Contributiof trop the Gates and Crellini Laboratories of Chemistry,
Calfformia Institute of Technoligy, No. 783

nel
A Theory of the Structure and Process of Formation JL, LT NG¢O
of Antibodies*

By Linus Pauling

 

* Some of the material in this paper was presented on April 17th, 1950, at
The Rockefeller Inetitute for Medical Research and on April 23rd, 1940, at
the meeting of the National Academy of Selences, Washington, D.C,

 

I. Int roduction

During the past four years I have been making an effort to understand and |
interpret ssrological vhenomena in terms of molecular atructure and molecular inter
actions, The field of immunology is ao extensive and the experimental observations
are so complex (and occasionally contradictory) that no one has found it possible to
induce a theory of the structure of antibodies from the observational material,

As an alternative method of attack we may pronound and attermt to answer the
following questions: What is the aimplest structure which can be sugrested, on the
basie of the extensive infornation now available about intramolecular and inter-

molecular forces’, for a molecule with the proverties observed for antibodies, and

 

See, for agwunmary of this information, L. Pauling, "The Nature of the
Chemical Bond Ԥnd the Structure of Molecules and Crystals," Cornell University
Precs, Ithaca, New York, Second Hditic.., 190.

 

what is the simplest reasonable process of formation of such a molecule?
Proceeding in this way, I have developed a detailed theory of the atructure and
process of formation of antibodies and the nature of serolorical reactions which
is nore definite and more widely avplicabdle than earlier theories, and which is
compatible with our present knowledre of the structure and properties of simmle
molecules ag well as with mest of the direct empirical information about anti-

vodies, This theory is described and discussed below,
Il. The Proposed Theory of the
Structure and Process of Formation of Antibodies

 

When an antigen is injected into an animal some of its molecules are

captured and held in the region of antibody production®, An antibody to this

2 There is some evidence that this 1s the cells of the reticulo-endothelial
syetem: see Florence R, Sabin, J, Exp. Med. 70, 67 (1939), and references
quoted by her.

 

 

antigen is @ molecule with a configuration which is complementary to that of a

portion of the antiren molecule” . Thie complementariness gives rise to specific

x The idea of complementary structures for antibody and antigen was wu
b

by (a) ¥, Breind and F. Haurowitz, Z. siol, Chem. 192, 45 (1930);

Stuart Mudd, J, Immunol. 23, 423 C153 oP and has come to de rather generally
accepted, There if some intimation of it in the early work of Zhrlich and of
Bordet.

 

 

forces of appreciable strength between the antibody molecule and the antigen

molecule; we may describe this as a bond between the two molecules, I assume,

with Marrack, Heidelberrer, and other investigators’, that the precipitate obtained

 

* J. 2, Marrack, "The Chemiatry of Antigens and Antibodies", Report Ho. 230 of
the Hedical Research Council, His Majesty's Stationery Office, London, 2338;
M. Heidelberger, Cham, Rev. 24, 323 (1939), and earlier pavers,

 

in the precipitin reaction ia a framework, and that to be effective in forming

 

5 The franework 1s sometimes called a "lattice by immunochemists; the use of
thia word in immunology 1s to be discouraged because of the implication of
regularity associated with it through its application in crystallography.

 

the framework an antibody molecule must have two or more distinct regions with
surface configuration complementary to that of the antigen. The rule of pareimony
(the use of the minimum effort to achieve the result) suggests that there are only
two such regions; that is, that the antibody molecules are at the most divalent.
The proposed theory is based on this rearconable assumption. It would, of course,

de posaible to expand the theory in such a way as to provide a mechaniem for the
formation of antibody molecules with valence higher than two; but this would
make the theory considerably more complex, and it is likely that antibodies with
valence higher than two ocour only rarely, if at all,

Antibodies are similar in amino-acid composition to one or another of the
fractions of serum globulin of the animal producing the serum, It is known
that there exist antibodies of different classes, with different molecular
weighte—-the molecular weights of rabbit antibody and of monkey antibody (to
pheumococeus polysaccharide) are about 157,000, whereas those of piz, cow, and

horse antibodies are about 930 ,000°, The following discussion is for antibodies

S 3, A. Kabat and K. 0, Pedersen, Science 81. 372 (1938); EB. A. Kabat,
J. Exp, Med. 69, 103 (1939).

 

 

with molecular weight about 160,000, and similar in constitution to the \

fraction of serum globulin’, the changes to be made to cause it to apply to

TT’, Tieelius, Biochem, J. a: 1464 (1937); Trans. Faraday Soc, 33, 52h
(1937); T. Swedberg, Ind. Sng, o+ BQ, 113 (i938),

antibodies of other classes are obvious.

 

 

The effect of an antigen in determining the structure of an antibody
molecule might involve the ordering of the amino-acid residues in the polypeptide
chains in a way different from that in the normal flotulin, as sugrested by Brein]
and Haurovite™ and Mudd”, I assume, however, that thie is not fo, but that all
antibody molecules contain the same polypeptide chains &8 normal globulin,
end differ from normal globulin only in the configuration of the chain; that is,
in the way that the chain is coiled in the molecule, There ia at preaant no direct
evidence supporting this assumption, The assumption is made because, although I
have found it impossible to formate in detail a reasonable mechanism whereby
the order of amino-acid residues in the chain would he determined by the antigen,
a simple and reasonable mechanian, described vdelow, can be advanced whereby the
antigen causes the polynentides chain to assume a configuration complementary te
the antigen. The number of configurations accessible to the polypeptide chain

is so great ae to provide an explanation of the ability of an animal to form
antibodies with considerable svecificity for an apparently unlimited number of

different antigens’, without the necessity of invoking also a warlation in the

g See K, Landateiner, “The Specificity of Serologiesl Reactions", Charles
C, Thomas, Baltimore, 2936.

 

 

amino-acid composition or amino-acid order’.

 

a It has been pointed out by A. Rothen and K. Landsteiner, Science 90, 65
(1939), that the possibility of different ways of folding the same polyneptide
chain to obtain different antibodies is worth considering.

