THE STRUCTURE OF ANTIBODIES AND THE NATURE OF SEROLOGICAL REACTIONS

By Linus Pauling

1942

The most striking and characteristic property of biological substances is the specificity of activity which they show - the power to combine selectivity with or to influence the behavior of one substance, rejecting others with a precision and certainty seen in few physical and chemical phenomena. This specificity of activity is shown by genes in their ability to reproduce themselves, usually unchanged, and to produce the enzymes or other products through which they determine characters; also by enzymes, which select from a mixture the molecules upon which they exert their catalytic action, by hormones, by therapeutic agents, and, in an especially striking way, by antibodies. A complete and reliable understanding of the physicochemical basis of biological specificity would bring us much nearer to the solution of the great fundamental problem of biology, that of the nature of life.

My own interest in biological problems was formed first by a study of hemoglobin made a dozen years ago, which led to the discovery that hemoglobin itself (ferrohemoglobin) is paramagnetic - is attracted to a strong magnetic field - whereas oxyhemoglobin and carbonmonoxyhemoglobin are diamagnetic - are repelled from a magnetic field. It was found that the magnetic properties of hemoglobin provide a simple physicochemical method of measuring equilibria and rates of reaction involving hemoglobin and its derivatives, and the magnetic technique has since been found useful also in study of cytochrome c and other iron-containing proteins. Dr. Alfred Mirsky and I developed a theory of the denaturation of proteins,¹ based upon the concepts that in a native protein the polypeptide chains are coiled together into a definite structure with a configuration which determines the specific properties of the protein, and that heat, alkali, urea, and other denaturing agents and conditions may cause the configuration of the polypeptide chain or chains to be altered, without necessarily breaking any peptide bonds or otherwise changing covalent-bond structure of the molecules. We pointed out that some protein molecules in the native state might have the configuration of coiled polypeptide chains which is the most stable of the configurations accessible to the chains, and that these proteins might be capable of undergoing reversible denaturation, with the chains coiling back into the stable configuration characteristic of the native protein as the denaturing agent or condition is slowly removed. Other native proteins, however, might be built originally into a configuration which is not the most stable of those accessible to the polypeptide chains; these proteins, when denatured by uncoiling their chains, would not, on removal of the denaturing agent, settle into the original native configuration but instead into the postulated more stable configuration; accordingly, a protein of this sort would not undergo reversible denaturation, regenerating the original native protein, although it might be denatured and then renatured to give a well-defined crystallizable protein, with properties different from those of the native protein.

When, in 1936, I became interested in the problem of the structure of antibodies, as the result of conversations with Dr. Karl Landsteiner, I found that the complex and at first confusing reported phenomena of immunology could be clarified and brought into order by a theory of the structure of antibodies based upon the idea of the folding of polypeptide chains into the most stable of the accessible configurations.²The theory of the structure and process of formation of antibodies developed in this way involved the acceptance of the suggestion that antibody and antigen have complementary structures, originally made by Breinl and Haurowitz, J. Alexander, and Stuart Mudd.³ The picture of the serological precipitate as a framework (lattice) of multivalent antibody and multivalent antigen, developed Marrack and by Heidelberger, also seemed to be so reasonable and so in accord with most of the observational data as to require its acceptance⁴.

This theory of the structure of antibodies and the process of antibody formation depends upon the assumption that the antibody precursor is a polypeptide chain of such a nature as to be able to fold into a large number of alternative configurations, which have nearly equal free energy, and hence nearly equal stability. In the absence of an antigen in the region in which the antibody precursor is being formed, the polypeptide chain will fold into one of the most stable of the configurations acceptable to it, producing a molecule of normal gamma globulin. However, if an antigen molecule is present, it must be considered as part of the environment acting upon the folding of the polypeptide chain, and the most stable configurations of those accessible to the folding chain now become different from those in the absence of the antigen: they are now configurations which take the greatest advantage of the opportunity of interaction with the antigen, of such a nature as to stabilize the system - that is, of the opportunity of assuming such a structure as to lead to attraction between the forming antibody and the antigen, and hence to the formation of an antigen-antibody bond. A structure of this sort would be one in which the surface atoms of the antibody molecule are able to come into the closest possible proximity to the surface atoms of the antigen molecule. This could be achieved in case that the folding antibody molecule were to mold itself over a portion of the surface of the antigen molecule, reproducing the configuration of the antigen in the same way as a coin does its die. The principal forces of attraction which are operative are the general van der Waals forces (electronic dispersion forces), the forces described as hydrogen-bond forces, and the electrostatic forces between positively charged and negatively charged ionized groups. A very high degree of specificity can be obtained if the surface area over which the complementariness in structure is exercised is great enough to include a good number of interacting structural units.

