Lymphocyte Survival and Lineage Determination through Positive Selection
Harald von Boehmer
In the early 1970s it was known that T and B lymphocytes proliferate and differentiate into effector cells upon stimulation with antigen. Throughout the life of the organism, T lymphocytes are generated from stem cells in the thymus whereas B cells develop from stem cells in the bone marrow. Each B lymphocyte expresses cell surface receptors (antibodies) with a single specificity for antigen. Some B cells develop into plasma cells that secrete antibodies. The antigen receptor on T cells was not yet known but it was clear that T cells could be divided into distinct subsets. Upon encountering antigen, they generate either cytolytic T lymphocytes (CTL) able to lyse other cells or T helper cells (TH) that stimulate B cells to produce antibodies. It was also known that T cells could not be stimulated directly by soluble antigen but that T cell activation required contact with other cells, such as macrophages and B cells, that somehow presented these antigens on their cell surface.
MHC-restricted antigen recognition
Between 1972 and 1974 several investigators discovered that the specific interaction of T cells with antigen-presenting cells was influenced by cell surface molecules encoded in the major histocompatibility complex (MHC). These molecules exist in two forms, class I and class II MHC molecules. In mice the former are encoded by K and D loci and expressed on practically all nucleated cells, whereas the latter are encoded in A and E loci and found on a subset of cells, including macrophages, dendritic cells, lymphocytes and thymic epithelial cells. MHC molecules are highly polymorphic, that is, different members of the same species express different alleles at each of the loci. The sum of different MHC loci as present on one chromosome is referred to as MHC haplotype (we call them a, b, c etc.). MHC molecules themselves act as antigens that elicit strong immune responses in organisms that do not express those MHC molecules. For example, mice of MHC genotype bb reject foreign skin grafts derived from mice of genotype aa, whereas mice of genotype ab or aa accept the grafts.
A biologically more relevant role for MHC molecules was discovered when immune responses to non-MHC antigens like viral or bacterial proteins were studied. The most cited experiment (Zinkernagel and Doherty, 1974) showed that the ability of CTL to lyse cells depended on the MHC genotype of target cells. Figure 1 shows the use of an in vitro assay in which CTL generated in response to a viral infection in a mouse with MHC genotype aa were able to lyse target cells (infected with the same virus) of MHC genotype aa and ab but not bb. This phenomenon is called MHC restriction.
The type of MHC molecule involved in restriction depends on the type of T lymphocyte. For CTL, at least one type of class I MHC molecules, but not the class II MHC molecules, of the immunized mouse had to match those of the target cells. Experiments similar to the one shown in Figure 1 showed that the interactions of activated TH cells with macrophages or B cells are also restricted by MHC molecules (cited in Zinkernagel and Doherty, 1974), although TH cells are restricted by MHC class II molecules.
At the time, two explanations for the molecular mechanism underlying MHC restriction were offered:
The former possibility would be ruled out if it could be shown that T cells were able to interact in a MHC-restricted and antigen-specific manner with cells expressing allogeneic MHC molecules (i.e., foreign MHC molecules that are not shared by the T cells). However, this was not an easy experiment to do because allogeneic MHC molecules themselves stimulate T cells, making it difficult to study a response to a defined non-MHC-antigen when presented in the context of allogeneic MHC molecules. This problem was solved by inducing specific immunological unresponsivness (tolerance) to allogeneic MHC molecules in hemopoietic chimeras.
One possibility was that in order for a T cell to interact specifically with a target cell, both cells must express the same MHC molecules, which might participate in a like-like interaction.
Alternatively, T cells might express receptor(s) that have the dual properties of recognizing polymorphic MHC molecules as well as antigens on target cells.
Figure 2 shows that these chimeras were produced by first destroying lymphocytes and other hemopoietic cells of MHC heterozygous (a × b) F1 (host) mice by irradiation. Then each host was reconstituted with equal numbers of MHC homozygous aa and bb (donor) stem cells that would generate lymphocytes in the new environment. Both donor-derived aa and bb lymphocytes, having developed in the presence of host and donor tissue expressing allogeneic a- and b-encoded MHC molecules, acquired specific tolerance of each other’s MHC molecules (cited in von Boehmer, Hudson, and Sprent, 1975). This made it possible to immunize with antigen and then to test whether a subset of purified aa T helper cells was able to interact in an antigen-specific manner with bb B cells (von Boehmer, Hudson, and Sprent, 1975).