 

The Post 8 ass of Formation o tibodies.—-Let us assuns that the
slobulin molecule consists of a single polypeptide chain, containing several
hundrad amino-acid reatdues, and that the order of amino-acid re idnes 16 such
that for the center of the chain one of the accessible configurations is much

more atable than any other, whereas the two end carts of the chain are of such

a nature that there exist for them many configurations with nearly the same energy.
(This voint is discussed in detail in Section IV.) Four steps in our postulated
process of formation of a normal globulin molecule are illustrated on the left
side of Figure 1, At atage I the polypeptide chain has been synthesized, the
amino-acid residues having been marshalled into the proper order, presumably with
the ald of nolynentidases and protein templates, and the two ends of the chain,

A and C, each containing verhaps two hundred residues, have been libernted with
the ~astable extended configuration. (The horizontal line in each draving
separates the rezion, velow the line, in which the polypeptide chain is not able

to change its conficuration from the region, above the line, where this is possible, )
Mach of these choin ende then coils wm into the most stable or one of the most
stable of the accessible configurations (stage II) and 16 tied into this
confiruration by the formation of hydrogen bonds and other weak bonds between

yarte of the chain, The central part B of the chain is then liberated (stage III)
and assumes its stable folded configuration (stage IV) to give the completed
globulin molecule,

There are also indicated in Figure 1 six stages in the process of formation
of an antibody molecule. In stage I there are shown an antigen molecule held at
& place of globulin preduction and a globulin molecule with its two ends A and C
liberated with the extended configuration. At stage II each of the ends has
assuned a stable coiled configuration. These stable configurations A’ and C! are
not, however, identical with those A and C assumed in the absence of the antigen.
The atoms and groups which form the surface of the antiren will attract certain
complementary parte of the globulin chain (a negatively-charged group, for example,
attracting a positively-charged group) and repel other parte; as a result of these
interactions the configurations A' and C’ of the chain ends which are stable in
the presence of the antigen and which are accordingly assumed in the vresence of
the antigen will be such that there is attraction between the coiled globulin
chain ends and the antigen, due to their complementarity in structure. The con-
figuration assumed by the chain end may be any one of a large number, depending
upon which vart of the surface of the antigen happens to exert its influence on
the chain end and how large a region of the surfa,e happens to be covered by it.

When the central part B of the globulin chain is liberated from the place of
ite synthesis (stage III), one of two processes may occur, If the forces of
attraction between the anticen and the portions A!’ and Gt are extrenely strong,
they will remain bonded to the antigen for an indefinite time, and nothing further
of interest will happen. If the forces are somewhat weaker, however, one will
ih time break away--dissociate from the antigen (stace IY). ‘Then the portion B
of the chain will fold up to achieve its normal stable configuration (stage Y),
making a completed antibody molecule. In time thie will dissociate from the

antigen and float away (stage VI). It is possible that an auxiliary mechanism
for freeing the active ends A' and C' from the anticen molecule comes into
operation; this is discussed in Section Yi.

The middle part of the antibody molecule thus nroduced would be like that
of a normal globulin molecule, and the two ends would have ‘eonfigurations more or
less complementary to parte of the surface of the antigen. These two active ends
are effective in different directions, 20 that, after the antibody is completely
formed, only one of them at a time oan grasp a particular antigen molecule.

The antigen molecule, after its desertion by the newly-formed antibody
molecule, may serve as the pattern for another, and continue to serve until ite
surface is covered by very strongly held antibodies or portions of antibodies or
until the concentration of antibodies becomes so great that even with weak forces
operating the antigen is combined with antibodies most of the time (as illustrated
in Pirure 1), or until the antigen molecule is destroyed or escapes from the region
of globulin formation.

11. Some Points of Comparison vith Bxperiment
a. The Heterogeneity of Immune Sera.—-The theory requires that the serum

 

homologous to a given antigen be not homogeneous, but heterogeneous, containing
antibody molecules of greatly varied configurations. Many of the antibody molecules
will be bivalent, with two active ende with configuration complementary to portions
of the surface of an antigen molecule. Great variety. in this complementary
configuration would be expected to result from the accidental approximation to

one or another @urface region, and further variety from variation in position of
the antigen molecule relative to the noint of liberation of the @lobulin chain

end and from aceldental cofling and linking of the chain end before it comes under
the influence of the antigen. Some of the antibody molecules would be univalent,
one of the chain ends having, because of its too great distance from the antigen,
folded into a normal globulin configuration.
These predictions are verified by experimental resulte. It 1s well known
that an immune serum to one antigen will, as a rule, ract with a related
heterologous antigen, and that after exhaustion with the latter there remains
a fraction which will still react with the original antigen. Landsteiner and

van der Scheer™°, using as antigens azoproteins carrying various haptens containing

 

10 K, Landsteiner and J. van der Scheer, J. Exo, Med. 63, 325 (1936).

 

the same active group, have shown that the antiserum for one antigen contains

various fractions aifferging in the strength of their attraction for the haptens.

By the quantitative study of presipitin reactions Heidelberger and Kendali +

Bs M, Heidelberger and F. =. Kendall, J. Exp. Med. 61, 559 (1935); 62, 467,
697 (1935).

 

 

reached the same conclusion, and showed in addition that even after prolonged
inmunization the antiserum studied (anti-eeg albumin) contained much low-grade
antibody, incapable by itaelf of forming a precipitate with the antigen, but with
the property of being carried down in the precipitate formed with a more reactive

fraction’.

 

12 fhe older experimental results bearing on this question 414 not permit a
Clear distinction between antibody fractions differing in being complementary to
different active groups in the antigen and fractions differing in the extent of
their complementarinese to the same group. The experiments are discussed in
Marrack's monograph.

 

0. The Bivalence of Antibodies and the Multivalence of Antigeng.--Our theory
4s based on the idea that the precipitate formed in the precipitin reaction is

 

a network of antibody and antigen molecules in which many or ali of the antibody
molecules grasp two antigen molecules apiece and the antizen molecules are

grasped by several antibody molecules. The direct experimental evidence for this
picture of the precipitate lee been ably discussed by its propounders and supporters,
Marrack and Heidelberger and Kendall, and need not be reviewed here. fo the

atructural chemist it is clear that this picture of the precipitate must be correct,
The great snecificity of antibody-antigen interactions requires that a definite
bond be formed between an antibody molecule and an antigen molecule. If anti-
yodies or anticens were univalent, this would lead to complexes of one antigen
molecule and one or more antibody molecules (or of one antibody molecule and one
or more antigen molecules), and we know from experience with proteins that these
aggregates would in general remain in solution. If both antibody and antigen
are multivalent, however, the complex will grow to an aggregate of indefinite
size, which is the precipitate.

This process is observed directly in the agglutination of cells. On the
addition of an agglutinin to a cell suspension the cells are seen to clum together.
It 1s obvious that the agelutinin molecules which are holding the cells together

are pivalent?--~each has two active ends, with configuration complementary 40 that

 

1 Following Heidelberger and Kendall, I use the terminology of chemical valence
theory in discnuseing the specific mutual attraction of antigen and antibody.

The antibody-antigen “valence bonds" are not, of course, to be confused with
ordinary covalent chemical bonds; they are due inatead to the integrated veak
forces discussed in Sec, IV,

 

of a portion of the surface of the cells; the agglutinin molecules hold the cells
together at their regions of contact, as shown in Figure 2.
asst
It seems probable that all antibodies have this structure--they they are
bivalent, with their two active regions opvositely directed. Heidelberger and
his collaborators and Marrack have emphasized the multivalence of antibodies and

antigens!", bot limitation of the valence of antibodies to the maximum value two

 

i" Professor Heidelberger has informed me that in their quantitative treatment
of data on the precipitin reaction he and Dr, Kendall have found no incompati-
bility with this restriction; in their papers they discussed the ceneral case
of multivalence of antibody as well as of antigen.

