7-17. Immature B cells that bind self antigens undergo further receptor rearrangement, or die, or are inactivated

Once an immature B cell expresses IgM on its surface (sIgM), its fate is guided by the nature of the signals it receives through its antigen receptor. This was first demonstrated by experiments in which antigen receptors on immature B cells were experimentally stimulated in vivo using anti-μ chain antibodies (see Appendix I, Section A-10); the outcome was elimination of the immature B cells. More recent experiments using mice expressing B-cell receptor transgenes have confirmed these earlier findings, but have also shown that immediate elimination is not the only possible outcome of binding to a self antigen. Instead, there are four possible fates for self-reactive immature B cells, depending on the nature of the ligand to which they are capable of binding (Fig. 7.25). These fates are cell death by apoptosis; the production of a new receptor by receptor editing (Fig. 7.26); the induction of a permanent state of unresponsiveness to antigen; and ignorance. An ignorant cell is defined as one that has affinity for a self antigen but that does not sense the self antigen, either because it is sequestered, is in low concentration, or does not cross-link the B-cell receptor.

Figure 7.25

Binding to self molecules in the bone marrow can lead to the death or inactivation of immature B cells. Left panels: when developing B cells express receptors that recognize multivalent ligands, for example, ubiquitous self cell-surface molecules such (more...)

Figure 7.26

Replacement of light chains by receptor editing can rescue some self-reactive B cells by changing their antigen specificity. When a developing B cell expresses antigen receptors that are strongly cross-linked by multivalent self antigens such as MHC molecules (more...)

Apoptosis and elimination of immature self-reactive B cells seems to predominate when the interacting self antigen is multivalent, for example the multiple copies of an MHC molecule on a cell surface. This antigen-induced loss of cells from the B-cell population is known as clonal deletion. The effect of encounter with a multivalent antigen was tested in mice transgenic for both chains of an immunoglobulin specific for H-2K^b MHC class I molecules. In such mice nearly all the B cells that develop bear the anti-MHC immunoglobulin as sIgM, because the presence of the already rearranged transgenes inhibits rearrangement of the endogenous immunoglobulin genes (see Sections 7-9 and 7-10). If the transgenic mouse does not express H-2K^b, normal numbers of B cells develop, all bearing transgene-encoded anti-H-2K^b receptors. However, in mice expressing both H-2K^b and the immuno-globulin transgenes, B-cell development is blocked. Normal numbers of pre-B cells and immature B cells are found, but B cells expressing the anti-H-2K^b immunoglobulin as sIgM never mature to populate the spleen and lymph nodes; instead, most of these immature B cells die in the bone marrow by apoptosis. This is because the anti-H-2K^b immunoglobulin on the immature B cells interacts strongly with the H-2K^b molecules on the bone marrow stromal cells.

Closer analysis of this experimental system and others like it revealed the surprising finding that clonal deletion was not the only outcome in these circumstances. There was, in fact, an interval before cell death during which the self-reactive B cell might be rescued by further gene rearrangements that replaced the self-reactive receptor with a new receptor that was not autoreactive (see Fig. 7.26). This mechanism for replacing receptors, termed receptor editing, works as follows. When an immature B cell first expresses a light chain, and produces sIgM, the RAG genes are still being expressed and RAG protein is being made. If the B cell does not interact with self antigen and there is no strong cross-linking of sIgM, gene rearrangement ceases and the B cell continues its development; RAG protein begins to decline but does not completely disappear until the B cell reaches full maturity in the spleen. However, in self-reactive B cells, strong cross-linking of sIgM, as a result of encounter with the self antigen, halts further development, and RAG gene expression continues at levels similar to that in pre-B cells undergoing light-chain gene rearrangement. As a consequence of the continued presence of recombinase, light-chain gene rearrangement continues, even though the cell has already made one productive rearrangement at this locus. The light-chain loci are able to make numerous successive rearrangements (see Fig. 7.16), and these rearrangements can rescue immature self-reactive B cells by deleting or displacing the rearrangement that encodes the self-reactive receptor and replacing it with another sequence.

This continuation of light-chain gene rearrangement has parallels with the continuation of α-gene rearrangement in developing T cells, but it should be emphasized that in B cells it only occurs if the receptor encounters a strongly cross-linking antigen. In T cells, in contrast, α-gene rearrangement and ‘receptor editing’ continue as part of the normal developmental program until the cell is positively selected or dies.

More recently, receptor editing has been shown unambiguously in mice bearing transgenes for autoantibody heavy and light chains that have been placed within the immunoglobulin loci by the homologous recombination method explained in Appendix I, Section A-47. The transgene imitates a primary gene rearrangement and is surrounded by unused endogenous gene segments. In mice that express the antigen recognized by the transgene-encoded receptor, the few mature B cells that emerge into the periphery have used these surrounding gene segments for further rearrangements that replace the autoreactive light-chain transgene with a nonautoreactive rearranged gene.

At a light-chain locus, the multiplicity of V and J segments allows the unused V and J gene segments to be selected for multiple further rearrangements (see Fig. 7.16). In addition, cells that have exhausted the J_κ regions can rearrange the λ locus. It is not clear whether receptor editing occurs at the heavy-chain locus. There are no available D segments at a rearranged heavy-chain locus, so a new rearrangement cannot simply occur by the normal mechanism and at the same time remove the preexisting one. Instead a process of V_H replacement may use embedded recombination signal sequences in a recombination event that displaces the V gene segment from the self-reactive rearrangement and replaces it with a new V gene segment. This has been observed in some B-cell tumors, but whether it occurs during normal B-cell development in response to signals from autoreactive B-cell receptors is not known.

It was originally thought that successful production of a heavy chain and a light chain caused the almost instantaneous shut down of further light-chain locus rearrangement and that this ensured both allelic and isotypic exclusion (see Section 7-10). The unexpected ability of self-reactive B cells to continue to rearrange their light-chain genes, even after having made a productive rearrangement, has raised questions about this supposed mechanism of allelic exclusion.