The assumption that antibodies are bivalent, or have still greater valence - that is, that each antibody molecule has two or more surface regions capable of combining specifically with the homologous antigen - is necessary in case that the framework theory of the serological precipitate is accepted. The general evidence, of varied nature, for the framework theory is summarized by Marrack and Heidelberger, is strong, but not complete. Further evidence was obtained by studies made by my collaborators (Professor Dan H. Campbell, Dr. David Pressman, Dr. Carol Ikeda, Dr. M. Ikawa, Dr. David H. Brown, Mr. A. L. Grossberg, Dr. Stanley M. Swingle, Dr. John T. Maynard) and myself by the study of the precipitation of antibodies with simple chemical substances of known structure. It was discovered by Landsteiner and van der Scheer⁵ that a precipitate is formed when a dye made by coupling two or more haptenic groups with resorcinol or tyrosine is added to an antiserum obtained by injecting an animal with an azoprotein containing the same haptenic group. We investigated the interaction of many substances containing the paraazobenzonearsonic acid group and anti-para-azobenzenearsonic acid serum, and found that all dyes containing two or more of these haptenic groups were able to form a precipitate with the serum, whereas those containing only one haptenic group were not. It was also found that under certain conditions the ratio of the number of molecules of antibody in the precipitate were equal to unity, and if the assumption is made that each of the two haptenic groups is operative in bond formation that bivalence of the antibody is proved. However, it was also found that the same molecular ratio of unity held for the precipitate formed by trihaptenic and tetrahaptenic dyes. This confusing result can be explained by the reasonable assumption that the steric interaction of the large antibody molecules about a small dye molecule is so great as to prevent more than two antibodies usually from combining with the haptenic groups of the same dye molecule, the steric repulsion thus effectively limiting the valence of the polyhaptenic substance to two, and the data then indicates bivalence of the antibody. A determinative experiment has also been carried out, involving the simultaneous precipitation of antibodies from two different antisera by a single substance, which is incapable of precipitating either of the antisera alone. The substance is a dye containing one haptenic group of each of two different kinds, and the two specific antisera which when mixed are simultaneously precipitated by the dye are those made by injecting separate rabbits with azoproteins containing, respectively, the two haptenic groups. This experiment provides very strong evidence for the framework theory of serological precipitation.⁶

The precipitation of antibody and antigen is closely similar to that of, say, silver ion by cyanide ion, and the similarity extends also to re-solution of the precipitate in an excess of one of the reactants (cyanide ion or antigen). The cyanide precipitate dissolves in an excess of cyanide ion because of the formation of the silver cyanide complex Ag(CN)₂ ^-, and similarly the antibody-antigen precipitate dissolves in an excess of antigen because the antigen molecules combine with both (or all) of the combining groups of the antibody, saturating them and forming a soluble complex. It would be of interest to physical chemists to investigate this reaction quantitatively, and to find whether the same simple laws of chemical equilibrium apply as to the silver cyanide precipitation and re-solution. It is found that these simple laws do not apply, but that instead the behavior of antisera and antigens is that which would be expected if the antiserum contained antibody molecules of many different kinds, with their combining groups differing by several kilocalories per mole in free energy of combination with haptens, corresponding to a several hundred-fold or thousand-fold range in equilibrium constants for the combination reaction. The data indicates clearly that natural antibodies are very heterogeneous. This is, of course to be expected form the theory of antibody production described above.

