t, Andreas Pliickthun2 and Robert Huber 1
1
Abteilung Strukturforschung, Max-Planck-Institut ju1· Biochemie, Am Klopjerspitz
8033 M artinsried, Germany
2
Genzentrum der Ludwig-Maximilians-Universitiit, cjo Max-Planck-Institute fur Biochemie
Am Klopferspitz, 8033 Martinsried, Germany
(Received 13 August 1991; accepted 15 January 1992)
We report the solution of the crystal structure of a mutant of the immunoglobulin VL
domain of the antibody McPC603, in which the complementarity-determining region 1
segment is replaced with that of a different antibody . The wild-type and mutant crystal
structures have been refined to a crystallographic R-factor of 14·9% at a nominal resolution
of 1·97 A. A detailed description of the structures is given. Crystal packing results in a
dimeric association of domains, in a fashion closely resembling that of an Fv fragment. The
comparison of this V L domain with the same domain in the Fab fragment of McPC603 shows
that the structure of an immunoglobulin V L domain is largely independent of its mode of
association , even in places where the inter-subunit contacts are not conserved between VL
and VH· In all three complementarity-determining regions we observe conformations that
would not have been predicted by the canonical structure hypothesis. Significant differences
between the V L domain dimer and the Fab fragment in the third complementaritydetermining region show that knowledge of the structure of the dimerization partner and its
exact mode of association may be needed to predict the precise conformation of antigenbinding loops.
Keywords: VL domain; domain interaction; CDR; canonical structures; antibodies
1. Introduction
(VL) combines with a variable domain of the heavy
chain (V Hl in the antibody molecule to form the
heterodimeric Fv fragment , the smallest immunoglobulin substructure that is fully competent to
bind the antigen . A number of structures of
immunoglobulin variable domains have been determined since the early 1970s. These proteins were
either light-chain dimers secreted by myelomas, or
the proteolytically accessible antigen-binding antibody fragment (Fab).
Research into the structures of specific CDR
sequences is of interest beyond basic science, where
they are regarded as a paradigm for local effects on
protein structure. The construction of recombinant
" humanized" antibodies, in which all the CDRs of a
specific murine antibody have been grafted onto a
human framework (e.g. see Riechmann et al. , 1988),
has recently become of interest for new therapeutic
approaches in medicine. As sequence information
for specific antibodies is nowadays readily obtainable, understanding of the sequence-structure
relationships of the CDRs has become the focus of
With immunoglobulins, nature has perfected a
protein engineering system permitting the generation of a seemingly unlimited repertoire of complementary molecular surfaces. The variable, antigenbinding immunoglobulin domain consists of a
conserved core structure formed by a /3-barrel, and
topped by three loops of highly variable length and
sequence, the so-called complementarity-determining regions (CDR:j:: for recent reviews, see
Alzari et al., 1988; Davies et al., 1990). These loops,
grafted on a structurally conserved platform, confer
antigen affinity and binding specificity to the antibody molecule. A variable domain of the light chain
t Author to whom correspondence should be
addressed.
t Abbreviations used: CDR, complementaritydetermining region; VL- D , homodimer of VL domains;
VcFab , VL domain in the context of the Fab fragment;
RMSD , root-mean-square difference; r.m .s. ,
root-mean-square.
0022-2836/92/110739- 15 803.00/0
739
© 1992 Academic Press Limited
B. Steipe et al.
740
Table 1
Aligned sequences fm· the VL CDRI of McPC603, MOPCI67 and the mutant M3
Position:
28
29
30
31
3l a
3lb
3lc
3ld
3I e
3lf
32
McPC603 VL
M3:
MOPCI67 VL
Ser
Ser
Ser
Leu
Leu
Leu
Leu
Leu
Leu
Asn
Tyr
Tyr
Ser
Lys
Lys
Gly
Asn
Asp
Asp
Glu
Gly
Gly
Lys
Lys
Lys
Asn
Asn
Thr
Phe
Phe
Tyr
Mutated residues relative to McPC603 are indicated by bold type.
theoreti cal (representative examples are: Stanford &
Wu , 1981 ; Snow & Amzel, 1986; Fine et al. , 1986;
Shenkin et al. , 1987 ; Bruccoleri et al., 1988) and
comparative efforts (representative examples are:
Mainhart et al. 1984; de Ia Paz et al., 1986; Chothia
& Lesk , 1987 ; Martin et al. , 1989 ; Chothia et al. ,
1989; Holm et al. , 1990) . A particularly attractive
approach is the " canonical-structure" hy pothesis
that has been forwarded by Chothia & Lesk (1987).
Their co mparison of published immunoglobulin
structures and sequences seems to indicate that the
fold of the CDRs critically depends only on the
nature of a small number of conserved residues and
the length of the loop . Unfortunately , the relative
scarcity of high -resolution structural information
slows progress in this fi eld , and it is not yet clear
whether this canonical-structure approach will be
generally app licable for accurate and reliable structural predictions .
With the successful functional expression of
immunoglobulin domains in Escherichia coli (Skerra
& Pliickthun , 1988; reviewed by Pliickthun , 1991), it
has become possible to easily obtain sufficient quan tities of specifi c immunoglobulin domains for
detailed study. We have previously reported the
crystallization and the so lution of the crystal structure of the recombinant K-VL dom ain of the phosphorylcholine-binding
murine
IgA
antibody
McPC603 (Giockshuber et al. , 1990). This structure
shows a dimeric crystal packing arrangement corresponding to VL dimers previously reported in the
literature (Fehlhammer et al. , 1975; Epp et al. , 1975;
Colman et al. , 1977) , resembling the arrangement of
VL and VH in t he antigen-binding Fv fragment. As
the structure of the whole Fab fragment of
McPC603 has been published previously (Satow et
al ., 1986) , we can now compare the structure of a VL
domain in the context of the homodimer (VL-D)
with the structure of the same domain in the heterodimeri c association (V c Fab).
Interestingly , none of the three CDRs of this
domain participates in interdomain contacts in the
homodimer structure. We thus predicted that this
system would be particularly useful for the structural study of different hypervariable loops. Indeed,
we were able to crystallize a mutant of the McPC603
VL , in which the first CDR sequence (CDR1) was
replaced by that of a different phosphorylcholinebinding antibody, MOPC167 (Potter, 1977;
Perlmutter et al., 1984) . The mutant domain, named
M3 , crystallized isomorphously under the same
conditions as the wild-type protein . We have now
solved the structure using the same model as
described previously (Giockshuber et al., 1990),
refined the structural model to a crystallographic Rfactor of 14·9 % and used the refined model for a
final refin ement of the wild-type domain .
2. Materials and Methods
(a) Preparation and crystallization
The preparation, crystallization and initial structural
determination of the McPC603 VL (subsequently called
M603) has been described in detail (Giockshuber et al.,
1990) . A mutant of this VL fragment, called M3, replaces
the CDR1 of McPC603 with that of MOPCI67 (Table I).
The plasmid encoding the Fv fragm ent of the mutant
M3 , constru cted by site-directed mutagenesis of the syntlleti c M603 gene , was obtained from J. Stadlmiiller. The
first I14 amino acid residues of the light chain, comprising
t he VL domain , are encoded. The M3 Fv fragm ent binds
phosphory lcholine with a binding constant of 3·5 x 10 5
M- 1 (Stadlmiiller , I99I) . Since this is similar to that
observed for the Fv fragm ent of McPC603 (Skerra &
Pliickthun , 1988) , the ex perimental protocol described
previously for the purification of M603 could ·be followed
for the preparation of M3 . The purified protein crystallized from 1·7 M-a mmonium sulfate, O·I M-acetate, (pH
4·0). Large, stable , macroscopically hexagonal crystals
grew within 2 days, and were morphologically indist inguishable from t he wild-type. They were completely
isomorphous to the M603 crystals with space group P6 1 22
and cell dimensions of a = b = 86·5 A, c = 74·6 A (I
A = 0·1 nm) . The asy mmetri c unit contains 1 VL domain.
