The question of how the immune system can have specificity for a given molecule out of an endless realm of possibilities, sometimes irreverently referred to as the G.O.D. (Generation of Diversity) problem, took nearly a century to answer. Physicians and scientists had long been aware that the immune system responds specifically to individual Ags and that subsequent exposure to the same Ag causes a more rapid and vigorous immune response. What possible mechanism could confer such a prodigious and versatile defense?
The first coherent hypothesis arrived in 1900 in the form of Paul Ehrlich’s (1) “side chain theory.” The idea was that Ab-synthesizing cells (T cells were not to be discovered for many decades) bore a large variety of receptors on their surfaces, each able to interact with a different Ag. When a particular Ag bound its corresponding receptor, much like a key finding the right lock, that receptor or side chain would fall from the cell surface. In compensation, the cell would then overproduce that particular side chain, and the excess would spill into the bloodstream as circulating Abs. The infectious agent thus was proposed to “select” the side chain that was then produced by the cell. This selective theory held sway until a few decades later, when Karl Landsteiner (2) found that he could raise Abs not only against nitrophenol but also against each compound that resulted from shifting the position of the nitro group around the phenol ring. Additional experiments showed that Abs could be raised against virtually any molecule, including ones that would never be found in nature. As well as proving that Abs are exquisitely specific, this work rendered the idea highly implausible that immune cells could have evolved a side chain for every conceivable compound.
What if the answer lay in protein structure itself? Could an Ag act as a template for Ab production, much like a die for casting coins? In the template model favored by Linus Pauling (3), among others, interactions with different Ags would induce Abs to adopt different folding patterns. Once folded into a specific shape, the Ab would subsequently interact only with that particular Ag. This template hypothesis, an instructive theory, obviated the requirement for an infinite number of side chains. Unfortunately, it could not explain affinity maturation, self/nonself discrimination, how an Ab continues to be produced in the absence of Ag, or why secondary responses are much more rapid and robust than the initial immune response. Further, the discovery of DNA and the attendant formulation of the central dogma of molecular biology (i.e., that information flows only from DNA to RNA to protein) made the template theory seem as implausible as the selective hypothesis; there was as yet no known mechanism to allow information to move from protein to protein or from a protein back to the gene. These difficulties led Jerne, Talmage, and Burnet to refine the earlier selective theories into what Burnet (4) in 1957 termed clonal selection. This model, now accepted, holds that there is a population of cells in the body that are already differentiated in their ability to bind specific Ag. Each of these cells bears only one type of Ag receptor and must await a chance encounter with an Ag that happens to fit. The Ag receptors are encoded in the genes of each lymphocyte by a random process that generates billions of receptor specificities before any encounter with Ag. When Ag binds to a receptor, that particular lymphocyte is prompted to proliferate, and each of the progeny bears the same receptor as the parent cell. The question now became how an unlimited number of specificities could possibly be encoded in a limited number of Ag receptor genes.
This genetic paradox was thrown into sharp relief by the discovery that Ab proteins consist of two regions: one relatively constant and the other highly variable. In other words, Abs gave every sign of being encoded by more than one gene. How could this be, given the one gene, one protein model? Following evidence rather than dogma, Dreyer and Bennett (5) proposed in 1965 that the V and C regions are the products of more than one gene. Their model was both elegant and subversive. Yet just over a decade later, Susumu Tonegawa’s Nobel Prize-winning experiments provided incontrovertible proof. Comparing DNA from mouse embryos with that from myeloma cells (corresponding to differentiated B cells), Tonegawa (6) showed that the genes encoding the V and C regions were separate in germline DNA but close together in mature B cells. The inescapable conclusion was that somatic DNA rearrangement had occurred.
By the time graduate student David Schatz joined David Baltimore’s lab, the Ag receptor loci were mapped in great detail; virtually everything was known about the substrates for the V(D)J reaction, but absolutely nothing had been gleaned about the enzyme(s) that select, cut, and paste the various gene segments. The only factor known to participate in the reaction at all was terminal deoxynucleotidyl transferase or TdT, whose role Baltimore had correctly hypothesized in an elegant Nature paper several years before the famous Tonegawa (7) experiments. As biochemical methods had proved unfruitful, the Baltimore lab turned to somatic cell genetics. In 1988, Schatz and Baltimore (8) reported in Cell that they had induced fibroblasts to carry out V(D)J recombination through the transfer of genomic DNA. This result suggested that the factors responsible for recombination existed at a single genetic locus that might be isolated through serial transfections of genomic DNA. It seemed unlikely that one enzyme could carry out all the various steps of recombination, but this was the hypothesis that led to a second Cell article (9), presented as this month’s Pillars of Immunology selection. Remarkably, the serial transfection approach worked, and recombination activating gene 1 (RAG-1) was identified. At the time it was not clear whether RAG1 was part of the recombinase itself or merely activated other enzymes that conducted recombination, but finally the field had a foothold for solving the V(D)J puzzle—or at least a toehold; expression of RAG1 cDNA in fibroblasts produced disappointingly inefficient recombination. In fact, the cDNA expression vector was little better at stimulating recombination than bulk genomic DNA. This perplexing observation led Schatz and colleagues to surmise that a second gene located very near RAG-1 was also required for V(D)J recombination; no more likely a possibility than the one that started the project, this one too proved correct. Continued searching rapidly located RAG-2 and led to the publication of the third article (10) in 3 years homing in on the V(D)J recombinase.
The G.O.D. questions did not end with this triumvirate, however. It took several more years of work to understand that RAG1/RAG2 are actually the enzymatic machinery of recombination in both B and T lymphocytes (11). Soon thereafter, the Gellert and Mizuuchi (12) laboratories showed that the mechanism of site-specific DNA cleavage by the RAG proteins bore striking similarities to the hydrolysis and transesterification reactions conducted by certain cut-and-paste transposases and retroviral integrases. This astute observation led to the current hypothesis that adaptive immunity evolved through transference of the RAG transposon into an ancestral Ag receptor gene in jawed vertebrates (13, 14).
In retrospect, the early attempts to account for the generation of diversity were prescient, all touching on aspects of biological truth. The insightful hypotheses that preceded, often by decades, the capacity to test them serve as a reminder that technological innovation is a mere handmaiden to scientific imagination and that there is still much room for theoretical ingenuity in immunology.