Undoubtedly, however, the decline in the level of RAG protein that occurs after a successful nonself-reactive rearrangement is crucial to maintaining allelic exclusion, as it will reduce the chance of a subsequent rearrangement. Furthermore, any additional productive rearrangement that did still occur would not necessarily breach allelic exclusion: if it occurred on the same chromosome it would eliminate the existing productive rearrangement, while if it occurred on the other chromosome it would be nonproductive in two out of three cases. Thus, the observed fall in the level of RAG protein may be sufficient to ensure that a second productive rearrangement is rarely made, and could be the principal, if not sole, mechanism behind allelic exclusion. Consistent with this idea, it appears that allelic exclusion is not absolute, since there are rare B cells that express two light chains.

We have so far discussed the fate of newly formed B cells that undergo multivalent cross-linking of their sIgM. Those immature B cells that encounter more weakly cross-linking self antigens of low valence, such as small soluble proteins, respond differently. In this situation, the self-reactive B cells tend to be inactivated and enter a state of permanent unresponsiveness, or anergy, but do not immediately die (see Fig. 7.25). Anergic B cells cannot be activated by their specific antigen even with help from antigen-specific T cells (see Section 1-15). Again, this phenomenon was elucidated using transgenic mice. When hen egg lysozyme (HEL) is expressed in soluble form from a transgene in mice that are also transgenic for high-affinity anti-HEL immunoglobulin, the HEL-specific B cells mature but are unable to respond to antigen. The anergic cells retain their IgM within the cell and transport little to the surface. In addition, they develop a partial block in signal transduction so that, despite normal levels of HEL-binding sIgD, the cells cannot be stimulated by cross-linking this receptor. It seems that signal transduction is blocked at a step before the phosphorylation of the Igα and Igβ chains (see Section 6-6), although the exact step is not yet known. The signaling defect may involve the inability of B-cell receptor molecules on tolerant B cells to enter regions of the cell in which important other signaling molecules normally segregate in order to transmit a complete signal subsequent to antigen binding. Furthermore, cells that have received an anergizing signal may upregulate molecules that inhibit signaling and transcription.

Anergic B cells are not only impaired in signal transduction and expression of sIgM. Their migration within peripheral lymphoid organs is altered and their life-span and ability to compete with immunocompetent B cells are compromised. Mature B cells recirculate through lymphoid follicles during their life-span. In normal circumstances, in which B cells binding soluble self antigen are in a minority, anergic B cells are detained in the T-cell areas of the peripheral lymphoid tissue and are excluded from lymphoid follicles. As anergic B cells cannot be activated by T cells, and T-cell help will in any case not be available for self antigens, to which the T cells will be tolerant, these self-reactive B cells will not be activated to secrete antibody. Instead they die relatively soon, presumably because they fail to get survival signals from T cells, thus ensuring that the long-lived pool of peripheral B cells is purged of potentially self-reactive cells.

The fourth potential fate of self-reactive immature B cells is that nothing happens to them; they remain in a state of immunological ‘ignorance’ of their self antigen (see Fig. 7.25). It is clear that some B cells with a weak but definite affinity for a self antigen, mature as if they were not self-reactive at all. Such B cells do not respond to the presence of their self-antigen because it interacts so weakly with the receptor that little if any, intracellular signal is generated when it binds. Alternatively, some self-reactive B cells may not encounter their antigen at this stage because it is not accessible to B cells developing in the bone marrow and spleen. The maturation of these B cells reflects a balance that the immune system strikes between purging all self-reactivity and maintaining the ability to respond to pathogens. If elimination of self-reactive cells were too efficient, the receptor repertoire might become too limited and thus not be able to recognize a wide variety of pathogens. Some autoimmune disease might be the price of this balance, as it is very likely that these low-affinity self-reactive lymphocytes can be activated and cause disease under certain circumstances. Thus these cells can be thought of as the seeds of autoimmune disease. Normally, however, ignorant B cells will be held in check by a lack of T-cell help, the continued inaccessibility of the self antigen, or the tolerance that can be induced in mature B cells, as described in the next section.

7-18. Mature B cells can also be rendered self-tolerant

It is likely that most self-reactive B cells will encounter their antigens while still immature, as many self antigens circulate through tissues in soluble form or are expressed by many different cell types, including those in the bone marrow. Nonetheless, some self antigens are not present in the tissues through which immature B cells pass, and thus B cells expressing receptors specific for these antigens will survive to become mature. Lack of T-cell help, or an inability of the antigen to strongly cross-link the B-cell receptor will usually prevent these cells from becoming activated and causing problems. However, mechanisms for eliminating or inactivating mature self-reactive B cells, and even activated self-reactive B cells, also exist.

For example, B cells expressing immunoglobulin specific for H-2K^b MHC class I molecules are deleted even when, in transgenic animals, the expression of the H-2K^b molecule is restricted to the liver by use of a liver-specific gene promoter. B cells that encounter strongly cross-linking antigens in the periphery undergo apoptosis directly, unlike their counterparts in the bone marrow, which attempt further receptor rearrangements instead. The different outcomes may be due to the fact that the B cells in the periphery are more mature and can no longer rearrange their light-chain loci. On the other hand, stromal signals may be important; recent data suggest that stromal cells in the spleen instruct B cells to respond in a different way to signals delivered through the antigen receptor compared with the stromal cells of the bone marrow.

Mature B cells that encounter and bind an abundant soluble antigen become anergized. This was demonstrated in mice by placing the HEL transgene under the control of an inducible promoter that can be regulated by changes in the diet. It is thus possible to induce the production of lysozyme at any time and thereby study its effects on HEL-specific B cells at different stages of maturation. These experiments have shown that both mature and immature B cells are inactivated when they are chronically exposed to soluble antigen.

Finally, there may be a mechanism to eliminate activated B cells if they mutate to self-reactivity while they are proliferating and undergoing somatic hypermutation in germinal centers of peripheral lymphoid tissues (see Section 4-9). This possibility has been demonstrated by first stimulating a germinal center response by immunizing with foreign antigen, and then infusing large quantities of this antigen in soluble form, a procedure that induced apoptosis in the responding B cells. This experment was designed to mimic the situation in which a germinal center B cell alters its specificity to express a mutated immunoglobulin with high affinity for self antigen; however, whether this form of self-tolerance operates under normal conditions is not yet clear.