The nature of specific forces operative between antigen and antibody has been investigated especially by the quantitative study of the phenomenon of hapten inhibition. A monohaptenic substance is able to combine with antibody, but not to form a precipitate. Through combination with the antibody, however, the formation of a precipitate by a polyhaptenic substance can be inhibited. This phenomenon of hapten inhibition was discovered by Landsteiner. Quantitative studies of the inhibiting power of different haptens of known structure have been made, and subjected to physicochemical interpretation by the use of a theory of heterogeneous antibody. This theory is based upon the assumption that the distribution function for the heterogeneous antibody is an error function in the free energy of interaction of antibody and haptenic group. The assumption of an error function in the free energy (that is, in the logarithm of the equilibrium constant) is seen to be reasonable one by the argument that the total free energy of combination of the combining group of an antibody with the hapten may depend upon several structural features, which may be present or absent independently of one another; if the number of structural features were large, there would result an error function distribution in the free energy of interaction with antibody, to which they make their independent additive contributions.⁷

It has been found that the hapten inhibition constants of different haptens depend very strongly upon the degree of conformity in shape of the haptens to the immunizing haptenic group. The requirement for similarity in shape is such that the conclusion can be drawn that the antibody reflects or reproduces, in a negative way, the shape of the haptenic group of the immunizing antigen to within about 1 Å.⁸ Moreover, it has been found for a series of related haptens containing substituent groups in the position para to the charged group (the arsonic acid group) the average equilibrium constants for combination with antibody depend significantly upon the optical polarizability of the para group, in the way indicated by the London theory of electronic dispersion forces, the magnitude of the effect being such as to indicate approximation of the antibody to within 1 Å of the haptenic group. A third test has been made, that of the contribution of an electrical charge to the antibody-antigen forces. ⁹ This test involves comparison of a hapten containing the trimethylammonium ion group and one containing the trimethylammonium ion group and one containing the uncharged tertiary butyl group, and the determination of their hapten inhibition constants. The difference in free energy of combination indicated by these hapten constants can be expressed in terms of distance between the positive charge of the charged haptenic group and a complementary negative charge of the antibody, with the use of effective dielectric constants as indicated by the investigations of Schwarzenbach.10 The distance so found is 7.0 Å. Since the radius of the phemyltrimethylammonium ion is 3.5 Å, and the minimum distance to which a negative charge could approach the surface of an antibody is 1.4 Å (the radius of an oxygen atom), the value of 7.0 Å shows that the complementary negative charge of the antibody is within 2.1 Å of the minimum possible distance from the positive charge of the immunizing haptenic group. This evidence also accordingly supports the thesis that the forces of specific attraction between antibody and antigen depend upon the very close approximation of the antigen and antibody molecules.

For a long time there remained unrecognized a striking analogy between the highly specific phenomenon of serological interaction and another highly specific phenomenon of the chemistry of simpler substances; namely, the phenomenon of crystallization. The process of the crystallization of a substance from a complex solution is in general highly specific - often a very pure substance can be grown as crystals from a complex mixture, as is shown by the example of the formation of pure crystals for cream of tartar from grape jelly. It is clear that the specificity of crystallization is the result of the same interatomic and intermolecular forces and the same striving toward complementariness that are responsible for the specificity of antibodies. A molecular crystal is stable because all of the molecules pile themselves into a configuration such that each molecule is surrounded as closely as possible by other molecules, in such a way as to make the forces of attraction of the molecules within the crystal as great as possible. This result is achieved if the cavity in the crystal into which each molecule fits conforms as closely as possible to the shape of the molecule, and if also there is a complementariness in structure, with respect to hydrogen bond formation and ionic interactions, between the molecule and the surrounding molecules. Other molecules, with different shape and structure, would not fit into this cavity nearly so well, and in consequence other molecules would not in general be incorporated in the growing crystal. Only if the other molecules were very similar to the molecules of the crystal would deviation from specificity occur, leading to the formation of solid solutions. It is well known, for example, that organic compounds containing methyl groups tend to form solid solutions with those containing chlorine atoms substituted in the corresponding positions. The replacement of methyl groups by chlorine atoms similarly leads to biological cross-reactivity - a hapten containing a methyl group interacts nearly as strongly with the serum homologous to a hapten containing a chlorine group in this position as does the chlorine-substituted hapten itself.

I shall now show some lantern slides illustrating the points raised in the discussion that I have just given.

