T-cell development is a distinctive branch of hematopoiesis as well as a crucial input to immune system function. T-cell precursors, like other blood cells, are continuously generated from hematopoietic stem cells through a process that involves loss of access to alternative hematopoietic fates as well as gain of lineage-specific characteristics. In addition, like B cells, T cells must undergo programmed rearrangement of gene segments in order to assemble the genes that will encode their receptors for antigen, the α, β, γ, and δ T-cell receptor (TCR) chains. T-cell development is more distinctive in other ways. Unlike almost all other hematopoietic cell types, T-cell precursors must leave the bone marrow in order to adopt their fate. Most of them must migrate to the thymus for their differentiation: a specialized epithelial organ that provides a framework for intensive signaling that guides the cells to become T cells at the expense of other developmental fates. The T-cell program itself is also remarkable because of the variety of different types of T cells that it produces. Despite indelibly establishing common T-cell features, the cells that pass through T-cell development retain great functional plasticity and eventually fan out into a spectrum of different effector subtypes. Some of their diversification begins in the thymus, even while their universal T-cell properties are being acquired. Remarkably, as described in the following, some of the same transcription factors simultaneously play roles both in universal T-cell gene expression and in competitive oppositional mechanisms that determine particular effector subtype choices.

Thus, the T-cell program is more than a simple progression toward maturity: it is a mosaic of developmental processes that on the one hand forge a durable T-lineage commitment and on the other hand open up many different routes to useful T-cell functionality. T-cell function depends on recognition of antigen, and so a crucial part of the arming of the T-cell precursor for function is assembly of the TCR complex. To generate cohorts of T cells with various different recognition specificities, this process depends on individual cells’ random choice of TCR gene segments for rearrangement, and on additional chance incorporation of somatic mutations in the TCR genes, potentially different for every cell. The process is capable of generating autoreactive receptors or defective receptors as well as useful ones. Thus, the T-cell development program also incorporates several discrete filtering steps to eliminate newly made cells if they have dangerous or useless TCR specificities. These “repertoire selection” checkpoints are major landmarks in the T-cell development process, and they often provide choice points in developmental fate as well.

Both TCR recognition specificity and effector function subdivide T cells, and one of the most notable features of T-cell development is that particular functional subtypes become preferentially associated with particular specificities of antigen recognition. For one major divergence, the cluster of differentiation (CD) 4+ “helper” versus CD8+ “killer” decision, a gene network explanation has emerged, but explanations for differentiation of TCRγδ versus TCRαβ and “innate-type” versus “conventional” T cells of both receptor types remain a challenge.

This chapter covers T-cell differentiation both as a hematopoietic developmental program and as an immunological process that trains new cohorts of cells for appropriate discrimination between recognition of those targets that should elicit attack and those that should elicit protection. It begins with an introduction to the molecules that define T-cell identity and distinguish stages in the T-cell development process; a global synopsis of intrathymic T-cell development is then presented as a framework for the rest of the chapter. Subsequent sections zero in on particular steps in the process, developmental branch points, and checkpoints, in which crucial aspects of mechanism have been illuminated. The chapter then closes with a survey of some open challenges that are faced in the next few years.

One goal of T-cell development is to arm mature T cells with a full set of the molecular criteria for T-cell identity. The mature T cell requires not only recognition apparatus, centered around the TCR complex, but also signal transduction apparatus to convey quantitative recognition information to the nucleus, and a poised transcriptional response system that can induce an appropriate combination of effector genes in response.

Mature T cells are armed for particular stereotyped responses to antigen recognition. Most CD4 helper cells make their first antigen response by turning on expression of the growth factor IL-2. Killer T cells respond to activation by expressing the membrane-perforating molecule perforin, granzymes A and B, and relatively toxic cytokines like interferon (IFN)γ. Th1 cells do so by expressing IFNγ and TNFα, but much less perforin. Th2 cells respond by expressing IL-4 and in some cases IL-5 and/or IL-13. Natural killer T (NKT) cells rapidly activate IL-4 expression together with IFNγ. Th17 cells, often found defending epithelia, express inflammatory cytokines such as IL-17A, IL-17F, and IL-22. Additional subsets, follicular helper T cells and Th9 cells, also have particular canonical effector programs that they use to respond to TCR signals. Whereas all these subtypes promote immune activation upon stimulation, regulatory T cells (T^reg cells) respond to stimulation by blocking the effector responses of neighboring T cells, at least in part through expression of IL-10, TGFβ, and specific cell-surface interaction molecules.

