Studies on the three-dimensional structure of immunoglobulins

Prog. Biophys.Molec.Biol., 1976,Vol. 31, pp. 67-93. PergamonPress. Printed in Great Britain.

S T U D I E S O N THE T H R E E - D I M E N S I O N A L STRUCTURE OF IMMUNOGLOBULINS R. J. POLJAK,L, M. AMZEL and R. P. PHIZACKERLEY Department of Biophysics, Johns Hopkins University School of Medicine, Baltimore, Maryland, U.S.A.

CONTENTS I. INTRODUCTION II. POLYPEPTIDE CHAIN STRUCTURE OF IMMUNOGLOBULINSAND ANTIBODIES Ill. THREE-DIMENSIONAL STRUCTURE OF LIGHT CHAINS IV. THREE-DIMENSIONAL STRUCTUREOF FAB FRAGMENTS

67 68 72 72

73 74

1. Low resolution studies 2. Hioh resolution studies V. STRUCTURAL STUDIES ON COMPLETE IMMUNOGLOBULINSAND FC FRAGMENTS VI. THE STRUCTURE OF THE ANTIBODY COMBINING SITE

1. General description 2. Subgroups and combining sites VII. STRUCTURAL STUDIES OF IMMUNOGLOBULIN-HAPTENCOMPLEXES

1. X-ray crystallographic studies of Fab-hapten complexes 2. Other studies on hapten binding 3. Multispecificity of antibody combining sites VIII. IMMUNOGLOBULINSTRUCTUREAND ANTIBODY DIVERSITY

1. Variable and constant sequences 2. Association of L and H-chains

IX. CONCLUSIONS

81 83 83 85 85 85 87 88 88 88 89

89

ACKNOWLEDGEMENTS REFERENCES

90 90

I. INTRODUCTION

Vertebrates possess an elaborate immune defense mechanism against invading foreign macromolecules or allogeneic cells. As part of this mechanism they are capable of producing highly specific antibody molecules which bind to these antigens. In recent years a major goal of immunochemists has been to define the activity, specificity, and physiological function of antibody molecules in terms of their three-dimensional structure. Experimental approaches such as amino acid sequence determination, affinity labeling, electron microscopy, optical rotatory dispersion, and circular dichroism have been successfully employed to provide important clues in the solution of this problem. It is, however, generally accepted that the method of single-crystal X-ray crystallographic analysis is the only technique currently available that can reveal the three-dimensional atomic structure of a protein molecule. A number of extensive and up-to-date reviews on the principles and techniques employed in X-ray crystallography are available (for example, Holmes and Blow, 1966; Dickerson, 1964; Cold Spring Harbor Symposium, 1971; Mathews, 1976). A necessary requirement for the solution of a protein structure is the availability of crystals which diffract well to high resolution and are stable for a reasonable period of time in an X-ray beam. Furthermore, a number of isomorphous heavy atom derivatives should be found for these crystals, by for example diffusing heavy atom compounds. It should then be possible to obtain both intensity and phase data from which a Fourier synthesis can subsequently be calculated. This synthesis generates a three-dimensional contour map which represents the distribution of electron density within the crystal. Under favorable conditions the diffraction patterns obtained from these crystals could extend to a resolution of at least 2 A. At this resolution the atomic structure can be derived from the electron density map if the amino acid sequence of the complete polypeptide chain is known. The ultimate test of a protein structure determination is the ease by which the electron density map can be interpreted in terms of the amino acid sequence. 67

68

R . J . POLJAK, 1,. M. AMZII and R. P. I)ttlZA('KIiRI.I ~

Most of the indirect evidence obtained to date suggests that thc structure of a protein in the crystalline state is very similar to its structure in solution, since many reactions which take place in solution can also be shown to take place in the crystal. Detailed analysis of all the available evidence could be briefly summarized by stating that tile structure in the crystalline form is at least one of several possible "ground states" of the molecule. During the past few years a number of laboratories have succeeded in obtaining atomic resolution models of immunoglobulins in this way and we intend here to review the major conclusions obtained by these studies. TABLE I.

Isotypes

Approximate tool. wt.

IgM IgA IgG IgD IgE

900,000 170,000 500,000 I50,000 180,000 180,0(}0

Percentage of total circulating immunoglobulin in serum of normal individuals 5 10 10 20 70 80
Carbohydrates percentage by weight 1% 12 10--12 2 3

The existence of a multichain structure of immunoglobulins was established about 12 years ago, but the fact that immunoglobulins were demonstrably heterogeneous during an immune response to an antigen seemed at that time to present an insurmountable obstacle to X-ray diffraction and amino acid sequence studies, since both require a homogeneous preparation of a single molecular species. This problem has been overcome by studying homogeneous pathological immunoglobulins produced by monoclonal neoplastic lymphocytic cells in mice and man. These myeloma proteins, associated with the spontaneous occurrence of multiple myelomatosis and other pathological lymphoproliferative disorders in man, and with experimentally induced tumors in mice, have been shown to be closely related to normal immunoglobulins and antibodies by a number of structural and functional properties. Myeloma proteins can be obtained in large quantities and can be readily purified. In general they exist as complete molecules although occasionally only a portion of the molecule is present, most frequently the "light chain" (see below). Bence-Jones proteins, which can be isolated from the urine of patients with multiple myelomatosis, are light chains which display an unusual thermal behavior: they precipitate at 40°-60°C, redissolve at 9 5 100~C and reprecipitate on cooling. (See Humphrey and Owens (1972) for a detailed review of plasma cell dyscrasias and pathological immunoglobulins). Immunoglobulins (Ig) can be divided into the major classes or isotypes shown in Table 1 which also includes their molecular weight and relative abundance in serum. IgM, lgA and IgG contain carbohydrates, which are covalently attached to the molecule, and are largely hexose and hexosamine molecules with smaller amounts of sialic acid and fucose. II. POLYPEPTIDE CHAIN STRUCTURE OF [MMUNOGLOBULINS AND ANTIBODIES Many excellent reviews on this subject have been written (Edelman and Gall, 1969; Milstein and Pink, 1970; Krause, 1970; Hood and Prahl, 1971; Natvig and Kunkel, 1973). The IgG class of immunoglobulins has been the most intensively studied; the diagrammatic structure of a human IgG1 molecule shown in Fig. 1 introduces some of the nomenclature that will be used in this review. The molecule consists of two identical "light" (L) polypeptide chains (each with a mol. wt. of 20,000-25,000 and consisting of approximately 214 amino acid residues) and two identical "heavy" (H) polypeptide chains (each with a mol. wt. of 50,000-55,000 and consisting of approximately 450 amino acid residues). Each chain can be divided into "homology regions" (see below),

Three-dimensional structure of immunoglobulins

69

each of approximately 110 amino acids. The L-chain consists of regions VL and CL and the H-chain consists of regions VH, C,1, CH2 and CH3. The four individual chains (2 light and 2 heavy) are linked covalently by interchain disulfide bonds. Owing to the existence of many non-covalent interactions between the L and H chains, drastic chemical conditions (such as acid pH, exposure to urea, etc.) are required for the separation of this structure into individual polypeptide chains after reduction of the interchain disulfide bonds.

/""...

Fab

,-'"~",

/. tj ,t

,,'

,

,"

Fab'

', VL

"'.~

os SS

,'..

"~

-

r

::

::

:i

~

_~

.2',,

~,¢'x

<7

i

',, ',

;

Fx~. 1. Diagram of a human immunoglobulin (IgG1) molecule. The light (L) chains (mol. wt. about 25,000) are divided into two homology regions, VL and CL. The heavy (H) chains (mol. wt. about 50,000) are divided into four homology regions, V., C.1, CH2 and CH3. The Cnl and CH2 are joined by a "hinge" region indicated by a thicker line. Cleavage of the IgG1 molecule by papain generates Fab fragments (mol. wt. about 50,000) consisting of an L and an Fd polypeptide chains, and Fc fragments (mol. wt. about 50,000). Cleavage by pepsin followed by reduction of inter-H-chain disulfide bonds generates an Fab' fragment consisting of an L and an Fd' polypeptide chains, lnterchain and intrachain disulfide bonds and the N-termini of the L and H-chains are indicated. (Reproduced from Poljak (1973) with kind permission of Plenum Press.)

Porter (1959) found that controlled enzymatic digestion of rabbit IgG produces three fragments (Fig. 1), each with a mol. wt. of approximately 50,000. Two of these fragments are identical and are called the Fab fragments and the third is called Fc. Each Fab fragment consists of a complete L-chain and the N-terminal half of an H-chain called Fd. The Fab fragment is so named because it retains the antigen binding activity of the parent molecule, although it can only behave as a monovalent antibody. No complement fixation activity can be observed in the immune Fab-antigen complex, indicating that the Fc region is required for complement fixation. Controlled digestion of a human or rabbit IgG protein by pepsin cleaves both H-chains on the C-terminal side of the inter-heavy-chain disulfide bond(s) and produces a major fragment called F(ab')2 (Nisonoff et al., 1960). By subsequent reduction and alkylation of the inter-heavy-chain disulfide bond(s), a fragment called Fab' (Fig. 1) is readily obtained. The N-terminal portion of the H-chain in this fragment is called Fd' and is about ten amino acid residues longer than Fd, in human IgG immunoglobulins. Similar Fab fragments have been obtained by the use of other proteolytic enzymes such as trypsin. All these proteolytic enzymes split peptide bonds in a region which appears to be openly accessible and which has been called the "hinge" (or "flexibility") region linking Fab to Fc. The hinge regions are indicated by thicker lines in Fig. 1.

70

R.J. POLJAK, L. M. AMZEL and R. P. PHIZACKFRLI~

The L-chains of human IgG can be antigenically classified into two isotypes called ~c and 2, each characterized by a unique sequence in their C-terminal regions. Human IgM, IgA, IgD and IgE also possess K and 2 light chains but their heavy chains are different and are specific to each class. Amino acid sequence studies of human myeloma L-chains have shown that L-chains of the same class (~c or 2) consist of a C-terminal half of constant amino acid sequence (CL) and an N-terminal half of variable sequence (VL). Because of the possible genetic implications, the patterns of variability of L-chain sequences have been extensively analyzed. It was observed that within a given class of L-chains there are sequences which are very similar to each other and these constitute a "subgroup". Three such subgroups have been recognized in human ~,--chains and at least four in human 2-chains. All chains within a subgroup are very similar in sequence except at certain positions within VL where extreme variability is observed (Wu and Kabat, 1970). Kabat and Wu (1971), proposed that these hypervariable sequences constitute the regions of the L-chain structure which are in contact with antigen, so that the presence of different sequences in these regions will modify the antibody specificity. Comparative studies on H-chains of the same class have shown that the sequence of the regions CHI, CH2 and CH3 remain constant whereas the VH region (Fig. 1) displays variability. Just as in the L-chain, the variable region of the H-chain occurs at the N-terminal end of the molecule and is approximately 110 amino acid residues in length and also contains hypervariable regions (Kabat and Wu. 1971: Capra and Kehoe. 1974). H - c h a i n sequences in the Fc region of rabbit IgG (Hill et al., 1966) revealed another important structural feature, namely the existence of sequence "homology regions". Two sequences are said to be homologous when chemically identical or related amino acids appear at corresponding positions in the two polypeptide chains, for example a serine in the first sequence and a threonine at the corresponding position in the second. Another criterion for homology between two sequences is to examine amino acid differences in terms of the minimum number of mutational events that are necessary to change the nucleotide sequence specifying the first chain to that specifying the second polypeptide chain. If the number of mutations is smaller than can be expected from random events then the sequences are said to be homologous. By either of the two criteria outlined above one can define four constant homology regions, namely C,1, CH2 and CH3 in the H-chains and CL in the L-chains. The N-terminal variable regions VL and VH are homologous to each other but have a weaker homology to CL, CH1, CH2 and CH3. It is interesting to observe that the pattern of a single intrachain disulfide loop of similar length exists in each of these regions (Fig. 1). In addition to their genetic implications, these findings also suggest that the existence of homology units greatly influences the overall three-dimensional folding of IgG molecules. Inspired by these and other observations, several L and H polypeptide chain folding schemes were proposed (Singer et al., 1967; Putnam et al., 1967; Edelman et al., 1969: Welscher, 1969: and others) which can be summarized by describing the tertiary structure of immunoglobulins as consisting of globular subunits each corresponding to a homology region. Electron microscopy has provided the first direct pictures of the general shape and structure of immunoglobulins. By using a divalent hapten ( b i s - N - D N P - o c t a m e t h y l e n e diamine, DNP: dinitrophenyl) as a link between anti-DNP antibody molecules, Valentine and Green (1967) were able to deduce from electron micrographs the general shape of an IgG molecule and the arrangement of the Fab and Fc regions (Fig. 2). When combined with antigen, the shape appears to be that of a letter "Y", the angle between the two Fab regions being dependent on the number of IgG molecules connected by the b i s - D N P haptens. The flexibility required to obtain this variable separation is assumed to reside in the "hinge" region connecting Fab to Fc. Electron micrographs of an IgA protein produced by the (laboratory induced) mouse plasma cell tumor MOPC 315 (Green et al., 1971) indicated that the IgA structure consists of globular units or domains. In this study on IgA, a divalent b i s - D N P hapten was also used, taking advantage of the fact that the MOPC 315 myeloma protein has the specificity of an anti-DNP