The first slide shows my picture of the process of formation of an antibody molecule. I think that the antibody precursor, which is also the precursor of normal gamma globulin, may consists of a single long polypeptide chain, containing more than a thousand amino acid residues. I assume that the two ends of this polypeptide chain are first extruded from the region of synthesis of the chain, and that they coiled up into the configuration of greatest stability. In the absence of an antigen, this configuration might be any one of a large number of configurations with almost equal energy and equal stability. In the presence of an antigen molecule, however, the most stable configurations accessible to the chain end would be different; they would be, indeed, those configurations which are stabilized the most by interaction with the surface atoms of the antigen molecule, and hence just those configurations which are most strongly complementary in structure to a portion of the surface of the antigen molecule. There is hence produced by this automatic process a configuration of the chain end of the molecule of such a nature as to lead to the formation of a chemical bond between the combining group of the antibody, formed in this way, and the antigen molecule. This bond, strong enough to offer effective resistance to the dissociating influence of thermal agitation, is not an ordinary chemical bond, involving the sharing of electrons between two atoms, but is the result of the cooperation of a large number of small interatomic forces. The automatic process of assumption of configurations of greatest stability by the chain end of the globulin molecule thus produces a structure in which a large number of atoms of this folded polypeptide chain are able to get into contact with atoms in the surface of the antigen, and to enter into electronic van der Waals attraction with them; also there will be positive charges in the antibody close by negative electrical charges in this antigen molecule, and hydrogen-bond-forming groups in the antibody adjacent to the complementary hydrogen-bond-forming groups in the antigen.

When the two combining regions have been formed, attached to the surface of the antigen molecule, by the folding of the two ends of the polypeptide chain, we may expect that sooner or later, as the result of thermal agitation, one of the combining regions thus formed will dissociate away from the antigen molecule, and that the intermediate portion of the polypeptide chain will then be able to coil up into its normal configuration, producing a completed antibody molecule, still attached at one end to the antigen molecule. Later on this antibody molecule may be dissociated away from the antigen molecule, under the influence of thermal agitation, producing a free molecule of antibody in the bloodstream. The antigen molecules can continue to act as the templates for the formation of additional antibody molecules, until as many as a thousand antibody molecules have been formed for each antigen molecule. This process may continue until the antigen molecule is completely covered by antibody molecules that are so firmly attached that they do not dissociate away, or until the antigen molecule has been destroyed by some metabolic process.

The next slide shows some antigen and antibody molecules combined together in a serological precipitate. We see the combining region of an antibody here, in contact with a portion of the surface of an antigen molecule, to which it is complementary in structure. Each antigen molecule can attach to itself as many antibody molecules as can be accommodated on its surface; its effective valence may accordingly be as great as 10 or 12, or even much larger for very large protein antigens. The antibody molecules are indicated as having two combining groups, and thus being effectively bivalent. A framework is in this way built up from the bivalent antibody molecules and the multivalent antigen molecules - the formation of a precipitate results from the same forces as those that cause a single antigen molecule and a single antibody molecule to combine with one another. The combination of these molecules occurs into larger and larger complexes, which ultimately become large enough to be seen, forming particles of the precipitate.

The next slide shows the formulas as formed of the simple substances of known structure that my colleagues and I prepared, in order to test the framework theory. We found that all of the substances in which there were two or more haptenic groups - paraazobenzenearsonic acid groups - in the same molecule formed precipitates with the antisera homologous to the paraazobenzenearsonic acid groups, but that none of the substances containing one haptenic group per molecule formed these precipitates. Instead, the monohaptenic substances combined with the antibody molecules to form soluble complexes, and thus inhibited their precipitation with polyhaptenic antigens.

The next slide shows the structures of some precipitating simple antigens, for which analyses were made of the precipitates. These analyses showed for bihaptenic substances that the ratio of antibody molecules to antigen molecules in the precipitate was approximately 1:1. If it is assumed that the bihaptenic substances are bivalent, using both of their bihaptenic groups for combination with antibody molecules then the 1:1 ratio in the precipitate requires that the antibody molecules also be bivalent. This result is, however, rendered less convincing by the fact that the trihaptenic and tetrahaptenic substances also gave the same 1:1 ratio. The conclusion might hence be drawn form the analytical results for those precipitates that the antibodies are tervalent or quadrivalent. The explanation of the paradoxical results is given in the next slide. Here we see a trihaptenic precipitating antigen attached ot the combining region of the two antibody molecules. The radius of curvature of the combining region of an antibody molecule is indicated as 30 Å, which is not an unreasonable value. It is seen that the seric interaction of the two combined antibody molecules with a third antibody molecule would keep it from approaching closely enough to the third haptenic group to form an effective bond. We believe that this steric inhibition effect is the explanation of our observation of a 1:1 molecular ratio for precipitates with trihaptenic and tetrahaptenic molecules.