Thus 12 VL domains, with a combined molecular weight of
148·8 kDa , occupy a unit cell of 4·8 x 10 5 A 3 . The ratio
volumefmolecular weight is 3·25 A3 /Da and the solvent
content is estimated to be 62 % (Matthews, 1968). This
value is at the upper end of those observed for protein
crystals and it is remarkable that, in spite of the high
solvent content, the crystals are very well ordered and
diffract to beyond I·9 A resolution.
(b) Data collection
A complete, native data set consisting of 96,963
measurements (Rmerge = (L-1 <I) )/"LI = 0·086) for I1 ,262
unique reflections was collected from a single crystal on a
FAST diffractometer (Enraf- Nonius , Delft) using a
Rigaku Rotaflex rotating-anode generator . Measurements
were made using the MADNES software (Messerschmidt
& Pflugrath, 1987). Measurements were corrected for
relative scale , temperature factors and absorption (Huber
& Kopfm ann , I969 ; Steigemann, I974; Messerschmidt et
al., 1990). The Rsym of averaged ¥riedel pairs was 0·029.
Of the possible reflections to I·97 A, 9I ·3 % were measured
(more than 2a a.bove background). The last shell of resolution from 1·99 A to I·97 A was complete to 50%.
741
High-1·esolution Stmcture of R ecombinant V L
Table 2
Co?Tespondence of 1·esidue numbers following Kabat
et al. (1987) and Chothia & L esk ( 1987) and a
sequential numbe1·ing for M 3 and M 603 VL domains
Kabat
et al. ( 1987)
Sequential i\1603
Seq uential i\13
1- 31
31a
31b
3 1c- 31f
32-109
1- 31
32
33
34- 37
38- 11 5
1-3 1
32
33-36
37-11 4
M603 has an insertion of 6 residues relative to the Kabat
alignment (31a to 311) and l\13 has an insertion of 5 residues
(deleting position 31 b).
(c) Structure solution
As the crystals were isomorphous to those described
previously for the wi ld-type domain , the structure was
readily solved by difference-Fourier techniques using the
model phases for the wild-type VL dom a in .
(d) N omenclatu1·e and definitions
The numbering of residues throughout this paper is
that proposed by Kabat et al. (1987) . The correspondence
between this numbering and a sequential numbering is
given in Table 2. Amino acids from sy mmetry-related
domains will be designated by a #. The numbers
denoting position of amino acid residues from the heavy
chain are preceded by the letter H. T he /1-strands wi ll be
designated as described by Marquardt & Deisen hofer
(1982), starting with strand A at t he N terminus and
ending with the C terminal strand H. To describe directions and orientations within a doma in , we wi ll consider
the dimerization inte rface of a domain to be t he " front" ,
the CDRs to lie on " top" a nd theN and C terminus to lie
on the " left " -hand side. Structural data for comparisons
has been taken from t he Brookhaven Protein Data Bank
(Bernstein et al. , 1977) using t he followin g entries: 1REI
(V L dimer REI, Epp et al. , 1975), 2FB4 (Fab KOL ,
Marquart et al. , 1980) , 2FBJ (Fab J539 , Suh et al. , 1986) ,
2HFL (Fab HyHel 5, Sheriff et al., 1987) , 2MCG (A chain
dimer MCG, Ely et al., 1989) , 2MCP (Fab McPC603,
Satow et al., 1986), 2RHE (V L dimer RHE , Furey et al.,
1983), 3HFM (Fab HyHel 10, Padlan et al ., 1989), 3MCG
(A chain dimer MCG, Ely et al., 1989) and 4FAB (Fab 4-420, H erron et al., 1989). All superpositions for structural
comparison were done using co-ord in ates for the residues
that we designate the core fi-strand region : 3 to 7 and 9 to
14 (not for}, chains), 16 to 24, 33 to 38, 45 to 50 , 52 to 55 ,
61 to 66 , 70 to 76 , 84 to 90 and 97 to 107. Superposition
matrices were calculated to minimize the root-meansquare co-ordinate difference (RMSD) for theN , C", C and
0 atoms of these residues. Hydrogen bonds were considered if the donor-H· ··acceptor distance was less than
2·7 A and the donor-H· ··acceptor angle was greater than
145°. The co-ordinates of the wild-type V L - D and the M3
mutant ha ve been deposited with the Brookhaven Protein
Data Bank and are avai lab le directly from the authors on
request until they have been processed and released.
(e)
Refinement of M3
An initial stru ctura l model of t he M3 protein was based
on t he co-ordinates of the V L domain from the McPC603
Fab fr ag ment. These were rotated into the orientation
found for the wild-type VL domain (Giockshuber et al.,
1990) and the mutated loop was built into the difference
electron density map calculated from the wild-type and
M3 reflection data and the preliminary wild-type VL
phases .
Refinement was started for th is initial model employing
the progra m EREF (energy restrained crystallographic
refinement: J ac k & Lev itt, 1978). R efin ement cycles were
run until the stru cture's R-factor had converged , followed
by interactive model building on an Evans and
Suther land graphics terminal using the program FRODO
(Jones, 1978).
As the refin ement progressed to an R-factor below
25% , solvent molecules were placed into regions of the
F 0 -Fc difference electron density map which were above
4·1 a of the mean and less than 4·0 A distance from the
protein atoms. Only such so lvent molecules that were
consistently observed, did not significantly increase their
B-va lues during furth er refinement and were in stereochemi cally plausible positions were retained . A strong peak of
residua l electron density located between 2 symmetryrelated arginine side-chains was interpreted as a su lfate
ion. A consistent residual electron density peak at the
surface of the protein , conta ining 4 maxima, was interpreted as a n acetate ion. A summary of the refinement
process is given in T a ble 3.
\\Thereas most regions of the protein refined without
major rebu ilding, t he site of the mutation itself, the
CDR1 , had to be remodelled several times. The electron
density of this loop is weaker than that of the rest of the
molecule a nd it was diffi cult to interpret unambiguously
the stru cture of the main chain. Atoms without defined
electron density were not used for calcul ation of structure
factors.
Table 3
Summary of M 3 crystallogmphic refinement
Cy·les
1-4
5-6
7- 12
13
14
15- 19
20
Procedure
Start refi nement, data from 8·0 to 2·5 A
Manual adjustments, 5 so lvent molecules added
54 solvent molecul es and I sulfate ion added
Data from 8·0 to 1·97 A, manual ad justments
48 solvent molecules added
Use un constra ined individual atomic B values and add
phase co rrection for disord ered so lvent space
Manual adjustments, 4 1 so lvent mol ecules added
Add I acetate mol ecul e, orient carboxamide groups
R-factor
34·0
25·4
2 1·5
19·7
18·1
16·5
15·0
14·9
Manual ad justments were foll owed by EREF cycles. The final model posesses a standard deviation of
bond lengths of 0·015 A, and a sta ndard deviation of bond a ngles of 2·23 ° from idea l val ues.
B. Steipe et al.
742
120
60
Ala51 •
'
'
''
'
''
''
''
'
-60 ~~~~:;-J
- 120
''
'
''
'
'
-- ----r--- - - ~ - - ----
0
- 180
-120
0
'
'
'
'
'''
''
------~ - -----r---- --
'
-60
'
'
''
-- - --- +---- - -~----- -
:o
0
4>
60
0
120
180
(0)
Figure 1. Ramachandran plot for the (c/J ,t/1) torsional
angles of the M3 V L domain. Glycine residues are plotted
with a square, all other residues are plotted with a circle.
The residues ly ing outside the predicted conformational
boundaries are labelled .
The final model shows clear electron density for all
residues from Asp1 to Lysl07. The side-chain of Argl08
and the last residue, Ala109 , are disordered. The solventexposed loop of CDR1 gives only poor electron density .