Three-dimensional structure of immunoglobulins

71

antibody. No such globular subunits or domains had been consistently observed in electron micrographs of IgG. Evidence has been obtained from fluorescence polarization and electron spin resonance studies (Zagyansky, 1975) which indicates that immunoglobulins from frog and tortoise have a fairly compact general structure with no marked intramolecular rotational freedom, in contrast to the pronounced flexibility of mammalian IgG. A corresponding difference in flexibility can also be observed, but to a lesser extent when comparing IgM molecules from these same species. As segmental flexibility increases on passing from lower vertebrate immunoglobulins to mammalian immunoglobulins it has'been suggested (Zagyansky, 1975) that structural evolution of antibodies has in part been directed to increasing this flexibility thus favoring the interaction of the polyvalent antibody with foreign antigens.

//~

~NO

60~'-~

2

NO2

~/40/~'

FIG. 2. Scale diagram of a hapten-linked trimer of IgG immunoglobulin molecules as obtained by electron microscopy (reproduced by permission from Valentine and Green, 1967).

Indeed, flexibility in the antibody molecule appears to favor the precipitation of foreign antigen, facilitating its elimination. IgE, whose general structure is more compact (Nezlin et al., 1973) is incapable of forming a precipitate after antigen binding (Stanworth, 1971). Furthermore, electron microscopy studies showed that the angle between the two Fab regions in carp macroglobulins does not vary (Shelton and Smith, 1970) and these macroglobulins only give rise to a weak precipitation reaction (Richter et al., 1972). In hen IgG, weak precipitation reactions (Hersh and Benedict, 1966) and limited flexibility (Zagyansky, 1975) have also been observed. Affinity labeling experiments have provided further insight into the location and topography of antigen binding sites. In these experiments, a haptenic group is specifically (reversibly) bound and covalently attached to an amino acid side chain on the antibody molecule by means of a chemically reactive group on the hapten. In principle, amino acid side chains that are part of the combining site in antibody molecules, or are close to it, can be specifically labeled and identified. Using a number of different reactive haptens and antibodies (and also myeloma proteins that behave like antibodies), it became evident that the antigen binding site was confined to the VL and the V. regions. Amino acid side chains in, or close to, the regions of hypervariable sequence in L and H-chains were labeled (Singer et al., 1967; Haimovich et al., 1970; Cebra et al., 1971), which supported the hypothesis that these regions contribute to (or determine) the antigen binding site of antibodies. Synthetic antigens have been used as an experimental tool for analyzing the specificity of antibodies, the role of immunodeterminant groups in the antigen-antibody reaction, and the dimensions of the combining sites. With different antigens, the most exposed end of an immunodeterminant group has consistently been found to make the larger contribution to the energy of the binding reaction (Kabat, 1966; Sela, 1969; Schechter, 1971). The dimensions of the binding site

72

R.J. POLJAK. [,. M. AMZ~t and R. P. PHIZACKI{RII'~

have been estimated to be of the order of 35 x 10-15 × 6~ 10 A by using antigenic polysaccharides (Kabat, 1966) and polypeptides (Maurer, 1964: Sage et al.. 1964: Haber et al., 1967). Ill. THREE-DIMENSIONAL STRUCTURE OF LIGHT CHAINS Detailed crystallographic analyses have been reported for human 2-chains (Schiffer et al., 1973; Edmundson et al., 1974) and human t,-chains (Epp et al., 1974). In addition

to these studies on crystalline L-chains, atomic resolution analysis of crystalline Fab fragments (reviewed in Section IV) have also provided structural models of human and murine light chains. A 3.5 A resolution Fourier map permitted Edmundson and his colleagues (Schiffer et al., 1973) to trace the polypeptide chain backbone of the two )°-chains which compose an Mcg Bence-Jones dimer. Although an atomic model was not built at this stage of the work, many amino acid side chains could be located. The most prominent features of the electron density map were identified as the four intrachain disulfide bonds. The density of the interchain disulfide bond was much lower, a feature which was attributed in part to the fact that this bond is near to the surface of the molecule and consequently has more freedom of motion. Two "domains", one containing the two variable homology regions (V-domain) and the other containing the two constant homology regions (C-domain) could be localized in this map. The region in each polypeptide chain called the "switch" region, which links the V and C domains, could also be identified in the two chains. Although both L-chains have an identical amino acid sequence, the angle between the major axes of the V and C homology regions is different in the two chains (110° in one and 70 ° in the other), an observation reminiscent of the Land H-chain structure of Fab fragments (see below). The V~, and CL homology subunits closely resemble each other in both shape and polypeptide chain folding, although residues 48 60, which includes a region of hypervariable sequence in the VL~ subunit. has no equivalent in the C~, subunits. At one end of the V-domain there is a cavity which resembles a truncated cone with an aperture width of approximately 15 /~ and a depth of approximately 10 /~. This cavity is lined by the hypervariable segments of both chains and is thought to constitute the equivalent of the antigen-binding site within the Fab region of antibody molecules. The structure of a crystalline dimer of the variable region V,, of a Bencedones protein (Rei) has been determined to a resolution of 2.8 A (Epp et al., 1974). From these studies a model was built with sufficient resolution to display the following secondary structural information. Two /3-sheets were found to be present in each V,, monomer in which approximately 50~o of all the amino acid residues occur. The two /~-sheets surround a hydrophobic interior formed by invariant or semi-invariant residues. Only one turn of a distorted c~ (or rr) helix could be found in the model which is in agreement with circular dichroism experiments on immunoglobulins and L-chains. Just as in the 2-chain dimers discussed above, a large cavity surrounded by the hypervariable regions of both chains, and lying on a local two-fold axis of symmetry, was found. In fact, the V,, monomers form a dimer which closely resembles the V~. domain of the ,;-chain dimer. indicating that the V-regions can associate independently of the interchain contacts that occur in the C-regions. Other features of this thorough structural analysis of the V~-dimer will be discussed below in connection with the Fab' structure. IV. THREE-DIMENSIONAL STRUCTURE OF FAB ERAGMENTS Fragments Fab and Fab' from human myeloma proteins were crystallized by Rossi and Nisonoff (1968) and by Rossi et al. (1969). X-ray diffraction patterns of the crystalline Fab and Fab' fragments (Avey et al., 1968; Humphrey et al., 1969) from IgG1 (2) New and IgG1 (2) Hil were obtained. The Fab and Fab' fragments of individual proteins gave identical intensity patterns indicating that they have the same three-dimensional structure. It would normally be expected that, since Fab' is about ten amino acid residues longer than Fab (in its Fd' chain), the X-ray intensity patterns will reflect this difference.

Three-dimensional structure of immunoglobulins

73

Subsequent h,igh-resolution studies of Fab' New indicated that the C-terminus of the Fd' chain does not contribute significantly to the diffracted X-ray intensities, probably due to random motion at the end of the Fd' polypeptide chain. 1. Low Resolution Studies

The 6-~ resolution study of Fab' New (Poljak et al., 1972) showed that the Fab' fragment consists of two discrete globular domains, V and C, containing the VL and Vn and the CL and Cr~l homology regions, respectively (see Fig. 3). Despite the limitations of a low-resolution study, the two polypeptide chains (L and Fd') were found to be two continuous, independent stretches of electron density, which originate in one of the globular subunits and, after a region of globular folding within that subunit, extend to the other subunit through a narrow outer bridge of polypeptide chain (the "switch" region). The overall scheme is that of a tetrahedral arrangement of four globular subunits, two of which (VL and CL) are formed by the L-chain and the other two (Vu and CH1) by the Fd' chain. The three-dimensional model obtained from the 6 A Fab' New structure can be used to explain some of the physicochemical properties of immunoglobulin molecules. For example, fragments of L-chains which appear to be of catabolic origin have been reported in human urine samples (Solomon and McLaughlin, 1969; Baglioni et al,, 1967). In vitro cleavage of human myeloma L-chains by controlled enzymatic digestion has been shown (Solomon and McLaughlin, 1969; Karlsson et al., 1972) to generate fragments which have a more compact shape, as determined by measurements of the radius of gyration and frictional coefficient, than the parent'L-chain. Studies of allotypic and idiotypic markers (Solomon and McLaughlin, 1969), peptide mapping, and partial amino acid sequence data (Karlsson et al., 1972) reveal that the fragments obtained

FIG. 3. Photographs of the 6 A resolution model of Fab' New. The lighter and darker parts of the model represent the two continuous chains of electron density. Two globular subunits can be easily recognized in all photographs. Reproduced from Poljak et al. (1972) with the permission of Nature.

74

R, J. POLJAK, L. M. AMZEL and R. P. PHIZACKERlt5

by enzymatic action on L-chains correspond to the VL and CL homology regions. Cleavage of the parent polypeptide chain occurs in the "switch" region connecting the VL and CL homology regions, which is probably easily accessible to proteolytic attack by enzymes. In the Fab' model, it is clear that both the L and the Fd'-chains are extended and exposed to solvent and proteolytic enzymes around their switch regions (half way along the polypeptide chains). Cleavage at the mid-point of the L-chain would consequently generate two globular L-chain fragments with a radius of about 15 to

FIG. 4. Schematic representation of the IgG molecule incorporating structural information obtained for the Fab' fragment and extrapolating to the Fc region. Only one of the possible orientations of the Fc subunits relative to those of Fab is shown here. Reproduced from Poljak et al. (1972) with permission from Nature.

