Correspondence: Professor EV Rothenberg, The Division of Biology 156-29, California Institute of Technology, Pasadena, CA 91125, USA. E-mail: evroth@its.caltech.edu

The essence of a developmental decision is to convert a temporary difference between two cells, for example in their current exposure to environmental signals, into a long-term, irreversible difference in phenotype and function. For years, one of the most striking examples of this mechanism has been the choice of developing T cells between maturation in the CD4⁺ CD8^- lineages or the CD4^- CD8⁺ lineages, but the molecular mechanisms underpinning this decision have remained poorly understood. Three elegant studies published recently in Nature Immunology establish a sophisticated hierarchy of transcriptional regulators in which expression of triggering factors is followed by expression of enforcement factors, including Thpok and Runx3, to determine CD4/CD8-lineage choice.^{1, 2, 3}

More than just a difference in cell surface co-receptor expression, the CD4/CD8 divergent pathways are linked with permanent differences in functional repertoires, growth control, and modes of homeostatic maintenance and memory. Yet, it has become increasingly clear that the two alternative fates are both triggered by signals through essentially the same receptor complex, delivered to essentially the same immature cells in the thymic cortex. This occurs during the T-cell receptor (TCR)-mediated 'positive selection', which rescues a select few thymocytes in each generation from death: cells with receptors that interact with class II MHC may be positively selected to become CD4⁺ cells, whereas those with receptors that interact with class I MHC may be positively selected to become CD8⁺ cells. It is now understood that the divergence between CD4⁺ and CD8⁺ cell fates is determined mainly through differences in signal intensity and duration.⁴ The challenge has been to understand the regulatory circuitry that can convert such a fleeting difference in experience into the stable, long-lived, heritable programming of mature T-cell subsets.

Major advances have been made in understanding the way such decisions are made in a variety of embryological systems. In each case, specific circuits within gene regulatory networks are involved.⁵ Key circuit elements are used for different biological jobs. One will be used to transform temporary to permanent states, as in positive selection generally. Another will be used to set a particularly stringent threshold for a given event, as in selection to the CD4 lineage. Yet another will be used to make one state permanently exclusive of an alternative state, as in the CD4 vs CD8 cell decision. Figure 1 shows three of these central motifs: the 'feed–forward' motif used to set sharp conditional thresholds, the 'lockdown' motif used to perpetuate a newly induced state and the 'mutual repression' motif used to make two states mutually exclusive. Note that in each case, one transcription factor's activity becomes significant based on its regulatory relationships with other factors, as a collaborator, an activator or a repressor. The basis of a lineage decision is clear only once these relationships have been revealed. It is this kind of illumination that is finally emerging for the CD4/CD8-lineage choice over the past year, in a confluence of elegant work from several groups.^{1, 2, 3, 6, 7, 8, 9}

Figure 1.

Classic gene network motifs used in developmental cell-fate decisions. (a) Feed–forward motif. In this circuit, if activation of regulatory gene c requires both the products of regulatory genes a and b, then the activity of c can exhibit a very sharp off-to-on transition behavior, causing a switch-like effect on targets. (b) Mutual regulatory reinforcement. This circuit converts a transient input signal to a sustained transcriptional response. In this circuit, regulatory gene j is turned on by an input signal initially, but then the product of j turns on another regulatory gene, k, which sustains the expression of j even when the initial signal is removed. (c) Mutual exclusion circuit. Two regulatory genes, p and q, can be turned on by related input signals (1 vs 1+2), yet induce mutually exclusive responses if they each repress each other's expression.

Full figure and legend (62K)

A major step toward solving the CD4/CD8 choice mechanism was the discovery 3 years ago of a central player in the mechanism, the transcription factor, Thpok (also known as Zbtb7b, cKrox and Zfp67).^{10, 11} Thpok is induced only by the kinds of signals that cause CD4⁺ cell positive selection, and both gain and loss of function experiments reveal a powerful role for this factor in choosing between the CD4 and CD8 developmental programs. In the absence of Thpok, cells that should become CD4 cells become CD8 cells instead; conversely, when Thpok is expressed forcibly in thymocytes, cells that should become CD8 cells are redirected to become CD4 cells. This has been interpreted to mean that Thpok is a 'master regulator' that is both 'necessary and sufficient' for the CD4⁺-cell fates. But in fact, Thpok has provided a very powerful tool for probing deeper into the regulatory network that represents the true lineage choice mechanism.