H ere, Lys3 1a is almost undefined . The main-chain atoms
from N Asp31c to C Gly31d a lso show weak electron
density . The whole loop seems to be rather flexible , with
main chain B-values above 50 A2 Thus the structural
model has to be interpreted with some caution in this
region.
(f) Refinement of the M603 V L domain
A model for the V L domain was constructed by building
the CDR1 loop between 31 and 31e from the co-ordinates
of the VL domain from the McPC603 Fab fragment superimposed onto the refined M3 co-ordinates. This model
immediately refined . to an R-factor of 15·8 %. The Rfactor dropped to 14·9% after a final round of minor
modifi cations. Most so lvent molecules refined into identical positions as in the M3 mutant with only minor
cha nges in their B-values.
3. Results
(a) The final models
The final models for M3 and M603 contains 864
and 866 protein ato ms as well as 121 and 123
solvent atoms, respectively. The two structures are
practically identica l. The root-mean-square (r.m. s. )
co-ordinate difference for all atoms (excluding
solvent and the mutated loop) between lVI3 and
M603 is 0·06 A. Thus, except where explicitly
stated , the followin g description is valid for both
structures. Th e fin al R-factor for both models is
14·9 % with a standard deviation of bond lengths
from ideal values of 0·015 A and of bond angles of
2·23 °. The maximum average co-ordinate error as
calculated from a Luzzati plot (Luzzatti, 1952) is
0.16 A. No residual peaks greater than 10·25 ef.t\.31
were found in the F o-F c maps.
The Ramachandran plot for M3 is shown in
Figure 1. The majority of residues are found in the
antiparallel {J-strand region (¢ = -60° to 150 °,
ljJ = 100 ° to 170 °) (Richardson, 1981) and all
residues but five are found within the energy boundaries postulated by Ramachandran et al. (1966).
Three exceptions are Ser52 (¢ = -141°, ljJ = 7 °),
His92
(¢ = -135 °,
ljJ = 14 °)
and
Thr69
(¢ = - 115 °, ljJ = 18 °), which are located in welldefined electron density and deviate only slightly
from the low energy contours . The most conspicuous deviation is seen at Ala51 (¢ = 67 °,
ljJ = - 34 °), which will be discussed below. Gly31d
(¢ = - 36 o, ljJ = -42 °) is poorly defined in electron
density (see above).
The refined V L domain shows the typical
immunoglobulin variable domain structure (Fig. 2).
It consists of two {J-sheets with four and five antiparallel strands, forming a typical {J-barrel. The
barrel is closed on the left-hand side by a compact
subdomain containing two strands forming the
second CDR (Fig. 3). It is closed on the right-hand
side by the first {J-strand , which is divided by the
cis-proline Pro8 into a part A, hydrogen bonding in
antiparallel fashion to the B {J-strand and a part A'
bonding in parallel fashion to the H strand.
(b) Crystal packing
A view of a c• trace of domains in the unit cell ,
along the 6-fold symmetry axis , clearly shows the
symmetry elements of the P6 1 22 space group (Fig.
4). Each monomer contacts four domains in the
crystal lattice. One of these domains forms the V L
dimer, which is discussed in more detail below.
Corresponding to the relative orientation of
domains to each other, these symmetry relationships will be designated syn, anti and meta. These
are generated through the following symmetry
operations: syn: (X- Y , Y , -Z); anti: (X , X - Y ,
f,--Z); and meta: (- Y , X- Y , -!+Z or -X+ Y ,
X- , t+Z).
The syn-interaction corresponds to the VL dimer
(Glockshuber et al. , 1990). The amino acids Tyr36,
Gln38 , Pro43 , Pro44, Leu46 , Tyr49, Glu55 , Tyr87 ,
Gln89 , Asp91, Tyr94, Pro95, Leu96 , Phe98 , Gly99
and AlalOO lose more than 10% of their solventaccessible surface upon association and thus form
the dimer interface.
The anti-symmetry partner binds to the V L
domain with an antiparallel {J-strand from the
amino acids Ser7 to Ser14 and Ser# 14 to Ser#7. As
this part of the first {J-strand (A') contributes to the
front {J-sheet of the domain , this {J-sheet is extended
in the crystal latti ce to twice its size.
The meta-association is non -symmetric. In this
mode of association , the amino acids Lys24 , Thr69
and Gly70 at the top of the domain contact
Ala#15, Gly#16 and Gln#79 at the bottom of the
meta-partner (- Y , X- Y , t+Z) . Equivalently, at
the bottom of the domain Ala15 , Gly16 and Gln79
743
High-Tesolution StTuctuTe of Recombinant VL
(a)
105
(b)
Figure 2. The refined structure of t he McPC603 V L domain . (a ) The 3-dimensional structure. Solvent molecules are
omitted , except for t he 2 internal water molecules. The view is from the right hand side. The main chain is plotted wit h a.
bold line, the side-chains are plotted with t hin lines. E very lOth
ato m is labelled . (b) Secondary structure of t he L
domain. Main-chain hyd rogen bonds a re indicated with a th in line.
c·
v
B. Steipe et a!.
744
Figure 3. The CDR2 subdomain of V L· Residues from Leu46 to Ser65 are shown together with the main stabilizing
residues of the protein core, Leu33 , Trp35 a nd Asp82. The view is from the front of the domain. The fram ework C" atom
co-ordinates are connected with a thin line , t he residues mentioned in the text are drawn with a bold line. Hydrogen
bonds are drawn with a broken line .
contact L ys # 24, Thr # 69 and Gly # 70 at the top of
the next meta-partner (-X+ Y , -X, f+Z). This
top-to-bottom arrangement can be extended in
space. The projection of the long axes of metaassociated monom ers onto the " a b" plane of the
crystal latti ce gives an angle of 60 o . Meta-associated
monomers turn around the 3-fold screw axis .
Thus the crystal lattice can be easi ly described. A
helix of meta-associated monom ers turns around a 3fold screw axis . A second helix , with a phase difference of exactly one half turn , associates through the
anti-interaction. Thi s double helix now has a period
length of the unit cell height. All monomers within
this helix point their syn-interfaces radially
outward. Finally , these columns aggregate in a
hexagonal packing via the syn-interaction to form
the crystal lattice . A .protein-free tunnel with a
radius of about 30 A, pa rallel to the protein
columns, is found in the crystal latti ce at sites
corresponding to every third position of a hexagonal
close packing. All three CDRs point into this tunnel.
Additionally , a domain is linked to the metapartner of the anti-domain , (-Y, -X , i-Z) ,
through a double-salt link : two symmetry-equivalent Arg60 residues bind a su lfate ion from the
solvent.
(c) Hydrogen bonding
The secondary structure ele ments, along with the
hydrogen bonds between main-chain atoms are
shown in Figure 2(b). The structure consists predominantly of antiparallel /3-strand s, a nd one turn
of a 3 10 -helix. The boundaries of the /3-strands and
the classification of the turns are summarized
in Table 4. The four /3-II and /3-II' turns observed
here, contain glycine in the third position
(Venkatachalam, 1968). We find that 97% of mainchain polar atoms form hydrogen bonds , 54% (118)
are involved in main chain to main-chain bonds,
12 % (26) are to side-chain atoms , the rest are to
solvent molecules.
All buried polar side-chain atoms make good
hy drogen bonds. This implies that the structure is
Table 4
{3-Stmnd bounda1·ies and turns
Elem ent
Type
{3-Strand A
{3-Strand A '
{3-II turn
rr-'urn
{3-Strand
{3-Strand
Turn
{3-Strand
Turn
{3-Strand
Turn
Turn
{3-St.-and
Turn
{3-Strand
H eli x
{3-Strand
(3-Strand
B
C
{3-ll turn
D
y-turn
X
{3-II turn
{3-l turn
E
{3-II' turn
F
3 10 -Heljx
G
H
Boundaries
Val3- Ser7
Ser9- Ser14
Ala14--Gly 15
Gly16-Ser26
Lys31e-Gln38
Pro4o--Gly41
Lys45--Gly 50
Gly5Q-Ser52
Ser52--Glu 55
Ser54--Gly 57
Asp6o-Arg61
Arg61--Ser6 7
Gly68- Thr69
Thr69--Ser74
Gln79- Leu82
Ala84--Asn90
Thr97- Lysl07
Boundaries of {3-Strands are defined by the hydrogen- bonding
patterns of the structural model and classi fi cation of turns was
performed according to the 4>- and 1/J-angles of consecutive
am in o- acids as given by Richardson ( 1981).