 

(ignoring the exceptional case of the attachmont of two or more antigens or

haptens to the same end region of an antibody) has not previously deen made,
The maximum valence of an antigen molecule would be civen by the ratio of
its surface area to the area effectively occupied by one antibody molecule, if
all regions of the antigen surface were active. In the especial case that the
antibody were able to combine only with one group (a hapten, say, with
immunization effected by use of another antigewith the same hapten attached)
the maxismm valence of the antigen would be equal to the number of groups per
molecule.

Cee Antibody-~Antigen Molecular Ratio in Frecipitates.--Our theory provides

an irmediate simple explanation of the observed antibody~anticen molecular ratios
in nrecipitates, Under optimum conditione a precipitate will be formed in which
all the valences of the antibody and antigen molecules are satisfied, An idealized
representation of a portion of such a precinitate is given in Figure 3, The figure
shows a part of a layer with each antigen molecule bonded to six surrounding
antibody molecules; this structure represents the -lue N = 12 for the valence

of the antigen, each antigen molecule being attached also to three antibody
molecules above the layer represented and to three telow. Bach antibedy molecule
is bonded to two antigen molecules, one at each end. An ideal structure of the
antibody-antigen precipitate for E= 12 may be described as having anticen
molecules at the positions corresponding to closest packine, with the twelve
antibody molecules which surround each antigen molecule lying along the lines
connecting it with the twelve nearest antigen neifhbors.

Similar ideal structures can be suggested for other values of the antigen
valence. The antigen molecules might be arranged for N= & at the points of a
dody-centered cuble lattice, and for F = 6 at the points of a aimple cubie
lattice, with antibody molecules along the connecting lines. For X= 4 the
antigen molecules, connected by antibody molemles, might lie at the points
eccupied by carbon atoma in diamond; or two auch frameworks might intervenstrate,

as in the cuprous oxide arrangement (copper and oxygen atoms being replaced by
10

antibody and antigen molecules, respectively ).

It da not to be inferred that the actual precipitates have the regularity
of structure of these ideal arrangements, ‘The nature of the process of antibody
formation, involving the uce of a portion of the antigen surface selected at
random ae the template for the molding of an active end of an antibody molemle,
introduces so much irregularity in the framework that a regular structure
analogous to that of a crystal is probably never formed, The precipitate is
to be commered rather with a clase such aos silica glass, in which ench silicon
etom 48 surrounded tetrahedrally by fou oxrzen ntoms and each oxyeen atom is
Donded te two eilicon atoms, but which lacks further orderliness of arrangement.
Additional disorder is introduced in the precipitate by variation in the effective
yalence of the antigen molecules and by the inclusion of antibody molecules with
only one active end.

The antibody-antizen molecular ratio R of a vrecinitate is given by the

equation

Re Bore, (antigen) /E np (antibody) (i)

in which Rope, (antigen) and Bore, (antibody) are the avorase effective valences
of antigen and antibody molecules, respectively. The maximum value of
x, gg, (antibody) is 2 (ignoring the exceptional possibility that two emall
haptens can attach themselves to the sane conbining region at one end of the
antibody; steric repulsion of antigen molecules would umally prevent this
occurrence), and under optimm conditions for formation of the most stable
precipitate we may expect this maximom value to be Closely approached. The
antibody-antigen molecular ratio then becomes

R= W2 (2)
in which Nis Hop, (antigen). Now a sphere can te brought into contact with
twelve surrounding spherea equal to it in size; hence a spherical antigen
molecule with molecular weight equal to that of the antibody (157,000) might
have the valence N= 12, if all regions of the antizen surface were active and

if the antibody molecules were gpherical; the assumption of elongated antibody
molecules would permit the valence to be somewhat larger. ‘The value 12 of ¥
corresponds to the value 6 for the ratio R. For larger antigens larger values of
R would be expected, and for amaller ones smaller values, Even for antigens with
molecular weight as emll as 11000 the predicted maximum value of R is 4 (for
spherical antibodies) or larger. In fact, a simla calculation based on the

mackine of mheres?? leads to the regults given in Table I 16 It is seen

 

3 See L. Pauling, Ref. 1, Sec, ‘ia.

16 It may de noted that values of 2 calenl-ted in the text chance with
molecular weight in about the same way as those calculated by ¥. C. Boyd and
S. B. Hooker, J. Gon. Physiol. 17, 341 (1934), on the assumption that each
antigen molecule 1s surrounded by a close-packed layer of (univalent) antibody
moleéuies, Tho Boyi-Hooker valuas agree roughly with exreriment (Marmck,
loc.cit. p. 161).

 

Table 1
Coordination of Spherfeal Antibody Molecules
about Spherical Antiren Moleculas

Nomber of anti- Minimum ratio of Minimum molecular “Maximum molecular Maximum mass

body molecules antigen radius to weight of antigen ratio Antibody ratio
about antigen antibody radius (antibesy 160,000) Anticen Antibody
in saturated pree Antizen

cipitate

12 1.000 160,000 6 6

8 0.732 63,000 4 10

6 oth 11,000 z iy

4 225 1,800 2 178

thet our theory provides a simple explanation of the fact that for antigens of
molecule weight equal to or less than that of the antibody the precipitate contains
considerably nore antibody than antigen. The values given in Table 1 are not

to be considered as having rigorous quantitative sienificance. The calculated
maximum molecular ratio would be lorger for elongated antibody molecules than

for cphertoal antibody molecules, and larger for non-spherical than for spherical
le

antigen molecules, and, moreover, in many sera the antibodies might be
complementary in the main only to certain surface regions of the antigen, the
number of theese determining the valence of the antigen. That this is so is

indicated by the observation”! that after long immunization of a rabbit with eg¢e

 

1 U, Heidelberger and VF, %, Kendall, J. Bxp. Med. 62, 697 (1935).

 

albumin serum was obtained giving a precipitate with a considerably larger

| molecular ratio than that for earlier pleedings. 72

ig The discussion of the nature of this phenomenon of change in the serum on
continued immunization must await the detailed treatment of intracellular
antibody-antigen interactions. The phenomenon may involve the masking of the
more effective surface regions of the injected antigen molecules by serum anti-
bodies produced by earlier inoculations, leaving only the less effective regions
available for template action,

 

 

Observed values of R for precipitates formed in the equivalence zone (with
amounts of antigen solution and serum so chosen that neither excess antibody
nor excess antigen can be detected in the supcrnate) for antigene with molecular
weights between about 4,000 and 700,000 lie between about 2.5 for the smaller

19

antifrens and 15 for the larger ones It is seen that the values of R are

 

ty It is our restriction of the valence of the antibody sf the maximum value
two which leads to our explanation of the antibody-antigen ratios. The general
observation of values of R considerably greater than 1 is not aceounted for by
a framework theory in which antibody and antigen molecules are both multivalent,
unless some auxiliary postulate is invoked to make the effective valence of
antigen considerably greater than that of antibody.