7-19. Only thymocytes whose receptors can interact with self MHC:self peptide complexes can survive and mature

As we learned in Chapters 3 and 5, α:β T-cell receptors recognize intracellular antigens after they have been processed into peptides that can be complexed with self MHC molecules. These antigen-presenting MHC molecules are encoded by a set of highly polymorphic genes found at the MHC locus, and their variability extends the range of peptides that can be presented to T cells in any one individual. However, MHC polymorphism also affects antigen recognition by T cells, since the T-cell receptor contacts both the bound peptide and the surrounding polymorphic surface of the MHC molecule itself. Each individual T-cell receptor is specific for a particular combination of MHC molecule and peptide antigen and is thus said to be MHC-restricted for antigen recognition. Not all the T-cell receptors generated by gene rearrangement will be able to recognize self MHC molecules and function in self MHC-restricted responses to foreign antigens; those that have this capability are selected for survival by a process of positive selection that occurs in the thymus.

Positive selection was demonstrated in experiments using mice whose bone marrow had been completely replaced by bone marrow from a mouse of different MHC genotype that was genetically identical except in the MHC gene region. These mice are known as bone marrow chimeras (see Appendix I, Section A-43). The recipient mouse is first irradiated in order to destroy all its own lymphocytes and bone marrow progenitor cells; after bone marrow transplantation, all bone marrow-derived cells will be of the donor genotype. These will include all lymphocytes, as well as the antigen-presenting cells they interact with. The rest of the animal's tissues, including the nonlymphoid stromal cells of the thymus will, however, be of the recipient MHC genotype.

In the experiments that demonstrated positive selection, the donor mice were F₁ hybrids derived from MHC^a and MHC^b parents, and thus were of the MHC^a×b genotype. The irradiated recipients were one of the parental strains, either MHC^a or MHC^b. Because of MHC restriction, individual T cells recognize either MHC^a or MHC^b, but not both. In normal circumstances, roughly equal numbers of the MHC^a×b T cells from MHC^a×b F₁ hybrid mice will recognize antigen presented by MHC^a or MHC^b. But in bone marrow chimeras in which T cells of MHC^a×b genotype develop in an MHC^a thymus, the T cells turn out to recognize antigen mainly, if not exclusively, when it is presented by MHC^a molecules, even though the antigen-presenting cells display antigen bound to both MHC^a and MHC^b (Fig. 7.27). These experiments showed clearly that the MHC molecules present in the environment in which T cells develop determine the MHC restriction of the mature T-cell receptor repertoire, and provided the first evidence for positive selection of developing T cells.

Figure 7.27

Positive selection is revealed by bone marrow chimeric mice. As shown in the top two sets of panels, bone marrow from an MHC^a×b F₁ hybrid mouse is transferred to a lethally irradiated recipient mouse of either parental MHC type (MHC^a or MHC^b). (more...)

A further experiment involving grafts of thymus tissue demonstrated that the thymic component responsible for positive selection is the thymic stroma. For this experiment, the recipient animals were athymic nude or thymectomized mice of MHC^a×b genotype who were given thymic stromal grafts of MHC^a genotype. Thus, all of their cells carried both MHC^a and MHC^b, except those of the thymic stroma. The MHC^a×b bone marrow cells of these mice also mature into T cells that recognize antigens presented by MHC^a but not antigens presented by MHC^b. Thus, what mature T cells consider to be self MHC is determined by the MHC molecules expressed by the thymic stromal cells that they encounter during intrathymic development.

The chimeric mice used to demonstrate positive selection produce normal T-cell responses to foreign antigens. By contrast, chimeras made by inject-ing MHC^a bone marrow cells into MHC^b animals cannot make normal T-cell responses. This is because the T cells in these animals have been selected to recognize peptides when they are presented by MHC^b, whereas the antigen-presenting cells that they encounter as mature T cells in the periphery are bone marrow-derived MHC^a cells. The T cells will therefore fail to recognize antigen presented by antigen-presenting cells of their own MHC type, and T cells can be activated in these animals only if antigenpresenting cells of the MHC^b type are injected together with the antigen. Thus, for a bone marrow graft to reconstitute T-cell immunity, there must be at least one MHC molecule in common between donor and recipient (Fig. 7.28).

Figure 7.28

Summary of T-cell responses to immunization in bone marrow chimeric mice. A set of bone marrow chimeric mice with different combinations of donor and recipient MHC types were made. These mice were then immunized and their T cells were isolated. These (more...)

7-20. Most thymocytes express receptors that cannot interact with self MHC and these cells die in the thymus

Bone marrow chimeras and thymic grafting provided the first evidence for the central importance of the thymus in positive selection, but more detailed investigation of the process has used mice transgenic for rearranged T-cell receptor genes. The rearranged α- and β-chain genes are molecularly cloned from a T-cell line or clone (see Appendix I, Section A-24) whose origin, antigen specificity, and MHC restriction are well characterized. When such genes are introduced into the mouse genome, rearrangement of the endogenous genes is inhibited (see Section 7-15), and most of the developing T cells express the receptor encoded by the transgenes. By introducing T-cell receptor transgenes into mice of known MHC genotype, the effect of known MHC molecules on the maturation of thymocytes with known recognition properties can be studied. It is routinely found that T cells bearing a transgenic receptor develop to maturity only within a thymus that expresses the same MHC molecule as that on which the original T-cell receptor was selected. In this situation the developing transgenic T cells can be positively selected.

Such experiments have also established the fate of T cells that fail positive selection. Rearranged receptor genes from a mature T cell specific for a peptide presented by a particular MHC molecule were introduced into a recipient mouse lacking that MHC molecule, and the fate of the thymocytes was investigated by staining with antibodies specific for the transgenic receptor. Antibodies to other molecules such as CD4 and CD8 were used at the same time to mark the stages of T-cell development. It was found that cells that fail to recognize the MHC molecules present on the thymic epithelium never progress further than the double-positive stage and die in the thymus within 3 or 4 days of their last division.