The results showed that aa T cells, which had matured in and were immunized in mice expressing both a-and b-encoded MHC molecules, could interact efficiently in an antigen-specific manner with bb B cells. Analogous observations were made with CTL. These results ruled out the possibility that MHC restriction required expression of the same MHC molecules on T cells and their target cells. More significantly, the experiments indicated that T cells of one particular MHC genotype had the genetic potential to interact in a MHC-restricted and antigen-specific manner with targets expressing any other MHC molecule. This fact became known as the "plasticity" of MHC-restricted antigen recognition.
Elaborating on the concept of "positive selection"
The chimera experiments (von Boehmer, Hudson, and Sprent, 1975) led us to the conclusion that T cells can express receptors that recognize a large variety of both antigens and MHC molecules. Two models were proposed to account for how this recognition might occur:
The dual recognition model, but not necessarily the altered MHC model, posed a dilemma when considered together with the observed "plasticity" of MHC-restricted antigen recognition. If the entire genetic potential for TCRs were to be expressed by mature T cells of a certain animal, the T cell population as a whole will have specificities that are restricted by a variety of MHC haplotypes. Only very few T cells will actually have a specificity that is restricted by a MHC molecule corresponding to the genotype of a particular animal. All the other T cells would in fact be useless since they would never encounter antigens in the context of allogeneic MHC molecules. For example, in an animal of MHC genotype aa, all those T cells that are restricted by allogeneic MHC molecules (b, c, d, etc.) would never be able to recognize a target because of the absence of the allogeneic MHC molecules.
The "altered MHC" model proposed that the collision of non-MHC and MHC molecules on the surface of cells resulted in specific conformational changes of the latter that were then recognized by the T cell receptor (TCR). Because of the putative specific interactions of vastly different cell surface molecules this model represented somewhat of a biochemist’s nightmare.
The alternative "dual recognition" model postulated that a given T cell had two distinct binding sites: one for (non-MHC) antigens, and another for MHC molecules.
As a way out of this dilemma, it was proposed by several investigators that T cells undergo a process of adaptive differentiation or "positive selection" mediated by self-MHC molecules and somehow resulting in the accumulation of T cells that can see foreign antigens in the context of self-MHC molecules. This process could explain the presence of allogeneic-MHC-restricted T cells in a chimeric environment where developing cells were exposed to allogeneic MHC molecules throughout their development. In the absence of allogeneic MHC molecules, however, allogeneic-MHC-restricted T cells would not accumulate even though some self-MHC-restricted TCRs may "cross-react" with allogeneic MHC molecules.
An experiment that was consistent with positive selection was published in 1977 by Bevan, who studied MHC restriction in chimeric mice (Bevan, 1977). MHC homozygous aa or bb (host) mice were X-irradiated, reconstituted with MHC heterozygous (a × b) F1 (donor) stem cells and subsequently immunized with non-MHC antigens. Figure 3 shows that in an in vitro assay, the majority of the CTL resulting from immunization of aa and bb hosts recognized these antigens presented in the context of host type a and b MHC molecules, respectively, as present on radioresistant tissue of the host. A minority of CTL in the aa and bb hosts was specific for antigens presented in the context of donor type b and a MHC molecules, respectively. These results suggested that the host MHC molecules expressed on radioresistant tissue can impose a bias in the specificity of CTL toward their haplotype.
These findings could be explained in different ways:
Zinkernagel was better able to distinguish between these explanations by clearly separating events that take place during T cell maturation in the thymus from events that occur once mature T cells are intentionally exposed to antigen (Zinkernagel et al., 1978). As shown in Figure 4, MHC heterozygous (a × b) F1 stem cells were first allowed to mature in a thymectomized and X-irradiated MHC heterozygous (a × b) F1 host that was also transplanted with an X-irradiated MHC a or b homozygous thymus. Next, the resulting mature T cells that had matured in and left the transplanted thymus were then immunized with antigen. Zinkernagel found that most of the T cells that responded specifically to the antigen were restricted by MHC molecules of the same type as those present on the radioresistant thymic tissue in which the T cells had developed. He therefore concluded that prior to their encounter with antigen, immature T cells were somehow positively selected by MHC molecules expressed on thymic tissue.