20 A. This value compares well with the 16 A obtained from measurements of the radius of gyration for several CL and VL fragments (Karlsson et al., 1972). The model of Fab' New suggests that it should be possible to cleave the Fab molecule into two fragments of molecular weight 25,000 each; one of which should correspond to the V-domain and which could presumably retain antigen binding activity. Givol and coworkers (Inbar et al., 1972) have demonstrated that in some cases cleavage of the Fab' fragment by pepsin generates a fragment (called Fv) consisting of the VL and the Vr~ regions. It is interesting to note that a similar type of cleavage can be made in the Fc region (Turner and Bennich, 1968) which produces an Fc' fragment which also has a molecular weight of approximately 25,000. The X-ray crystallographic model of Fab' New shows a definite correlation between the internal homologies in amino acid sequence and the tertiary structure of irn_munoglobulins. It verifies that the homology regions fold into compact globular structures linked by linear connecting regions. The existence in the Fab fragment of two globular domains (V and C), each comprising a portion of two distinct polypeptide chains, suggests that gene duplication (Hill et al., 1966) determined a structural duplication which resulted in increasingly larger precursors of immunoglobulin molecules by the addition of globular domains. Structural features observed in the Fab' New model can therefore be extended to provide a schematic model of a whole IgG molecule as shown in Fig. 4. In this schematic representation, the V-domain consists of the homology regions VL and VH, C1 consists of CL and CH1, and C2 and C3 consist of (Cn2)z and (Cn3)z, respectively. C2 and C3 constitute the Fc region in which an exact two-fold axis of symmetry relates two Cn2 and two CH3 subunits (Goldstein et al., 1968). 2. High Resolution Studies

The conclusions outlined above were confirmed by high resolution studies on the crystalline fragment Fab' New (Poljak et al., 1973, 1974). In these studies atomic models were built using amino acid sequence data and Fourier maps at 2.8-A and 2.0-A resolution. The Fab fragment from the murine myeloma protein McPC 603 has also been analyzed to a resolution of 3.1-A (Segal et al., 1974). As stated above, the Fab' and

Three-dimensional structure of immunoglobulins

75

FIG. 5. Schematicdiagramof the Fab' New fragment.The L-chain(solidline) and the Fd'-chain (open line) are each shownlabeled with a number of amino acid residues. A centrallylocated cleft divides the fragment into V (left) and C1 (right) domains. Four homologysubunits VL, Vu, CL and Cnl can be dearly differentiated.The CL and Cnl subunits interact very closely and appear more tightlypacked than the VL - Vn subunits. Fab fragments are equivalent for crystallographic studies and therefore the results discussed below will apply to either fragment. The overall dimensions of the Fab molecule are 80 x 50 x 40 A. A centrally located cleft divides the molecule into two structural domains, V and C (as shown in the schematic diagram Fig. 5 and the Fab' New 2.8-A model Fig. 6). Four "homology subunits" which correspond to the homology regions VL, VH, CL and CH1 can be clearly differentiated. The CL and CH1 subunits interact very closdy and appear more tightly packed than the VL - Vn subunits. Approximate local two-fold axes of symmetry relate the C~ and CH1 subunits and also the VL and VH subunits. The angle between these two independent two-fold axes (shown by rods in Fig. 6) is approximately 145°. The C~, CH1, VL and Vn subunits are strikingly similar in their three-dimensional folding as would be expected from the homology in their amino acid sequences. However, although VI and V~ share the basic "immunoglobulin-fold" of C~. and CH1, they include an additional length of polypeptide chain in the form of a loop which is not present in CL and CH1 (see Fig. 7). In the L (4) chain of IgG New a deletion of seven amino acids in Vt. (residues 53-59, Fig. 10) shortens this additional loop of polypeptide chain so that the VL structure of IgG New appears more similar to that of the CL and Cnl subunits than to VH. The deleted amino acid residues can be readily placed in the model of VL, by comparison with the folding of the VH subunit. The structure of the homology subunits can be described as consisting of two fl-pleated sheets formed by several strands (seven in CL and Cnl) of antiparallel polypeptide chain. The CL homology subunit, for example, consists of a fl-pleated sheet composed of four hydrogen-bonded antiparallel chains (residues 116-120, 132-140, 160--169, and 173-182) and another E-pleated sheet composed of three antiparallel chains (residues 147-151, 193-199, and 202-208, see Figs. 8 and 9). These two twisted and roughly parallel sheets surround a tighly packed interior of hydrophobic side chains including the intrachain

76

R.J. f'OLJAk. I.. M. A'~1/:I:I and R. 1). PtlIz.,~,( kl:~
FiG. 6. View of the 2.~-k resolution model ~,1 t a b Next. The backbone>; oI the L arid t d (H) polypeptide chains can be seen above and belm~ the tx~o ~hilc rod~,, respectivel 5. l'hc V (left) and C (rightt domains are scpt,,rated h~ a central cleft. Labels shm~ the "swileh region" tit the midpoint of both chains. The approxmlatc local 2-1k~ld axes of symmetry are indicated by two white rods. The four intrachain disullide bonds and Ihe intcrchain disullide bond are marked by white spheres. Tapes connect the > c a r b o n positions of residues that lbrm disulfide bonds in other immunoglobnlin molecules. "Fhc numbered labels (round labels m V~ and rectangular labels in VIII at the left end of thc model indicate the "'hypcrv~lriablc positions". The homologous tryptophan residues near the intrachain distllfide bonds in e'lch subunit can he clearly seen. Arrows a! the right cnd of the model point to Set 154 and Lys 191 (Kern and Oz serological markers, rcspectivelyl. This view of the molecule closely corresponds 1o the "'top" xiew of the Iow-fesolution (6 <,\) model (Fig. 3).

disulfide bond which links the two sheets in a direction approximately perpendicular to their planes. Greater than 50°,i of the CL residues are included in these two fi-pleated sheets. The homologous subunits CHI, V~,, and VH can be described in similar terms (Fig. 9). In addition to the features of secondary structure just discussed, the Fab' New VL residues 26 to 27c appear in a helical conformation close to that of a ~t-helix. The VL residues 78 to 82, and the homologous V~ residues 87 to 91 also fold with the conformation of a distorted 7t or a-helix. In general, the precise conformation of aminoacid residues that do not belong to the regions of secondary structure discussed above is more difficult to define, especially when they correspond to regions of lower electron density on the surface of the molecule. Amino-acid sequence comparisons have been extensively used in several laboratories to align the primary structures of different immunoglobulin molecules and their homology regions. The alignment of V~, and VH sequences with those of the C~, and CH1 homology regions is less straighforward than alignments between VL and V~ or between CL and CHI regions. This problem has been approached by matching their similar three-dimensional structures and aligning residues that occupy an identical or similar

Three-dimensional structure of immunoglobulins

NH~

FIG. 7. Diagram of the basic "immunoglobulin-fold". Solid trace shows the folding of the polypeptide chain in the constant subunits (CL and CH1). Numbers designate L ().)-chain residues, beginning at "NH~-" which corresponds to residue 110 for the L-chain. Broken lines indicate the additional loop of polypeptide chain characteristic of the Vt. and VH subunits.

131 190~ 212 /

161

203

116

FIG. 8. Schematic diagram of the ~-carbon backbone of the CL homology subunit showing two planes of #-pleated sheet (one containing four hydrogen-bonded anti-parallel chains shown by white arrows and another containing three hydrogen-bonded antiparallel chains shown by striated arrows). These two twisted and roughly parallel sheets surround the intrachain disulfide bond (shown in black) which links the two sheets in a direction approximately perpendicular to their planes. (Reproduced by permission from Edmundson et al. (1975).)

77

78

R.J. POLJAK, L. M. AMZEL and R. P. PHIZA('KERII'* l&l 142 143 144 145 I46 ~

11o

CL

95 90 .-. 31 67 68 - . . 24 4 98 :== 89 32 66 ::=69 2 ~ :=: 5 99 ~1~ =:: 33 65 70 .'.,~12 ~ 6 100 = : : [~'71 34 ..-46 64 ===Pl 21 7. 102 . . . . . . . . . . 101 ~ = = : 35 45 63 72 = : : 20 8."103 ........... " " - - ' - = : : = 85 36 : : : 44 62 = : : 73 19 9 ~106 84 : : : 37 43 61 74:=: 18 10,~. 105 . . . . . . . . . . . . . . . . . . . . . 83 58 ===4142 6 0 " " 75 17 II ~I06 I 39 40

16.

/

12:[."108 lO7

15 ~ 13 14

109

,..~::::,--q 76 77 78 79 80 81 82

|

111 171 112 170 172 113 169---173 114 168 174...140 115 167=:=175 139=::I16 166 176===138 I17 165164---I 77 7~57~=-'II 8 163 178" 119 162=::179 "-:'120 161 180===134 121 160---181 133 122

201 200 ! 202--.199 I 203 198:==147 204::=197 148 205 196===149 206=::~ 150 207 194===151 208===193 152 ] 209 192 153 : 210 lSZ : 211 155 J

131 124 130 125 129 126 128 127

/

21~

183 184 185 186 187 188 189 190 191 ~

52 51 50 49 48 47

159 158 157 156

151 152 153 154

VH

f

26 27 28 29 30 31 32

i 2 75 76 25 . - . 3 74 77 ... 24 4 73 78 23 :== 5 72 79 ==: ~-~ 6 71 ===80 ~ ==: 7 70 81 ==:20 8

69 :== 82 19 68 03 === 18 67 - " 8 4 17 85 "--16

118 r--r---'~ i 86 87 88 89 90 91 66 65 64 63 - -

I 119 11~

177 120 116 178 150:==121 I75:==179 149 122 174 180.--148 123 173:::181 147::=124 172171 182===146 125 170...183 ==2126 169 184==~ 127 168:::185 143=:=128 167 186===142 129 166-"187 141 130 188 140 131 I 139 132 138 133 137 134 136 135

55 56 57 58

~T :::37 1.,110 4748 ~tI12=:= 94 38 ==: 46 113 93 === 39 45 114 ::= 92 40 / 41 42

9 I0~ II ~ I 1 5 12~" 116 14

,o l

103 101 104...100 53 5 4 1 0 5 1 0 6 ~ 99 : : = 3 3 52.'~.. 107 ... 98 34=== 51 108 9 7 : 2 2 35 50=== 109 :== ~-~ 36 :::&,

CH1

J

207 206 208 205 209===204 210 203::=155 211:==202 156 212 2 :::157 213::~ 158. 214 199"--159 215"--198 160 217 216 218 219

1

162 J 161 163

189 190 191 192 193 194 195 196 197

61 60 59

165 164

FIG. 9. Diagram of the hydrogen bonds (broken lines) between main chain atoms for V,, C,, VH, and CHI. The hydrogen-bonded clusters correspond to the two /7-sheets of each subunit (see text). Cysteine residues that participate in the intrachain disulfide bonds linking the two /7-sheets are enclosed in rectangles: CL residue 213 and CHI residue 220 form the interchain disulfide bond and are also enclosed in rectangles. Numbers refer to residues as given in Fig. 10.

position in the constant pattern of the tertiary structure of the homology subunits. Alignments obtained by this procedure, which is independent of amino-acid sequence homologies, are shown in Fig. 10. Some of the variable and constant features of VL sequences can be discussed in terms of the three-dimensional structure of V~. New. Hairpin bends in the polypeptide chain of V~ New occur around positions 14-15, 27-30, 39-40, 67-68, and 92-93 and an approximate 90 ° bend around residues 75-76. Except for the bend at positions 92-93 VL VH CL

....

1 i0 20 27 a b c 30 ZSVLTQPPSVSGAP-GQRVTISCTGSSSNIGAGNHVKWYQQLPGTAPK-LLIFHNNA

....

1 iO 2O 3O ZVQLPESGPELVSP-RQTLSLTCTVSGSTFSNDYY--TWVRQPPGRGLEWIGYVFYHGT$['~JT

ii0 120 130 140 QPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAV-TVAWK--ADSS

vL

.....

VH

70 -PLRSRVTMLVNT-S

CL

160 170 180 190 -PVKA--GVETTTPSKQSNNKYAASSYLSLTPEQWKSHKSYSCQVTH--EGST-VEKT-VAPTEC$

CHI

170 180 190 -ALTS--GVHTFPAVLQSSGLYSLSSVVTVPSSSLGT-QTYICNVNHKPSNTK-VDKR-VEPKSC

.......

...... 4O

5O

,6~

..................

CHI

..........

5o

150

120 130 140 ASTKGPSVFPLAPSSKSTSGGTAALGCLVKDYFPEPV-TVSWN---SG 60 RFSVSKSG

40

150

160 ...................

70 80 90 SSATLAITGLQAEDEADYYCQSYDRS--LR-VFGGG~KLTVL~ 80 90 KNQFSLRLSSVTAADTAVYYCARBLIAG-CIBVW

ioo

i00

GI~0G

IIH ~LVqV~[:

200 200

io9

210 210

FIG. 10. Amino-acid sequences of the VL, CH, VH, and CHI homology regions of Fab New, aligned by comparison of their three-dimensional structures (see text). The tentative VH sequence given here was obtained by Drs. Y. Nakashima, W. Konigsberg, R. J. PoOak and B. L. Chen (in preparation). The Vt. and C1. sequences are taken ~om Chen and Poljak (1974). Abbreviations are ~om Dayhoff (1972).