In a normal thymus, the fate of each thymocyte depends on the specificity of the receptor it expresses and, as we saw in Section 7-16, the specificity can undergo several changes as the α-chain genes continue to rearrange. The ability of a single developing thymocyte to express several different rearranged α-chain genes during the time that it is susceptible to positive selection must increase the yield of useful T cells significantly; without this mechanism many more thymocytes would fail positive selection and die. However, this continued rearrangement of α-chain genes also makes it likely that a significant percentage of T cells will express two receptors, sharing a β chain but differing in their α chains. Indeed, one can predict that if the frequency of positive selection is sufficiently low, roughly one in three mature T cells will have two α chains at the cell surface. This was confirmed recently for both human and mouse T cells.

T cells with dual specificity might be expected to give rise to inappropriate immune responses if the cell is activated through one receptor yet can act upon target cells recognized by the second receptor. However, only one of the two receptors is likely to be able to recognize peptide presented by a self MHC molecule as, once the cell has been positively selected, α-chain gene rearrangement ceases. Thus, the existence of cells with two α-chain genes productively rearranged and two α chains expressed at the cell surface does not seem to challenge the importance of clonal selection, which depends on a single functional specificity being expressed by each cell.

7-21. Positive selection acts on a repertoire of receptors with inherent specificity for MHC molecules

Positive selection acts on a repertoire of receptors whose specificity is determined by a combination of germline gene segments and junctional regions whose diversity is randomly created as the genes rearrange (see Section 4-11). The unselected receptor repertoire must be capable of recognizing all of the hundreds of different allelic variants of MHC molecules present in the population, as the genes for the T-cell receptor α and β chains seregate in the population independently from those of the MHC. If the binding specificity of the unselected repertoire were completely random, only a very small proportion of thymocytes would be expected to recognize an MHC molecule. However, it seems that the variable CDR1 and CDR2 loops of both chains of the T-cell receptor, which are encoded within the germline V gene segments (see Section 4-12), give the T-cell receptor an intrinsic specificity for MHC molecules. This is evident from the way these two regions contact MHC molecules in crystal structures (see Section 3-18). An inherent specificity for MHC molecules has also been shown by examining mature T cells that represent an unselected repertoire of receptors. Such T cells can be generated in fetal thymic organ cultures, using thymuses that do not express either MHC class I or MHC class II molecules, by substituting binding of anti-β-chain antibodies and anti-CD4 antibodies for the receptor engagement responsible for normal positive selection. When the reactivity of these antibody-selected CD4 T cells is tested, roughly 5% are able to respond to any one MHC class II genotype and, because they developed without selection by MHC molecules, this must reflect a specificity inherent in the germline V gene segments. This germline-encoded specificity for MHC should significantly increase the proportion of receptors that can be positively selected in any one individual.

7-22. Positive selection coordinates the expression of CD4 or CD8 with the specificity of the T-cell receptor and the potential effector functions of the cell

At the time of positive selection, the thymocyte expresses both CD4 and CD8 co-receptor molecules. At the end of the selection process, mature thymocytes ready for export to the periphery express only one of these co-receptors. Moreover, almost all mature T cells that express CD4 have receptors that recognize peptides bound to self MHC class II molecules and are programmed to become cytokine-secreting cells. In contrast, most of the cells that express CD8 have receptors that recognize peptides bound to self MHC class I molecules and are programmed to become cytotoxic effector cells. Thus, positive selection also determines the cell-surface phenotype and functional potential of the mature T cell, selecting the appropriate co-receptor for efficient antigen recognition and the appropriate program for the T cell's eventual functional differentiation in an immune response.

Experiments with mice made transgenic for rearranged T-cell receptor genes show clearly that it is the specificity of the T-cell receptor for self MHC molecules that determines which co-receptor a mature T cell will express. If the T-cell receptor transgenes encode a receptor specific for antigen presented by self MHC class I molecules, mature T cells that express the transgenic receptor are CD8 T cells. Similarly, in mice made transgenic for a receptor that recognizes antigen with self MHC class II molecules, mature T cells that express the transgenic receptor are CD4 T cells (Fig. 7.29).

Figure 7.29

Positive selection determines co-receptor specificity. In mice transgenic for T-cell receptors restricted by an MHC class I molecule (top panel), the only mature T cells to develop have the CD8 (red) phenotype. In mice transgenic for receptors restricted (more...)

The importance of MHC molecules in this selection is illustrated by the class of human immunodeficiency diseases known as bare lymphocyte syndromes, which are caused by mutations that lead to an absence of MHC molecules on lymphocytes and thymic epithelial cells. People who lack MHC class II molecules have CD8 T cells but only a few, highly abnormal, CD4 T cells; a similar result has been obtained in mice in which MHC class II expression has been eliminated by targeted gene disruption (see Appendix I, Section A-47). Likewise, mice and humans that lack MHC class I molecules lack CD8 T cells. Thus, MHC class II molecules are required for CD4 T-cell development, whereas MHC class I molecules are required for CD8 T-cell development. (MHC Class I Deficiency, in Case Studies in Immunology, see Preface for details)

In mature T cells, the co-receptor functions of CD8 and CD4 depend on their respective abilities to bind invariant sites on MHC class I and MHC class II molecules (see Section 3-12). Co-receptor binding to an MHC molecule is also required for normal positive selection. as shown for CD4 in the experiment discussed below (see Section 7-23). Thus positive selection depends on engagement of both the antigen receptor and co-receptor with an MHC molecule, and determines the survival of single-positive cells that express only the appropriate co-receptor. However, the exact mechanism whereby lineage commitment is coordinated with receptor specificity remains to be established. At present, it seems as though the developing thymocyte integrates the signals that it gets from both the antigen receptor and the co-receptor in order to determine its fate. Co-receptor-associated Lck signals are most effectively delivered when CD4 is engaged as a co-receptor rather than CD8, and these Lck signals play a large role in the decision to become a CD4 mature cell. Other signaling molecules are clearly important for thymocyte survival and development, but appear to be common to the commitment of both lineages. A few of these may, nonetheless, preferentially influence the commitment of thymocytes to one lineage over the other: for example, the MAPK signaling pathway initiated by the antigen receptor (see Section 6-11) biases cells toward the CD4 versus the CD8 lineage. It is a general principle of lineage commitment that different signals must be created in order to activate lineage-specific factors and generate a divergence of developmental programming. While much remains to be discovered about this process in developing α:β thymocytes, it is clear that the different signals that are created result in a divergence of functional programming, so that the ability to express genes involved in the killing of target cells, for example, develops in CD8 T cells, whereas the potential to express various cytokine genes develops in CD4 T cells.