In mature T cells, the poised state of distinct sets of cytokine genes that distinguishes each subset is maintained, in between rounds of stimulation, by transcription factor expression patterns and site-specific chromatin modification. Particular transcription factors provide distinctive, crucial regulatory inputs in particular cell types. Thus, Runx3, T-bet, and eomesodermin are important to maintain killer function, T-bet to maintain Th1 function, GATA-3 to promote Th2 function, PLZF to promote NKT-cell function,
RORγt to promote Th17 function, Bcl6 for follicular helper T-cell function, and Foxp3 to promote T_reg function. Of particular interest is the fact that most of these factors also play pivotal earlier roles in the generation of T cells in general.

The differentiation of particular effector subsets is usually considered to start with antigen-inexperienced (naïve) mature T cells, after they have finished with their intrathymic development. The specialization itself is part of the mature T cell’s portfolio of possible responses to antigen, especially among CD4 T cells. Conventionally, one may consider these antigen-driven events to be a completely different type of developmental process from the intrathymic events that mold hematopoietic precursors into T cells. However, we will see in the following that effector programming can also originate in the thymus for subsets like NKT cells, certain classes of TCRγδ cells, and some T_regs: the distinction is not absolute.

The major known site for T-cell development in all jawed vertebrates is the thymus.¹ The thymus does not harbor its own long-term source of precursors but is continuously populated throughout life by new precursors that are replaced as their descendants complete maturation and selection. These precursors migrate to the thymus from bone marrow or fetal liver when they are still multipotent. The thymus not only enforces their commitment to a T-cell fate and T-lineage differentiation but also provides a proliferation-inducing environment. It is estimated that the numbers entering the thymus per day are quite modest, on the order of < 100 per day for a typical young adult mouse.^2,3 The descendants of that small cohort of input cells expand over a 2-week period to a cohort of cells that have undergone T-lineage commitment, TCR gene rearrangement, and TCR expression, eventually yielding ˜50 × 10⁶ new thymocytes a day before their proliferation stops.

The three-dimensional structure of the thymus was for many years a challenge to the full definition of T-cell developmental mechanisms. It was clear after many attempts that dissociated thymic epithelial cells could not reconstitute thymic function if plated in monolayer culture, and thus it seemed that the migration of precursors through organized three-dimensional domains would be a crucial aspect of T lymphopoiesis. Redox conditions also seemed crucial for thymic function. There were two early breakthroughs in experimental dissection of T-cell development. First was the generation of fetal thymic organ cultures, which were found to require a high oxygen concentration to work, either by plating a thymic lobe at an air-medium interface, or by pumping up to 70% oxygen into the incubator in which a submerged fetal thymic lobe was cultured.^24,25 Second was the discovery that in these excellent culture conditions, even dissociated, purified thymic stromal elements could regenerate a competent thymic microenvironment if mixed together in a reaggregate fetal thymic organ culture. The development of reaggregate fetal thymic organ culture has allowed the separate roles of distinct epithelial and hematopoietically derived elements of the cortical and medullary environments in positive and negative selection to be rigorously demonstrated, as discussed in later sections.²⁶ Both fetal thymic organ culture and reaggregate fetal thymic organ culture systems promote T-cell development from prethymic precursor stages all the way through positive and negative selection to generate functional mature T cells.

The fetal and reaggregate thymic organ cultures both make use of natural thymic stromal elements, which makes them close to physiological conditions, but also a logistical challenge. Furthermore, the organ cultures still re-form into closed structures, making the thymocytes developing within them hard to count, track phenotypically, or experimentally manipulate. Finally, these organ cultures remain very small, hampering use for molecular biology.