Nevertheless, there might be some doubt about the interpretation of these results. We accordingly carried out another experiment, in which a substance was synthesized which had the power of forming a serological precipitate with the serum from two rabbits, differently immunized, but not with the serum from any one rabbit alone. This is a very striking experiment, which can be explained very simply on the bases of bivalent antibody and the framework theory, but not on any other basis. The explanation of the experiment is shown in the next slide. This is the substance used to form the precipitate with the two different antisera. It is the substance RX, with one haptenic group paraazobenzenearsonic acid, and the other paraazobenzoic acid. These two haptenic groups, though very closely similar in structure, are enough different to prevent cross reaction from taking place; that is, antibodies complementary to one do not combine significantly with the other haptenic group. The next slide shows the formation of a precipitate with the two antisera, one an anti-R serum and the other an anti-X serum, and the substance RX. A molecule of anti-R could combine with two molecules RX, but this soluble complex could then not undergo further combination, if only anti-R serum is present, because of the haptenic group X would not combine with anti-R molecules. If, however, anti-X molecules are also present, then these cannot add onto the X group, and the process can continue until a framework is built up. Hence the precipitate appears only when the substance RX is mixed with the two different antisera, anti-R serum and anti-X serum. I believe that this experiment provides very strong evidence for the framework theory and for the bivalence of antibodies.

The next slide illustrates a recent application that we have made of serological precipitates to the determination of the configuration of haptenic groups and of certain molecules in solution. An antiserum was made by inoculating rabbits with an azoprotein made from para-aminosuccinanilic acid. This azoprotein contains the haptenic groups shown in the slide, the para-azosuccinanilate ion group. It should be pointed out that the succinic acid part of this haptenic group is shown with the cis configuration, whereas it might be expected that this group would have the trans configuration. The single bond between the two methylene carbon atoms is, of course, susceptible to rather free rotation, and accordingly it is not possible to predict reliably whether one configuration or another, involving rotation around this bond, will exist for a particular substance. Crystals of the succinic acid itself show that the molecules have the trans configuration around the single bond. We have found, however, that this haptenic group has the coiled up or cis configuration around the single bond. The evidence for this is the following. The antiserum is strongly inhibited from precipitating with a homologous antigen in the presence of the maleanilate ion, shown here on the slides, whereas the fumaranilate ion, shown here, has only a very small power of precipitation. This means that the maleanilate ion, which combines 100 times more strongly with the antibody than the fumaranilate ion does, must be more closely complementary in structure to the combining region of the antibody molecule than is the fumaranilate ion. That is, the combining region of the antibody thus has such a shape as to permit the maleanilate ion to fit into it very snugly; it is accordingly complementary in structure to the cis configuration about the double bonds, as shown in the maleanilate ion. Since the antibody was made in response to injection of azoprotein containing the azosuccinaniltae haptenic group, we conclude that the succinanilate haptenic group must also have the cis configuration. We have found it possible to determine whether various substituted succinic acids exist in aqueous solution at physiological pH's mainly in the cis configuration or in the trans configuration by determining the combining powers of these substances with the antibodies against para-azosuccinanilate ion. It has turned out, for example, that the succinate ion itself has the trans configuration, whereas its monoalkyl esters and monoamides have the cis configuration.

The last slide shows the combining region of an antibody homologous to the para-azosuccinanilate ion, and the succinanilate ion haptenic group in place within it. It is seen that the closeness of fit, of about 1 Å , means that the atoms of the antibody and the atoms of the antigen are in essentially atomic contact. Moreover, there is shown here a positive charge in the antibody in close proximity to the negative charge of the carboxyl group of the hapten, and a hydrogen-bond-forming group of the antibody in combination with the carbonyl group of the hapten. Experimental evidence has been obtained for all of these structural features.

