Molecular Immunology,Vol. 28, No. 12, pp. 1341-1345, 1991 0161-5890/91 $3.00 + 0.

00
Printed in Great Britain. Pergamon Press plc

STRUCTURE OF ANTIBODIES AND THEIR COMPLEXES

WITH ANTIGENS
ROBERTOJ. POLJAK
U.R.A. 359 CNRS, Departement d’Immunologie, Institut Pasteur,
75724 Paris Cedex 15, France

(Received 1 December 1990)

INTRODUCTION et al., 1968) provided Fab and Fab’ crystals of good

This is a narrative account and a personal recollec- quality for X-ray analysis. The determination of its
tion of some of the studies on the three-dimensional three-dimensional structure to 6-A resolution (Poljak
structure of antibodies conducted in my laboratory. et al., 1972) was a preliminary step for a study at
It is not intended to be a critical or detailed account higher resolution. For this study it was necessary to
of developments in this field. determine the amino acid sequence that had to be
In the 1950s and subsequently, research on the fitted to the electron density map to build a three-
molecular basis of immune responses and their dimensional model. With my colleague B. L. Chen
diversity attracted the attention of many molecular we chose to start on the sequence of the L chain. This
biologists, immunologists and protein chemists. The choice proved fortunate because the L chain Newm
observations that stimulated my first attempts in contains what appears to be an unusual deletion
this fascinating field were: (1) the classical demon- of seven amino acids, an unexpected feature which
stration by Rodney Porter (1959) that polyclonal delayed the interpretation of the 2.8-A resolution
rabbit antibodies could be cleaved into Fab and Fc electron density map. With the combined approach
fragments of which the Fc could be readily crystal- of X-ray crystallography and amino acid sequencing
lized; (2) the demonstration (Hill et al., 1966) of we could be sure that there really was an unusual
internal homologies in immunoglobulin sequences deletion. Thus, the 2.8-A resolution electron density
indicating gene duplication events and suggesting the map (Poljak et al., 1973) and a subsequent map
presence of discrete three-dimensional domains in calculated with part of the X-ray intensity data
antibody molecules. It is perhaps unnecessary to state between 2.8 and 2.0-A (Poljak et al., 1974), inter-
here that in the 1960s protein crystallography had preted with the aid of the amino acid sequence
not reached the high level of automation that we enabled us to build models at atomic resolution. The
know today. In addition, monoclonal myeloma pro- observations that were made on these models have
teins were not universally accepted as good models been reviewed (Amzel and Poljak, 1979). The deter-
of antibodies. mination of this three-dimensional structure allowed
The first analysis of immunoglobulins and their us to align the sequence of the V and C domains by
fragments by X-ray diffraction in my laboratory were matching residues that occupy similar positions in
done on the crystalline Fc fragments from rabbit the tertiary structure of the different domains. These
antibodies and from human myeloma proteins. When alignments clearly showed the great amino acid
the Fab and Fab’ from human immunoglobulins sequence homology between the V and the C domains,
were first crystallized by Rossi and Nisonoff (1968) supporting the proposal of gene duplication (Hill
and subsequently analysed by us by X-ray diffraction et al., 1966) as the mechanism that gave rise to
(Avey et al., 1968; Humphrey et al., 1969) it became immunoglobulin genes. The immunoglobulin-fold
clear that they were highly ordered and better suited was later found to occur in a large number of proteins
for X-ray analysis than the Fc crystals. Since then, grouped under the general name of the immuno-
our efforts have been almost exclusively directed to globulin superfamily, including proteins which have
the study of Fabs, since these fragments retain the no evident function in immune responses (reviewed in
antigen-binding properties of the antibody molecules Williams and Barclay, 1988).
from which they are derived. The electron density of the L and the heavy (H, Fd)
chains of FabNew was much lower near the inter-
THREE-DIMENSIONAL STRUCTURE OF FABS
chain disulfide bond than in the rest of the C or V
domains. This result was consistent with the widely
The human immunoglobulin IgGl(1) New, some- accepted notion of flexibility in the hinge region
times called Newm to avoid confusion with the of immunoglobulins. About ten years later Drs C.
human myeloma (a) light (L) chain New, (Langer and E. Mihaesco provided us with samples of the