Against this background, the development of stromal monolayer coculture systems for T-cell differentiation has revolutionized the field. Two lines of research converged to make this possible. First, Nakano, Kodama, and Honjo developed OP9 bone marrow stromal lines from myeloid colony stimulating factor-deficient mice (Op, osteopetrotic mice),²⁷ which proved to be ideal for supporting lymphoid development. However, only B and NK lymphocytes, not T cells, develop on these original OP9 stromal cell lines. Second, crucial experiments by the groups of Freddy Radtke and Warren Pear showed that environmental activation of Notch signaling is crucial to induce hematopoietic precursors to become T cells.^28,29 Engineering OP9 cells to express a potent Notch ligand, Schmitt and Zuniga-Pflucker developed the OP9-DL1 stromal line to support T-cell development.³⁰

This created robust, highly efficient, high-yield conditions for driving and supporting T-lineage development in vitro from a wide range of uncommitted hematopoietic progenitor types, while keeping the cells accessible, easily monitored, and easily transferable throughout.

The OP9-DL1 system and variations of it have made possible enormous advances in understanding of the early, Notch-dependent stages of T-cell development. This system exposes to experimental dissection all the stages from initiation of the T-cell developmental program through commitment and up to the DP stage (ie, all the stages that DN precursors would normally undergo during their “outward bound” migration through the thymic cortex). It has also opened up the ability to track and experimentally perturb early stages of human T-cell development, starting from cord blood precursors, in a direct parallel to early mouse T-cell development.²³ However, the “inward bound” phases of T-cell development after β-selection or γδ-selection appear to depend on special properties of thymic epithelial cells that remain to be duplicated by OP9-DL1 or any stromal coculture.

Notch signaling is crucial for induction of the T-cell program, for progression from stage to stage of the specification process, and for maintenance of viability of the developing T cells throughout the DN stages, all the way to β-selection. DN thymocytes gain privileged access to Notch ligands in the environment despite the large excess of DP cells. This is because the DP cells “deny themselves” the chance to bind efficiently with Delta-class Notch ligands in the environment by downregulating key Notch modifier genes, Lfng and Mfng, during β-selection,³¹ which are needed to compete for stromal DLL4 binding. Thus, in a healthy normal adult thymus, the masses of DP cells actually insulate individual DN cells from competition with each other, allowing Notch triggering to be renewed repeatedly from ETP to DN3 stage.

Notch-DLL4 signaling activates the regulatory cascade that turns on T-lineage genes from ETP stage to DN2 stage to DN3 stage. T-lineage-specific signaling mediators and invariant TCR components are upregulated starting by the DN2 stage or in the DN2-DN3 transition. The IL-7R (Il7r gene) is fully turned on by DN2 stage to support proliferation of the precursors as they are acquiring their T-cell identity. Between DN2 and DN3 stage, RAG1 and RAG2 recombinases and multiple TCR signaling complex components and mediators of TCR signal transduction, including the kinases Zap70 and Itk, are all induced or upregulated. The specialized surrogate T-cell receptor α-like chain, pre-TCRα (pTα, encoded by the Ptcra gene), is also turned on to provide a partner for complex formation with any successfully-encoded TCRβ chain. Thus, the cells that reach the DN3 stage are already fully armed with substantial levels of the T-cell specific components that will be needed to transduce signals through the TCR, as soon as they have made a productive TCR gene rearrangement.

Notch signaling accomplishes this large-scale transcriptional mobilization by first activating the expression of several T-lineage transcription factors that can help the cells to open up previously silent genetic loci.^32,33,34 The most important T-cell-specific members of this group are GATA-3 (encoded by the Gata3 gene) and TCF-1 (encoded by the Tcf7 gene). These T-cell factors collaborate with the Notchactivated transcription factor RBPJκ (also called CBF1 – suppressor of hairless – Lag-1 = CSL), and together they interact with a set of transcription factors that is already present and active in multilineage hematopoietic progenitors. These preexisting factors include “E protein” basic helix-loop-helix factors E2A (encoded by Tcf3 = Tcfe2a) and HEB (Tcf12), Ikaros family members like Ikaros (Ikzf1) and Helios (Ikzf2), Runx family factors like Runx1 and Runx3, Ets family factors including PU.1 (SPI1 in human, Sfpi1 in mouse) and Ets1, and other key hematopoietic factors like Myb and Gfi1.³⁵ The knockout of any one of these factors or factor families is sufficient to abort early T-cell development.