Other factors have been well known to be involved in the choice, but with less easily categorized roles. On the other side, Runx3 has been deeply implicated in CD8⁺-cell development for years, initially through its involvement in the silencing of CD4 expression itself.^{12, 13, 14, 15, 16, 17} But the role of Runx3 has appeared less simple than that of Thpok, as it has not always seemed capable of imposing the CD8⁺-cell fate on cells that should be selected as CD4⁺ cells. Runx3 has also appeared less convincing as a reciprocal 'master regulator' because of uncertain redundancy with Runx1,^{14, 15, 16, 17, 18, 19, 20, 21} a closely related factor that is expressed in a completely different developmental pattern. In addition to Runx3 and Thpok, at least two other transcription factors were implicated. Early evidence showed that the essential T-lineage transcription factor GATA-3 was specifically required for CD4⁺-cell development and was capable of blocking CD8⁺-cell development,^{22, 23, 24} but GATA-3 appeared to fail the 'master regulator' test of redirecting positively selected cells from one lineage to the other. Furthermore, the high-mobility group box factor TOX was found to be implicated in both CD4⁺ and CD8⁺ positive selection,^{6, 25} with a specific catastrophic loss of all CD4⁺ cell lineages in the case of TOX knockout mutants.⁶

The new work now distinguishes between triggering factors and enforcement factors in the CD4/CD8-lineage choice, and shows that Thpok is mostly an enforcement factor for the CD4⁺ cell lineage. As hinted by some early results, it is actually not induced until a mid to late stage of CD4⁺-cell development, subsequent to the lineage-specific upregulation of GATA-3 and the lineage-nonspecific upregulation of TOX. GATA-3 levels in particular depend on the integrated magnitude of TCR signals over time,^{22, 26} and both GATA-3³ and TOX⁶ are required to trigger Thpok (Zbtb7b) expression. Yet, new results show that even forced Thpok expression cannot promote CD4⁺-cell development unless GATA-3 can also be induced.³ GATA-3 is indispensable for the earliest stages of CD4⁺ positive selection, entry stages that do not depend on Thpok.³ Indeed, this result sheds new light on the reasons why loss of GATA-3 may not always yield visible lineage redirection: it may severely reduce the number of cells going through the very entry point for positive selection, so that any subset that is then eligible to switch pathways could be too small to detect. Thus, the CD4⁺ differentiation that depends on GATA-3 as well as Thpok emerges from a classic feed–forward circuit, in which Thpok induction requirements multiply the assurance that CD4⁺ development will only be elicited by strong, sustained TCR-Lck signaling events. Then, as Thpok can also induce GATA-3 expression in thymocytes,¹⁰ these two factors may become engaged in a mutually reinforcing positive cross-regulation that sustains the CD4⁺-cell differentiation program to maturity (Figure 2a).

Figure 2.

Proposed network for regulatory interactions in CD4/CD8-lineage choice. The figure summarizes the interactions described in the text. Here, (a) indicates the position of the activating inputs required for Thpok (Zbtb7b) induction and the mutual positive feedback of Thpok on GATA-3; (b) shows the Runx repression of Thpok and its antagonism by Thpok itself, in parallel with the similar action on the Cd4 gene silencer; (c) shows the repressive input of Thpok on the distal promoter of Runx3, which is one key element of the prohibition of CD8⁺-cell differentiation by Thpok. It is not clear whether the Thpok effects on GATA-3 or on the Runx3 promoter are direct, and the exact mechanism for Runx3 repression of Thpok is also under intense study. Additional effects of Thpok on CD8⁺-cell-associated genes are shown based on Jenkinson et al.²⁷ and Wang et al.²⁸ The exact regulatory control that positively establishes CD4⁺-cell gene expression, especially in the absence of Thpok, is not established. Note, however, that effects on the expression of CD4 and CD8 themselves can feed back on the strength and duration of TCR signaling experienced during positive selection (left of figure), depending on the major histocompatibility complex restriction of the TCR expressed.⁴

Full figure and legend (57K)