1341
1342 ROBERTOJ.POLJAK

pathologic human H chain disease protein Riv which ANTIGEN-ANTIBODY COMPLEXES

is equivalent to an Fc plus the hinge region of IgGl.
The X-ray diffraction analysis of the Riv crystals Hoping to characterize specific antigen-antibody
showed that they are isomorphous with Fc and interactions we attempted in 1973, in a collaboration
consequently, that the hinge region does not assume with the laboratory of Dr Ed Haber, then at the
a unique conformation in the crystals (Mariuzza Massachusetts General Hospital, to crystallize anti-
et al., 1983). pneumoccocal rabbit antibodies of restricted hetero-
Perhaps the most interesting observation that was geneity. This resulted in the crystallization of the Fab
made in the model of the three-dimensional structure from the antibody designed Ab3322 for which the
of FabNew was that the segments of polypeptide unit cell dimensions and space group symmetry were
chain corresponding to the regions of hypervariable fairly close to those of FabNew. The crystals of
sequences (defined by Wu and Kabat, 1970) occur in Fab3322 had a unit cell with a = 104.3 A, b = 67.0 A,
close spatial proximity, fully exposed to solvent at c = 73.1 A, /I = 128.5”, space group monoclinic C2
one end of the molecule. At the center of this site, (unpublished work), while those of FabNew have
between the H and the L chains there is a cavity a = 114.4A, b =56.7& c =90.3& /I= 116.5”, and
of about 15 x 6 A, with a depth of 6 A. This was the same space group, C2. Wanting to continue this
inferred to constitute the antigen binding site. The study I telephoned Ed Haber with great excitement.
subsequent study of a specific crystalline complex He told me, first, that the preparation he had sent
between a derivative of vitamin K, and FabNew me contained three or four different antibodies.
(Amzel et al., 1974) showed that the binding site was Then, he informed me that the donor rabbit was dead
indeed the cavity at the center of the hypervariable and consequently, no more of that material could be
regions (or complementarity determining regions, obtained. Since there were one or two large crystals
CDRs). in the preparation, present-day techniques would
Recently, the possibility of cloning and expressing have allowed the measurement of X-ray intensities
immunoglobulin genes and the protein engineering and the determination of the structure, but this was
of chimeric human-mouse antibodies has revived not then technically feasible. I believe this anecdote
interest in the detailed three-dimensional structure of illustrates the state of the field at the time and the
human myeloma proteins. This has prompted my subsequent impact that the introduction of the
colleague Fred Saul to remeasure X-ray diffraction hybridoma technique by Kohler and Milstein (1975)
data for FabNew using area detectors to increase the had in this as well as in many other fields. It also
resolution and the accuracy of the structure. illustrates the enormous improvement in X-ray
Crystallographic studies in other laboratories diffraction techniques in the last 20 years.
confirmed and extended the conclusions summarized Having moved to the Pasteur Institute in 1981, my
above. Thus, Allen Edmundson and colleagues laboratory started a program to obtain monoclonal
(Schiffer et al., 1973) independently determined the antibodies with the aim of crystallizing antigen-
immunoglobulin-fold from a study of the (1) L-chain antibody complexes suitable for X-ray diffraction
dimer Meg at 3.5 8, resolution. Epp et al. (1974) work (Harper et al., 1987). Since hen egg-white
showed that the same type of folding was present lysozyme (HEL) had been used as a model antigen in
in a human V, homodimer. David Davies, Michael many laboratories and since I had previously worked
Potter and colleagues (Segal et al., 1974) determined on the determination of its three-dimensional struc-
the three-dimensional structure of the Fab from the ture, it seemed to be an ideal antigen model for X-ray
mouse myeloma IgA McPC603 as well as that of diffraction studies.
the complex between the Fab and the specific ligand From the point of view of the three-dimensional
phosphorylcholine. Since the mouse myeloma system structure of antibodies, several important questions
developed in the laboratory of Michael Potter was remained to be answered. Some of these can be
an important model system for the study of anti- summarized as follows: (1) What is the complete
body structure and function, the structure of structure of the antigen combining site; do all CDRs
FabMcPC603 has had a great impact in the field. contribute to the binding of an antigen? (2) Are there
No conformational changes could be seen in the Fabs conformational changes in the combining site and
complexed to the ligands vitamin K, and phosphoryl- all along the antibody molecule when a whole
choline. This observation, together with immuno- antigen has been bound? (3) What are the chemical
chemical evidence suggested that there was no specific bonds between the antigen and the antibody?
mechanism of signal transmission from the combin- (4) What is the structure of an epitope of a pro-
ing site to the Fc region of an antibody for triggering tein antigen, is it made up of contiguous residues
secondary functions such as complement activation or is it “topographical”, made up from discon-
(reviewed in Davies and Metzger, 1983). A lingering tinuous segments of the sequence of the antigen?
doubt, however, was that if a specific antigen rather (5) How can we explain at the molecular level that a
than a small ligand such as a hapten was bound to single amino acid variation in an antigen can some-
the combining site, then, would a conformational times prevent its recognition by a specific monoclonal
change be observed? antibody?
Structure of antibodies 1343