7-23. Thymic cortical epithelial cells mediate positive selection of developing thymocytes

The thymic cortical epithelial cell is the stromal cell type critical for positive selection. These cells form a web of cell processes that make close contacts with the double-positive T cells undergoing positive selection (see Fig. 7.8) and T-cell receptors can be seen clustering with MHC molecules at the sites of contact.

Direct evidence that thymic cortical epithelial cells mediate positive selection comes from an ingenious manipulation of mice whose MHC class II genes have been eliminated by targeted gene disruption (Fig. 7.30). Mutant mice that lack MHC class II molecules do not normally produce CD4 T cells. To test the role of the thymic epithelium in positive selection, an MHC class II gene was placed under the control of a promoter that restricted its expression to thymic cortical epithelial cells, and was introduced as a transgene into these mutant mice. CD4 T cells then developed. A further variant of this experiment shows that, in order to promote the development of CD4 T cells, the MHC class II molecule on the thymic epithelium must be able to interact effectively with CD4. Thus, when the MHC class II transgene expressed in the thymus contains a mutation that prevents its binding to CD4, very few CD4 T cells develop. Equivalent studies of CD8 interaction with MHC class I molecules show that co-receptor binding is necessary for normal positive selection of CD8 cells as well.

Figure 7.30

Thymic cortical epithelial cells mediate positive selection. The expression of MHC class II molecules in the thymus of normal and mutant strains of mice is shown by coloring the stromal cells only if they are expressing MHC class II molecules. In the (more...)

The critical role of the thymic epithelium in positive selection raises the question whether there is anything distinctive about the antigen-presenting properties of these cells. This is not clear at present; however, thymic epithelium may differ from other tissues in the proteases used to degrade the invariant chain (Ii) (see Section 5-6). The protease cathepsin L dominates in thymic cortical epithelium, whereas cathepsin S seems to be most important in other tissues. Consequently, CD4 T-cell development is severely impaired in cathepsin L knockout mice. Thymic epithelial cells do seem to bear on their cell surfaces a relatively high density of MHC class II molecules that retain the invariant chain-associated peptide (CLIP) (see Fig. 5.7). However, they also present a range of other peptides, and it remains to be seen whether the MHC:peptide complexes presented by these cells have any special characteristics that are important for positive selection.

Artificial manipulation of the peptides presented by thymic epithelium has a profound effect on positive selection. The effect of selecting thymocytes in the presence of a single peptide:MHC class II complex has been demonstrated in experiments in which the α chain of H-2M, the mouse homologue of human HLA-DM, is disrupted (Fig. 7.31). In mice lacking a functional H-2Mα chain, the CLIP fragment of the invariant chain is not released from the newly synthesized MHC class II molecules (see Section 5-7). These CLIP-associated MHC class II molecules are unable to bind the self peptides present in the endosomes, and the main self peptide presented by MHC class II molecules on the surface of the thymic epithelium is therefore the invariant CLIP peptide. In these mice, the total number of CD4 T cells is reduced twofold to threefold. A diversity of T-cell receptor β chains is expressed in these cells, but of seven T-cell receptor transgenes examined so far, none is positively selected in H-2Mα knockout mice despite the presence of typical levels of the MHC class II molecule that normally selects them.

Figure 7.31

The peptides bound to MHC class II molecules can affect the T-cell receptor repertoire. The left panels show the normal situation, in which a range of peptides is presented by antigen-presenting cells (APCs) to immature T cells in the thymus, with the (more...)

Thus it seems that a significantly reduced repertoire of T cells is selected in the presence of a single predominant peptide:MHC class II complex. In addition, a high proportion of the T cells that are positively selected in H2-M-deficient mice and reach maturity are self-reactive, as shown by their activation by antigen-presenting cells from wild-type mice of the same MHC genotype. These T cells would have been eliminated by negative selection in a nonmutant mouse (see Fig. 7.31). These self-reactive cells might have been positively selected through an interaction dominated by contacts between the T-cell receptor and MHC class II molecule, to which the bound peptide made a minimal contribution. Such T cells would thus be more likely to react with the same MHC molecule complexed with a different self peptide than would cells that were positively selected through an interaction dominated by contacts with the bound peptide.

The influence of peptide diversity on positive selection is seen even more clearly in experiments using H2-M-deficient mice that have been bred to carry a rearranged T-cell receptor β-chain transgene. Endogenous β-gene rearrangements are suppressed in such mice and this limits the range of antigen receptor specificities on which positive selection can act. In control mice that are MHC-identical and carry the same transgene but are not H2-M-deficient, positive selection is able to generate mature T cells that express a large repertoire of α-chain genes. By contrast, mature T cells in the H-2M-deficient animals have a highly restricted repertoire of endogenous T-cell receptor α chains, and most of the selected T cells react against self MHC molecules when these are combined with the normal range of self peptides. These experiments illustrate the results of positive selection for α-gene rearrangements when there is only a single β-chain rearrangement with which they can combine, and when selection occurs on only a small set of self peptide:self MHC complexes. They therefore suggest that a diversity of bound peptides is important for the positive selection of a diverse repertoire of self MHC-restricted T cells.