There is a highly specific phenomenon of the chemistry of simpler substances that is closely analogous in its nature and its cause to the highly specific phenomenon of serological interaction; namely, the phenomenon of crystallization. Chemists are accustomed to using the process of crystallization as a method of purification: a crystal growing in a complex mixture of molecules is able to select from the mixture just the molecules of one kind, rejecting all others. Thus pure crystals of sugar may deposit from a jam in which there are molecules of thousands of different substances. The specificity of crystallization is the result of the same striving toward complementariness and the operation of the same interatomic and intermolecular forces that are responsible for the specificity of antibodies. A molecular crystal has the structure that gives it the greatest stability, which would result from the maximum amount of attraction for each molecule in the crystal and the surrounding molecules. Each molecule in the crystal is then in a cavity that conforms in shape to the shape of the molecule itself. The molecule may be described as complementary in structure to the remainder of the crystal, and other molecules, with different shape and structure, would not fit into this cavity nearly so well, and in general would not be incorporated in the growing crystal. We may hence say that life has borrowed from inanimate processes the same basic mechanism used in producing these striking structures that are crystals, with their beautiful plane faces, their unfailing constant interfacial angles, and their wonderfully complex geometrical forms.

I believe that the same mechanism, dependent on a detailed complementariness in molecular structure, is responsible for all biological specificity. I think that enzymes are molecules that are complementary in structure to the activated complexes of the reactions that they catalyze, that is, to the molecular configuration that tis intermediate between the reacting substances and the products of reaction for these catalyzed processes. The attraction of the enzyme molecule for the activated complex would thus lead to a decrease in its energy, and hence to a decrease in the energy of activation of the reaction, and to an increase in the rate of the reaction. Although convincing evidence is not yet at hand, I believe it will be found that the highly specific powers of self-duplication shown by genes and viruses are due to the same intermolecular forces, dependent atomic contact, and the same processes of replica-formation through complementariness in structure as are operative in the formation of antibodies under the influence of an antigen. I believe that it is molecular size and shape, of the atomic scale, that are of primary importance in these phenomena, rather than the ordinary chemical properties of the substances, involving their power of entering into reactions in which ordinary chemical bonds are broken and formed.

Even though the general picture of some important biological processes is becoming clear, our present knowledge of the detailed structure of the complex substances of biological importance is vague. We may expect that as more precise information about the structure of these molecules is obtained in the future a more penetrating understanding of biological reactions will develop, and that this understanding will lead to great progress in the fields of biology and medicine.

¹ A. E. Mirsky and L. Pauling, Proc. Nat. Acad. Sci., XXII, 439 (1936).

² L. Pauling, J. Am. Chem. Soc., LXII, 2643 (1940).

³ F. Breinl and F. Haurowitz, Zeit. Physiol. Chem., CXCII, 45 (1930); J. Alexander, J. Protoplasma, XIV, 296 (1931); S. Mudd, J. Immunol., XXIII, 423 (1932).

⁴ J.R. Marrack, "The Chemistry of Antigens and Antibodies," Report 230 of the Medical Research Council, H.M. Stationery Office, London, 1934; 2nd ed., 1938; M. Heidelberger et al., J. Exp. Med., LXI, 563 (1935); Chem. rev., XXIV, 323 (1939).

⁵ K. Landsteiner and J. van der Scheer, Proc. Soc. Expt. Biol. Med., XXIX, 747 (1932); J. Exp. Med., LVI, 399 (1932); LVII, 633 (1933); LXVII, 79 (1938).

⁶ L. Pauling, D. Pressman, and D. H. Campbell, J. Am. Chem. Soc., LXVI, 330 (1944).

⁷ L. Pauling, D. Pressman, and A.L. Grossberg, J. Am. Chem. Soc., LXVI, 784 (1944), and later papers.

⁸ L. Pauling and D. Pressman, J. Am. Chem. Soc., LXVII, 1003 (1945).

⁹ D. Pressman. A.L. Grossberg, L.H. Pence,, and L. Pauling, J. Am. Chem. Soc., LXVIII, 250 (1946).

¹⁰ G. Schwarzenbach, Z. Physik. Chem., A CLXXVI, 133 (1936).