High-resolulion Structure of Recombinant V L
Figure 4. VL doma ins in the crystallographic unit cell
viewed along t he 6 -fold axis of symmetry. A C" atom trace
is shown, with the CDRs plotted with a bold line. The
symmetry elements of the P6 1 22 space group are clearly
visible.
at a conformational optimum , in the sense that a
mutation at any of these sites would increase the
conformational energy or necessitate some (local)
refolding. The buried polar side-chain atoms are all
associated with some distinct conformational
feature of the domain , and all are among the most
highly conserved amino acid residues in the
immunoglobulins. (I) Gln6 hydrogen bonds to the
structural water molecule W2 (see below) and to
amide nitrogen atoms ofGlyiOI and CysSS. Thus, it
may be important in the formation or stabilization
of the P-bulge of AlaiOO in strand H. This P-bulge
generates the twist in the front P-sheet, which characterizes the immunoglobulin variable domain
dimerization (Chothia et al., 19S5). Gln6 is almost
completely conserved in the V L domains and found
to be mutated to Glu in only half of the VH domains.
(2) 0 1 Ser25 bonds to 0 Gln27 (2·73 A) , stabilizing
the first turn from P-strand B into the CDRL It is
not part of the core stmcture and, indeed, an
alanine is found in this position in the structures of
REI , J539, HyHel5 and H yHellO without significant structural consequences. (3) N£1 Trp35 bonds
to the buried solvent WI (see belmy). (4) 0" TyrS6
hydrogen bonds to 0 AspS2 (2·61 A) , thus affixing
745
the 3 10 -helix , which cwsses under the domain , to
the core of the structure. Again, this residue is
almost completely conserved, and it is significant
that a mutation to Phe seems to be strongly selected
against in the immunoglobulin sequence. (There are
only 2 Phe residues in this position in all the VL
sequences, and all the rest are Tyr.) This hydrogen
bond is conserved in all the published immunoglobulin VL structures. (5) Thr97 is the first
conserved residue after the CDR3. This position is
also frequently taken by a valine and thus the
requirements seem to favor a small , P-branehed
amino. acid. The 0 71 Tlu97 to 0 Ile2 hy drogen bond
(2·72 A) may stabilize theN terminus of the protein .
As can be seen fwm Figure 2(a) and (b) , the first
two residues cannot participate in the regular antiparallel P-strand, since strand B turns at the height
of these residues to cross over the top of the domain
and form the CDRL In the stmctures of KOL, MCG
and RHE , this hydrogen bond is absent, since Val ,
Val and Gly are found , respectively, in position 97 .
This correlates with an N terminus that is somewhat displaced from the framework in these structures. On the other hand , wherever Thr is found , the
N terminus is integrated into the rest of the domain
in a manner comparable to the VL-D- (6) Thrl02
forms a hydrogen bond to 0 ProS (2·7S A) to
provide the cm cial interaction that breaks the
regular main-chain hydrogen-bonding pattern , as
the first P-stmnd makes its transition from A' (antiparallel P-stmcture) to A' (parallel P-structure).
This conformation is conserved in all the published
immunoglobulin VL stmctures.
Additionally, the stereochemistry of the pola1·
side-chain atoms buried in the domain core
conforms well to the spatial preference regions
found by Ippolito et al. (I990) in their recent analysis of side-chain hydrogen bonding in well-refined
high resolution protein stmcturt.;s. The hydrogen
bond 0" Ty rS6 to 0 AspS2 (2·6I A) is found almost
perfectly in plane and at a 120 ° angle, consistent
with an sp 2 -hybridization of the tyrosine hydroxyl
group. N£1 Trp35 bonds WI only 4 o out of plane
and ll·2 ° out of line. 0 7 1 Thr97 bonds 0 lle2 I3 °
out of trans and 0 7 1 Thrl02 bonds 0 ProSSer only
13 o out of gauche-. Finally, the salt link between
Arg6I and Asp82 also displays ideal stereochemistry
(type I: Singh et al., I987) , 9 ° out of the guanidinium group plane and 3 ° out of the carboxylate
group plane, with the oxygen atoms approaching
the nitrogen atoms N' and N" 1 at 2·78 and 2·77 A.
Again this important interaction is conserved in the
other immunoglobulin structures examined. Thus ,
the confm·mation of the whole domain seems highly
optimized towards optimal stereochemist1·y of the
conserved buried polar side-chain atoms.
(d) Orientation of Nand 0 atoms in carboxamide
groups of Asn and Gln
Assuming, that the side-chain carboxamide atom
B-values should be approximately equal , those that
showed large differences in their B-values, with
746
B. Steipe et al.
values for 0 atoms larger than those for N atoms,
were rotated by 180° from their initial orientation
taken from the VcFab. The solvent-exposed sidechain of Gln27 was modelled better this way, indicating a conformational preference, while Gln79 and
Gln89 seem statistically oriented. More significant
differences are the hydrogen bonding carboxamide
groups of (1) Gln37, which hydrogen bonds with its
N' 2 to the lone pair of o~ Tyr86, which in turn
hydrogen bonds to the carbonyl group of Asp82, and
(2) Asn90, which has a stereochemically implausible
conformation in the VcFab.
(e)
The disulfide bridge
Epp et al. (1975) have described two alternative
conformations of the intramolecular disulfide bond
between Cys23 and Cys88 in the REI VL dimer.
Here, clearly only a single conformation is seen; it
possesses the torsional angles:
170 ° 121 ° 79° 167°-58 °
23: N-Ca-cP-sr-sy_cP-ea-e: 88.
Thus it would be considered a right-handed disulfide spiral (Richardson, 198I). The disulfide bond
bridges the large ca-ea distance of 6·6 A, typical
for immunoglobulin domains.
(f) The structural water molecules Wl and W2
Two water molecules are found within the V L
domains that are not part of the hydration shell,
but are integral parts of the domain architecture.
These are designated WI and W2. WI is found in a
cavity between the CDR2 subdomain and the core
of the protein. It makes three hydrogen bonds of
near-ideal geometry with the indole nitrogen atom
of the conserved core residue Trp35, and the backbone carbonyl oxygen atoms of Ala5I and Ser65.
This water molecule is found in all K-domain structures solved at sufficient resolution, REI and J539.
Thus, we postulate that this water molecule is a
universally conserved structural feature of K-VL
domains. W2 is found in a cavity between the
P-strands A, B and H, some 5 A from the protein
surface: Again, we find three hydrogen bonds with
near-ideal geometry. They connect the water molecule . with the buried glutamine carboxamide
nitrogen atom, N' 2 Gln6, with the carbonyl 0 atom
of Ser22 and with the buried hydroxyl group of
Thr102. Interestingly, this water molecule is not
found in the structure of REI even though all three
potential ligands are present. The only local structural difference is a change from methionine (Met2 I)
to isoleucine at the bottom of the cavity that
contains this water molecule. The water molecule is
present in the structure of J539, despite an isoleucine in position 21.