7-24. T cells that react strongly with ubiquitous self antigens are deleted in the thymus

When the T-cell receptor of a mature naive T cell is ligated by peptide plus MHC antigen displayed by a professional antigen-presenting cell in a peripheral lymphoid organ, the T cell is activated to proliferate and produce effector T cells (see Section 1-12). In contrast, when the T-cell receptor of a developing thymocyte is similarly ligated by antigen on stromal or bone marrow-derived cells in the thymus, it dies by apoptosis. The response of immature T cells to stimulation by antigen is the basis of negative selection. Elimination of these T cells in the thymus prevents their potentially harmful activation later on, if they should encounter the same peptides as mature T cells. Negative selection has been demonstrated in mice that express a transgenic T-cell receptor specific for a peptide of ovalbumin bound to an MHC class II molecule. As explained in Section 7-15, all the T cells developing in such a mouse will express the transgenic receptor. When these mice are injected with the ovalbumin peptide, the mature CD4 T cells in the periphery become activated, but most of the intrathymic T cells die (Fig. 7.32). Similar results can be obtained in thymic organ culture with T cells from both normal and transgenic mice, showing that secondary effects caused by the induction of cytokines or corticosteroids due to simultaneous activation of peripheral T cells in vivo cannot account for this cell death of immature thymocytes.

Figure 7.32

T cells specific for self antigens are deleted in the thymus. In mice transgenic for a T-cell receptor that recognizes a known peptide antigen complexed with self MHC, all of the T cells have the same specificity; in the absence of the peptide, most thymocytes mature (more...)

The deletion of developing T cells that recognize self peptides synthesized naturally in the thymus has also been demonstrated experimentally. The negative selection of such thymocytes was observed in mice made transgenic for rearranged genes encoding T-cell receptors specific for self peptides expressed only in male mice. Thymocytes bearing these receptors disappear from the developing T-cell population in male mice at the CD4⁺ CD8⁺ double-positive stage of development, and no single-positive cells bearing the transgenic receptors mature. By contrast, in female mice, which lack the male-specific peptide, the transgenic T cells mature normally. This initial observation has been confirmed using T-cell receptor transgenes that recognize other antigens, with similar results.

These experiments illustrate the principle that self peptide:self MHC complexes encountered in the thymus purge the T-cell repertoire of immature T cells bearing self-reactive receptors. Not all self proteins are expressed in the thymus, however, and those that appear in other tissues, or are expressed at different stages in development, such as after puberty, will encounter mature T cells with the potential to respond to them. However, there are mechanisms that prevent mature T cells from responding to such antigens, and these will be discussed in Chapter 13, when we consider the problem of autoimmune responses and their avoidance.

7-25. Negative selection is driven most efficiently by bone marrow-derived antigen-presenting cells

Negative selection in the thymus can be mediated by several different cell types. The most important are the bone marrow-derived dendritic cells and macrophages. These are professional antigen-presenting cell types that also activate mature T cells in peripheral lymphoid tissues. The self antigens presented by these cells are therefore the most important source of potential autoimmune responses, and T cells responding to such self peptides must be eliminated in the thymus.

Experiments using bone marrow chimeric mice have shown clearly the role of thymic macrophages and dendritic cells in negative selection. In these experiments, MHC^a×b F₁ bone marrow is grafted into one of the parental strains (MHC^a in Fig. 7.33). The MHC^a×b T cells developing in the grafted animals are thus exposed to the thymic epithelium of the MHC^a host strain. Bone marrow-derived dendritic cells and macrophages will, however, express both MHC^a and MHC^b. The bone marrow chimeras will tolerate skin grafts from either MHC^a or MHC^b animals (see Fig. 7.33), and from the acceptance of both types of skin grafts we can infer that the developing T cells are not self-reactive for either of the two MHC antigens. The only cells that could present self peptide:MHC^b complexes to thymocytes, and thus induce tolerance to MHC^b, are the bone marrow-derived cells. The dendritic cells and macrophages are therefore assumed to have a crucial role in negative selection.

Figure 7.33

Bone marrow-derived cells mediate negative selection in the thymus. When MHC^a×b F₁ bone marrow is injected into an irradiated MHC^a mouse, the T cells mature on thymic epithelium expressing only MHC^a molecules. Nevertheless, the chimeric mice are (more...)

In addition, both the thymocytes themselves and thymic epithelial cells also have the ability to cause the deletion of self-reactive cells. Such reactions may normally be of secondary significance compared to the dominant role of bone marrow-derived cells in negative selection. In patients undergoing bone marrow transplantation from an unrelated donor, however, where all the thymic macrophages and dendritic cells are of donor type, negative selection mediated by thymic epithelial cells can assume a special importance in maintaining tolerance to the recipient's own antigens.

7-26. Endogenous superantigens mediate negative selection of T-cell receptors derived from particular V_β gene segments

It is virtually impossible to demonstrate directly the negative selection of T cells specific for any particular self antigen in the normal thymus because such T cells will be too rare to detect. There is, however, one case in which negative selection can be seen on a large scale in normal mice and the point at which it occurs in T-cell development can be identified. In the most striking examples, T cells expressing receptors encoded by particular V_β gene segments are virtually eliminated in the affected mouse strains. This occurs as the consequence of the interaction of immature thymocytes with endogenous superantigens present in those strains. We learned in Chapter 5 (see Section 5-15) that superantigens are viral or bacterial proteins that bind tightly to both MHC class II molecules and particular V_β domains, irrespective of the antigen specificity of the receptor and the peptide bound by the MHC molecule (see Fig. 5.18).

The endogenous superantigens of mice are encoded by mouse mammary tumor virus (MMTV) genomes that have become integrated at various sites into the mouse chromosomes. Different mouse strains have different complements of inherited MMTV genomes, and therefore express different viral antigens. Like the bacterial superantigens, these MMTV superantigens induce strong T-cell responses; indeed, they were originally designated minor lymphocyte-stimulating (Mls) antigens (see Section 5-15). Mice that carry these endogenous superantigens are said to be Mls⁺, and a series of Mls antigens (Mls-1^a, Mls-1^b, …) have been identified by their ability to stimulate primary T-cell responses when T cells from a strain lacking the superantigen are mixed with B cells from MHC-identical mice that express it.

In Mls⁺ strains, T cells bearing V_β regions to which the Mls proteins bind, die by apoptosis during intrathymic maturation. For example, one variant of the Mls antigen (Mls-1^a or MTV7) deletes all thymocytes expressing the V_β6 V gene segment (and also those expressing V_β8.1 and V_β9), whereas such cells are not deleted in mice that lack Mls-1^a. Thus, the expression of endogenous superantigens in mice has a profound impact on the T-cell receptor repertoire. This sort of deletion has not yet been seen in any other species, including humans, despite the presence of endogenous retroviral sequences in many mammals.