214 22O

Three-dimensionalstructure of immunoglobulins

79

(a hypervariable region) all others involve a Gly residue that is constant in human ).-chain sequences. Most of these bends also involve a constant or nearly constant ProGly or (Ser, Thr}-Gly sequence. A similar conclusion has been obtained from the study of a crystalline V~ fragment (Epp et al., 1974). Glycine residues also contribute to a constant sequence Phe--Gly-Gly-Gly (positions 99-102) which is not part of a bend. The constant character of this seqfience in all ).-chains can be explained by the following observations: (a) Phe 99 is located in an internal, interchain hydrophobic pocket that includes the homologous constant Trp 108 in Vn, related to Phe 99 by a local pseudo two-fold axis of symmetry, and therefore it can be assumed that Phe 99 (and Trp 108) make an interchain contact that is important for VL-VR assembly; (b) Gly 100 in Va (or Gly 109 in VH) is tightly packed between the intrachain disulfide bond and Leu 4 (a constant residue in VL and VH); (c) Gly 101 is relatively close to the constant Gin 6, although there is room here for a side chain as observed in V~(Gln 101) or in Vn (position 110); (d) Gly 102 (111 in Vn) is very close to a constant aromatic residue (Tyr/Phe, 86 in V~ or 95 in VH) which occupies most of the space available in this region, thus only a Gly residue can be accommodated. Some other residues that are constant in V~, or that show only limited variability, such as Tyr 35, Gln 37, Ala 42, Pro 43, and Asp 84 are involved in close contacts with the Vn subunit. Other constant residues such as Gin 37, Glu 82, Tyr 85 in VL and Tyr 142 in CL make internal hydrogen bonds. In addition to the residues just discussed, most of the nonpolar, hydrophobic amino acids that occur in the interior of the structure between the two fl-sheets are invariant or are replaced by other hydrophobic residues. They are Leu 4, Gin 6, Val 10, Val 18, lie 20, Cys 22, Val 32, Trp 34, Leu 46, Phe 61, Val 63, Ala 70, Leu 72, lie 74, Leu 77, Ala 83, Tyr 85, Cys 87, Ser 89, Val 98, Thr 103, Leu 105, and Val 107. Thus residues that appear to have an important structural role, such as those at bends and those which contribute to intra- and inter-subunit bonds, are found to be invariant or semi-invariant. In the CL subunit, Ser 154 and Lys 191, which correlate with the serologic nonallelic human 2-chain markers Kern- and Oz ÷, respectively, appear on the surface of the molecule approximately 8 /~ from each other. The lnv allotypic markers of human x-chains have been shown (Milstein et al., 1974) to involve Ala/Val and Val/Leu substitutions at positions 153 and 191, respectively, which closely correspond to the positions of the Kern- and Oz ÷ markers in human ).-chains. Replacements at positions 153 and 191 in x-chains will alter the antigenic determinants of the molecule that are recognized by anti-allotypic antisera. Since the distance between the or-carbon atoms of the homologous residues in Fab' New is approximately 8 /~ replacements involving both positions can be simultaneously recognized by a single anti-allotypic antibody molecule. As the folding pattern of each homology region in the Fab molecule is similar and the allotypic markers of human x-chains can be correlated with structural features of a human ).-chain, it is thought to be likely that the structural model of the Fab fragment is valid for other immunoglobulins of different isotype or even different animal species. Supporting evidence for this postulate was obtained by an analysis of the interchain and intrachain disulfide bonds (Poljak et al., 1973), which will be reviewed below. The L and H-chains of all immunoglobulins are covalently linked by a disulfide bond (Fig. 1), (with the exception of some IgA molecules where two L-chains are linked to each other by a disulfide bond (Grey et al., 1968)). The L-chain cysteine residue that contributes to this bond is at the C-terminus of the chain in human x-chains and penultimate to the C-terminus in human ).-chains, see Fig. 9. In different isotypes of H-chain, and in different animal species, the cysteine residue that completes the S--S bond is either at position 214 or at about position 131 (Fig. 11). The bonding scheme illustrated in Fig. 1la applies to human IgG1 (Steiner and Porter, 1967) immunoglobulins such as IgG New. The interchain disulfide bond illustrated in Fig. l lb, in

80

R.J. POIJAK. L. M. AMZti[ and R. P. PHIZA( KI:Rlf~ L

213 s

H -'

214

L

213.

a)

I

b) H FIG.

11. Diagram

showing

two

131 different patterns immunoglobulins.

of

interchain

disulfide

bonds

of

which the H-chain cysteinyl residue occurs at position 13l, is found in human IgG2, human IgG3, human IgG4 (De Preval et al., 1970), human IgM (Putnam et al., 19711, rabbit IgG (O'Donnel et al., 1970), guinea pig IgG2a (Birshtein et al., 1971) and in mouse IgG2a and IgG2b (De Preval et al., 1970). In the three-dimensional model of Fab' New, the H-chain Cys 214 is approximately 6 A from L-chain Cys 213 to which it is linked by a disulfide bond. However, position 131 in CHI also occurs at a distance of 6 A from L-chain Cys 213, so that its replacement by a cysteinyl residue could lead to an alternative interchain disulfide bond as found in the immunoglobulin molecules listed above. Unusual intrachain disulfide bonds that have been observed in several immunoglobulins can be explained on the basis of the model of Fab' New (Poljak et al., 1973). One of these bonds is the intra-H-chain disulfide observed in rabbit IgG, linking the polypeptide chain between residues 131 and 221 (O'Donnel et al.. 1970). An unusual disulfide bond which is observed in the VH region of a human 71-chain from IgG Daw (Press and Hogg, 1970) can also be explained by the close spatial proximity (about 6 A) of the homologous residues in Fab' New. Perhaps the most interesting interchain disulfide bond that has been reported is the one that links V~, position 80 to CL position 171 in rabbit antibodies of restricted heterogeneity (Poulsen et al., 1972; Appella et al., 1973). A comparison of the sequences of these rabbit K-chains and the human 2-chain from IgG New indicates that Cys 80 and Cys 171 in rabbit K-chains correspond to Ala 79 and Asn 172, respectively, in IgG New. The side chains of V~, Ala 79 and CL Asn 172 face each other, and the distance between their ~.-carbon atoms is about 5.5/~. This is compatible with the presence of a disulfide bond linking the two homology subunits as observed in some rabbit K-chains. Thus, the Fab' model provides an adequate structural framework for the various patterns of interchain and intrachain disulfide bonds that have been established by sequence analyses, therefore giving further support to the postulate that K- and 2L-chains have the same overall three-dimensional structure and that the VH and CH regions of different classes of H-chains t~, 7, #, etc.) also have the same overall pattern of polypeptide chain folding. Further support for this postulate was given by the results of the crystallographic analysis of the V,, Rei dimer and the murine McPC 603 IgA-Fab fragment (Segal et al., 1974) in which the overall three-dimensional folding of the polypeptide chains is the same as that observed in the human Fab fragment. At the present stage of the structural analysis of immunoglobulins, the postulate that they all have a common pattern of three-dimensional structure can be accepted with reasonable confidence. As mentioned above, the Fab structure can be described as a tetrahedral arrangement of homologous subunits, covalently linked in pairs (V~ to C~, and Vn to C H I t by linear stretches of polypeptide chain ("switch" regionst bent to a greater extent in the Fd'-chain than in the L-chain (Poljak et al., 1973; Segal et al.. 1974). Furthermore, two identical L-chains in a dimer assume different conformations (Schiffer et al., 1973) such that one chain appears similar to the L-chain of the Fab fragment whereas the other L-chain of the dimer resembles the Fd' chain. Although no gross conformational changes have so far been observed in Fab fragments, these findings suggest that the Fab fragment

Three-dimensional structure of immunoglobulins

.81

V

C i.

/ / FIG. 12. A view of the a-carbon backbone of Fab' New. The V and C1 domains, the L-chain (solid line), the Fd' chain (open line) and the local, approximate two-fold axes (broken lines) relating the VL to the VH subunit and the CL to the C.I subunit are shown. The two short arrows indicate the switch region of both chains. The longer arrow indicates a possible relative motion of the V and C1 domains.

could display flexibility in the switch regions allowing a conformational change to take place by a hinge-like movement at one or both switch regions. Since a disulfide bond linking VL to CL has been found in some rabbit I g G molecules (described above), the flexibility of the m o r e open L-chain may be more limited than that of the H-chain. An "opening" of the Fd' chain, as illustrated in Fig. 12 would lead to a relative movement of the structural subunits exposing some of the VH and CH1 side chains that were not previously exposed. v. S T R U C T U R A L S T U D I E S ON C O M P L E T E I M M U N O G L O B U L I N S FC F R A G M E N T S

AND

Fc fragments from h u m a n and rabbit I g G immunoglobulins have been studied by X-ray diffraction in several laboratories. M a n y problems have been encountered however, such as pronounced crystal damage after X-irradiation and difficulty in obtaining isomorphous heavy atom derivatives. In view of these difficulties it is not surprising that a structural model of Fc has not yet been obtained. A number of important antibody

82

R.J. POLJAK, L. M. AMZEL and R. P. PHIZACKt7RI.I:5

properties have been associated with the Fc region, for example, interactions with cell membranes and fixation of complement, the active site for the latter being thought to reside in the CH2--CH2 domain. Therefore a considerable effort is being made to obtain a structural model of the Fc fragment. It is very probable that the observed structural homologies between the subunits in the Fab fragment extends to the Fc fragment, as proposed schematically in Fig. 4. In an attempt to verify this postulate Colman et al. (1974) used a crystallographic rotation correlation function (Rossmann and Blow, 1962) on data obtained from (l) a crystalline human Fc fragment and (2) a crystalline human myeloma IgG protein which was reported to be suitable for high

I I

t

I

I

t

#\

IB

0 -~--. ×

..-~)

b

0

m

t

--~x

j

~

16A

t

iv

~

--~X

BL;

t

~

FIG. 13. Four possible ways of joining the Fab region (C and D) to a nearby Fc region (A and B). Model IV where the hinge lies between regions B and C is preferred (Sarma et al., 1971 ) (with permission).

resolution studies (Palm and Colman, 1974). In the latter, the two halves of the IgG molecule are related by a crystallographic diad. A crystallographic two-fold axis relating the two H-chains moieties has also been demonstrated in human Fc crystals (Goldstein et al., 1968). From their studies on the Fc fragment using rotation function techniques, Colman et al. (1974) propose that the CH2-CH2 contacts and relative orientation are different from those that occur in the CH3-Ctt3 pair. This is consistent with studies on the properties of free CH2 and CH3 dimers (Turner and Bennich, 1968; Ellerson el al., 1972) which imply a stronger contact for CH3-CH3 than for CH2~CH2. A crystallographic study on a complete human IgG molecule (IgG Dob) has been reported by Sarma et al. (1971). This molecule, a cryoglobulin, crystallizes spontaneously on cooling. One L-chain and one H-chain are contained in the asymmetric unit of the crystal and are related to the other L and H-chains by a two-fold axis of symmetry which runs through the inter-H-chain disulfide bonds. Owing to structural disorder within the crystal, the X-ray diffraction pattern does not extend beyond a resolution of about 6 /~. Furthermore the crystals are damaged by relatively short exposures to X-rays making such an analysis an experimentally difficult problem even at low resolution. However, an electron density map at a nominal resolution of 6 ~ was calculated and two globular regions could be isolated (Sarma et al., 1971). One of these globular regions surrounds a two-fold axis of symmetry and was therefore assigned to the Fc part of the molecule. The other globular region should correspond to the Fab part of the molecule. Four possible ways of assembling a complete IgG Dob molecule were found in this study (as shown in Fig. 13), and of these, one was chosen as a tentative

