This paper argues that immunological memory is in the same class of associative memories as Kanerva's Sparse Distributed Memory, Albus's Cerebellar Model Arithmetic Computer, and Marr's Theory of the Cerebellar Cortex. This class of memories derives its associative and robust nature from a sparse sampling of a huge input space by recognition units (B and T cells in the immune system) and a distribution of the memory among many independent units (B and T cells in the memory population in the immune system). Keywords: Immunological Memory, Associative Memory, Cross-Reactive Memory, Original Antigenic Sin, Sparse Distributed Memory. 1 Introduction Cowpox vaccination, used to protect humans from smallpox, was the first known deliberate use of associative recall in the immune response (Jenner, 1798). The modern investigation of associative recall began with the observation that antibodies induced during an influenza infection often have greater affinity to prior strains of influenza than t...

The immune response to repeated antigenic stimtflation differs in a number of respects from the response to the initial exposure to immunogen. For example, several weeks (or even years) after the first injection, a second exposure to the immunogen often evokes an unusually vigorous response: the serum antibody concentration typically increases rapidly to reach and remain at a level higher than that attained after the primary stimulus (1-4). In addition, the antibodies produced in the secondary response often form more stable complexes with antigen than those formed at a comparable time after the first injection (5-9). Recent studies of antibodies isolated from serum at various intervals after immunization with 2,4-dinitrophenylated proteins have shown that the affinity of these antibodies for simple dinitrophenyl haptens (expressed as the average intrinsic association constant) is greater the longer the period between immunization and bleeding, and the smaller the dose of immunogen (10). These properties of the serum antibodies are the direct result of a sequential change in the nature of the antibodies synthesized by lymph node cells (11). It seemed possible, therefore, that the affinity of the antibodies formed at any time after a single injection of antigen is inversely related to the amount of antigen remaining in the animal. Accordingly, we wished to determine the effect on antibody affinity of a second injection of antigen, given when the animal is already forming high affinity antibody in response to the initial stimulus. In experiments reported here, it is demonstrated that the typical response to a second injection of antigen is a burst of synthesis of antibodies with much greater a/~ity for the dinitrophenyl group than those formed at a comparable time after the primary immunization with the same dose of immunogen. Indeed, the earliest antibodies formed after restimulation are already high in affinity. *A preliminary report of some of this work was presented at an annual meeting of the American

When an antigen is injected into a higher animal, a sequence of reactions occurs which culminate in the manifestations of immunity, e.g., antibody formation. A number of experimental approaches are available for obtaining information on this sequence of reactions. Specificity studies on the process of tolerance induction, on the one hand, and on secondary stimulation on the other, carried out by using pairs of well-defined cross-reacting antigens have revealed phenomena which are probably connected with the early phase of the immune response. Tolerance to an antigen could be overcome by injecting antigens which cross-reacted with the "tolerogen " (1-9); several workers have demonstrated that an antigen may stimulate the production of antibodies directed to another though related antigen with which the organism had previously been in contact (10-15). Both the overcoming of tolerance and the phenomenon of "original antigenic sin " (16) have been explained (5, 9, 14) by postulating certain receptors for antigen whose specificity did not necessarily seem to be identical with the specificity of the antibodies produced. Experimental

The "original antigenic sin " phenomenon (OAS) 1 challenges the dogma of the specificity of the immunological memory: when there is sequential infection with two different but antigenically related strains of influenza A virus, the antibody stimulated by the second infection reacts more strongly with the primary virus than with the one actually eliciting the response (1-3). This phenomenon is now well documented with many viral and nonviral cross-reacting antigens (4-8). However, its study with influenza antigens is of special interest because of its wide implications in the sero-epidemiology of influenza and the response to vaccination. In the accompanying paper (9) we have shown that purified hemagglutinin (HA) extracted from related influenza viruses share cross-reacting antigenic determinants, but differ in strain-specific determinants. We have now analyzed the antibody response to each of these groups of determinants after sequential exposure of mice to two related HA's, and have carried out cell-transfer experiments in an attempt to elucidate the cellular mechanisms responsible for the aberrant immunological recall of the OAS phenomenon.

To obtain anamnestic antibody synthesis to a second antigenic stimulus the second antigen must share some immunogenic determinant with the primary sensitizing material. Generally the secondary response takes place only to the common determinants (1-3). Occasional animals have been reported to undergo a secondary antibody response to antigenic determinants specific only to the first antigen when injected with a cross-reacting antigen as the secondary stimulus (1, 4). While the antibodies produced appeared to be specific for the first antigen based on absorption of the secondary sera with the second antigen, the complexity of the antigens (albumins) has limited the interpretation of these results. Albumins are multispecific antigens (5, 6) and the extent to which hidden determinants of one species might cross-react with surface determinants on a cross-reacting species has not been investigated. An artificial haptenic group can represent a significant portion of a specific determinant. Subtle alterations in the spacial configuration of haptens cause corresponding alterations in the elicited antihapten antibodies and are distinguishable

The identification of cross-reactions between related but distinct antigens during anamnestic responses led to the concept of "original antigenic sin" (1-3). Such cross-reactivity noted first for antibodies to different influenza virus strains was also observed for antibodies to the haptens dinitrophenyl (DNP) 1 and trinitrophenyl (TNP) (4), and to sulfanilate-metanilate haptenic groups (5). In the system employing two strains of influenza virus as cross-reactlng antigens, the phenomenon was interpreted as a restricted anamnestic response to the primary antigen mediated by a trapping mechanism deflecting antigen from one kind of cell to another (3). Chronologically analyzed, this process involved an early production of cross-reactive secondary response antibody followed by the primary response to the boosting antigen (3). Antibodies stimulated by either DNP or TNP protein conjugates had a high degree of immunological specificity. The anamnestic recall phenomenon in this system was explained as a process of degeneracy of the immune response reflected by high affinity antibodies with low specificities (4). The serological group specificity of streptococci results from distinct polysaccharide moieties characteristic for every group of streptococci (6, 7). For example, the group specific polysaccharides of A (A-CHO) and A-variant (Av-CHO) streptococci share a branched backbone of rhamnose. They differ, however, by the terminal N-acetylglucosaminide residues in Group A streptococci linked fll ~ 3 glycosidically to the rhamnose side chains. This hexosamine moiety confers serological specificity to the Group A carbohydrate (6, 7). The A-variant polysaccharide, composed predominantly of rhamnose, has its cross-reactive serological representation in di- and trisaccharides of rhamnose of unknown linkage (6, 7). Since antibodies of restricted electrophoretic mobility are inducible against both the streptococcal Groups A and A-variant polysaccharide (8, 9), the a Abbreviations used in this paper: A-CHO, streptococcal Group A polysaccharides; A-CHO tyr 125I and A-CHO tyr 131I, streptococcal Group A polysaccharides tyraminated and either