(g) The V L dimer compared to the
M cPC603 Fab structure
As has been noted previously, the V L monomer is
seen to associate as a homodimer in the crystal
lattice. The second domain can be generated
through the symmetry operator (X- Y , - Y, -Z)
and a translation vector of (A , 2B, C) in the nonorthogonal co-ordinate system of the unit cell. This
association is very similar to the VL dimer REI and
places the two V L domains in a relative spatial
arrangement corresponding closely to the structure
of the heterodimeric Fv fragment. A structural
superposition of the VL domains of this study and
the published McPC603 Fab shows only small differences. The core P-strand region ca atoms of the two
structures ean b~ superimposed with an r .m.s.
deviation of 0·32 A, which is within the limits seen
for other proteins crystallized under non-identical
conditions (Chothia & Lesk, 1986) . M603 and REI
have only 74% amino-acid identity in this region
(63% overall). Still, the RMSD of superposition of
these two different proteins is the same as that
between the VcD and the V cFab. As the resolution of the Fab fragment is only 3·I A, the significance of conformational differences between the
VcD and the VcFab is difficult to judge in some
cases. The more conspicuous differences are: (I) the
side-chain of Leull packs differently in the VcD ;
this allows the main chain to be modelled into more
favorable stereochemistry. (2) The loop between
Serl4 and Glu17 was rebuilt in the VL-D , the orientation of the Serl4-Alai5 peptide plane was
changed by I80 ° and the rest of the residues moved
accordingly . The Vc D conformation corresponds
more closely to the conformation seen in other highresolution VL structures. (3) Pro43 and Pro44 move
back in the VL- Fab, possibly due to different interface packing interactions. (4) The peptide plane
between Gly50 and Ala51 was turned by 180 °. (5)
The conformation of the CDR3 is significantly
changed, as discussed in more detail below.
In accordance with the observation that residues
at the interface between the two domains are highly
conserved and their spatial location is preserved
among VLand VH domains (Chothia et al., 1985), the
structure of the VcD and the VcFab superimposes
particularly well in this region. There are no major
rotational or translational movements seen . Evolutionary pressure · seems to have allowed only very
conservative changes from the time VH and V L
diverged from a common precursor. It is only in one
place, where TrpH103 of VH is replaced by Phe#98
of the VL dimer, that a small cavity would be
created . But the interface region in this three-layer
packing is sufficiently flexible to fill this space
through a slight shift in the side-chain positions of
Phe98 and Tyr36 (Fig. 5). A Gln-Gln hydrogen
bond within the interface, a structural feature found
in all immunoglobulin variable domain dimers, is
also seen in VcD. Gln38 lies at the 2-fold symmetry
axis of the dimer and forms two hydrogen bonds of
good geometry with Gin# 38. We conclude that
crystal packing and domain association play only a
minor role in the generation of this structure.
A comparison of the mode of association of the
VcD and the Fv fragment shows that the VL
domain loses 30% less of its solvent-accessible
High-Tesolution Stmctu1'e of Recombinant V L
surface through association in the VL dimer than in
the complex with the Fv fragme~t. The soh;~n~­
accessible surface of the VL domam of 25620
A IS
.
reduced to 5050 A 2 {a difference of 570 A ) .1~ t.he VL
dimer, but to 4720 A 2 (a difference of900 A ) m the
Fv fragment. Carbon atoms contribute 65(o ~f the
interface surface. The absence of the extensive mteractions that the V H CDR3 makes with V Lin the Fab
fragment does not cause significant structu:·al
changes in the V L dimer. But a number of specific
interactions between the VL domains can be
discerned, each of which would be predicted to
cause a decreased association energy of the V L dimer
as compared to the Fv fragment. (I) Asp9I , which
hydrogen bonds to AsnH95 of the VH domain in t~e
Fab fragment, is close to the 2-fold symmetry axis
in the VL dimer. Here the side-chain of Asp9I is
turned, so that the carboxyl groups of Asp9I and
Asp#9I are I2 A apart. Even so, repuls~ve int~r. actions of this paired, negatively charged side-cham,
close to the 2-fold axis of symmetry, might serve to
explain the fact that V L crystallizes only at acidic
pH. (2) The carboxyl group of Glu55 is seen to make
a close contact to the hydrophobic ring of Pro# 95
of the CDR3. This might come in conflict with the
hydration shell of the carboxyl gr~u~. ·In. the Fab
fragment, Glu55 is modelled pomtmg mto the
solvent. (3) The hydroxyl group of Tyr94 is hydrogen bonded to the carboxyl group of .GluH35 in the
Fab fragment . This hydrogen bond IS absent fro~
the V L dimer and a well-defined solvent molecule IS
found in place of the carboxyl group. (4) The most
conspicuous difference in the association is the
absence of the long CDR3 of the VH domain, which
packs against the descending part of the VL CDRI,
the C strand in the V L domain. Residue PheHIOOc is
absent from the V L dimer. This residue plugs deeply
into the interface, in the heterodimer. The top of the
interface which is the bottom of the hydrophobic
hapten-b,i nding pocket of the Fv fragment , is
formed by a hydrogen bond between Asp9I of VL
and AsnH95 of VH· The absence of this bond and
the PheHIOOc side-chain leave a deep cleft in the
interface that extends the solvent-accessible surface
almost IO A deep into the interior of the dimer (Fig.
5). Additionally, the hydrogen bond from o~ Tyr36
toN PheHIOOc has no counterpart in the VcD.
0
0
(h) St1'uctu1'e of the CDRs
(i) CDRJ
The long , solvent-exposed loop of the M603 CDRI
between Asn3I and Lys3Ie makes few interactions
with the main body of the domain. It is thus not
surprising that it is found to be partially disorder;,ed
or flexible with main chain B-values above 50 A 2 .
Although the M3 sequence is one residue shorter,
the loop is found to occupy approximately the sam.e
region of space as the M603 loop. The M3 loop IS
found similarly disordered , with the weakest electron density at the first " corner" of the loop (Ser3Ia
in M603 and Lys3Ia in M3).
747
(ii) CDR2
The CDR2 is very well defined and the electron
density permits an unambiguous definition of the
position of all involved atoms (Fig. 6(a)). It is
bridged by four hydrogen bonds. A "classical "
y-turn (Rose et al ., I985; class 3 y-turn ,
Milner-White et al. , I988) is found at the apex of the
two /3-strands D and X, and Ala5I, bracketed by
this turn, displays the unusual {¢, 1/1) combination
of(¢= 66 °, 1/J = -35 °). This region of the (¢ , 1/1)
map was originally "forbidden " by Ramachandran
et al. (I966) but was found to be favorable m more
refined calculations which allow some flexibility of
atomic bonds (e.g. see Weiner et al., I984). The
conformation around Gly50 and Ala5I is different
from the one modelled by Segal et al. (I974) in the
V -Fab (Fig. 6(b)). But the clearly defined electron
d~nsity and more favorable stereochemistry in t.his
region prompt us to postulate that the conformatiOn
observed here is the correct one. This makes the
conformation of the CDR2 uniformly conserved
among immunoglobulin VL domains .
.
The main chain continues with two /3-turns
(Ser56-Gly57, Asp60-Arg6I) after the CDR2. They
fold the main chain back onto itself to form one of
the most compact subdomains observed so far in
proteins (Zehfus & Rose , I986). This compact subdomain , running from Leu46 to Ser65 wraps around
the side-chain of Ile48 and forms a plug, sitting
against the hydrophobic core of the protein and
closing it off to the side (Fig. 3). Only theN- and Cterminal segments of this subdomain are anchored
to the main body of the protein with: (I) a peculiar
sequence of double-single-double hydrogen bonds
between /3-strands C and D (these are: N Leu47 (2·8
A) and N Ile48 (3·2 A) with 0Trp35; 0 lle48 with N
Trp35 (2·9 A); N Gly50 and N Ala5I with 0 Leu33
(both 3·0 A)) ; (2) the hydrogen bond of 0 Ala5I to
the core tryptophan Trp35 via the internal water
molecule WI; (3) a salt link between Arg6I and
Asp82 , two residues that are almost universall!
conserved in the VL domains; and (4) , regular antiparallel /3-structure of the strands E and F , beginning at Arg6I (Fig. 2(b)). Thus, the stabilization of
this subdomain largely relies on non-covalent interactions with only three residues , Leu33, Trp35 and
Asp82 , besides van der Waals' interactions.