This complexity is important to acknowledge because despite the essential role of Notch, no one transcription factor alone turns on the whole T-cell program in a single stroke. In fact, different factors collaborate with each other in a balanced way throughout the DN stages, to coordinate differentiation with the right degree of proliferation before the cells reach the DN3 stage. Furthermore, the E proteins E2A and HEB not only regulate T-cell identity genes, but also serve throughout T-cell development to enforce TCR signaling checkpoints, undergoing cycles of inhibition and reactivation in response to TCR stimulation. We will see others of these factors that also have different, dose-dependent roles later on, working with or against each other to cause one T-cell lineage to diverge from another.

The cells that populate the thymus are not yet committed to a T-cell fate when they arrive. Elegant single-cell culture experiments using either adapted organ culture systems or stromal coculture systems show that the same individual cells that can give rise to T cells would still be able to generate NK cells, myeloid cells, and DCs under other conditions, when they arrive in the thymus.^36,37,38,39 Several lines of research also imply that many of them are initially able to give rise to B cells.^40,41 The commitment process is the process through which the developing T cells decide permanently against these other options.⁴²

Given how similarly T and B cells rearrange their immunoreceptor genes and observe developmental checkpoints, it is not surprising that B-cell potential should be present in T-cell precursors. It is notable how early B-cell potential is lost—soon after the thymus-settling precursors enter the thymus; many experiments even imply that access to the B-cell fate can be lost even before thymic entry.^43,44,45 However, after losing the B-cell option, DN thymocytes maintain a clearly demonstrable ability to shift to a myeloid fate, either macrophage or granulocyte, and this persists into the DN2 stage.^37,38,39 As long as the cells remain in an intact thymus,
these possibilities remain latent and development progresses overwhelmingly toward a T-cell fate, but all that is needed to reveal these options is to remove the cells from the thymus and offer them cytokines that support myeloid cell survival. Even into the DN2 stage, some T-cell precursors also appear to be able to develop into mast cells, yet another nonlymphoid cell type.⁴⁶ The cells that start the T-cell program thus begin with many alternative possibilities.

A great deal of work has been carried out to link the developmental repertoire of newly arrived thymic immigrants with uncommitted cells in the bone marrow that are likely to be their precursors. This lively field of research has shown that thymic immigrants can be derived not only from cells that were mostly committed to some kind of lymphoid fate prethymically (ie, common lymphoid precursors), but also from cells that were broader-spectrum lymphoid-myeloid multilineage precursors (lymphoid-primed multipotent precursors).^{43,47,48,49,50} Furthermore, even cells that have already begun to be restricted to a nonlymphoid, myeloid pathway (granulocyte-macrophage precursors) can also develop into T cells, if transferred from the bone marrow into a thymic environment or OP9-DL1 coculture.⁵¹ Many T-cell precursors in the thymus in fact show evidence of having earlier activated a myeloid lineage differentiation gene, lysozyme M.⁵² Thus, the dominant developmental program imposed by the thymic environment is a kind of boot camp that can force cells of different developmental origins to converge on a common T-cell fate, erasing signs of their previous distinct identities.

The thymus uses different mechanisms to eliminate various fate alternatives for the cells. One of these is the environmental signal through Notch itself. As long as thymic immigrants and even ETP and DN2 cells are being exposed to Notch ligands in the thymic microenvironment, they cannot exhibit any alternative potentials efficiently. As differentiation progresses, however, the cells also give up the ability to access these alternatives under any conditions (Fig. 13.4).