The determination of the three-dimensional struc- The results described above, together with those
ture of the complex between the Fab from the obtained in the laboratories of D. R. Davies and
anti-HEL MAb D1.3 (of BALB/c origin) and HEL P. M. Colman (summarized in Davies et al., 1989,
(Amit et al., 1985, 1986) provided answers to the and in Tulip et al., 1989, respectively) form a coherent
above questions. The most relevant observations that ensemble. The antigenic determinants of the model
were made are summarized below. antigen lysozyme are distributed all over its surface,
In the FabD 1.3-lysozyme complex the Fab moiety contrary to the notion that there are unique areas
appears in an elongated conformation. This goes which are recognized by antibodies. The results show
against the postulate that binding of an antigen would that in terms of their specificity, complementarity,
provoke a contraction within the Fab to bring the V, area of interaction, hydrogen-bonding patterns, etc.,
domain closer to the Cnl and thus transmit a confor- the antigen-antibody complexes resemble those
mational signal along the molecule towards the Fc between protease enzymes and their inhibitors. As
region. In the complex, the antigen has not under- in the case of the antigens, in the enzyme-inhibitor
gone any noticeable change of conformation com- complexes the mobility of the residues at the con-
pared to its unbound state. In addition, comparison tacting surfaces is average or just below average.
of the Fab with other Fabs whose three-dimensional A possible difference is given by the fact that in
structure has been determined suggested that there antibody-combining sites there is a concentration of
are no major conformational changes within the Fab aromatic residues (His, Phe, Tyr and Trp) which
itself. The contacting surfaces between antigen and favors van der Waals interactions and hydrogen
antibody are very complementary, they have an bonding while decreasing the loss of entropy in going
area of about 600-700 AZ, with about 14 amino acid from the free to the antigen-bound state. But, as we
residues from the antibody and a similar number know, antibodies seem to provide “impromptu” sol-
from the antigen-making atomic contacts. Two or utions to the specificity problem posed by the large
three water molecules are buried at the interface. The number of variable antigens that an immune system
antigenic determinant of HEL is clearly discontinu- may have to face. Thus, it is amazing that their
ous or topographical, and it can be expected that a specificity and affinity are so high. Here there is a
typical epitope from a protein antigen will be discon- lesson in chemistry to be analysed from which we
tinuous in nature. The amino acid residue Gln121 of may learn much about chemical complementarity
HEL makes closer contacts with the antibody. This and specificity in the biological world.
position is occupied by a His residue in most of the
avian lysozymes that are not recognized by MAb AN IDIOTOPEkANTI-IDIOTOPE COMPLEX
D1.3. Steric hindrance prevents the entering of a His
residue into the antibody cavity penetrated by Gln121 The idiotypic antigenic determinants of antibodies
and explains the fine specificity of the antibody. have been the object of intensive studies aiming at
Through a collaborative arrangement the mono- a definition of their molecular structure and their
clonal antibody D1.3 was the object of structural postulated role in the regulation of the immune
studies at the Laboratory of Molecular Biology, system. Anti-idiotypic antibodies may mimic external
MRC, Cambridge, U.K. There, in the laboratory of antigens at the molecular level. The use of this
Greg Winter its cDNA was cloned and sequenced, potential property has been suggested for practical
integrated into a bacterial plasmic and expressed in applications, including vaccine development (Nisonoff
the bacterium Escherichia coli as an Fv (Vi, + V,) and Lamoyi, 1981; Roitt et al., 1981; Sacks et al.,
fragment. We drew an important benefit from this 1983). Idiotypes had been first defined and conceptu-
work in Cambridge since we were able to produce ally analysed by the late Jacques Oudin at the Pasteur
and crystallize the free as well as the antigen-bound Institute. A. Nisonoff did the first work in character-
Fv (Boulot et al., 1990). The determination of their izing idiotypes in specific antibody immune responses,
three-dimensional structures (Bhat et al., 1990) pro- showing that specific hapten binding could inhibit
vided an answer to the question concerning confor- idiotype recognition. My laboratory became inter-
mational changes that might occur at the combining ested in defining idiotopes at the molecular level,
site after antigen binding. Small changes were detected which led us to the crystallization (Boulot et al., 1987)
and shown to consist of a relative displacement of the and the determination of the three-dimensional struc-
Vn and V, domains which otherwise keep the tertiary ture (Bentley et al., 1990) of an idiotope-anti-idiotope
structure of the free Fv. Thus, the binding of the complex between the anti-lysozyme FabD1.3 and the
antibody is accompanied by a small molecular move- anti-idiotope, FabE225. The private idiotope thus
ment, in agreement with an induced fit mechanism, defined consists of some 13 amino acid residues,
although it could also be said that it is nearly mostly from the CDRs of D1.3 but including three
compatible with a lock and key fit since the antibody residues from the third framework region of its V,
variable domains undergo a minimal movement and domain. This idiotope shares about one-half of its
their amino acid side chains do not change much in amino acid residues with the anti-HEL paratope of
conformation in going from the free to the antigen- D1.3. Antibody E225 mimics the external antigen
bound state. in a very general way, i.e. by its binding to the
1344 ROBERTO J. POLJAK