 

somewhat less than the corresponding values from Table 1, which indicates that
not all of the surface regions of the antigens are effective. The data given
in Table 2 are those of Heidelberger and his collaboratore; the values reported

by other investigators are sinilar in magnitude.
13

Table 2
Values of Antibody-Antigen Molecular Ratio

for Precipitates from Rabbit Antisera*

Antigen Molecular uqui valence axtrene antie Antigen Solsbie
weight zone body excess excess compound

Bee albumin 42000 2.5 + 3 5 2 1

Dye egg albumin®* 46000 25-3 5 0.75 i/2

Serum albumin 67000 344 6 2 1

Thyroglobulin 700000 10 ~ 14 40 2 1

 

The experimental waluas are thoxe obtained by Heidelberger and collaborators,
and quoted by H. Heidelberger, THIS JOURNAL 60, 2h2 (1938),
“e

Regalt-asobiphenylazo egg albumin,

 

In a precipitate formed from a solution containing an excess of antibody not
all of the antibody valences will be saturated, At the limit of antibody excess
the precipitate will be a network of linear aggregates with a structure such as
that represented in Figure 4, Here each antigen molecule (with an occasional
exception) 1s surrounded by F antibody molecules, only two of which bond it to
neighboring antizen molecules. The padded strings formed in thie way are tied
together by an oocasional cross-link to form the precipitate. The antibody~-
antigen molecular ratio is seen to be close to H- 1, which is one lees than twice
the value N/2 for the valence-satursted precipitate. The predicted relation
between these ratios

R = 2k

=—antibody excess equivalence zone ~
is seen from the data in Table 2 to be verified anproximately by experiment for

1

the antigens other than thyroglobulin.

The discrepancy shown by thyrozlobulin is, indeed, to be expected for an
antigen with molecular weight greater than that of the antibody. The requirements
of geometry are such that an arrangement in which each antigen is bonded

equivalently by antibodies to more than twolve surrounding antigens is impossible,
14

Hence for large antigen molecules the molecular ratio can exceed 6 in the valence-
saturated precipitate only if two or more antibody molecules are shared between
the same pair of antiren molecules, whereas in the antibody-~excese rezion the
entire surface of the large antigen may be covered by antibody molecules.

With antigen excess the precivitate formed will have the limiting structure
shown in Figure 5, in which (with an occasional exception) both antigen and
antibody are bivalent, the molecular ratio approaching unity. The reported
experimental valuea for this ratio (Table 2) lie between 2 and 3/4.

With great excesa of antigen finite complexes which remain in solution are
forme), with structures such as shown in Figure 6, For these the molecnlar ratio
varies between 1 and the minimum value 1/2. It is observed that in general no
precipitate forms in the region of great antigen excess, and Heidelberger and his
collaborators have in fact assigned values 1 and 1/2 to the molecular ratios for
the complexes in solution.

Vhereas precipitation is inhibited by antigen excess, it usually occurs even
with great antibody excess, although soluble complexes with molecular ratio 5
and the structure shown in Figure 7 are expected to exist. It seems probable
that the difference in behavior of systems in the excess antigen region and excess
antibody region is to be attributed to the fact that the molecular ratio for
precipitate and soluble complex differs by a factor aa great as two for the former
case, and by only N/(H - 1) for the latter.

d._ the Use of a Single Antigen Kolecule as the Template for an Antibody Molecule.--

There are two ways in which an antibody molecule with two opposed active regione

 

commlementary to the antigen might be produced, One is the way described in
Section II, The other would involve the manufacture of the antibody molecule

in its final confircuration hetween two antizen molecules, one of which would serve
ag the pattern for one antibody end and the other for the second, Ne attemnt to

decide between these alternatives seens to have been made before; there exists
15

evidence, however, some of which is mentioned below, to indicate that the first
method of antibody production, involving only one antigen molecule, occurs
predominantly. It is for thie reason th..t I have develoved the r.ther complicated
theory described above, with the two end >ortions of the antibedy forming firet,
one (or both) then separating from the antigen, and the central part of the anti-
vody then aseuming its shape and holding the active ends in rosition for
attachment to two antigen molecules,

Thie theory requires that the formation of antibody be a reaction of the
first order with respect to the anticen, whereas the other alternative would
require it to be of the second order, There exists very little evidence as to
vhether on immunization with smell amounts of antigen the antibody production is
proportional to the amount of antigen injected or to its square. Some support for
the one-antigen-molecule theory 1s provided by the experiments dealing with the
injection of a mixture of antigens. If two antigen molecules were required for
antibody form:.tion, it would be expected that antibodies A'~B', A'-C!, BYct...
complementary to two different antifens A and B, A and C, Band €,--- as well as
those A'-~A', BY-Bt, Ct.C',.-- complementary to a single antigen would be formed,
The evailable evidence speaks etrongly cgainat this, Thus Dean, Taylor, and
Adair? hive reported that the serum produced by immunization with a mixture

—

2 H. R. Dean, G, L. Taylor, and M. B, Adair, J. Hygiene 35, 69 (1935).

 

of egg albumin and serum albumin contains distinct antibodies homologous to the
two antigens, and that vrecipitation with one antizen leaves the amount of the
heterologous antibody unaltered, An even more rigorous demonstration was

furnished by Heidelberger and Kabat

ai

» who, from the serum of a cow which had

 

M, Heidelberger and EB, A, Kabat, J. Exp. Me’. 67, 181 (1938).

 

 

been injected with types I, II, and III pnewmococci, isolated in euccession, with
the corresponding ssecific polysaccharide, the three anticarbohydrates, each in

an apparently pure state and with no sopreciable cross-reactivity as to
16

2

pneumoceccus tyne. In another atriking experiment Hektoen and Boor” found

22

 

L, Hektoen and A. K. Boor, J. Infect, Dis. 48, 588 (1931).

 

that the serum obtained on injecting a rabbit with a mixture of 35 antigens
reacted with 34 of the antizens, and that absorption with any one had in the main
little effect on subsequent reaction with another. Since on the two-antigen-
molecule theory the amounts of antibodies A'-A', B!-B!,... capable of causing
precivitation with 2 single antigen vould be szr11 compared with the total

amount of antibody (of the order of 1/n, for n antigens-—about 3% in this case),
these qualitative observations provide significant evidence in favor of the
alternative theory.

2. Criteria for Antigenic Power.--There has been extensive discussion of the
question of what makes a substance an antigen, but no vendrally accented
conclusions have been reached. Our theory nermits the formulation of the follow
ing reasonable criteria for antigenic activity:

1. The antigen molecule must contain active crouns, canable of sufficiently
estrone inter-ction with the globulin chain to influence its configuration.

2. The configuration of the antizen molecule must be well-defined over
surface regions large enough to give rise to an integrated antibody~antiren
force sufficient to hold the nolecules together.