There could be some sort of selection in the process of T cell maturation resulting in a bias toward one type of MHC molecule. Bevan formulated a model of this type that based positive selection on Jerne’s idea that TCR diversity is generated by somatic mutation (Jerne, 1971). This model proposed that immature T cells, upon specific recognition of MHC molecules expressed on radioresistant thymic epithelium, proliferate and accumulate mutations in their antigen-receptor genes such that the resulting receptors recognize "altered self-MHC" molecules (i.e., an altered form of MHC a-type or b-type molecules).
- At the time, an alternative and more trivial explanation of the data could have been that the bias in terms of MHC-restricted cytolytic activity stemmed from the fact that antigen was more efficiently presented in the context of the much more numerous MHC molecules that were present on both host and donor tissue, than in the context of the less numerous MHC molecules that were present on donor tissue only.
Since the experiments of Bevan and Zinkernagel measured specificity of effector T cells after considerable expansion rather than directly analyzing cellular selection of nonimmunized precursor cells, it is perhaps not surprising that their interpretations were challenged in subsequent years by many investigators, some of whom claimed to readily detect a high frequency of allogeneic-MHC-restricted T cells in normal mice (Nagy and Klein, 1981). Others concluded that the bias in MHC restriction observed in Bevan’s and Zinkernagel’s experiments resulted from suppression of the response restricted by donor-type MHC molecules by specific regulatory T cells rather than from positive selection of T cells by MHC molecules expressed on host tissue (Smith and Miller, 1980). Thus, from the observed biases in MHC restriction of effector cells, it was impossible to conclusively demonstrate the existence of positive selection of nonimmunized developing T cells let alone show that it relied on a mechanism such as clonal expansion, mutation or selective survival. It was just as impossible to demonstrate from the observation of acquired immunological tolerance in the chimeric mice the existence of negative selection (i.e. the deletion of certain lymphocytes).
Progress in immunology helps the design of more conclusive experiments
From the ensuing debates on the existence of positive selection, it became clear that conclusive experiments required identification of antigen-specific TCRs so that it would be possible to directly track nonimmunized T cells expressing specific receptors throughout their development. A variety of approaches during the late 1970s and early 1980s identified TCRs and their ligands. It was shown that cells can break down proteins into peptides that can then bind within a groove in class I and class II MHC molecules, and that these MHC-peptide complexes are transported to the cell surface. Crystallographic studies revealed that polymorphic MHC residues form part of this groove and part of the surface of MHC molecules that is in contact with TCRs (reviewed in von Boehmer, 1990). It also became evident that certain antigen presenting cells can efficiently internalize and degrade all kinds of proteins and can also express co-stimulatory molecules on their cell surface that enable them to activate na?ve, nonimmunized T cells ( Figure 5). Furthermore, it was found that as a general rule class I MHC molecules usually present peptides derived from proteins made inside the cell (e.g. virally-encoded proteins), whereas class II MHC molecules mostly present peptides from proteins that enter the cell (e.g. bacterial proteins).
It was also established that TH cells and CTL express CD4 or CD8 coreceptors, respectively, and that CD4 and CD8 bind to non-polymorphic parts of class II or class I MHC molecules, respectively ( Figure 5). Most significantly, in 1984 the bold strategy of Davis resulted in the identification of TCR genes (Hedrick et al., 1984), which were shown to rearrange much like antibody genes and to encode and subunits that form a TCR heterodimer. Thus, a decade after the introduction of molecular techniques into immunology, one could approach old questions with new experimental tools.