Three-dimensionalstructureof immunoglobulins

:83

structure. In this model the IgG molecule is T-shaped, however, an electron microscopic study of the same crystalline material (Labow and Davies, 1971) showed a Y-shaped molecule. In this respect this molecule appeared to be similar to rabbit anti-DNP antibody preparations observed by Valentine and Green (1967), which also showed a Y-shaped structure. Given the experimental difficulties mentioned above, the tentative nature of the model should be emphasized; furthermore it should be considered that Dob may not be a typical IgG molecule since it contains a deletion in its hinge region (Lopes and Steiner, 1973), a factor which could influence the shape of the molecule. It is also possible that in this particular crystal form, lattice forces may induce a T-shape, a shape which may be quite different in solution or after glutaraldehyde fixation prior to electron microscopy. Small-angle X-ray scattering studies have been conducted on a human myeloma immunoglobulin (Pilz et al., 1970) and on rabbit antibodies (Pilz et al., 1975). The X-ray scattering curves obtained in these experiments can be analyzed to provide information about the shape of the molecule. From these studies it was concluded (Pilz et al., 1970) that IgG molecules are T-shaped in solution. In considering this conclusion it is difficult to decide whether there are other possible approximations to the shape of the molecule which could fit the data equally well since molecular shape could be a rather complex function even at low resolution. The importance of determining the overall conformation of the individual components of antibody molecules, with and without antigen attached, should be emphasized in view of the fact that complement fixation or binding to effector cells may depend on a conformational change in the molecule after antigen binding. In another small-angle X-ray scattering study (Pilz et al., 1975), a volume contraction of about 10~o and a decrease in the radius of gyration of about 8~o was found when 90~o of the antigen binding sites of rabbit anti-poly-o-alanyl antibodies where titrated with a tetra-D-alanine hapten. This finding is in agreement with a potential for conformational change which resides in the "segmental flexibility" of the molecule around its hinge regions (Yguerabide et al., 1970). VI. THE STRUCTURE OF THE ANTIBODY COMBINING SITE 1. General Description

Evidence obtained from both amino acid sequence determinations and affinity labeling studies was used to identify the amino acid residues that are involved in the definition of the antigen binding sites of immunoglobulins. The presence of regions of "hypervariable" sequence around positions 30, 50 and 95 were recognized by statistical analysis of L-chain sequences (Wu and Kabat, 1970; Kabat and Wu, 1971). A similar conclusion was obtained by comparison of H-chain sequences, however, fewer H-chain sequences were available and therefore the location of the hypervariable regions was less definitive (Kabat and Wu, 1971). Hypervariable regions were also recognized in the amino acid sequences of guinea pig H-chains (Cebra et al., 1974) and human H-chains (Kehoe and Capra, 1971!. These findings lead to the postulate that the hypervariable regions specify the conformation of the antigen binding site of immunoglobulin molecules (antigen complementarity regions). Independent identification of antigen complementarity regions has been attempted by affinity labeling studies (Singer et al., 1971; Goetzl and Metzger, 1970; Franek, 1971; Ray and Cebra, 1972; Givol et al., 1971; Fleet et al., 1972). These studies could be interpreted as showing that, in general, the labeled amino acids occur at, or adjacent to, the hypervariable regions of the L and the H-chains. Definitive evidence that hypervariable regions of the H and L-chains determine the conformation of antigen combining sites was obtained in the study of Fab' New (Poljak et al., 1973). The hypervariable regions occur at adjacent bends of the polypeptide chains defining a crevice or pocket at one end of the Fab' fragment. A very similar arrangement of hypervariable regions was observed in the Fab fragment of the murine IgA McPC

84

R. ,J. POIJAK, L. M. AMZH and R. P. PHIZA(KERI,]'~

603 myeloma protein (Segat et al., 1974). The human L-chain dimer Mcg (Schiffer et al., 1973) and the human V,, dimer Rei (Epp et al., 1973) show a similar arrangement of hypervariable regions in the L-chains. The combining site of Fab' New extends over a large area surrounding a central groove or pocket which is 15 /~ long, 6 A wide and 6 /~ deep. Residues 27 to 30 of the L-chain (first L-chain hypervariable regionl form the "upper" end of the groove (see Fig. 14) while H-chain residues 55 to 60 (second H-chain hypervariable region) and 30 to 33 (first H-chain hypervariable region) form

GLY27C~90

fiR6

92 L

98 31

FIG. 14. View of some of the amino-acid residues at the combining site of IgG New.

the "lower" end. The "sides" are formed by L-chain residues 90 to 95 (third L-chain hypervariable region) on the left and H-chain residues 102 to 107 (third H-chain hypervariable region) on the right. The pocket is predominantly lined by the side chains of these residues of the VL and VH regions. It is therefore evident that amino acid replacements at these positions will alter the conformation and consequently the specificity of the antibody combining site. Furthermore, hypervariable regions of immunoglobulins contain insertions and/or deletions that alter the length of the polypeptide chain in these regions. Sequence modifications of this kind have a very significant effect on the size and shape of the combining site. For example, the L-chain (~c) of murine myeloma protein McPC 603 is longer than other human and murine ~c- and k-chains due to an insertion of six amino acids in its first hypervariable region (as in human V,,.~ sequences). Partly due to this insertion, the structure of the Fab fragment of McPC 603 (Segal et al., 1974) has a "'deeper" combining region than Fab' New. H-chain sequences also show a variable number of amino acids in their hypervariable regions. In particular the length of the third hypervariable region of VH has been found to range from 13 to 20 amino acids when counted from Cys 96 to Trp 107. In Fab' New the loop defined by this region has a different overall conformation to that found in the equivalent regions of the other homology subunits. It is bent towards the local two-fold axis of symmetry, which relates VH and V,, forming a narrow groove at the

Three-dimensionalstructureof immunoglobulins

85

active site. Varying the number of amino acids in this region probably results in drastic changes in the size and shape of this groove. In conclusion, the pattern of insertions and deletions observed in the sequences of hypervariable regions will be a major determinant of the general dimensions of the active site. In addition, these variations complemented by single amino acid replacements will define a chemical environment within the active site which is characteristic of a given immunoglobulin molecule. In this way it is possible to accomodate a large number of antibody specificities while retaining a constant overall pattern of three dimensional folding.

2. Subgroups and Combining Sites The description of the combining site of immunoglobulins presented in the previous section can be used to assign a structural and functional meaning to the notion of "subgroup" in ~-- and 2-chains. Since "subgroups" correlate with the presence of insertions and/or deletions in the hypervariable regions (in particular the first hypervariable region) they correspond to different conformations of the active site. For example, human k-chains of subgroup III (V~m) have insertions of up to six amino acids in their first hypervariable region. These chains will define a "deeper" active site (such as that observed in McPC 603) than those defined by either a K-chain of subgroup I (V~) or a human 2-chain (such as that in IgG New). It is interesting to observe that murine 2-chains appear to lack the pattern of insertions and deletions found in the hypervariable regions of murine ~-chains and in human ~- and 2-chains. Thus, it can be expected that murine R-chains will make a smaller contribution to the diversity of antibody specificities so that the expression of 2-chains or the number of 2-chain genes in mice would be limited in relation to h--chains as reflected in the ~-/2 serum ratio (Weigert et al., 1970). VII. STRUCTURAL STUDIES OF IMMUNOGLOBULIN-HAPTEN COMPLEXES

1. X-Ray Crystallographic Studies of Fab-Hapten Complexes Crystallographic studies of Fab-hapten complexes provided a direct characterization of the combining site of immunoglobulin molecules. It is important to point out that human and murine myeloma proteins were used in these studies and not induced antibodies, however, myeloma proteins have been shown to bind a variety of haptens and antigens in reactions that closely resemble those of induced antibodies (reviews by Metzger, 1969; Krause, 1970; Potter, 1973; Seligman and Brouet, 1973). For example, myeloma proteins have been screened for hapten binding activity and frequently found to bind some ligands with association constants comparable to those of induced antibodies (i.e. greater than 104-1051/mol). Furthermore these binding reactions give linear Scatchard plots with intercepts that show that two haptens are bound to one immunoglobulin molecule. These and other criteria indicate that hapten binding by myeloma proteins can be directly correlated to antibody-hapten reactions. However, as expected for small ligands, structural studies of haptens bound to immunoglobulin molecules show interactions involving only a fraction of the amino acids present in the combining site. This was true, for example, in the study of the structure of the complex between the Fab fragment of murine IgA McPC 603 and the hapten phosphoryl-choline (Padlan et al., 1973). Phosphoryl-choline binds to McPC 603 with an affinity constant of 1.7 x 1051/mol. A derivative of the hapten labeled with a heavy atom was diffused into McPC 603 Fab crystals and its position was determined by a difference Fourier map. Phosphoryl-choline was found to bind to the combining region of the protein in the cavity defined by the hypervariable regions of the L and H-chain. The interactions of the hapten with the combining site involve mainly Arg 52, Lys 54 and Glu 35 of the H-chain in agreement with the presence of both a positive and a negative charge on phosphoryl-choline at neutral pH.

86

R.J. POLJAK, L. M. AMZEL and R. P. PHIZACKERLtiY

o

o

iLH

FIG. 15. Chemical formula of Vit. K I O H (3-[3'-hydroxy-3',7',lt',15',tetramethyl hexadecyl]2-methyl-l,4-naphtoquinone).

Varga et al. (1974) screened the human myeloma protein IgG New for the binding of a large number of haptens. Several compounds such as uridine, orceine, menadione and others were found to bind to this protein with low affinity constants (approximately 10 31/mol). However a 7-hydroxyl derivative of vitamin K1 (vit. K~OH, Fig. 15) was found to bind with an affinity constant of 1.7 x 105 1/mol. Crystallographic studies on the complexes of these compounds and Fab' New showed that all of them bind to the "upper" part of the combining site in the region surrounded by the third hypervariable region of the H-chain and the first and third hypervariable regions of the L-chain (Amzel et al., 1974). The naphtoquinone moiety of Vit. K1OH is in close contact with residues Tyr 90, Gly 29 and Asn 30 of the L-chain and residues 100 to 102 of the H-chain (Fig. 16). The phytyl chain extends over a larger area making contact with L-chain residues 47 (a constant Trp), 50, 58 and 101. Since the affinity constant for the binding of Vit. KIOH (1.7 x 105 1/mol) is larger than that of menadione (approximately 10 3 1/mol) it was concluded that the increased affinity is due to the interaction of the phytyl chain with the combining site of Fab' New. GL'r27C t~_

..£3.GL Y 29

,HIS 31

QSN27Q ALAI02 GLTI03 QRG 92

TTR 90

LEUIO0

IETSlO~

LEU 9q

ARG 95 RP q7

SER 93

ASP TTR 33

RSP 58

98 TTR 50 31 PHE 52 SER 57 TTR 53 DLT

5~

F1G. 16. View of Vit. K I O H bound to the combining site of Fab' New. The orientation of the drawing is the same as Fig. 14.