In mice that express the superantigen and thus are tolerant to it, cells expressing receptors responsive to superantigens are found among the double-positive thymocytes and are abundant in thymic cortex. They are, however, absent from the thymic medulla and tissues outside the thymus. This suggests that superantigens might delete relatively mature T cells as they migrate out of the cortex into the medulla, where a particularly dense network of dendritic cells marks the cortico-medullary junction (Fig. 7.34). Although clonal deletion by superantigens is a powerful tool for examining negative selection in normal mice, superantigen-driven clonal deletion might not be representative of clonal deletion by self peptide:self MHC complexes. What is clear is that clonal deletion by either superantigens or self peptide:self MHC complexes generates a repertoire of T cells that does not respond to the self antigens expressed by its own professional antigen-presenting cells.

Figure 7.34

Clonal deletion by Mls-1^a occurs late in the development of thymocytes. T cells with Mls-1^a-responsive receptors encoded by V_β6 are seen in both the cortex and medulla of Mls-1^b mice (top panel, cells stained brown with anti-V_β6 antibody). (more...)

7-27. The specificity and strength of signals for negative and positive selection must differ

We have described some of the experiments that contributed to the large body of evidence that T cells are selected for both self MHC restriction and self tolerance by MHC molecules expressed on stromal cells in the thymus. We now turn to the central question posed by positive and negative selection: how can engagement of the receptor by self MHC:self peptide complexes lead both to further maturation of thymocytes during positive selection and to cell death during negative selection? The answers to these questions are still not known for certain, but two possible mechanisms have been suggested. We will briefly describe each, before going on to discuss some recent experiments that bear on the mechanism of positive selection.

There are two issues to be resolved. First, the interactions that lead to positive selection must include more receptor specificities than those that lead to negative selection. Otherwise, all the cells that were positively selected in the thymic cortex would be eliminated by negative selection, and no T cells would ever leave the thymus (Fig. 7.35). Second, the consequences of the interactions that lead to positive and negative selection must differ; cells that recognize self peptide:self MHC complexes on cortical epithelial cells are induced to mature, whereas those whose receptors might confer strong and potentially damaging autoreactivity are induced to die.

Figure 7.35

The specificity or affinity of positive selection must differ from that of negative selection. Immature T cells are positively selected in such a way that only those thymocytes whose receptors can engage the peptide:MHC complexes on thymic epithelium (more...)

Two main hypotheses have been proposed to account for these differences between positive and negative selection. The first is the avidity hypothesis. This states that the outcome of MHC:peptide binding by thymocyte T-cell receptors depends on the strength of the signal delivered by the receptor on binding, and that this will, in turn, depend upon both the affinity of the T-cell receptor for the MHC:peptide complex and the density of the complex on a thymic cortical epithelial cell. Thymocytes that are signaled weakly are rescued from apoptosis and are thus positively selected, whereas thymocytes that are signaled strongly are driven to apoptosis and are thus negatively selected. Because more complexes are likely to bind weakly rather than strongly, this will result in the positive selection of a larger repertoire of cells than are negatively selected.

Alternatively, the delivery of incomplete activating signals by self peptides could account for positive selection: we will call this the differential signaling hypothesis. Under this hypothesis, it is the nature of the signal delivered by the receptor, not just the number of receptors engaged, that distinguishes positive from negative selection. According to the avidity hypothesis, the same MHC:peptide complex could drive positive or negative selection of the same receptor, depending on its density on the cell surface. This could not occur under the differential signaling hypothesis, because it proposes that the signals leading to positive and negative selection are qualitatively different.

A new approach to testing these hypotheses has opened up with the recent description of antagonist peptides. These are known as antagonist peptides because they inhibit the response of mature T cells to their normal stimulatory, or agonist, peptide. On binding to the antigen receptor of a mature T cell, antagonist peptide:MHC complexes generate some, but not all, of the intracellular signaling events that are associated with full agonist-driven T-cell activation (see Section 6-12). Recognition of antagonist peptides by thymocytes has been shown to induce positive selection, whereas recognition of agonist peptides induces negative selection. Thus the differences between the ways agonist and antagonist peptides interact with the receptor and signal to the cell are likely to be relevant to the issue of how positive selection works.

The experiments shown in Fig. 7.36 used a T-cell receptor of known specificity whose in vivo behavior with different variants of its peptide antigen had been characterized. The affinities of this T-cell receptor for the agonist peptide and for its antagonist variants (all bound to the appropriate MHC class I molecule) were measured using an affinity sensor (see Appendix I, Section A-30). Initially, affinity measurements were carried out at room temperature and revealed very slight differences in affinity for the agonist versus antagonist peptides. At physiological temperatures, the differences were larger, with the antagonist peptide:MHC complexes having lower affinities for the T-cell receptor. However, the affinities were still broadly similar and this parameter failed to capture a more radical difference in the binding of the T-cell receptor that was observed under these conditions. The complexes of MHC class I molecules with agonist peptides induced dimerization of the T-cell receptors, whereas the complexes with antagonist peptides did not. This suggests that in vivo the agonist peptides induce the receptor clustering required for generating the signals that lead to activation (see Section 6-2) and, during thymocyte development, to deletion. The antagonist peptides, in contrast, bind but fail to induce receptor clustering, and therefore deliver a qualitatively different signal; during thymocyte development, recognition of the antagonist peptide delivers a survival signal to the cell. As the agonist and antagonist peptides bind the T-cell receptor with similar affinity, but only the antagonist induces positive selection, the implication is that the signals delivered by the agonist peptides that cause deletion and by the antagonist peptides that mediate positive selection are fundamentally different. This suggests that the differential signaling hypothesis of T-cell selection is correct. These experiments have, however, only been carried out in a single laboratory, so one should keep an open mind until other laboratories produce similar results in other systems.

Figure 7.36

The differences between negative and positive selection may be due to differences in the aggregation of T-cell receptors upon ligand binding. Agonist, antagonist, and null peptides can be identified for a particular T-cell receptor by assaying the responses (more...)