(iii) CDR3
The CDR3 conformation is significantly different
from the V cFab conformation. Again , the electron
density is very well defined and the main chain and
side-chains of this region can be unambiguously
placed (Fig. 7(a)). The main chain is moved approximately 2 A towards the front and the side-chain of
His92 is tilted in the same direction , so that the
His92 is at a distance of 3·0 A from the VL-Fa~
conformation. The imidazole group moves by 6·8 A
(Fig. 7(b)). The RJ\!SD of t~e 44 atoms from Asn90
to Tyr94 is 2·76 A (2·05 A for the I9 backbone
atoms). Additionally, the hydrogen bonding topology of this loop changes: Asn90 is found to make
ca
748
B . Steipe et al.
( 0)
(b)
Figure 5. Stereoplot of t he. dimerizatio n interface of t he VL dimer and t he Fv frag ment. (a) The view is from t he top
a long t he 2-fold ax is of symmetry , only t he front /1-sheets and in teracting side-chains a re shown. One domain of the
Vc D dimer is drawn with bold lines, t he other is drawn with t hin lines . Solvent molecules of t his region are in cluded.
The hydrogen bonding Gln 38 a nd Gin # 38 a re clearly seen at t he bottom of t he interface. (b) View from t he right-hand
side , only the front /1-sheets a nd interacting side-cha ins are shown. Both domains of t he VL - D d imer are draw n with bold
lines, t he F v fragm ent is draw n with t hin lines . Solvent molecules for Vc D a re in cluded. The overa ll geo metry of t he
interface, which forms a /1-barrel, can be well appreciated . Note that this /1-b arrel is not closed through secondary
stru cture. The VH CDR3 ext ends over t he top of the domains . Note , by comparison, how deep t he solvent molecules
penetrate into the dimer interface in the VL - D . The side-cha ins of Phe98 and Trpl03H are clearl y seen in the center of
t he interface.
two weak hydmgen bonds with N Ser99 (3·06 A)
and 0 Ser99 (3·06 A), instead of t he strong bond to
N His98 (2·54 A) seen in VcFab.
so lvent molecul es . This shows t hat t he co nformation of t he mutated loop has no signifi cance for t he
remaining dom a in stru cture.
(i) Compa1·ison of Jlll603 and Jlll3
4. Discussion and Conclusions
Th e stru ctures of M603 a nd M3, excluding t he
mutated loop from Leu30 to Asn3lf, can be superimposed with a n RMSD of 0·06 A. Th ey are
virtually identi cal, dow n to t he B -values of most
Chothia & Lesk (1987) have put forwa rd the
in teresting hypothesis t hat for t he accm ate pred ict ion of CDR co nform ations, knowledge of t he
nature of a limi ted number of key residues a nd a
High-1·esolution StmctuTe of R ecombinant V L
749
( Q)
(b)
Figure 6. Stereoplot of the CDR2 region of VL· (a ) The residues from T y r49 to Ser62 a re show n. Ala5 l is at the top of
loop a nd t he view is from the side of the domain. An electron density map is displayed , conto ured at a. level of 0·4
ef A 3 (l ·25£T). Density for the carbon y l oxygen atoms is clearl y seen. (b) Comparison of t he VcD and VcFab CDR2. The
view is from the front as in Figure 3. V cD co-ordinates are draw n with bold lines ; V cFab co-o rdinates are drawn with
t hin lines .
th ~
representative structure mig ht be suffi cient. Th ese
key residues would co ntain t he relevant inform ation
to generate a typical "canoni cal" fold , regardl ess of
the rest of the sequen ce. By th is co mparative
modelling strat egy , it has been possibl e to predi ct
t he conformation of som e CDR loops prior to th e
publi cation of the experim ental stru cture (Chothi a
et al., 1989). If t he implied causal relationship
between the observed key residues a nd the t hree-
dim ension a l loop stru cture was of general significance, t hi s would provide us with a n instance of
bein g able to localize essentia l elements of folding
information a mong the prim ary structure. Besides,
t he knowl edge of precisely whi ch residu es a re necessary and suffi cient to generate a certain can on ical
fold , would prove extremely helpful in efforts a im ed
at predi ctin g a ntibody stru cture from seq uence data
a lone, with a view to elu cidating molecu lar inter-
750
B. Steipe et a!.
(a)
(b)
Figure 7. Comparison of the VcD and V L -Fab CDR3. (a) The residues from Asn90 to Ser93 are shown. His92 is at t~e
top and the view is from the front of the domain. An electron density map is displayed , contoured at a level of 0·4 efA3
(1·25u) . (b) The residues from Asn90 to Tyr94 are shown . VcD co-ordinates are drawn with bold lines; VcFab
co-ordinates are drawn with thin lines.
act ions with a given hapten or antigen. Indeed, once
an antibody structure is precisely known , a docking
procedure may provide the correct mode of association with the antigen in favorable cases (Goodsell &
Olson, 1990) .
How well is the canonical structure hypothesis
supported by the findings of this X-ray crystallographic study?
(l) The CDR1 , arching across the top of the
{1-barrel, is found to be pinned into the core of the
structure, in a fashion typical for CDR l s previously
observed , by a conserved hydrophobic amino acid
(Leu29 in this case). Leu29 is tightly packed into
the core with low B-values and the CDRl loop up to
Leu30 is well stabi lized. On the other hand , the long
solvent-exposed loop , continuing from Asn31 is
hardly stabilized at all. It had been necessary to
postulate a separate canonical structure for 4-4-20
(Chot,h ia ·et al. , 1989) , which is one amino acid
residue shorter than McPC603 and has a conformation very different from that of McPC603 (Herron et
al. , 1989). M3 has the same length of the CDRl as 44-20, but the conformation is found again to be
quite different, so that yet another canonical structure should have to be introduced. It is interesting
to note that the CDRl B-values of 4-4-20 are much
lower than those observed for either M603 or M3.
Since the 4-4-20 loop does not seem to have significantly more stabilizing interactions with the
domain scaffold than M603 or M3, other than the
approach to 3·5 A from an arginine residue of a
symmetry-related domain in the crystal and a
possible participation of His31 in hapten binding,
this argues for an important role for internal
stabi lization of the loop itself. Three hydrogen
bonds in 4-4-20 cou ld be important in this context:
the first bonds 0 Gln 31 b to 0'1 Tyr32 (which is
Phe32 in M3). The other two bridge the back of the
I-ligh-1·esolution StTuctuTe of Recombinant V L
751
Figure 8. Comparison of the M3 and 4-4-20 V cFab CDRl conformations. Residues from 29 to 32 are shown. The view
is from the right-hand side as in Fig. 2. M3 co-ordinates are drawn with bold. lines, 4-4-20 co-ordinates are drawn with
thin lines. The RMSD for the 36 backbone atoms from Leu29 to Phe32 is 3·75 A. The VcD CDRl conformation is drawn
with thin lines for comparison. It closely corresponds to the M3 Ll conformation. Hydrogen bonds that possibly stabilize
the loop conformation of 4-4-20 are also shown.
loop with N 62 Asn31e (which is Lys31e in M3) to 0
Ser31a and Ob 1 Asn31e toN Gly31d , in a conformation reminiscent of the stabilization of some CDR3
loops (Fig. 8). In the face of some residual uncertainty about the accuracy of the 4-4-20 model in
this region and until experimental data on the
relative importance of Tyr32, Asn31e or other
residues within the loop for stabilizing this structure
are available, we would suggest that the structure of
loops inserted between position 30 and 32 should be
considered to be unpredictable.
(2) In the structure of the CDR2, Chothia & Lesk
(1987) had postulated that the variant conformation observed in McPC603 was linked to the presence of a glycine residue (Gly50). Our data on the
VcD structure indicate, however, that the CDR2
conformation is highly conserved among all V L
domains, thus unifying the observed canonical
structures, and a glycine residue at position 50 does
not generate an exception.