3. The antigen molecule must be large encugh to have two or more such
surface regions, and in case that the antigenic activity denends upon a particular
group the molecule must contain at least two of these groups, (This criterion
applias to antibodies effective in the vrecipitin and agrlutinin reactions and
in anaphylazis. )

These criteria ore satiefied by substances mow to have antisenic action.
Many proteins, some carbohydratea with high molecular weight (bacterial poly-

saccharides, invertebrate glycogen”), and some lipids and carbohydrate-lipid

 

a D, H. Campbell, Proc. Soc. Exp, Biol, Med. 36, 511 (1937); J. Parasitol.
23. 348 (1937). Se

 
17

complexes are antigenic. ‘The simple chemical substances so far studied have
been found to be inactive, excent those which are capable of combining with
proteins in the body. Non-antigenic substances have been reported to become

antigenic when adsorbed on narticles (Forssman antigen on kaolin’); in this

 

e+ P, Gonzales and H. Armangue, Cont, Rend. Soc. Biol. 105, 1006 (1971);
K, Landsteiner and J, Jacobs, Proc. Soc, ax>, Biol, Med. 30, 1055 (1933).

 

 

case the varticle with sdsorbed hapten is to be donsidered the anticen "molecule"
of our theory, I predict that relatively eimmle molecules containing two or
more hantens will be found to be antizenic; exoriments to test this prediction

are now under way.

Iv, A More “etailed Discussion of the St
of Antibodies and Other Proteins
There h-g been gathered eo far very little direct evidence rerardinge thea
detailed structure of protein molecules, Chemical information is commatible
with the volynentide-chain theory of »rotein structure, and this theory is also

mxpported vy the rather small amount of rertinent x-ray evidence->. It «me pointed

 

25 A brief statament of the situ:tion has been made by L. Pauling and C,
Hdenann, THIS JOURKAL 61, 1860 (1939); see also R. 3. Corey, Chem. Rev.
20
* le ee °

 

26

out some years ago” that the well-defined propertics of native proteins require

* a. 3, Mirsky and L, Pauling, Proc, Hat. Acad, Sel, 22, 439 (1936); B. Wu,

Chinese J. Physiol. 5, 321 (1931).

that their molecules have definite configurations, the polypeptide chain or chains

 

 

in a molecule being coiled in a definite way and held in nosition by forces acting
between narts of the chains. The phenomenon of denaturation involves the loes of
configuration through the partial or complete uncolling of the chains, Of the
forces involved in the retention of the native configuration those @oseribed ag
hydrogen bonds are probably the most important. Our knowledge of the properties

of the hy‘rogen bond has increased to such an extent during the vast five years
18

as to justify some enecul:tion «2 to the niture of the stable configurations
of protein molecules. |

Hydrogen bonds can be formed by the peptide carbonyl and imino groupe of
& polypeptide chain, and also by the carboxy, amino, hydroxy, and other oxyzen-
and nitrogen-containing groups in the side chains of the amino acid residues,
In a stable configuration as many strong hydrogen bonds as possible will be
present. (One configuration in which all of the peptide carbonyl and imino
eToups are forming strong hydrogen bonde is that shown in Figure &, Here
extended chains are bonded together to form a compact layer, with the side chains
extending alternately ahove and below the nlane of the layer (provided that the
levo configuration is the only one represented by the amino-acid residues),
This configuration has been afeigned to B-keratin and other fibrous proteins by

Astbury! on the basis of x-ray data, Although tthe structure has not bean

oF

 

W. 2, Astbury, Trang. Faraday Soc. 29, 193 (1933), and other vapors.

 

verified in de§all by the analysis of the x-ray data, the agreement in the

dimensions found experimentally for the naeudo-unit cell of fekeratin ané those

28

predicted from the complete etructure determinations of flycine”” and diketovineragine”?

2B G, Albrecht ond R, B, Corey, THIS JOURNAL 61, 1087 (1939).
29 2. B, Corey, ibid. 60, 1598 (1938).

 

 

makes it very probehle that the etructure is essentially correct, with, however,

the chains somevh:.t ‘distorted from the completely ertended configuration” .

 

» R, B, Corey, ref. 25.

 

It is to be noted that the ~NH-CO-CHR- sequence alternates in alternate
lines in euch a layer, so that a layer of finite size could be constructed by
running a sinsle polypeptide chain back and forth. A globular protain could then

be made by building several such layers parallel to one another and in contact,
19

like a stack of pancakes, the layers being held together by side-chain inter~
actions as well as by the volypentide chain itself, A vrotein molecule with,
for example, roughly the shape and size of a cube io A on edge might contain
four layers, each with about eight strings of about twelve residues each.

A few yoore ago I noticed, by studying molecular modele, that a proline or
hydroxyproline residue in the chain would interfere with the structure shown in
Pipure § in such a way as to cause the chain to tend to turn through 180°; hence
if these residues were suitably distributed along the chain during its synthesis
the chain would tend to assume the configuration discussed above.

layer structures other than this one might aleo be assumed, in which the
chaing are not extended, Some fibrous »roteins, such as a-keratin, are known to
have structures of this general type, hut the miture of the folding of the chains
has not yet been determined.

We have postulated the existence of an extremely large number of accessible
configurations with nearly the cane energy for the end varts of the globulin
polypeptide chain, A layer etructure, with variety in the type of folding in the
layer, would not, it seems to me, give enouch configurational possibilities to
explain the great observed versatility of the antibody precursor in adjusting
itself to the antiren, and I think that skew configurations amst be invoked.

But simple considerations show that 14 would be difficult for the chain to assume
a skew configuration in which most of the peptide carbonyl and imino bonds take
part in forming hydrozen bonds, 22 they do in the layer structures; and in
consequence the skew configurations would be amuch less stable than the layer
configurations. The way out of this difficulty is provided by the nectulate that
the end parts of the globulin polypeptide chains contain a very large proportion
(perhaps one-third or one-half) of proline and hytrozyproline residues and other

residues wiich prevent the assumption of a stable layer configuration.

 

We may, indeed, anticinate that globular proteins may be divided into two

main classes, comprising respectively those in which there is a layer structure
and thoee in which the stable confiscuration of the chains is more complex; the
latter my show the high proline and hydroxyproline content postulated for the
flobulin chain ende.

About one-third of the residues in gelatin are proline and hydroxyproline.
We expect accordingly that there are many confieuretions with nearly the same
energy accessible to a gelatin molecule, and that czelatin is not characterized

?
by a single well-defined native molecular configuration’-. Since the substance

 

3A Collagen has a definite fiber structure, as shown by x-ray photogranhe,
A possible atomic arrangement has been sugeested by W, T. Aatbury and FP. O, Bell,
Nature 345, 421 (1940).

 

contains no strong antigenic groups, a definite configuration is a requisite for

antigenic activity. These considerations thus provide 2 possible explanation of

the well-known fact that relatin is net e?fective as ar antien-.

 

32 Gelatin with haptens attached is anticenic (for references see Landetciner,
loc.cit., 7. 102, and ©, 2, Harington, J. Chem, Soc. 1940, 119). The pre-
sumption is that the haptens interact with the antibody so effectively that one
haoten forna the antigen-antibody bond, and the lack of definite configuration
of the celatin is of no significance.