combining site of D1.3, occupying the same place Amzel L. M., Poljak R. J., Saul F., Varga J. M. and
Richards F. F. (1974) The three-dimensional structure of
that is taken by the antigen in the antigen-D1.3
a combining region ligand complex of immunoglobulin
complex. It is perhaps not surprising that the anti- NEW at 3.5 8, resolution. Proc. natn. Acad. Sci. U.S.A.
idiotopic antibody E225 does not mimic the epitope 71, 1427-1430.
of the external antigen recognized by D1.3, first Avey H. P., Poljak R. J., Rossi G. and Nisonoff A. (1968)
because it may not represent the best choice of Crystallographic data for the Fab fragment of a human
myeloma immunoglobulin. Nature 220, 1248-1249.
“internal image” anti-idiotope and second, because
Bentley G. A., Boulot G., Riottot M. M. and Poljak
it may not be possible to mimic an epitope which R. J. (1990) Three-dimensional structure of an idiotooe-
is partly helical with the open loop structure of anti-idiotope complex. Nature 347, 1050-1053. A
the CDRs of antibodies. Further exploration of the Bhat T. N., Bentley G. A., Fischmann T. O., Boulot G. and
mimicking of external antigens is needed to ascertain Poliak R. J. (1990). , Small rearrangements in structures
of Fv and Fab fragments of antibody Dl.3 on antigen
the structural bases of this phenomenon. binding. Nature 347, 483-485. _
Boulot G.. Roias C.. Bentlev G. A.. Poliak R. J.. Barbier E..
PERSPECTIVES Le G&n C. and Cazehave PI A.* (1987) ‘Preliminary
crystallographic study of a complex between the Fab
One hundred years have elapsed since the discovery fragment of a monoclonal anti-lysozyme antibody (D 1.3)
of antibodies, however, we do not yet know every- and the Fab fragment from an anti-idiotopic antibody
against Dl.3. J. molec. Biol. 194, 577-579.
thing about them. Dramatic advances have recently
Boulot G., Eisele J. L., Bentley G. A., Bhat T. N., Ward
been made in the protein engineering of antibodies, E. S., Winter G. and Poljak R. J. (1990) Crystallization
for example in obtaining murine-human chimeric and preliminary X-ray diffraction study of the bacterially
antibody molecules. These antibodies were con- expressed Fv from the monoclonal anti-lysozyme anti-
structed with CDRs of murine origin, grafted onto body Dl.3 and of its complex with the antigen, lysozyme.
J. molec. Biol. 213, 617-619.
human frameworks. They retain the antigen-binding
Davies D. R. and Metzger H. (1983) Structural basis of
specificity of the parent murine antibodies, and have antibody function. A. Rev. Immun. 1. 87-117.
proven useful in human therapy (for a review see Davies D.-R., Sheriff S. and Padlan E. (1989) Comparative
Waldman, 1989). study of two Fab-lysozyme crystal structures. Co/h Spring
Concerning structure and function of antibodies Harbor Symp. &ant. Biol. 54, 233-238.
Epp O., Colman P M., Fehlhammer H., Bode W., Schiffer
there are many important problems that await sol- M., Huber R. and Palm W. (1974) Crystal and molecular
ution. For example, the putative role of confor- structure of a dimer composed of the variable portions
mational signals in the activation of secondary of the Bence-Jones protein REI. Eur. J. Biochem. 45,
functions, such as activation of complement, is still 5 13-524.