Early B-cell development in particular is intensely inhibited by Notch signaling. Not only are precursors prevented from initiating B-cell development in the presence of Notch- Delta signals, but also exposure to Notch signals rapidly strips them of the capability to enter the B-cell pathway even if Notch-Delta signaling is removed.⁴⁰ Perhaps because early B and T cells share many other early developmental requirements —the cytokine IL-7, the role of stroma, prominent use of E proteins among their crucial lymphoid transcription factors—the Notch signal is uniquely important to make these two developmental programs mutually exclusive. The thymus is too good an environment for B-cell development in every other way; in fact it fills up with developing B cells if Notch-DLL4 signaling is defective.^41,53,54

The choice against other alternatives takes longer and involves more conditionality. The myeloid and DC-enabling transcription factor PU.1 is normally expressed into the DN2 stage and probably explains the access to myeloid potential up to this point. Under normal conditions, however, the thymus is not a good source of myeloid-cell growth factors or differentiation factors, and so myeloid development is a low-efficiency path in the thymus regardless of the programming of the cells. In addition, active Notch signaling appears to add a further conditional obstacle to myeloid development through interference with the action of myeloid transcription factors.^52,55 This mechanism still allows cells to keep myeloid potential as a latent option through several cell divisions, accessible if they are removed from the thymus. But eventually the T-cell developmental program reaches a stage when the myeloid-enabling transcription factors themselves are repressed. This occurs in the later DN2 stage, and it marks the point when the myeloid paths too are permanently cut off.

The most persistent fate alternative is the NK fate, which falls into a somewhat different category. The choice against becoming an NK cell appears to be made about the same time as the choice to forgo the DC fate option,^36,56,57 but in fact one could argue that many CD8+ T cells never fully give up the NK possibility. There are many differences between NK cells and T cells: most obviously, the NK cells do not rearrange their TCR genes and utilize a substantially different antigen recognition system than T cells. Thus, any TCR+ cell would be classified as T rather than NK no matter how NK-like it was in other respects. In an apparent dichotomy, too, developing T cells require a class of transcription factors, the E proteins (E2A, HEB) that NK cells must inhibit: the E protein antagonist Id2 is an NK developmental requirement. However, NK cells mobilize effector programs that are almost completely congruent with those of effector CD8+ killer T cells. On persistent antigen stimulation, CD8 killer T cells turn on expression of NK-cell activation receptors as well as TCR and behave even more like NK cells in terms of promiscuity of killing. A recent gene knockout model suggests that only one transcription factor, Bcl11b, may stand between mature CD8 cells and NK-like behavior normally throughout their lifetimes as mature T cells.⁵⁸ The loss, or reduction, of access to the NK program occurs at the specific stage of T-cell development, also in the DN2 stage, when Bcl11b is turned on for the first time.^58,59,60

Bcl11b is turned on in a process that requires Notch signaling and TCF-1,^61,62 possibly restrained until DN2 stage by other timing factors. DN2 cells that cannot turn on Bcl11b do not reach a normal DN2b or DN3 stage: instead, they keep prolonged access to myeloid fates and ongoing self-renewal programs as well as the NK program as long as cytokine levels permit.^59,60 They are profoundly defective in generating TCRαβ-lineage thymocytes, DP, and later stages, as well.^63,64 Thus, in addition to its special role in suppressing the NK-cell program, Bcl11b also appears to work in a gatekeeper capacity at the crux of the commitment process.

Commitment, therefore, involves at least three distinct mechanisms: the suppression of B-cell potential, the silencing of myeloid transcription factors, and the induction of Bcl11b. The last two of these happen roughly at the same point, in a transition that is now recognized to split the DN2 stage between uncommitted DN2a cells and committed
DN2b cells^56,65 (see Fig. 13.4). DN2a cells have access to other fates and can survive when removed from Notch ligandrich environments; DN2b cells are committed and become acutely Notch signaling-dependent for viability. From commitment on, DN cells cannot survive deprivation of Notch ligands until they have successfully undergone β-selection or γδ-selection. Transition to the DN2b stage also appears to prepare the way for shutting down precursor expansion and establishing the molecular conditions for enforcing the β-selection checkpoint.

All TCRαβ cells, regardless of functional subset, are survivors and products of β-selection. The β-selection checkpoint is the first TCR-dependent control checkpoint that these cells encounter, and the β-selection process is the first TCR-dependent signaling response that they undergo. The outcome shapes the whole population upon which later selection and differentiation mechanisms must operate.