 

Serological experiments with artifical conjusated antipens””, especially

 

35 K, landsteiner and H, Lampl, Biochem. 2. £6, 342 (1918): K. Landeteiner,
fpid. 204, 280 (1920).

 

azoproteins, have provided resulte of great significance to the theory of anti-
body structure, Many of the arguments based on these results are >resented in
the books of Lanésteiner and Marrack, The data obtained recarding cross~reactions
of azoprotein sera with related azoproteins show that electrically charred groups
(carboxyl, eulfonate, arsenate) interact stroncly with homologous antibodies, and
that someyhat weaker interactions are oroduced by hydregen-bond-forning croups
ané groups with large electric dipole moments (hydroxy, nitro). The principal
action of a weak group such ss alkyl, phenyl, or halocen is steric; this is shown

clearly by the strong cross-reactions between similar chloro and methyl haptens.
21

The data on apecificity of antibodies with reapect to haptens indicate strongly
that the hapten group fits into a pocket in the antibody, and that the fit isa
close one, It ehowld not be concluded that all antibody-antigen bonds are of
just this tyne; for example, the fitting of an antibody group into a pocket in
the antigen may also be often of importance. Extensive work will be needad to
determine the detailed nature of the antibody structures complementary to

particular haptens and antigene.

Y, Further Comarison of the Theory with Experiment.
ogsible tal Test Predictions.

 

a. Hethods Detern the Vale tibodies.--The following methods may
be proposed to determing the valence of antibodies. Yirst, let a serum be
produced by injection of an azoprotein of the following type: its hanten is to
be sufficiently strong (that is, to interact sufficiently strongly with the
homologous antibody) that one uapten group forms a satisfactory antibody-antigen
bond, and the number of hapten groups ner molecule 1s to be eamall enough so that
in the main only one grow will be present in the area serving as a vattern
for an antibody end. ‘The same hapten ie then attached to another protein, and
this azoprotein is precipitated with the serum. If it be assumed that the
precipitate 1s valence-saturated, the ratio of hapten groups to antibody molecules
in the precipitate gives the average valence of the antibody. |

Data of essentially this sort, obtained with arsanilic acid as the hapten and
the rabbit as the experimental animal, have been published by Haurowitz and

his collaborators?” in a paver reporting many interesting experiments. The

3E

 

F, Haurowitz, F. Kraus, and ¥. Marx, %. physiol, Chem. 245, 23 (21936).

 

equivalent weight of antibody per arsanilic acid residue was found to vary from
23,000 for anti-ens with very many attached haptena to 51,000 for those with
only a few. The expected value for saturation of all haptens by bivalent anti-~-

body molecules is one-half the antibody molecular weight, that is, about 79,000.
22

The low experimental values are probably due to failure of some haptens to
combine with antibodies. In particular, if two haptens are attached to the
same tyrosine or histidine residue steric interactions of antibodies may permit
only one of the hantens to be effective”,

 

bo From the data discussed above, which in our Opinion indicate that two
hapten groups combine vith a bivalent antibody molecule, Haurowitz drew the
different conclueion that one areenic-containing group in the antigen combines
with one antibody molecule. He reached this result by assuming 100,000 (rather
than 157,000) for the molecular weight of the antibody and by assuming that the
active group in the antigen consists of two haptens attached to a tyrosine or
histidine residue, It is, of course, likely that this occura in antigens with
high arsenio content, but it seens probable that the haptens are mainly attached
to sevarate residues in the antigens con taining only a few haptens ner molecule.

 

The divalence of antibodies and our postulate that only one antifzen molecule
is involved in the formation of an antibody molecule require that a vrecinitin-
effective antihapten be produced only if the injected antigen contain at least
two hapten ¢roups. Pertinent data have been obtained on this point by
Haurowits and hie collaborators, who found that effective antihapten precipitin
serum was produced by an azoprotein, made from arsanalic acid and horse globulin,
containing 0.24% arsenic (4,8 haptens per average molecule of molecular
weight 157,000), and a trace by one containing 0.13% arsenic (2.6 haptens per
molecule).

b, the Possible Anticenic Activity of Simple Substances.--The criteria given
above for antigenic activity would be satiefied by a substance of relatively
Low molecular veight in which several hapten groups (such as several arsanilic
acid residves) ere present in the molecule. A substance of this sort vould

be expected also to show the precipitin reaction with its own serum or vith
serun homologous te an azopretein containing the same hapten, and to be capable
aleo of producing anaphylaxis.

A substance with only two hapten groups per molecule might be expected not
to give a precipitate with the homologous antibedy, but rather to form long

strings with antibody and antigen molecules alternating. These strings would
23

remain in solution, and would confer on the solution the proverty of pronounced
birefringence of flow, If, however, there were in one end of some of the antibody
molecules complementary regions for two hapten groups, making theee molecules
effectively trivalent, the etrinzs would be tied tozether and a precipitate would
be formed. It is probably significant in this connection that landsteiner and

van der Scheer have observed both the precipitin reaction and anaphylactic

 

3 K, Landsteiner and J, van der Scheer, Proc. Soc. Exo. Biol. Med. 29,
TH7 (1932); J. Exp. Med. 56, 399 (1932).

 

shock (in guinea pigs sensitized with the corresponding azoproteins) with azodyes,
such as resorcinoldiazo=p-suberanilic acid, (OH) aCgHa(NNC gH, NHCO( CHa) gCOOH) 9,
forned by coupling two anilic acid molecules with resorcinol. Landsteiner
himself has explained these observ-tions as resulting from the low solubility

of the dyes, but the explanation advanced above seems more probable.

c. Experiments with Two or More Different Hanteng in the Same Antigen.--We would
predict that if there were used as an antigen a molecule to which several hanten
groups A and several different hapten groups B were attached, the serum obtained
would contain three kinds of bivalent antibodies, A'-A', B!-B!', and A'~Bt (as well
as the univalent antibodies A'~ and B'-), Our picture of the process of formation
of antibodies permits little or no correlation between the two chain onde of the
globulin in the selection of surface regione of the antigen molecule to serve as
templates, except that the two regions suet not overlap. We accordingly predict
that for ny and By Groups A and B respectively ner antigen molecule the numbers of
divalent antibody molecules of different kinde in the serum would be

Surgr @ on (a, - 2) (3a)
Ryrge = S08 nny (3b)
Syvp * B*n, (a, +1), (3c)

in which ao and — are coefficiente which give the vrobabilities that the groups

serve ae templates. It 46 seen that the amount of A'~B! antibodies is predicted
24

to de equal to or Slightly gre:ter than twice the geometric mean of the amounts
of A'-A' and Bi-BS, The only data permitting a quantitative test of this
relation which have come to my attention are those obtained by Haurowits and
collabormtorsa (loe. cit.) by use of azoproteins made from arsanilic acid,

A serum produced by injecting rabbite with an antigen mde from arsaniliec acid
and horse globulin was found to contain antibodies to the arsanilic hapten
(precipitating an azoprotein made from arsanilic acid and rabbit flobulin),
others to horse globulin itself, and still othere to the homologous antigen,
these last vreaumbly being complementary in structure both to the hapten and
to the active groups of horse globulin. ‘The quantitative results obtained are
the following: 10 ml, of immune seru gave 17-18 mg. of precipitate with the
maximal precipitating amount of asoprotein from arsanilic acid and rabbit
flobulin, and 8 mg. of precipitate with horse globulin, these amounts being
independent of the order of the two precipitations. After exhaustion with these
two antigene the serum gave 17 mg, of precipitate with the homologous antigen,
If we assum that the conditions of each precipitation were such that only
suitable bivalent antibody molecules were incorporated in the precipitate, the
equation above would require the third precipitate to weigh about 24 mg.; the
experiment accordingly provides some support for the theory. The quantitative
discrepancy may possibly be cue to the incorporation of some effectively
univalent antibody molecules in the first two precipitates.