The coherence of the CD4 and CD8-lineage choices now seems to result from an asymmetrical cross-repression between Runx3 and Thpok. Elegant knockin fluorescent reporter alleles of Runx3 and Zbtb7b display this cross-regulation at the single-cell level.^{1, 2} Zbtb7b has complex regulatory elements that include a silencer, which keeps expression off in CD8⁺ lineage cells.^{7, 8} There is some evidence that Runx3 can participate in this repression complex directly,^{1, 7} in parallel with its repression activity on the Cd4 silencer element (although the mapping of sites for this repression is not fully clear).⁸ Thus, if Runx3 is activated, it can block CD4⁺ T-cell development in two ways: by downregulating the CD4 coreceptor, on which the sustained signaling for Thpok induction depends, and perhaps by directly repressing the Thpok (Zbtb7b) gene expression as well. New evidence shows that when induced successfully, Thpok can protect its own expression actively by binding to the same silencer element vicinity as the Runx factor, and in some way antagonizing the repression activity of Runx.¹ This anti-repression activity of Thpok seems to be exerted in parallel on its own silencer element and on the Cd4 silencer element^{1, 29} (Figure 2b). This kind of modulatable repression may be a general feature of Runx factors: it is interesting that Runx transcription factors are implicated now in at least three cases where anti-repression is used to protect a target gene by changing the effect of Runx binding without necessarily eliminating or displacing it.^{1, 29, 30} In cells undergoing CD4⁺ positive selection, Thpok needs to be able to protect itself and CD4 specifically against Runx-factor repressive activities at the target gene level, not just through repression of Runx3, because CD4⁺ cells depend on continuous expression of the Runx3 relative, Runx1, for their own optimal maturation.¹²

However, Thpok also seems to cement CD4⁺-cell fate by working, directly or indirectly, to block Runx3 expression from the promoter that yields the most active form of that protein² (Figure 2c). When Thpok is conditionally deleted, Runx3 expression is readily detected in cells undergoing CD4⁺ lineage selection. The results confirm that the early events of CD4⁺ cell positive selection might normally induce Runx3 gene activation, on a leisurely timescale, just like the early events of CD8⁺ cell positive selection,^{2, 31} but that this is aborted normally by the rise of Thpok.² Therefore, there is a kinetic competition between Thpok and Runx3 upregulation, which becomes biased in favor of Thpok by sustained TCR-Lck signals, through the special circuitry that activates Thpok expression. As a corollary, Thpok's effectiveness is dose dependent. Two of the new reports show that if cells start CD4⁺-cell development with artificially lowered Thpok expression, they are unstable and many of them switch over to the CD8⁺ pathway.^{1, 2}

In fact, Thpok seems most important as an antagonist of Runx3 and the competing CD8⁺-cell gene expression program. Indeed, even in mature CD8⁺ T cells, forced expression of Thpok can cause loss of CD8 and CD8⁺-cell-associated gene expression and appearance of some CD4⁺-cell-associated functions.²⁷ A new report shows that the downregulation of Runx3 and another CD8⁺-cell transcription factor, eomesodermin, are central to the effects of Thpok as a guardian of CD4⁺-cell identity.²⁸ A remarkable finding is that, if Runx activity overall is sufficiently reduced during positive selection, then Thpok too can be deleted and still allow a significant number of CD4⁺ cells to develop.² The CD4⁺ phenotype and even some hints of CD4⁺ cell function can be generated without Thpok involvement, as long as the cells can escape redirection.

The regulatory basis of the CD4/CD8-lineage choice is thus more complex than a simple 'master regulator' commanding the acquisition of a particular fate assignment. Instead, distinct subcircuits of a TCR-triggered regulatory network control upregulation of two mutual antagonists, through a system that incorporates a time-and-intensity measurement subcircuit with feed–forward architecture, stabilization subcircuits and fate–exclusion subcircuits that work through a combination of transcriptional activation and modulated repression linkages. These complexities, seen with hindsight, are completely in accord with the biology of the lineage choice itself, which is based on an integration of signal strength over time⁴, not a stop/go signal dichotomy. There are many questions that remain, about details in the experimental designs used in the new studies, and there are many mechanistic issues that still need to be resolved by future work. The cis-regulatory analysis is still in an early stage, and the detailed links between the lineage choice factors described here and the stable features of CD4⁺ or CD8⁺-cell phenotype that they elicit still need to be forged. But the questions that lie ahead are increasingly targeted to reveal the full architecture of a network that is worthy of the biologically sophisticated developmental choice mechanism itself.