7-28. The B-1 subset of B cells has a distinct developmental history and expresses a distinctive repertoire of receptors

A minority subset of B cells (comprising about 5%) in mice and humans, and the major population in rabbits, arises during fetal development and has a restricted receptor repertoire (Fig. 7.37). The B cells belonging to this subset were first identified by surface expression of the protein CD5 and are also characterized by high levels of sIgM with little sIgD, even when mature. They are termed B-1 cells, because their development precedes that of the conventional B cells whose development has been discussed up to now and which are sometimes termed B-2 cells. B-1 cells are also known as CD5 B cells, although CD5 itself cannot be essential for their function because cells that have similar traits develop normally in mice lacking the CD5 gene, and B-1 cells in rats do not display CD5. B-1 cells are of interest to clinicians, as they are the origin of the common B-cell tumor chronic lymphocytic leukemia (CLL). CLL cells often display CD5, which is a useful diagnostic clue.

Figure 7.37

A comparison of the properties of B-1 cells and conventional B cells (B-2 cells). B-1 cells can develop in unusual sites in the fetus, such as the omentum, in addition to the liver. B-1 cells predominate in the young animal although they probably can (more...)

There is great debate about the origin of B-1 cells, and it is not yet clear whether they arise as a distinct lineage from a unique precursor cell, or instead differentiate to the B-1 phenotype from a precursor cell that could also give rise to B-2 cells. In the mouse, fetal liver mainly produces B-1 cells, whereas adult bone marrow generates predominantly B-2 cells, and this has been interpreted as support for the unique precursor hypothesis. However, the weight of evidence favors the idea that the antigen specificity of the B-cell receptor determines whether the precursor becomes a B-1 or a B-2 cell. Under this hypothesis, commitment to the B-1 or B-2 subset would be due to a selection step, rather than being a distinct lineage difference like that between γ:δ and α:β T cells. It is, however, difficult to rule out the idea that cells are committed before this but only survive if their receptor specificity matches the predetermined fate.

The two hypotheses of B-cell commitment might be reconciled by proposing that fetal cells tend to generate B-1 cells because of the particular antigen specificities their immunoglobulin gene rearrangements tend to generate. Indeed, fetal B-cell precursors preferentially rearrange certain V_H gene segments, and also do not incorporate N-nucleotides; thus their receptors have a restricted and particular range of specificities, analogous to the T-cell receptors of the early waves of γ:δ T cells (see Section 7-14). The V gene segments that are commonly used to encode the receptors of B-1 cells seem to have evolved to recognize common bacterial and self antigens. These specificities may then be positively selected so that the cell matures with the phenotype of a B-1 cell.

Regardless of how B-1 cells originate, they are certainly expanded and maintained by interaction with self antigens or nonself antigens normally present in the body, such as those of the bacterial gut flora. Expansion by a relatively small number of ubiquitous antigens is a form of positive selection that would also tend to restrict the receptor repertoire of the B-1 cell population. In adult animals, the population of B-1 cells is maintained by continued division in peripheral sites such as the peritoneal and pleural cavities, a process that requires the cytokine IL-10 in addition to stimulation through the B-cell receptor. Some of the antigens that drive B-1 cell expansion, such as the phospholipid phosphatidylcholine, are encountered on the surface of bacteria that colonize the gut. Interestingly, a self antigen that can drive the expansion of B-1 cells has recently been identified on the surface of thymocytes. Why a ligand for B-1 cells should be expressed on T cells is unknown at present.

Little is yet known about the functions of B-1 cells. Their location suggests a role in defending the body cavities, while their restricted repertoire of receptors appears to equip them for a function in the early, nonadaptive phase of an immune response (see Section 2-28). Indeed, the V gene segments that are used to encode the receptors of B-1 cells might have evolved by natural selection to recognize common bacterial antigens, thus allowing them to contribute to the very early phases of the adaptive immune response. In practice, it is found that B-1 cells make little contribution to adaptive immune responses to most protein antigens but contribute strongly to some antibody responses against carbohydrate antigens. Moreover, a large proportion of the IgM that normally circulates in the blood of nonimmunized mice derives from B-1 cells. The existence of these so-called ‘natural antibodies,’ which are highly cross-reactive, and bind with low affinity to both microbial and self antigens, supports the view that B-1 cells are partially activated as they are selected for self-renewal by ubiquitous self and foreign antigens.

Summary

Initially, random receptor rearrangements and junctional diversity create a broad repertoire of antigen receptors. Only some of these will not be dangerously self-reactive and yet be useful to the immune system. Cells meeting these criteria are selected, a process that begins as soon as immature lymphocytes express antigen receptors. Lymphocyte development involves both negative and positive selection. Negative selection includes deletion from the repertoire, receptor editing, and anergy, which most often are imposed on immature self-reactive lymphocytes in the thymus or bone marrow. However, even mature lymphocytes can be subject to negative selection when presented with a strong antigen signal without the usual co-stimulation needed for activation. Positive selection is best defined for T cells and B-1 cells; as we will discuss in the next part of this chapter, B-2 cells may also be positively selected as they mature to populate the periphery, although this is less well-established. For developing T cells, recognition of self MHC:self peptide complexes on thymic epithelial cells provides an as yet poorly defined positive survival signal. T-cell receptors capable of responding to foreign peptides presented by self MHC molecules are thus selected from a primary repertoire that has an inherent specificity for all the different MHC molecules in the population. Positive selection also ensures the functional matching of receptor, co-receptor, and the class of MHC molecule recognized. The paradox that recognition of self MHC:self peptide ligands by the T-cell receptor can lead to two opposing effects, namely positive and negative selection, is one of the central mysteries of immunology. Its solution will come from a full understanding of the ligand-receptor interactions, the signal transduction mechanisms, and the physiology of each step of the process. An antigen-driven selection process during development may also lead to the generation of B-1 cells, a minor subset of B cells. These cells originate early in fetal life and are self-renewing in the periphery. They reside mainly in the pleural and peritoneal cavities and have B-cell receptors with a limited range of antigen specificities compared with conventional B-2 cells.