Harper M., Lema F., Boulot G. and Poljak R. J. (1987)
a matter of debate. The structural correlations of
Antigen specificity and cross-reactivity of monoclonal
the secondary functions of antibodies have not pro- anti-lysozyme antibodies. Molec. Immun. 24, 977108.
gressed as much as the definition of the molecular Hill R. L., Delaney R., Fellow R. E. Jr, and Lebowitz H. E.
basis of their specificity. The study of secondary (1966) The evolutionary origins of the immunoglobulins.
functions will be a more difficult area for detailed Proc. natn. Acad. Sci. U.S.A. 56, 1762-1769.
Humphrey R. L., Avey H. P., Becka L. N., Poljak R. J.,
structural analyses such as those made by X-ray Rossi G., Choi T. K. and Nisonoff A. (1969) X-ray
diffraction. The role of antibodies as cellular antigen- crystallographic study of the Fab fragments from human
receptors in conjunction with other B-cell com- myeloma proteins. J. molec. Biol. 43, 223-226.
ponents may eventually become a topic for that type Kohler G. and Milstein C. (1975) Continuous cultures of
fused cells secreting antibody of predefined specificity.
of studies. At this time it is neither necessary nor
Nature 256, 495-498.
perhaps desirable to search further for interesting Langer B., Steinmetz-Kayne M. and Hilschmann N. (1968)
topics for future research. Die vollstlndige aminosauresequenz des Bence-Jones-
proteins New (lambda-typ) subgruppen im variablen teil
Acknowledgements-I would like to thank all the former bei immunoglobulin-L-ketten vom lambda-typ. Hoppe-
and present colleagues that took part in the work performed Seyler Z. Phys. Chem. 349, 945-951.
in my laboratory. I also would like to thank the granting Mariuzza R. A., Poljak R. J., Mihaesco C. and Mihaesco
agencies that supported our research, in particular th: E. (1983) Crystals of the human heavy chain disease
USPHS, the Pasteur Institute. the CNRS. the EEC and the protein Riv and human Fc fragment are isomorphous:
Louis Jeantet Foundation. further evidence for conformational flexibility in the
hinge regions of immunoglobulins. J. molec. Biol. 165,
REFERENCES 559-56 1.
Nisonoff A. and Lamoyi E. (1981) Implications of the
Amit A. G., Mariuzza R. A., Phillips S. E. V. and Poljak presence of an internal image of the antigen in anti-
R. J. (1985) Three-dimensional structure of an antigen- idiotypic antibodies: possible application to vaccine
antibody complex at 6.0 A resolution. Nature (Lond.) 313, production. Clin. Immun. Immunopath. 21, 397-406.
156-158. Poljak R. J., Amzel L. M., Avey H. P., Becka L. N. and
Amit A. G., Mariuzza R. A., Phillips S. E. V. and Nisonoff A. (1972) The structure of Fab’ New at 6A
Poljak R. J. (1986) Three-dimensional structure of an resolution. N&ure ‘New Biol. 235, 137-140.
antigen-antibody complex at 2.8 8, resolution. Science Poljak R. J., Amzel L. M., Avey H. P., Chen B. L.,
233, 141-753. Phizackerley R. P. and Saul F. A. (1973) Three-
Amzel L. M. and Poljak R. J. (1979) Three-dimensional dimensional structure of the Fab’ fragment of a human
structure of immunoglobulins. A. Rev. Biochem. 48, immunoglobulin at 2.8 8, resolution. Proc. natn. Acad.
961-997. Sci. U.S.A. 70, 3305-3310.
Structure of antibodies 1345