Three-dimensional structure of immunoglobulins

87

No major conformational changes were observed after hapten binding in the studies of both Fab' New-Vit. K~OH and Fab McPC 603-phosphoryl-choline complexes. 2. Other Studies on Hapten Binding The Bence-Jones dimer Mcg(2) was screened for hapten binding in the crystalline state by diffusing ligands into preformed crystals (Edmundson et al., 1974). A number of compounds such as colchicine, e-dansyl-lysine, 1-10 phenantroline, menadione and others were found to bind to the dimer. The ligands bind to the cavity that corresponds to the combining site of the Fab fragments. However, in the L-chain dimer, the cavity is "wider" and "deeper" than in the Fab fragments permitting some of the ligands to diffuse further into the molecule in the region of contact between the two VL subunits. It was proposed that the Bence-Jones dimer can be considered as a model of a primitive antibody which could have existed before the divergence of independent L- and H-chains. MOPC 315 is a well studied D N P binding murine myeloma protein. Even though the three-dimensional structure of this molecule was not determined by crystallographic analysis a correlation between its binding properties and the structure of its active site has been made using the model of Fab' New as a basic structural framework (Poljak et al., 1974). The L (2) chains of IgG New and MOPC 315 (Francis et al., 1974) are highly homologous and contain an equal number of residues in the third hypervariable region between the constant amino acid residues Cys 89 and Phe 99. Furthermore, a comparison of the tentative sequence of VH New (Fig. 10) and that of VH M O P C 315 indicates that the third hypervariable regions of VH in both chains between the constant Cys 96 and Trp 108 include the same number of amino acid residues. It is, therefore, feasible to fit the MOPC 315 sequence to the basic VL and VH structures obtained in the 2.0-A Fourier map of Fab' New. The model of the MOPC 315 binding site that results from this comparison includes a crevice similar to that shown in Fig. 14, in which L-chain Tyr 34 forms the "upper" limit, H-chain Trp 47 and Phe 50 form the "lower" limit, L-chain Phe 99, H-chain Tyr 104 and Phe 34 contribute to the "sides" and L-chain Trp 98 and Phe 103 and H-chain Phe 105 form the 6- to 10-A deep "floor". The high density of adjacent aromatic side chains that line this crevice, at the center of the active region, is striking and correlates with the observed specificity of MOPC IgA immunoglobulin for D N P and other haptens that include benzene and naphthalene aromatic rings. The relatively shallow depth of the active center is in agreement with the electron microscopic study of a complex between MOPC 315 IgA and the bifunctional hapten bis(DNP-~-alanyl)-diaminosuccinate (Green et al., 1971) in which the D N P groups that join the Fab arms of different IgA molecules end to end are only 15 A apart. Several calorimetric studies provided additional information about the interactions involved in the binding of antigens and haptens by antibodies. Barisas et al. (1972) used pooled anti-DNP rabbit antibodies and measured the enthalpy of binding of DNPlysine. Despite the fact that the pooled antibodies were heterogeneous, with affinity constants ranging over more than one order of magnitude, it was possible to conclude that the reaction was enthalpically driven (AH = - 2 1 kcal mol-1). Halsey and Biltonen (1975) also used pooled rabbit anti-DNP antibodies, but in this case the antibodies were fractionated according to their affinity for DNP by precipitation with limiting amounts of antigen. By this procedure five pools of antibodies were obtained with affinity constants covering a seventeen-fold range. The measured enthalpies for the antibody-DNP-lysine reaction for the five pools did not show significant differences. The average value for the enthalpy of binding in the reaction was -13.9 _ 0.4 kcal mol-1. Since the affinity constants for the different pools varied from 2.24 x 107 to 1.37 × 10 6 the free energies and entropies of binding varied from -10.0kcal mo1-1 and -5.1 cal mol- 1 deg- 1 to - 8.35 kcal mol- 1 and - 10.6 cal mol- 1 deg- 1 respectively. Evidently the reactions are enthalpically driven in all the pools and in all cases entropic effects favor the dissociation of the antibody hapten complexes. Furthermore, the differences

88

R.J. POLJAK. L. M. AMZEI. and R. P. PHIZA('KI-RLI'~

in affinity for different antibody pools seem to be due entirely to changes in the entropy of binding. Several tentative explanations of these effects are proposed by the authors but it is clear that more detailed studies, combined with three-dimensional structural information, are necessary to understand the thermodynamics of binding in antibodx reactions.

3. Multispecificity of Antibody Combining Sites The capacity of individual immunoglobulins to bind different ligands has been taken as support for the postulate that an antibody molecule can be multispecific (Eisen et al., 1967; Richards et al., 1975; Poljak et al., 1975). X-ray crystallographic studies (Amzel et al., 1974) have shown that IgG New can bind at the same locus in the active site of a number of ligands with different chemical properties. The comparison made between IgG New and the MOPC 315 IgA immunoglobulin can also be used to illustrate this concept. MOPC 315 has high affinity for DNP and also binds competitively with high affinity other ligands, such as menadione. As pointed out in a previous section the amino acid sequences of IgG New and those of IgA MOPC 315 are highly homologous and contain a similar number of amino acid residues in their hypervariable regions. Both immunoglobulins have nearly equal avidities for vitamin K1, and each binds additional ligands that are not bound by the other. These results can be taken as an indication that antibodies are multispecific and can bind different haptens at the same general site in the active region. The mode of binding, for example by formation of charge transfer complexes involving aromatic side chains of the antibody could be operative for a given chemical class of ligands such as DNP, menadione, flavin mononucleotide, etc. (Michaelides and Eisen, 1974) for which one would expect different affinity constants reflecting different degrees of chemical interaction. In addition, small conformational changes at the active site could play a significant role in promoting the binding of different haptens to the same antibody molecule. Independent of the role of somatic mutation mechanisms of diversification, a gene pool encoding multispecific antibodies would be smaller than one encoding monospecific antibodies and would naturally be subject to repeated induction and expression. The advantages of carrying such a genetic pool would insure its maintenance against genetic drift and deleterious mutations. VIII. IMMUNOGLOBULIN STRUCTURE AND ANTIBODY DIVERSITY

1. Variable and Constant Sequences The three-dimensional structure of the Fab fragment of immunoglobulins was used to correlate the variable and constant features of V~, and V, sequences with structural characteristics of the molecule (Poljak et al., 1974). As described in a previous section, all the constant or nearly constant residues that appear at bends or contribute to intraand inter-subunit bonds seem to be important for the preservation of the structure. Mutations that alter any of these residues, which represent more than 507,,~iof the total number of residues of the V~ sequence, can not be considered neutral or modulating. Different combinations of these invariant and semi-invariant residues that are compatible with the requirements of the constant three-dimensional structure of the homology subunits can be better explained by a process of evolutionary germ line gene divergence than by random somatic mutations. By contrast, the nature of the residues that occur in the regions of hypervariable sequence, at one end of the molecule and fully exposed to the solvent, is not limited by any visible structural constraints. Insertions and/or deletions frequently observed in immunoglobulin sequences can also be closely correlated with the three-dimensional structure. A striking example of this correlation is that of positions 27a, 27b and 27c in human 2-chains. These residues form part of a single turn helical loop which can be deleted without creating a gap or other major alteration in the path of the polypeptide chain (Poljak et al., 1973). Deletions also occur in segments of the polypeptide chains of immunoglobulins outside

Three-dimensionalstructure of immunoglobulins

89

the hypervariable regions. Two examples with a clear structural correlation can be given here. The first is that of the deletion of'residues numbered 53 to 59 in the amino acid sequence of the L-chain of IgG New (Poljak et al., 1973). These residues very nearly constitute the loop of polypeptide chain which is characteristic of the variable regions (VL, VH) of immunoglobulins and which does not oc6ur in the constant regions CL and CH1, Fig. 7. An integral deletion of this loop can be easily visualized in a three-dimensional model of VL (Fig. 7) and evidently results in a stable tertiary structure similar to that of CL. The second example to be considered here is that of the deletion of an entire homology region, which occurs in H-chains synthesized by human tumor cells in the "heavy chain disease" (Cooper et al., 1972; Wolfenstein-Todel et al., 1974) and in murine myeloma H-chains (Milstein et al., 1974; Preud'homme et al., 1975), beginning at the start of the CH1 region (position 119) or before, and ending at position 215, close to the beginning of the CH2 region. In the three-dimensional structure the sequence... Val-Ser-Ser shared by ~ and/~ chains (Florent et al., 1974) clearly constitutes the end of VH, and after a sharp bend, position 119 marks the beginning of the CH1 homology subunit. Complete deletion of Cnl should not affect the molecular interactions of CH2 and CH3, leading to a compact, stable Fc-like molecule with an N-terminal appendix of variable length. 2. Association of L and H-Chains The structural design of immunoglobulin molecules as described can incorporate a large number of antibody specificities while maintaining a constant overall three-dimensional structure. However, if L and H-chains could recombine to form new immunoglobulin molecules the economy of this design could be further extended. The permutation of L and H-chains would give rise to a large number (N z) of antibody specificities with a minimum number (N) of L and H-chains. Experimental evidence of in vitro recombination of L and H-chains has been obtained in several laboratories. Subgroup determination in L and H-chains of human immunoglobulins (Capra and Kehoe, 1975) and in vitro recombination studies (Stevenson and Mole, 1974) do not give clear indications of subgroup preferences in L and H-chain pairing. The study of the three-dimensional structure of Fab' New (Poljak et al., 1975) indicates that the close interactions between VL and V, involve the side chains of residues at positions 35, 37, 42, 43, 86, and 99 of the L-chain and positions 37, 39, 43, 45, 47, 95 and 108 of H-chain. These positions are occupied by constant residues or by fairly conservative replacements in VL and VH sequences of all species studied so far. It is striking that human ~: and 2-chains are very similar in their amino acid sequences at these positions. Almost all human L-chain contain the following residues: Tyr 35, Gin 37, Ala 42, Pro 43, Tyr 86, and Phe 99. Positions 88 and 90 in L-chains occur at the active site and also appear to provide for VL--Vn interactions but the importance of these interactions is more difficult to evaluate, because they could be affected by the conformation of the third hypervariable region of VH. However, some amino acids are more frequently found at these positions namely, Gin 88, and Trp/Tyr 90. Undoubtedly preferred associations could occur in vivo by mechanisms under genetic control, but the interactions between VL and Vn which should provide the stereochemical basis for the unique association of L and H-chain pairs appear to be independent of L and H-chain "subgroups" or even L-chain class.

IX. CONCLUSIONS The three-dimensional structure of immunoglobulin fragments is now available to atomic resolution as a result of X-ray diffraction studies. The models obtained in these studies provide a structural framework for the analysis of biological and physical chemical properties of immunoglobulins. The structures determined to high resolution contain ~c and 2 L-chains and Fd-fragments of 7 and a H-chains from murine and human myeloma proteins. Comparison of these structures indicate that the Fab fragments of immunoglobulin molecules from all classes have the same overall three-dimensional