TCRβ rearrangement and TCR-complex signaling are crucial for both the proliferation and differentiation events of β-selection. Thus recombinase-deficient (Rag1 or Rag2 knockout), DNA repair-deficient (eg, Prkdc^scid, “SCID”), TCRβ constant region knockout, or CD3γ or ε knockout mice all leave T-cell development blocked at DN3 stage (see Fig. 13.2). In normal development, successful TCR Vβ joining to Dβ-Jβ joints in a productive reading frame is the limiting event, as CD3 chains begin to be synthesized and expressed on the surface at a low level irrespective of rearrangement. It is the ability to deliver a signal through these complexes that really drives β-selection, as both in vivo and in vitro a response that strongly mimics natural β-selection can be triggered even in RAG-deficient thymocytes if monoclonal antibodies are used to crosslink the CD3 chains on their surfaces to deliver an artificial signal.

Pre-TCR spontaneous signaling triggers a dramatic cascade of events. In hours, the cells not only begin to proliferate again but also set in motion a profound transformation. The cells become large blasts, almost 10 times the volume of most thymocytes. As they turn on CD4 and CD8αβ expression, and express the coreceptor CD28 for the first time, they shed their DN cell markers and everything that had sustained their viability as DN cells, losing CD25, IL-7R, and the last vestiges of Kit expression and also inactivating the Notch pathway as well.

The proliferation unleashed by β-selection is intense. In their 2 to 3 days of proliferation, the selected DN3 cells generate approximately 10²-fold more DP cells. During this expansion, there are ample opportunities for clearing the cell surface of obsolete membrane proteins (eg, old CD25 molecules), clearing the nucleus of obsolete transcription factors (eg, the Notch-induced transcription factor Hes-1), and clearing the genome of obsolete states of chromatin modification.

During the proliferation triggered by β-selection, TCR gene rearrangement is temporarily suspended, as the cells downregulate Rag1 and Rag2 gene expression and inactivate RAG2 protein by cell cycle-dependent phosphorylation.⁷⁰ RAG1 and RAG2 expression is restored by the time proliferation arrests in the DP stage, but by that time the cells have transformed the chromatin landscape in which the RAG1/RAG2 recombinase must find its targets. TCRβ is now closed, establishing rigorous allelic exclusion for this TCR chain. TCRα is now open for rearrangement, pioneered by some germline transcription as early as the late DN3 stage. At the same time, the potential for expressing TCRγδ receptors is eliminated. TCRγ genes become inaccessible, so that both germline transcripts from unrearranged loci and transcripts from any previously rearranged TCRγ genes are silenced.^71,72,73 TCRδ genes, whether previously rearranged or not, are deleted as the TCRα segments that straddle them rearrange. Thus, by the time proliferation stops, the DP cell is irreversibly confirmed in a TCRαβ lineage.

With the first productive rearrangement of TCRα, the developing DP cell will acquire the capacity to recognize unpredictable antigenic targets, including self-antigens. Until the new receptor specificity has passed quality control, a DP cell is therefore functionally disabled and “contained,” both within the thymic cortex and within a short default lifespan. DP cells not only lack emigration receptor expression but also acquire features that would cause them to be killed quickly if they did escape. As they pass through β-selection,
they downregulate their own MHC class I molecules and their surface glycoprotein sialylation⁷⁴ pathways: they can even be purified on the basis of their resulting low MHC class I expression and preferential binding by the lectin peanut agglutinin. These changes make them into potential NK cell targets and ensure that they would be scavenged by macrophages via asialoglycoprotein receptors if they escaped. Also as β-selection begins, the antiapoptotic factor Bcl2 is repressed, replaced with the weaker or more conditional survival factors Bcl-xL and Bcl2A1.^75,76 Even while they are still rapidly dividing, the emerging cells acquire an intense vulnerability to death signals, both glucocorticoid signaling and other strong signals.⁷⁷