Poljak R. J., Amzel R. J., Chen B. L., Phizackerley R. J. (1973) Structure of a I-type Ben-Jones protein at 3.5 8,
and Saul F. A. (1974) The three-dimensional structure of resolution. Biochemistry 12, 4620-463 1.
the Fab’ fragment of a human myeloma immunoglobulin Segal D. M., Padlan E. A., Cohen G. H., Rudikoff S.,
at 2.0A resolution. Proc. natn. Acad. Sci. U.S.A. 71, Potter M. and Davies D. R. (1974) The three-dimensional
3440-3444. structure of a phosphorylcholine-binding mouse immuno-
Porter R. R. (1959) The hydrolysis of rabbit y-globulin globulin Fab and the nature of the antigen binding site.
and antibodies with crystalline papain. Biochem. J. 73, Proc. natn. Acad. Sci. U.S.A. 71, 4298-4302.
119-126. Tulip W. R., Varghese J. N., Webster R. G., Air G. M.,
Roitt I. M., Cooke A., Male D. K., Hay F. C., Guarnotta Laver W. G. and Colman P. M. (1989) Crystal structures
G., Lydyard P. M., de Carvalho L. P., Thanavala Y. and of neuraminidase-antibody complexes. Cold Spring
Ivanyi J. (1981) Idiotypic networks and their possible Harbor Symp. Quant. Biol. 54, 257-263.
exploitation for manipulation of the immune response. Waldman H. (1989) Maninulation of T-cell resnonses with
Luncet 1, 104-1045. monoclonal‘antibodies. A. Rev. Immun. 7, 407-444.
Rossi G. and Nisonoff A. (1968) Crystallization of fragment Williams A. F. and Barclay A. N. (1988) The immuno-
Fab of human IgG myeloma proteins. Biochem. biophys. globulin superfamily-domains for cell surface recognition.
Res. Commun. 31, 914-917. A. Rev. Immun. 6, 381-405.
Sacks D. L., Kelsoe G. H. and Sachs D. H. (1983) Induction Wu T. T. and Kabat E. A. (1970) An analysis of the
of immune responses with anti-idiotypic antibodies: sequences of the variable regions of Ben-Jones pro-
implications for the induction of protective immunity. teins and myeloma light chains and their implications
Springer Semin. Immunopath. 6, 79-97. for antibody complementarity. J. exp. Med. 132,
Schiffer M., Girling R. L., Ely K. R. and Edmundson A. B. 21 l-250.