(3) The canonical structure of this CDR3 has been
postulated to _be determined by the conserved cisproline Pro95 and two hydrogen bonds from the
carboxamide group of Asn90 to the back of the loop
(Tramontano et al. , 1989). We observe a different
topology of hydrogep bonds (which are, in addition ,
all longer than 3·0 A) from Asn90 in this structure.
Thus, we would suggest that t he conformational
importance of the hydrogen-bond stabilization of
this solvent-exposed loop is not yet completely
clear. Furthermore, the significant structural differences between the VL- D and the VcFab CDR3
indicate that the basic premise of the canonicalstructure hypothesis has to be viewed with some
caution. A strong influence of VH (or the second V d
on the conformation of a CDR (possibly mediated in
this case through electrostatic effects involving
Asp91) would mean that this conformation would
depend both on the precise mode of association
between the two domains in the homo- or heterodimer . This would be especially problematic if a
CDR conformation would depend on the precise
conformation of the VH CDR3. Both cannot be ruled
out and both may be impossible to predict with
sufficient accuracy (Stevens et al. , 1988; Colman et
al., 1987) .
In conclusion, the potential of the canonicalstructure hypothesis to predict antigen-binding
loops from sequence data alone with sufficient accuracy to permit modelling of binding interactions,
must still be viewed with some caution, and it
becomes clear that much furth er structural work
will be necessary. -An attempt to rationalize the
structural differences between the M3 and 4-4-20
CDR1 illustrates the difficulty of drawing conclusions from the observation of mere sequence-structure correlations, as too little is known about true
causal relationships. The conformation of the CDR2
and its lack of correlation with a glycine residue at
position 50 illustrates the limitation of such knowledge-based conformational predictions that rest in
the accuracy and reiiability of the underlying
experimental structures. Finally, the difference in
CDR3 conformation between the V cD and the
V L-Fab illustrates the difficulties for conformational predictions in cases where t he local structure
may be significantly influenced by long-range interactions or quarternary structure.
The finding that a CDR mutation such as in M3
can have negligible structural consequences for the
rest of the domain is enco uraging for further engineering: experiments to transplant and combine
loops of known structure to generate new binding
properties and efforts to solve th e structure of new
752
B. Steipe et al.
sequences may one day lead to an improved understanding of sequence-structure relationships in
immunoglobulin domains .
References
Alzari , P. M. , Lascombe, M. B. & Poljak, R. J. (1988) .
3-dimensional structure of antibodies. Annu. R ev.
Immunol. 6 , 555-580.
Bernstein, F . C., Koetzle, T. F. , Williams, G. J . B. , Meyer,
E. F. , Brice, M. D., Rodgers, J. R. , Kennard , 0. ,
Shimanouchi , T. & Tas umi , M. (1977). The protein
databank: a com puter based archival file for macromolecular structure. J. Mol. Biol. 112, 535-542.
Bruccoleri , R . E. , Haber, E. & Novotny, J. (1988). Structure of antibody hypervariable loops reproduced by a
conformational search algorithm . Nature (London) ,
335, 564-568.
Chothia, C. & Lesk, A. M. (1986) . The relation between
the divergence of sequence and structure in proteins.
EM BO J. 5, 823- 826.
Chothia, C. & Lesk , A.M. (1987) . Canonical structures for
the hypervariable regions of immunoglobulins. J.
Mol. Biol. 196, 901- 917.
Chothia, C., Novotny , J. , Bruccoleri , R. & Karplus, M.
(1985) . Domain association in immunoglobulin molecules. J . Mol. Biol. 186, 651-633.
Chothia, C. , Lesk , A. M., Tramontano, A. , Levitt, M.,
Smith-Gill , S. J. , Air, G., Sheriff, S. , Padlan , E.,
Davies, D. R. , Tulip, W. R ., Colm an , P ., Spinelli , S. ,
Alzari, P . M. & Poljak , R. J . (1989) . Conformations of
immunoglobulin hy pervariable regions. Nature
(London) , 342, 877- 883.
Colman, P. M., Schramm , H . J . & Guss, J. M. (1977) .
Crystal and molecular structure of the dimer of vai·iable domains of the Bence-Jones Protein ROY. J .
Mol. Biol. 116, 73- 79.
Colman, P. M. , Laver, W. G., Varghese, J . N. , Baker,
A. T ., Tulloch , P . A. , Air, G. M. & Webster, R. G.
(1987) . Three-dimensional structure of a complex of
antibody with influenza-virus neuraminidase. Nature
(London) , 326, 358- 363.
Davies, D . R. , Padlan , E . A. & Sheriff, S. (1990). Antibody-antigen complexes. Annu. Rev. Biochem. 59,
439-473.
de Ia Paz, P. , Sutton , B. J ., Darsley, M. J . & Rees, A. R .
(1986) . Modelling the combining sites of three antilysozy me monoclonal a ntibodies and of the complex
between one of the antibodies and its epitope. EJJfBO
J. 5, 415-425.
Ely, K . R. , H erron , J. N. , Harker, M. & Edmundson,
A. B. (1989) . Three-dimensional structure of a light
chain dimer crystallized in water. J . Mol. Biol. 210,
601-615.
Epp, 0., Lattman , E. E., Schiffer, M., Huber, R. & Palm ,
W. (1975) . The molecular structure of a dimer
composed of the variable portions of the Bence-Jones
protein REI refined at 2·0 A resolution. Biochemistry,
14, 4943-4952.
Fehlhammer, H. , Schiffer, M. , Epp. , 0., Colman , P. M. ,
Lattman , E. E. , Schwager, P . & Steigemann , W.
(1975). The structure determination of the variable
portion of the Bence-Jones protein AU . Biophys.
Struct. Mechanism , 1, 139- 146 .
Fine, R. M. , Wang, H ., Shenkin , P. S., Yarmush , D. L . &
Levinthal , C. (1986). Predicting antibody hypervariable loop conformations. II. Minimization and
molecular dynamics stud ies of McPC603 f1·om many
randomly generated loop conformations. Proteins, 1,
342-462.
Furey, W. , Wang, B. C., Yoo, C. S. & Sax , M. (1983) .
Structure of a novel Bence-Jones protein (RHE)
fragment at 1·6 A resolution. J. Mol. Biol. 167,
661 - 692.
Glockshuber, R ., Steipe, B., Huber, R . & Pliickthun , A.
(1990). Crystallization and preliminary X-ray studies
of the V L domain of the antibody McPC603 produced
in E. coli. J. Mol. Biol. 213, 613-615.
Goodsell , D. S. & Olson , A. J . (1990) . Automated docking
of substrates to proteins by simulated annealing.
Proteins, 8 , 195-202.
Herron, J. N., He, X .-H. , Mason, M. L. , Voss, E. W . &
Edmundson , A. B. (1989). Three-dimensional structure of a fluorescein- Fab complex crystallized in 2methyl-2,4-pentanediol. Proteins, 5, 271-280.
Holm , L ., Laaksonen , L ., Kaartinen, M., Teeri , T. T. &
Knowles, J . C. (1990). Molecular modelling study of
antigen binding to oxazolone-specific antibodies: the
Ox 1 idiotypic IgG and its mature variant with
increased affinity to 2-phenyloxazolone. Protein Eng.
3, 403-409.
Huber, R. & Kopfmann , G. (1969) . A method of absorption correction by X-ray-intensity measurements.
Acta Crystallogr. sect. A , 25, 143- 152.
Ippolito , J. A. , Alexander, R. S. & Christianson , D . W .
(1990). Hydrogen bond stereochemistry in protein
structure and function. J. Mol. Biol. 215, 457-471.
Jack, T . & Levitt, M. (1978). Refinement of structures by
simultaneous minimization of energy and R-factor.
Acta Crystallogr . sect. A , 34, 931- 935.
Jones, T . A. (1978). A graphics model building and refinement system for macromolecules. J. Appl. Crystallogr. 11 , 268-272.
Kabat, E. A., Wu, T. T ., Reid-Miller, M. , Perry, H . M. &
Gottesmann , K. (1987). Sequences of proteins of
immunological interest, Public Health Service, Natl.