The qualitative experinental reaulte which have been reported are in part

compatible and in part incompatible with the theory, The most interesting of the

experiments are those of Landesteiner and van der Scheer”, who prepared aro-

 

37 K, landsteiner and J, van der Scheer, 3. Uxr. Hed, SJ, 709 (1938).

 

proteins containing two different kinds of haptens sn@ studied the antibodies
produced by them. In some cases only one of the haptens was effective in antibody

formation. With the azoprotein made from S~amino-5«suceinylaminobenszoyl-~p~
25

aminonhenylarsenic acid, however, & serum was obtained which would combine

not only with the homologous antigen but aleo with azoproteins formed sither
from m-aminoauccinanilic acid or from peaminophenylarsenic acid, After the serun
was exhausted by interaction with stromata coupled with either one of these two
simple haptens, it reacted ss strongly (as measured by the estimated amount of
nrecipitate) with the azoprotein containing the other simle hapten as before
exhaustion. From this experiment and others the investigatore concluded that
there avpeared to be present in the sera, if any, only small amounts of anti-
bodies with two combining groups capable of interaction with the two different
haptene in the antigen. It seems possible that the conclusion is not justified
by the data, and, with the kind cooperation of Dr. landateiner, we are continuing
this investigation. .

a. The Anticentc Acti bibodies.——Our picture of an antibody molecule
requires that the configuration of its middle portion be the same ag that of
normal serum globulin. Hence antibodies should have antigenic activity, with
eseentially commlete cross-reactions with normal globulin. This is in agreement

with exveriment} Iandsteiner and Prasek™ found that prectpitine which

 

#8 K. Landsteiner and &, Prasek, Z. Immunforech. 10, 68 (1911).

 

precipitated the serum of an animal precipitated also the agglutinins in the

7
serum, and Bisler-? showed that a nrecinitin to horse serum would precipitate

 

By, Bisler, Zent. dakt. $4, 46 (1920).

 

tetanus antitoxin. On the other hand, according to our picture the active end
regions of the antibody molecules wonld not have affective antisenic power, since
their configurations would be different from molecule to molecule (depending

on the accidentally selected template region and accidental wey of coiling),

and an antibody complementary to one antibody end would as a rule not combine

with another. The antibody ends would hence in the main be left free in the
precipitate formed by an antibody and its precipitin, as well ae in that

formed by an antibody and the precipitin to normal globulin, In agreement with

this, Smith and Marrack © found that a precipitate formed by a precipitin and

WO

 

F, C, Smith and J, Marrack, Brit, J, Bx, Path. 12, 494 (1930).

 

& serum containing diphtheria antitoxin has the power of combining with
diphtheria toxin; and similar reqults have also been obtained recently by

uy

Heidelberger and Treffers ~ in the case of specific predipitates formed by

pneumoceeous antibody with homologous antiserum,

4h Personal commnication of unpublished material by Profeesor XM, Reidelberger
and Dr. H., P. Treffera, ,

e. Factors Affecting the Rate of Antibody Production apd the Specificity of

Antibodies.-—In order that an antibody be effective, the surface region of the

 

 

antisen covered by an antibody end must be large enough so that the integrated
attractive forces constitute an antibody-antigen bond of significant strength.
With a dond of average strength the antibody molecule will be dissociated fron
its anticen template, perhaps with the aid of the auxiliary mechanien mentioned
in Section VI, until an equilibrium or steady-state concentration is built un
in the serum. Wow if the antigen surface contains a large number of strong
groups, capable of interacting strongly with complementary structures in the
antibody, the antibody-antigen bond may be so strong that the antibody ie not
able to separate itself from the antigen, and only a small antibody concentration
can be built up in the serum. We hence conclude that an antigen containing weak
sgroups will in general be a cod antizen, whereas one containing many strong
groups will be a poor antigen, with respect to antibody production.

Thie prediction, which at firat thought seens paradoxical, is in fact borne

 

out by experiment. Thus a bland protein such 2s egg albumin is a good antigen,
as are also conjugated proteins with weak groups attached. An axoprotein with
many strong groups attached (arsenic acid, sulfonic acid, nitro, etc; the azo

group itself 18 a rather strong group, capable of forming hydrogen bonds) is
eT

a poor antigen; in order to obtain serum to such haptens an azoprotein containing

a limited mumber of the groups muat be used. I am told by Dr, Landsteiner that

this observation was made in the saarly days of the study of exovroteine’@,

 

12 see K, landsteiner and H, Lampl, loc.cit.

 

Pertinent data have been reported in recent years by Haurowitz and his
collaborators (loc.cit.), whe found the optimum arsenic content for the production
of antihapten by azoprotein made from arsanilic acid to be between 0.5 and 1.0%;
very little antibody is produced by antigens with over 2% arsenic, although strong
precipi¢in reaction 1s shown by azoprotein with an arsenic content as great as 10%.
A second deduotion, relating to specificity, can also be made. To achieve
a sufficiently strong antibody-antigen bend with an antigen containing only weak
groups a large surface region of the antigen must come into play, whereas with
an antigen containing strong groups only a small region (in the limit one group)
ia needed. Hence antibodies to antigens containing atrong groups show low
specificity, and those to antigens containing weak groups show high specificity.
Thies vredioction is substantiated by many observations. Bez albumin, hemoglobin,
and similar proteine give highly specific sera, whereas azoproteins produce sera
which are less specific, etrong cross-reactions being observed among various
proteins with the same hapten attached. This shows, indeed, that a single
hapten group gives a sufficiently strong bond to hold antibody and antigen
together. In such a case the approximation of the antibody to a strong hapten
4s very close, and great specificity 1e shown with regard to the hapten iteelf,
this specificity being the greater the stronger the hapten. Many examples of
these effects are to be found in Landateiner's work.
f. The Sffect of Denaturing Agents.—-We made the fundamental postulate that the
end parts of the polypeptide chains of the globulin molecule are characterized

by having a very large number of accessible confieurotions with nearly the sane
energy, whereas there is only one stable configuretion for the central part.

It is accordingly probable that the end configurations, giving characteristic
proverties to the antibodies, would be destroyed before the central part of

the molecule is affect+d and, moreover, that the sensitivity to denaturing agente
or conditions of antibodies to different antigens would be different. The
available meager experimental information seems to be compatible with these

tea},

 

1
‘7 See Marrack, loc.clt., op. 48-53,

 

Some remarks may ve made regarding the difference in behavior of anti-
bodies and antigens in the presence of denaturing cerents. An antigen molecule
may undergo a considerable change in configuration without losing comletely its
power of reveting with the homologous serum; if some of the surface regions
remain essentially unchanged after partial denaturation of the pretein, the
antibody molecules complementary to these regions will retain the power of
combining with them, whereas the antibody molecules comlementary to the regions
vhich have been greatly changed by denaturation will no longer be effective.