Cell surface expression of TCRαβ complexes are kept low on DP cells even when TCRα chains are first expressed, partly due to uniquely high levels of Src-like adaptor protein (Sla gene product), which targets TCRζ protein for rapid destruction.⁷⁸ Nevertheless, DP thymocytes are, if anything, more sensitive to TCR-ligand interactions than mature cells with the same TCRαβ if an appropriate assay is used. Contributing factors may include their relative lack of sialic acid residues, thus reducing electrostatic repulsion, and their low expression of the negative feedback molecule CD5 that could otherwise damp their response to TCRαβ signals.^79,80,81,82 Even though it is sensed, TCR signaling cannot turn on functional response genes like Il2 in DP cells. In part, this is because these cells have downregulated key activationtransducing transcription factors, c-Jun and c-Rel (for more information, see www.immgen.org/).⁸³ β-selection also induces expression of the transcription factor, RORγ, from a distinctive promoter (RORγt, also called RORC2), which actively blocks conventional effector responses even while maintaining Bcl-xL expression throughout the DP stage.^84,85 RORγt expression sets DP cells apart from all other major classes of thymocytes; however, it provocatively links them with Th17 cells in the periphery, with lymphoid tissue-inducing cells that organize lymph nodes in fetal life, and certain classes of newly recognized innate effector cells.^86,87,88,89 The DP cells that emerge from the complex regulatory events of β-selection thus acquire a highly specialized physiology, and they are poised for selection.

The β-selection response is driven by the same major signaling pathways used in the activation of mature T cells. Lck, Ras pathway, PI3-kinase, and protein kinase C activation all play roles in the process, using virtually all the same adaptor molecules [LAT, Slp76 (Lcp2), GADS (Grap2)] that are used to coordinate signaling in positive selection later and in mature T cells.^90,91,92 As in later stages of T-cell development, TCR signaling temporarily antagonizes effective activity of E proteins (E2A/HEB). A transient wave of immediate-early Egr factor activation results in transient upregulation of the E protein antagonist Id3, a response that will be echoed in positive selection later.^93,94,95,96 As the cells reach the resting DP state, high-level E protein activity is restored.⁹⁷ These factors cooperate with some of the same regulatory factors that are used in early T-cell development: initially Notch signaling, then the important T-cell transcription factor TCF-1.^{98,99,100,101} The fact that the cells emerging from this signaling response are idiosyncratic, functionally disarmed, short-lived DP thymocytes rather than robustly activated effector T cells must be explained as the result of the characteristic transcriptional and epigenetic regulatory state of the cells in this stage—the context within which the signaling pathways are sensed—rather than a unique biochemical pathway for the signals themselves.

The cell biology of the β-selection response is an obvious hazard for the organism. Intense polyclonal proliferation of this magnitude brings the cells close to a malignant state, and the program must include brakes to stop the proliferation as well as constraints to prevent it from being triggered inappropriately. In fact, most T-cell acute leukemias have precisely the phenotype of cells that cannot stop the β-selection response, continuously generating immature DP cells or DN-DP intermediates. Under normal conditions, a limiting factor for the process is the programmed downregulation of Notch responsiveness, as the extent of population expansion during β-selection is strongly influenced by Notch signaling.^102,103 Indeed, by far the most dominant and consistent feature of T-cell acute lymphoblastic leukemias is the presence of gain of function mutations in the Notch1 locus, especially mutations that make Notch signaling ligand-independent.¹⁰⁴ The result is particularly grave when the cells can join viability signals from pre-TCR complexes and Notch alike to proliferate without restraint.

Another dangerous combination is if the very rapid proliferation program induced by β-selection is superimposed on any continuing expression of hematopoietic progenitor self-renewal genes, inherited from earlier stages of T-cell specification, including SCL (Tal1), Lyl1, Lmo1, Lmo2, and Hhex.^105,106 These genes are normally turned off during the lineage commitment transition, between DN2a and DN2b, but may easily act as T-cell oncogenes if kept on. Thus, β-selection may not be safe until these genes have been successfully repressed. The implication is that progressing into a normal DN2b-DN3 stage sequence may be important to establish the correct control over future growth.