Inst. of Health , Bethesda, MD, U .S.A.
Luzatti , V. (1952). Traitement statistique des erreurs
dans Ia determination des structures cristallines.
Acta. Crystallogr. Sect. A , 5, 802-810.
Mainhart, C. R. , Potter, M. & Feldman , R. J . (1984) . A
refined model fo1· the variable domains (F v) of the
J539 p(l , 6)-D-galactan-binding immunoglobulin .
Mol . Immunol. 21 , 469-478.
Marquart, M. & Deisenhofer, J. (1982). The three-dimensional structure of antibodies. hnmunol. Today , 3,
160-166.
Marquart, M., Deisenhofer , J. , Huber, R . & Palm , W.
( 1980). Crystallographic refinement and atomic
models of the intact immunoglobulin molecule KOL
and its antigen-binding fragment at 3·0 A and 1·9 A
resolution . J. Mol. Biol. 141 , 369-39 1.
Martin , A. C. R. , Cheetham , J . C. & Rees, A. R. (1989).
Modelling antibody hypervariable loops: a combined
algorithm. Proc. Nat. Acad. Sci., U.S.A. 86,
9268- 9272.
Matthews, B. W . (1968). Solvent content of protein
crystals. J. Mol. Biol. 244, 491-497.
Messerschmitt, A. & Pflugrath , J. W. (1987). Crystal
orientation and X -ray pattern prediction routines for
area-detector diffractometer systems in macromolecular crystallography. J . Appl. Crystallogr. 20,
306-315.
Messerschmidt, A., Schneide1·, M. & Huber, R . (1990) .
ABSCOR: a scaling and absorption correction
program for the FAST area detector diffractometer.
J. Appl. Cryst.allogr. 23 , 436-439 .
High-resolution Structure of Recombinant V L
Milner-White, J. E., Ross, B . M. , Isma il , R. , Kha led ,
B.-M. & Poet, R. (1988) . One type of gamma-turn ,
rather than the other gives rise to cha in-reversal in
proteins. J. Mol. Biol. 204, 777- 782 .
Padlan, E. A. , Silverton , E. W., Sheriff, S. , Cohen, G. H. ,
Smith-Gill , S. J . & Davies, D. R. (1989) . Stru cture of
an a nt ibody- antigen complex: crystal stru cture of
the Hy HEL-10 F a b- lysozy me complex. Proc . Nat.
Acad. Sci ., U.S.A. 86 , 5938- 5942.
Perlmutter, R . M. , Crews, S. T ., Douglas, R. , Sorensen ,
G. , Johnson, N. , Nivera, N. , Gearhart, P. J . & Hood ,
L . (1984) . The generation of diversity in phosphory lcholine- binding antibodies. Advan. Irnrnu-rwl. 35 ,
1- 37 .
Pliickthun , A. (1991) . Ant ibody engineering: advances
from t he use of Escherichia coli expression systems.
Bio /tech-rwlogy, 9 , 545- 551.
Potter, M. ( 1977) . Ant igen-binding myeloma proteins of
mice. Advan. hmnU?wl. 25 , 14 1- 211.
Ramachandran, G. N. , Venkatachalam , C. M. & Krimm,
S. (1966). Stereochemical criteria for polypeptide and
protein conformations. Biophys. J . 6 , 849-872.
Richardson , J . S. (1981) . The a natomy and taxonomy of
protei n structure. Advan. Protein Chern. 34, 167- 339.
Riechmann , L. , Clark , M. , ' Valdmann , H. & Winter, G.
(1988) . Reshaping huma n antibodies for therapy.
Nature (London) , 332, 323- 327.
Rose, G. D ., Gierasch, L. M. & Smith , J. A. (1985). Turns
in peptides and proteins. Advan. Protein Chern. 37,
1- 109.
Satow, Y ., Cohen , G. H ., Padlan , E. A . & Da.vies, D. R.
(1986) . Phosphocholine binding immunoglobulin Fab
McPC603. J . Mol. Biol. 190, 593--604.
Segal, D. M., Padlan , E. A. , Cohen, G. H. , Rudikoff, S. ,
Potter, M. & Davies, D. R . (1974) . The t hree-dimensiona l structure of a phosphorylcoline-binding mouse
immunoglobulin Fab and the nature of t he antigen
binding site. Proc. Nat. Acad. Sci ., U.S.A. 71 ,
4298-4302.
Shenkin , P. S. , Yarmush, D. L. , Fine, R. M., Wang, H. &
Levinthal , C. (1987) . Predicting anti body hypervariable loop conformations. I. Ensembles of random
conforma.tions for ringlike stru ctures. Biopolyrners,
26, 2053- 2085.
Sheriff, S. , Silverton , E. W. , Padlan , E. A. , Cohen , G. H. ,
Smi t h-Gill , S. J ., Finzel , B. C. & Davies, D. R. (1987) .
Three-dimensional stru cture of a n antibody/antigen
complex . P roc. Nat. Acad. Sci. , U .S.A. 84,
8075- 8079.
753
Singh , J ., Thornton , J . M., Snarey , M. & Campbell , S. F.
(1987) . The geo metries of interacting argininecarboxy ls in proteins. FEES Letters, 224, 161 - 171.
Skerra, A. & Pliickthun , A. (1988) . Assembly of a functional Fv fragm ent in Escherichia coli. Science , 240 ,
1038- 1041.
Snow, M. E. & Amzel, L . M. (1986) . Calculating t hreedimensional changes in protein structure due to
amino acid substitutions: the variable region of
immunoglobulins. Proteins, 1, 267- 279.
Stadlmiiller, J. (1991) . Das Fv-Fragment des Phosphorylcholin-bindenden Antikorpers McPC603: Mutanten ,
Bindungseigenschaften und ka talytische Aktivitat.
Ph .D
thesis,
Ludwig-Maximilians-University ,
Munich , F.R.G.
Stanford , J . M. & Wu , T. T. (1981) . A predictive method
for determining possible three-dimensional foldings of
immunoglobulin
backbones
around
antibody
combining sites. J . Theoret. Biol. 88 , 421-439.
Steigemann, W . (1974) . Die Entwicklung und Anwendung
von Rechenverfahren und Rechenprogrammen zur
Strukturanalyse von Proteinen a m Beispiel des
Trypsin- Trypsininhibitor Komplexes, des freien
Inhibitors und der L-Asparaginase. Ph.D thesis,
Technical University , Munich, F.R.G .
Stevens, F . J. , Chang, C.-H. & Schiffer, M. (1988). Dual
confor mations of an immunoglobulin light-chain
dimer: heterogeneity of antigen specificity and idiotype profile may result from multiple variabledomain interaction mechanisms. lrnrnu-rwlogy, 85 ,
6895--6899.
Suh , S. W. , Bhat, T. , N. , Navia, M. A. , Cohen, G. H ., Rao,
D. N. , Rudikoff, S. & Davies, D. R. (1986) . The
galactan binding immunogl<?bulin Fab J539: an Xray diffraction study at 2·6 A resolution . Proteins , 1,
74-80.
Tramontano , A. , Chothia , C. & Lesk, A.M. (1989) . Stt·uct ural determinants of the conformations of mediumsized loops in proteins. Proteins, 6 , 382- 394.
Venkatacha la m, C. M. (1968). Stereochemical criteria for
polypeptides and proteins.V. Conformation of a
system of t hree linked peptide units. Biopolynwrs, 6 ,
1425- 1436.
Weiner, S. J ., Sing h, U. C. , O' Donnell, T . J. & Kollman ,
P. A. (1984). Quantum and molecular mecha ni cal
studies on a lany l dipept ide. J. Anwr. Clwrn. Soc. 106,
6243--6245.
Zehfus, M. H. & Rose, G. D. (1986) . Compact units in
proteins. Bioclwmisll-y , 25 , 5759- 5765.
Edited by A. Klug

Similar documents

×

Report this document