In particular some native proteins may be built of cuperimposed layers, as
described in Section IV, the antigenic regions on top of the top layer and on
the bottom of the bottom layer would still be effective after the partial
denaturation of the molecule uy the umleafing of the layers, whereas the
anticenic regions at the sides of the original molecule would in large part

loge their effectiveness by this unleafing. The observation by Rothen and
Landsteiner’ that ecg albumin spread into surface films 10 i thick retaing

ay

 

A, Rothen and K, Landsteiner, Science 99, 65 (1939).

 

the ability to combine with anti-ezg-albumin rabbit seraa is most simply
explained by the assumptions that the native egg albumin molecule hag the layer

structure suggested above and that the process of surface denaturation of this
29

molecule involves the unleafing of the layers without the loss of their structure.
As mentioned above, it is probable that for nost antibodies the end rezions
are affected by denaturing agents more eseily than the centr] rerion, and that
the firet step in denaturation of an antibody involves these end regions and
leads to loss of their specific properties. It has been shown by Danielli,
Dandelli, and Marrack!® that the reactivity of antibodies ie destroyed by

 

we J. F. Danielli, M, Danielli, and J. R, Marrack, British J. Exp, Path.
29. 393 (1938).

 

46

surface denaturation ~.

nae Rothen and Landeteiner (los.cit.) have pointed out that from these facts
regarding surface denaturation the conclusion can be drawn that "the specifie
reactivity of antibodies is to a large extent dependent upon structures
aifferent from those which mainly determine the specificity of antigens".

 

 

An interesting possible method of producing antibodies from serum or
flobulin solution outside of the animal is suggented by the theory. The globulin
would be treated with a denaturing agent or condition aufficiently strong to cause
the chain ends to uncoil; after which this agent or condition would be ranoved
slowly while antigen or hapten is present in the solution in considerable
concentration. The chain ends would then coil up to assume the configurations
stable under these conditions, which would be configurations complementary to
thoee of the antigen or hapten.

Many of the experiments suggested above are being undertaken in our
Laboratories, with the collaboration of Dr. Dar Campbell.

VI. Processes Auxiliary to Antibody Formation

It seems not unlikely that certain processes auxiliary to antibody formation
oceur, ‘The reported increase in globulin (aside from the antibody fraction) after
immunization suggests the operation of a mechanisem whereby the presence of
antigen molecules accelerates the synthesis of the globulin polypeptide chains.

There is little basis for suggesting possible mechaniems for this process at present.
The occurrence of the anamestic reaction—-the renewed production of
antibodies to an antigen caused by injection of a second antigen--may be explained
by the asaumtion that following the synthesis of an antibody a nechani sm cones
into operation in the cell to facilitate the removal of the antibody from the
antigen, perhaps by changing the hydrogen-ion or salt concentration or
dielectric constant. This would assiet in removing antibodies not only from the
second antigen but aleo from those molecules of the first antigen which had
remained, covered with homologous antibody attached too firmly for spontaneous
removal, in the cell. The evidence indicates that the anamnestic reaction is
not in general strong. At a tice after inoculation with typhoid bacillus or
erythrocytes long enough that the corresponding agglutinins are no longer
detectable in the serum injection of ancther antigen gives rise to the presence |
of there agglutinins in amounts detectable by the very sensitive ag:lutination

test; but Kabat and Heidelberger’? found that the amount of additional antibody

 

WT 8, A, Kabat and M. Heidelberger, J. Exp. Med. G8, 229 (3957).

 

to serum albumin produced by injection of ege albumin or typhoid toxin was too
amall to be detected by their method of analysis.

aus The mechanian for catching the antibody molecule and holding it in the
region of clobulin aynthesia may be closely related to that of antibody
production--possibly a partially liberated clobulin chain which forms a bond
or two bonds with an antigen molecule directly above it fie rrevented from
freeing ite central part from the cell wall, and eo serves as an anchor,

fhe renewed production of antibody in the serum after bleoting ie to be

attributed to the presence of trapned antigen molecules in the celis. The greater
duration of active than of passive irmmmnigation may be attributed to this or to

the presence of complexes of antigen and surrounding antibodies, the outer ends

of which could combine with additional antigen.
Acknov. ent

My interest in immunology was awakened by conversations with Dr. Karl
landeteiner; I am glad to express ny gratitude to him, and to acknowledge ay
indebtedness to him for ideas as well ae for facts, I wish also to thank
Professors Michael Heidelberger and Dan Campbell for advice and assistance.

Summary

It 18 assumed that antibodies differ from normal seran globulin only in the
way in which the two end parte of the globulin polypeptide chain are coiled, these
parte, as a result of their amino-acid composition and order, having accessible
a very grest many configurations with nearly the same stability; under the
influence of an antigen molecule they assume configurations complementary to
surface regions of the antigen, thus forming two active ends. After the frasing
of one end and the liberation of the central part of the chain this part of the
chain folds up to form the central part of the antibody molecule, with two
oppositely-directed ends able to attach themselves to two antizen molecules.

Among the points of comparison of the theory and experiment are the
followin:: the heterogeneity of sera, the bivalence of antibodies and multi-~
valence of antigens, the framework structure and molecular ratio of antibody-
antigen precipitates, the use of a single antizen molecule as template for an
antibody molecule, criteria for antigenic activity, the behavior of antigens
containing two differant haptens, the antigenic activity of antibodies,
factors affecting the rate of antibody production and the specificity of

antibodies, and the effect of denaturing agents. It is shown that most of the

reported experimental results are compatible with the theory. Some new

experiments suggested by the theory are mentioned,

Pasadena, California Recel ved

 
Fig.

Fig.

Fir,

Fir.

Fig.
Fig.
Fig.

Fig.

1.

2.

4,

5.

32
Legende for figures

Diagrams representing four stages in the process of formation of a
molecule of normal serum globulin (left side of figure) and six stages
in the process of formation of an antibody molecule as the result of
interaction of the globulin polypeptide chain with an antigen
molecule, There is also shown (lower right) an antigen molecule
surrounded by attached antibody molecules or parts of molecules and
thus inhibited from further antibody formation.
(A) Diagram representing agclutinated cells. (B) Diagram of the
region of contact of two cells, showing the postulated structure
and mode of action of agglutinin molecules,
A portion of an ideal antibody-antigen framework, One plane of the
atructure corresponding to the value twelve for the valence of the
antigen molecules is shown.
4 portion of an antibody-antigen network formed in the region of
antibody excess,
A portion of the network formed in the region of antigen excess.
Repwesentative soluble complexes formed with excess antigen,
A soluble complex formed with excess antibody.
The folding of polypeptide chains into a layer held together by

imino-carbonyl hydrogen bonds.