In 1986 we reported that the MHC-restricted specificity of a T cell clone could indeed be transferred by transfecting its TCR and genes into another cell (Dembi et al., 1986). We then proceeded to construct TCR transgenic mice in collaboration with several other labs in the hope of obtaining mice in which a significant fraction of nonimmunized T cells would express the transgenic receptor of defined specificity. Such animals would allow us to follow the developmental fate of T cells expressing this receptor under different experimental conditions. The goal was to test a detailed hypothesis, formulated in 1986 (see von Boehmer, 1988 and references therein) and proposing that positive selection in the thymus does not simply reflect the MHC-induced expansion of certain clones but reflects rescue from programmed cell death. Specifically, it was suggested that immature CD4–8– and CD4+8+ thymocytes undergo sequential TCR and TCR rearrangement and that the vast majority of immature and not yet functional CD4+8+ TCR -expressing cells is destined to die because they express allo-MHC-restricted receptors unable to bind self-MHC molecules. Only CD4+8+ thymocytes with self-MHC restricted TCRs binding with low affinity to thymic MHC molecules would be rescued from cell death. Furthermore, depending on the specificity of the TCR for class I and class II MHC molecules, the cells would be selected to develop along the CD4+8– TH or CD4–8+ CTL pathway, respectively. In contrast, T cells with receptors binding with high affinity to self-MHC molecules would be immediately deleted ( Figure 6).
The experiment: lymphocyte survival and lineage determination through positive selection
We isolated the receptor genes of a T cell clone recognizing a specific peptide (derived from the male-specific antigen HY), which was presented by class I MHC Db molecules (i.e. the class I molecule encoded in the D region of the MHC b haplotype). We then made transgenic mice that expressed this TCR on a large fraction of their T cells. These T cells could be identified using flow cytometry with fluorochrome-labeled monoclonal antibodies directed against CD4 or CD8 cell surface molecules or the transgenic receptor. The choice of receptor allowed us to readily assess positive and negative selection in female and male mice, respectively. It turned out that the introduction of the transgenes did not completely prevent the expression of endogenous receptor genes. This situation was avoided in experiments in which the transgenes were introduced into SCID mice that because of a genetic defect could not rearrange endogenous receptor genes, in which case the transgenic TCR was the only expressed receptor.
The results in the transgenic mice entirely supported the hypothesis described above. The observations in male TCR transgenic mice of the MHC b haplotype that naturally express MHC Db molecules containing the HY peptide were in agreement with observations made by Kappler and colleagues (Kappler, Roehm, and Marrack, 1987). These authors studied T cells specific for super-antigens, which bind TCRs at sites that are different from those contacted by peptide-MHC complexes. Figure 6 shows that the numbers of immature CD4+8+ thymocytes were severely reduced, and mature CD4+8– or CD4–8+ T cells were absent. We concluded that HY peptides presented by Db MHC molecules deleted HY-specific T cells with the transgenic TCR before acquiring functional maturity. Thus, negative selection served as a plausible mechanism of acquired immunological tolerance, at least to antigens that are present in the thymus (Kisielow et al., 1988).
The analysis of TCR transgenic female mice showed what happens when the Db MHC molecules are expressed in the absence of the HY peptide ( Figure 6). The mice contained a high proportion of immature, nonfunctional CD4+8+ thymocytes and mature functional CD4–8+ cells. This indicated that in the presence of the restricting MHC molecules, but in the absence of the cognate peptide, cells with the transgenic TCR could mature completely, but only as CD4–8+ and not CD4+8– cells (Kisielow et al., 1988; Teh et al., 1988; Scott et al., 1989). It was then analyzed whether the expression of Db MHC molecules on radioresistant thymic tissue was required for positive selection of mature CD4–8+ T cells with the transgenic TCR. We injected hemopoietic stem cells from MHC b-type, female transgenic mice into X-irradiated female mice that expressed different MHC haplotypes as indicated in Figure 7 and analyzed the appearance of CD8+ positive T cells with high levels of the transgenic TCR (representing mature CD4–8+ thymocytes). The experiment showed that expression of Db MHC molecules by radioresistant thymic tissue was both essential and sufficient to positively select CD4–8+ T cells with the transgenic TCR (Kisielow et al., 1988; Teh et al., 1988; Scott et al., 1989). Since the injected hemopoietic cells expressed Db MHC molecules themselves, our data indicated that their expression by thymic hemopoietic cells was not sufficient for efficient positive selection.