90

R.J. POLJAK. k. M. AMZEL and R. P. PHIZA('KtiRI.I~

structure. Furthermore, all the homology regions in a given immunoglobulin share a common pattern of polypeptide chain folding (immunoglobulin fold). The immunoglobulin fold consists of two irregular, roughly parallel/;-pleated sheets surrounding a tightly packed core of hydrophobic residues. In each homology subunit more than 50°0 of the amino acid residues are included in the fl-sheets. In the VL and V,-homology regions these residues are found to be constant or nearly constant when comparing different immunoglobulins. The hypervariable residues of V~ and V. occur in close spatial proximity at one end of the molecule and are fully exposed to the solvent. These residues determine the conformation of the combining site of antibodies. Amino acid replacements, insertions and deletions that occur in these regions confer unique antigen binding specificities to individual immunoglobulin molecules. Human L-chain subgroups correlate with the presence of insertions and deletions in hypervariable regions, therefore the notion of subgroup can in this case be directly correlated with the conformation of the combining site of immunoglobulins. The design of antibody molecules is such that a large number of specificities can be accommodated while maintaining a constant overall three-dimensional structure. Furthermore, amino acid residues involved in VL--V. interactions appear to be constant or nearly constant indicating that, in principle, all combination of L and H-chains can assemble to give functional immunoglobulin molecules. It is clear from this review that many of the properties of antibodies can now be interpreted in terms of the three-dimensional structure of Ig's. Studies on haptens bound to Fab fragments provide a model for the interactions involved in antibody-antigen reactions. However, other important questions related to these reactions will require the study of induced antibodies and their antigens by crystallographic techniques. Furthermore the knowledge of the structure of the Fc fragment and of an intact Ig will probably be necessary for the interpretation of the effector functions of antibody molecules. ACKNOWLEDGEMENTS The research carried out in the authors' laboratory was supported by Grants AI 08202 from the National Institutes of Health, NP-141A from the American Cancer Society, and N,I.H. Research Career Development Award AI 70091 to R.J.P. REFERENCES AMZEE,L. M., POLJAK.R. J., SAUL,F., VARGA,J. M. and R~CHARDS,F. F. (1974) The three-dimensional structure of a combining region ligand complex of IgG New at 3.5-A. resolution. Proc. Natn. Acad. Sci. U.S.A. 71. 1427-1430. APPELLA, E., ROHOLT, O. A., CHERSI, A., RADZIMINSKI, G. and PRESSMAN, D. (1973) amino acid sequence of the light chain derived from a rabbit anti-p-azobenzoate antibody of restricted heterogeneity. Biochem. Biophys. Res. Commun. 53, 1122-1129. AVEr, H. P., POLJAK, R. J., ROSSI, G. and NISONOFF, A. (1968) Crystallographic data for the Fab fragment of a human myeloma immunoglobulin. Nature 220, 1248-1249. BAGLIONI, C., CIOLI, D., GORINI, G., RUFFILLI, A. and ALESCIO-ZONTA, L. (1967) Studies on fragments of light chains of human immunoglobulins: genetic and biochemical implications. Cold Sprin9 Harh. Syrup. Quant. Biol. 32, 147-159. BARISAS, B. G,, SINGER, S. J. and STURTEVANT, J. M. (1972) Thermodynamics of binding of 2,4-dinitrophenyl and 2,4,6-trinotrophenyl haptens to the homologous and heterologus rabbit antibodies. Biochemistry 11, 2741-2744. BARSTAD, P., RUD1KOFF, S., POTTER, M., COHN, M., KONIGSBERG, W. and HOOD, L. (1974) Immunoglobulin structure: amino terminal sequences of mouse myeloma proteins that bind phosphorylcholine. Science 183, 962-964. BIRSHTEIN, B. K., HUSSAIN, Q. Z. and CEBRA, J. J. (1971) Structure of heavy chain from strain 13 guinea pig immunoglobulin-G(2). III. Amino acid sequence of the region around the half cystine joining heavy and light chains. Biochemistry 10, 18-25. CAPRA, J. D. and KEHOE, J. M. (1974) Variable region sequences of five human immunoglobulin heavy chains of the VHIII subgroup: definitive identification of four heavy chain hypervariable regions, Proc. Natn. Acad. Sci. U.S.A. 71. 845-848. CAPRA, J. D. and KEHOE, J. M. (1975) Distribution and association of heavy and light chain variable region subgroups among human IgA immunoglobulins. J. lmmunol. 114, 678-681. CEBRA, J. J., RAY, A., BENJAMIN, D. and BIRSFITEIN,B. K. (1971) Localization of affinity label within the primary structure of 72 chain from guinea pig IgG(2) in Progress in Immunology (B. A~os, ed.), pp. 269 284, Academic Press, New York.

Three-dimensional structure of immunoglobulins

91

CEBRA, J. J., KOO, P. H. and RAY, A. (1974) Specificity of antibodies: primary structural basis of hapten binding. Science 186, 263-265. CRIES, B. L. and POLJAK, R. J. (1974) Amino acid sequence of the (2) light chain of a human myeloma immunoglobulin (IgG New). Biochemistry 13, 1295-1302. Cold Spring Harbor Symposium, Vol. 36, (1971) Structure and function of proteins at the three-dimensional level. COLMAN, P. M., EPP, O., FEHLMAMMER,H., BODE, W., SCHIFFER,M., LATTMAN,E. E. and JONES, T. A. (1974) X-ray studies on antibody fragments, FEBS Lett. 44, 2, 194-199. COOPER, S. M., FRANKLIN,E. C. and FRANGIONE,B. (1972) Molecular defect in a gamma-2 heavy chain, Science 176, 187-189. DAYHOFL M. O. (1972) Atlas of Protein Sequence and Structure Vol. 5, National Biomedical Research Foundation, Silver Spring, Maryland. DEPREVAL, C., PINK, J. R. L. and MILSTEIN,C. (1970) Variability of interchain binding of immunoglobulins. Nature 228, 930-932. DICKERSON,R. E. (1964) X-ray analysis and protein structure, in The Proteins (H. Neurath, ed.), pp. 603-778, Academic Press, New York. EDELMAN, G. M., CUNNINGHAM,B. A., GALL, W. E., GOTTLIEB,P. D., RUTISHAUSER,U. and WAXDAL, M. J. (1969) The covalent structure of an entire ~,G immunoglobulin molecule, Proc. Nam. Acad. Sci. U.S.A. 63, 78-85. EDELMAN, G. M. and GALL, W. E. (1969) The antibody problem. A. Rev. Biochem. 38, 415--466. EDMUNDSON, A. B., ELY, K. R., GIRLING, R. L., ABOLA, E. E. SCHIFFER,M., WESTHOLM,F. A., FAUSCH, M. D. and DEUTSCH,H. F. (1974) Binding of 2,4-dinitrophenyl compounds and other small molecules to a crystalline ).-type Bence-Jones dimer. Biochemistry 13, 3816-3827. [I)MUNDSON. A. B., ELY, K. R., ABOLA, E. E., SCHIFFER,M. and PANAGIOTOPOULOS,N. (1975) Rotational allomerism and divergent evolution of domains in immunoglobulin light chains. Biochemistry 14. 3953-3961. EISEN, H. N., LITTLE, J. R., OSTERLAND,C. K. and SIMMS, E. S. (1967) A myeloma protein with antibody activity. Cold Spring Harb, Syrup. Quant. Biol. 32, 75-81. ELLERSON, J. R., YASMEEN,D., PAINTER, R. H. and DORRINGTON, K. J. (1972) A fragment corresponding to the CH2 region of immunoglobulin G (IgG) with complement fixing activity. FEBS Lett. 24, 318-322. EPP, O., COLMAN,P., FEHLHAMMER,H., BODE, W., SCHIFFER,M. and HUBER, R. (1974) Crystal and molecular structure of a dimer composed of the variable portions of the Hence-Jones protein Rei. Eur. d. Bioche~ 45. 513-524. FLEET, G. W. J., KNOWLES,J. R. and PORTER, R, R. (1972) The antibody binding site. labeling of a specific antibody against the photo-precursor of an aryl nitrene. Biochem. J. 128, 499-508. FLORENT, G., LEHMAN,D. and PUTNAM, F. W. (1974) The switch point in /~ heavy chains of human IgM immunoglobulins. Biochemistry 13, 2482-2498. FRANCIS, S. H., LESLIE, R, G. Q., HooD, L. and ElSEN, H. N. (1974) Amino Acid sequence of the variable region of the heavy (ct) chain of a mouse myeloma protein with anti-hapten activity. Proc. Natn. Acad. Sci. U.S.A. 71, 1123-1127. FRANEK, F. (1971) Affinity labeling by M-nitrobenzenediazonium fluoroborate of porcine anti-dinitrophenyl antibodies. Eur. d. Biochem. 19, 176-183. G1VOL, D., STRAUSBAUCH,P. H., HURWITZ,E., WILCHEK,M., HAIMOVICH,J. and EISEN, H. N. (1971) Affinity labeling and cross-linking of the heavy and light chains of a myeloma protein with anti-2,4-dinitrophenyl activity, Biochemistry 10, 3461-3466. GOETZL, E. J. and METZGER,H. (1970) Affinity labeling of a mouse myeloma protein which binds nitrophenyl ligands. Sequence and position of a labeled tryptic peptide. Biochemistry 9, 3862-3871. GOLDSTEIN,n. J., HUMPHREY,R. L. and POLJAK, R. J. (1968) Human Fc fragment: crystallographic evidence for two equivalent subunits. J, Molec. Biol. 35, 247-251. GREEN, N. M., DOURMASHKIN,R. R. and PARKHOUSE,R. M. E. (1971) Electron Microscopy of human and mouse myeloma serum IgA. J. Molec. Biol. 56, 203-208. GREY, H. M., ABEL, C. A., YOUNT, W. J. and KUNKEL, H. G. (1968) A subclass of human 7A-globulins (7A2) which lacks the disulfide bonds linking heavy and light chains, J. Exp. Med. 128, 1223-1236. HABER, E., RICHARDS,F. F., SPRAGG,J., AUSTEN,K. F., VALLOTTON,M. and PAGE, L. B. (1967) Modifications in the heterogeneity of the antibody response. Cold Spring Harb. Syrup. Quant. Biol. 32, 299-310. HALSEY, J. F. and BILTONEN, R. L. (1975) The thermodynamics of hapten and antigen binding by rabbit anti-dinitrophenyl antibody. Biochemistry 14, 800-804. HAIMOVICH,J., GIVOL, 0. and EISEN, H. N. (1970) Affinity labeling of the H and L-chains of a myeloma protein with anti-2,4-dinitrophenyl activity, Proc. Natn. Acad. Sci. U.S.A. 67, 1656-1661. HERSH, R. T. and BENEDICT,A. A. (1966) Aggregation of chicken ~G immunoglobulin in 1.5 i sodium chloride solution. Biochim. Biophys. Acta 115, 242-244. HILL, R. L., DELANEY,R., FELLOWJR., R. E. and LEBOWITZ,H. E. (1966) The evolutionary origins of the immunoglobulins. Proc. Natn. Acad. Sci. U.S.A. 56, 1762-1767. HOLMES, K. C. and BLOW, D. M. (1966) The Use of X-Ray Diffraction in the Study of Protein and Nucleic Acid Structure, lnterscience, New York. HOOD, L. and PRAHL, J. (1971) The immune system: A model for differentiation in higher organisms, Adv. lmmunol. 14, 291-351. HUMPHREY, R. L., AVEY,H. P., BECKA,L. N., POLJAK, R. J., ROSSI, G., CHOL T. K. and NlSONOFF,A. (1969) X-ray crystallographic study of the Fab fragments from two human myeloma proteins. J. molec. Biol. 43, 223-226. HUMPHREY, R. L. and OWENS, A. H., JR. (1972) Immunoglobulins and the plasma cell dyscrasias, in The Principles and Practice of Medicine (A. H. HARVEY, R. J. JOHNS, A. H. OWENS, JR. and R. S. ROSS, eds.) pp. 1206-1240, Appleton~entury~Crofts, New York.