The novel aspect of these findings was not merely the existence of positive selection of immature T cells, but the realization that it represented a step essential for the formation of mature T cells (Scott et al., 1989). The most exciting aspect of being able to follow developing cells with a single MHC-restricted receptor was, however, the conclusion that the receptor specificity for class I or class II MHC molecules indeed determined whether the positively selected cells would be of the CD4–8+ killer or CD4+8– helper phenotype, respectively (Teh et al., 1988). Thus, positive selection conferred not only survival, but also determined lineage fate. This was further supported by the study of mice with class II MHC-restricted TCRs (Kaye et al., 1989) in which positive selection resulted in CD4+8– but not CD4–8+ T cells. It was the clarity of the results obtained in TCR transgenic mice that led to the somewhat euphoric and populistic formulation that "The thymus selects the useful, neglects the useless and destroys the harmful." The positively selected T cells are useful because their receptors, selected on the basis of a low affinity interaction with self-MHC molecules have a higher probability of recognizing a different self-MHC-peptide complex with high affinity when compared to receptors selected by a different MHC molecule, and presumably when compared to non-selected receptors. The results also clearly showed that T cells recognizing allogeneic MHC molecules are self-MHC-selected and thus potentially useful. Thus, positive and negative selection represent evolutionarily selected mechanisms that increase the fitness of the individual by optimizing recognition of foreign antigens presented by self-MHC molecules and by avoiding autoaggression.
The experiments in TCR transgenic mice provided a coherent picture of the principles that govern the immune system’s adaptation to self-MHC molecules and antigens by positive and negative selection and highlighted the important role of programmed cell death in this process. The results also raised questions with regard to the molecular mechanism of cell death, the signaling pathways that are involved in positive and negative selection as well as the nature of the ligands that mediate positive selection. Although cell death and signaling pathways have been identified, we still have no clue of how positive selection by low affinity TCR-MHC molecule binding selects TCRs that have a higher probability of binding with high affinity to peptides presented by the same as opposed to different MHC molecules. In this context we also need to understand how self-MHC-selected TCRs recognize allogeneic MHC molecules. Other experiments concerning the mechanisms by which TCRs determine CD4/CD8 lineage fate have yielded intriguing but not yet completely conclusive results (von Boehmer, 2000). Our own studies in TCR transgenic mice have identified yet another, earlier checkpoint in T cell development at which thymocytes which have succeeded in productive TCR chain rearrangement are positively selected by the pre-TCR (Groettrup et al., 1993), which also is involved in the early lineage fate decision between and T cells. Thus, survival and lineage fate determination through positive selection represent fundamental principles that govern the development of both T and B lymphocytes (reviewed in von Boehmer, 1994) at distinct developmental stages and may well prove to represent fundamental principles in developmental biology.
The close collaboration with my friend and colleague Pawel Kisielow was essential for the successful outcome of these studies. It is also clear that the generation of several transgenic lines by Anton Berns at the Netherlands Cancer Institute (later produced by Horst Blüthmann’s laboratory at Hoffmann-LaRoche) resulting in the identification of the T cell receptor enhancer and the skills of Michael Steinmetz’s group at the Basel Institute for Immunology in cosmid cloning were crucial to the success of these studies.
Harald von Boehmer was born in 1942 and studied Medicine in Goettingen, Freiburg and Munich where he received his M.D. in 1968. In 1974 he received a Ph.D. from the University of Melbourne, Australia. Postdoctoral stays were at the Max Planck Institute for Biochemistry, Munich and the Walter and Eliza Hall Institute, Melbourne. He became a member (1973) and a permanent member (1976-1996), of the Basel Institute for Immunology, then Director of Unité 373 INSERM, Paris (1996-2000) and moved to Harvard Medical School(1999). Academic appointments include Adjunct Professor, Medical College, Gainesville, Florida (1981), Professor of Immunology, University of Basel (1992) and University Descartes, Paris (1996-2000), and Professor of Pathology, Harvard Medical School (1999). von Boehmer is a member of the Academia Europaea, the European Molecular Biology Organization, the Institut Universitaire de France and the German Immunological Society and honorary member of the American Association of Immunologists and the Scandinavian Society of Immunology. Because of American roots, he is also a lifetime member of the Massachusetts Society of Mayflower Descendants. Harald von Boehmer received the Avery-Landsteiner Prize (Aachen 1990) and shared with Gottfried Schatz and Nicole le Douarin the Louis Jeantet Prize for Medicine (Geneva 1990), with Philippa Marrack and John Kappler the Paul Ehrlich Prize (Frankfurt, 1993) and with Pawel Kisielow and Klaus Rajewsky the Koerber Prize for European Science (Hamburg 1997).