92

R.J. POLJAK, L. M. AMZEL and R. P. PHIZACKERLI~

INBAR, D., HOCHMAN, Y. and GIVOL, D. (1972) Localization of antibody-combining sites within the xariable portions of heavy and light chains. Proc. Natn. Acad. Sci. U.S.A. 69, 2659 2662. KABAT, E. A. (1966) The nature of an antigenic determinant. J. lmmunol. 97, 1 1 I. KABAT, E. A. and Wu, T. T. (1971) Attempts to locate complementarity-determining residues in the variable positions of light and heavy chains. Ann. N.E Acad. Sei. 190, 382-391. KARLSSON, F. A., PETERSON, P. A. and BERGGORD, 1. (1972) A structural feature of human immunoglobtdin light chains. J. Biol. Chem. 2,17, 1065-1073. KEHOE, M. J. and CAPRA,J. D. (1971) Localization of two additional hypervariable regions Jn immunoglobulin heavy chains. Proc. Natn. Acad. Sei. U.S.A. 68, 2019-2021. KRAUSL R. M. (1970) The search for antibodies with molecular uniformity, Adl:. lmmunol. 12, I 56. LABAW, L. W. and DAVIES, D. R. (1971) An electron microscopic study of human 7G1 immunoglobulin crystals. J. Biol. Chem. 246, 3760-3762. LOPES, A. D. and STEINER, L. A. (1973) A structural defect in a crystallizable myeloma protein. Fedn. Proc. Fedn Am. Socs. Eup. Biol. 32, 1003. MATTHEWS, B. W. (1976) in The Proteins (H. NEURATH and R. L. HILL, eds), Vol. 3, 3rd edn. Academic Press, New York, in press. MAURER, P. H. (1964) Use of synthetic polymers of amino acids to study the basis of antigenicity. Prog. Aller(4y 8, 1-40. METZGER, H. (1969) Myeloma proteins and antibodies. Am. J. Med. 47, 837 844. MICHAELIDES, M. C. and EISEN, H. N. (1974) The strange cross-reaction of menadione (vitamin K3! and 2,4-dinitrophenyl ligands with a myeloma protein and some conventional antibodies. J. Exp. Med. 140, 687-702. MILSTEIN, C. and PINK, J. R. L. (1970) Structure and evolution of immunoglobulins, Prog. Biophys. Molec. Biol. 21, 209-263. MILSTEIN, C., ADETUGBO, K., COWAN, N. J. and SECrtE~_, D. S. (1974) Clonal variants of myeloma cells. In Progress in Immunology It, VoL 1 (L: BRErq'rand E. J. HOLaOROW. eds.) pp. 157-168. North Holland. Amsterdam. M1LSTEtN, C. P. STEINBERG, A, G., McLAUGHLIN, C. L. and SOLOMON, A. (1974) Amino acid sequence change associated with genetic marker Inv (2) of human immunoglobulin. Nature 248, 160-161. NAKASHIMA, V. T., KONIGSBERG, W., CHEN, B. L. and POLJAK, R. J. (1975) in preparation. NATV1G, J. B. and KUNKEL, H. G. (1973) Human Immunoglobulins: Classes, Subclasses, Genetic Variants, and Idiotypes, Adv. lmmunol. 16, 1-59. NEZLIN, R. S., ZAGYANSKI, Y. A., K~IV~R~INEN, A. 1. and STEEANI, D. V. (1973) Properties of myeloma immunoglobulin E(Yu). Chemical fluorescence polarisation and spin-labeled studies, lmmum~ehemi.stry 10, 681 688. N1SONOFF, A., WISSLER, F. C,, LIPPMAN, L. N. and WOER, D. L. (1960) Separation of univalent fragments from the bivalent rabbit antibody molecule by reduction of disulfide bonds. Archs Biochem. Biophy.~. 89, 230-244. O'DONNELL, I. J., FRANG1ONE, B. and PORTER, R. R. (1970) The disulfide bonds of the heavy chain of rabbit immunoglobulin G. Biochem. J. 116, 261-268. PADLAN, E. A,, SEGAL, D. M., SPANDE, T. F., DAVIES, D. R., RUDIKOEF, S. and POTTER, M. (1973) Structure at 4.5 k resolution of a phosphorylcholine-binding Fab. Nature New Biol. 245, 165 167. PALM, W. and CoLragN, P. M. (1974) Preliminary X-ray data ]'rom well-ordered crystals for a human immunoglobulin G molecule. J. Molec. Biol. 82, 587-588. PILZ, I., PUCHWEIN, G., KRATKY, O., HERBST, M., NAAGER, O., GALL, W. E. and EDELMAN. G. M. (1970) Small angle X-ray scattering of a homogeneous 7GI immunoglobulin. Biochemistry 9, 211 219. PILZ, I., KRATKY, O., LICHT, A. and SELA, M. (1975) Shape and volume of fragments Fab' and (Fab')2 of anti-poly (D-alanyl) antibodies in the presence and absence of tetra-D-alanine as determined by small-angle X-ray scattering. Biochemistry 14, 1326-1333. POLJAK, R. J. (1973) X-ray crystallographic studies of immunoglobulins. In Contemporary Topics in Molecular Immunology, Vol. 2 (R. A. REISEELD and W. J. MANDY, eds.) pp. 1-26, Plenum Press, New York. POLJAK, R. J., AMZEL, L. M., AVEY, H. P., BECKA, L. N. and NISONOFF, A. (19721 Structure of Fab' New at 6 A resolution. Nature New Biol. 235, 137-140. POLJAK, R. J., AMZEL, L. M., AVEY, U. P,, CHEN. B. L., PHIZACKERLEY, R. P. and SALE, F. (1973) Three-dimensional structure of the Fab' fragment of a human immunoglobulin at 2.8-A resolution. Proc. Nam. Acad. Sci. U.S.A. 70, 3305-3310. POLJAK, R. J., AMZEL, L. U., CHEN. B. L., PHIZACKERLEY, R. P. and SAUL, F. (1974) The three-dimensional structure of the Fab' fragment of a human myeloma immunoglobulin at 2.0-A rcsolution. Proe. Natn. Acad, Sci. U.S.A. 71. 3440-3444. POLJAK, R. J., AMZEL, L. M., CHEN, B. L., PHIZACKERLEY, R. P. and SAUL, F. {1975) Structural basis for the association of heavy and light chains and the relation of subgroups to the conformation of the active site of immunoglobulins. Immunogenetics 2, 393-394. POLJAK, R. J.. AMZEL, L. M., CHEN, B. L., PHIZAf'KERLEY, R. P. and SAUL, F. (1975) Structure and Specificity of Antibody Molecules, Phil. Tran,s. R. Soe. Lond. B. 272.43 51. PORTER, R. R. (1959) The hydrolysis of rabbit 7-globulin and antibodies with crystalline papain. Biochem. J. 73, 119-126. POTTER, M. (1973) The developmental history of the neoplastic plasma cell in mice: A brief review of recent developments. Semin. Hematol. 10, 19-32, POULSEN, K., FRASER, K. J. and HABER, E. (1972) An active derivative of rabbit antibody light chain composed of the constant and the variable domains held together only by a native disulfide bond. Proe. Natn. Aead. Sci. U.S.A. 69, 2495 2499. PRESS, E. U. and HOGG, N. M. (1970) The amino acid sequences of the Fd fragments of two human 71 heavy chains. Biochem. J. 117, 641 660.

Three-dimensional structure of immunoglobulins

93

PREUD'HOMME,J. L., BIRSHTEIN,B. K. and SCBARFF, M. D. (1975) Variants of a mouse myeloma celF line that synthesize immunoglobulin heavy chains having an altered serotype. Proc. Natn. Acad. Sci. U.S.A. 72, 1427-1430. PUTNAM, F. W., SHIMIZU, A., PAUL, C., SHINODA, T. and KOHLER, H. (1971) The amino acid sequence of human macroglobulins. Ann. N.Y. Acad. Sci. 190, 83-102. PUTNAM, F. W., TITANI, K., WIKLER,M. and SHINODA,T. (1967) Structure and evolution of kappa and lambda light chains. Cold Spring Hath. Syrup. Quant. Biol. 32, 9-30. RAY, A. and CERRA,J. J. (1972) Localization of affinity-labeled residues in the primary structure of anti-dinitrophenyl antibody raised in strain 13 guinea pigs. Biochemistry 11, 3647-3657. RICHARDS, F. F., KONIGSBERG,W. H., ROSENSTEIN,R. W. and VARGA, J. M. (1975) On the specificity of antibodies. Science 187, 130-137. RICHTER, R., NUHN, P., AMBROSIUS,H,, ZAGYANSKI,Y. A.) TUMERMAN,L. A. and NEZLIN,R. S. (1972) Restricted flexibility of carp 15S immunoglobulin molecules as revealed by fluorescence polarisation. FEBS Lett. 27, 184-186. Rossl, G. and NISONOFF,A. (1968) Crystallization of fragment Fab of human IgG myeloma proteins. Biochem. Biophys. Res. Commun. 31, 914-918. RossI, G., CHOI, T. K. and NISONOFE,A. (1969) Crystals of fragment Fab': Preparation from pepsin digests of human IgG Myeloma proteins. Nature 223, 837-838. ROSSMANN,M. G. and BLOW,D. M. (1962) The detection of sub-units within the crystallographic asymmetric unit. Acta cryst. 15, 24-31. SAGE,H. J., DEUTSCH,H. F., FASMAN,G. and LEVINE,L. (1964) The serological specificity of the poly-alanine immune system. Immunochemistry 1, 133-144. SARMA, V. R., SILVERTON,E. W., DAVIES, D. R. and TERRY, W. D. (1971) The three-dimensional structure at 6/~ resolution of a human ~G1 immunoglobulin molecule. J. Biol. Chem. 246, 3753-3759. SCHECHTER, I. (1971) Mapping the combining sites of antibodies specific to polyalanine chains. Ann. N.Y. Acad. Sci. 190, 394-418. SCUIEFER,M., GmLrNG, R. U, ELy, K. R. and EDMUNDSON,A. B. (1973) Structure of a lambda-type Bence-Jones protein at 3.5-A resolution. Biochemistry 12, 4620-4631. SEGAL, D. M., PADLAN,E. A., COHEN, G. H., RUOmOEE,S., POTTER, M. and DAVIES,D. R. (1974) The three-dimensional structure of a phosphorylcholine-binding mouse immunoglobulin Fab and the nature of the antigen binding site. Proc. Natn. Acad. Sci. U.S.A. 71, 4298--4302. SELA, M. (1969) Antigenicity: Some molecular aspects. Science 166, 1365-1374. SELIG~ANN, M. and BROUEi",J, C. (1973) Antibody activity of human myeloma globulins, Semin, Hematol. 10, 163-177. SHELTON, E. and SMITH,M. (1970} The ultrastructure of carp (Cyprinus carpio) immunoglobulin: A tetrameric macroglobulin. J. Molec. Biol. 54, 615-617. SrNGER, S. J., MARTIN, N. and THORPE, N. O. (1971) Affinity labeling of the active sites of antibodies and myeloma proteins. Ann. N,Y. Acad. Sci. 190, 342-351. SrNGER, S. J., SLOBrN, L. L, THORPE, N. O. and FENTON, J. W. (1967) On the structure of antibody active sites. Cold Spring Hath. Symp. Quant. Biol. 32, 99-109. SOLOMON, A. and MCLAUGHLIN,C. L. (1969) Bence-Jones proteins and light chains of immunoglobulins. J. Biol. Chem. 244, 3395-3404. STANWORTH,D. R. (1971) Immunoglobulin E (Reagin) and allergy. Nature 233, 310-316. STERNER,L. A. and PORTER,R. R. (1967) The interchain disulfide bonds of a human pathological immunoglobulin. Biochemistry 6, 3957-3970. STEVENSON, G. T. and MOLE, L. E. (1974) The specificity of chain interactions among immunoglobulins. Combination of 7 chains with ~ chains of the same subgroup as in the parent immunoglobulin G. Biochem. J. 139, 369-374. TURNER, M. W. and BENNICH, H. (1968) Subfragments from the Fc fragment of human immunoglobulin G. Isolation and physicochemical characterization. Biochem. J. 107, 171-178. VALENTINE, R. C. and GREEN, N. M. (1967) Electron microscopy of an antibody-hapten complex, d. Mole. Biol. 27, 615-617. VARGA, J. M., LANDE, S. and RICHARDS,F. F. (1974) Immunoglobulins with multiple binding functions. J. Immunol. 112, 1565-1570. WEIGERT, M. G., CESARI, I. M., YONKOVlCE,S. J. and COHN, M. (1970) Variability in the lambda light chain sequences of mouse antibody. Nature 228, 1045-1047. WELSCHER, H. D. (1969) Correlations between amino acid sequence and conformation of immunoglobulin light chains. Int. J. Protein Res. 1, 253-265. WOLFENSTEIN-TODEL,C., MIHAESCO,E. and FRANGIONE,B. (1974) "Alpha chain disease" protein def: Internal deletion of a human immunoglobulin A t heavy chain. Proc. Natn. Acad. Sci. U.S.A. 71, 974-978. Wu, T. T. and KABAT, E. A. (1970) An analysis of the sequences of the variable regions of Bence-Jones proteins and myeloma light chains and their implications for antibody complementarity. J. Exp. Med. 132, 211-250. YGUERAaIDE,J., EPSTEIN,H. F. and STRYER,U (1970) Segmental flexibility in an antibody molecule. J.-Molec. Biol. 51, 573-590. ZAGYANSKY,YU. A. (1975) Phylogenesis of the general structure of immunoglobulins, Archs Biochem. Biophys. 166, 371-381.