Effector T cell plasticity: flexibility in the face of changing circumstances

Abstract

As additional states of CD4 T cell differentiation are uncovered, their flexibility is also beginning to be recognized. Components that control the plasticity of CD4 T cell populations include their cellular conditions, clonality, transcriptional circuitry and chromatin modifications. Appearance of cellular flexibility may arise from truly flexible genetic programs or alternately heterogeneous populations. New tools will be needed to define the rules that allow or prohibit cellular transitions.

The differentiation of the CD4⁺ T cell lineage into effector cells underlies successful adaptive immune responses aimed at distinct categories of pathogens. Their functional specialization is coordinated by genetic programs that use different transcription factors to direct expression of distinct soluble mediators and surface molecules that support interactions with other immune cells (Fig. 1). The first paradigm for this functional diversification was the description of T_H1 and T_H2 CD4⁺ effector subsets by Mosmann and Coffman in 1986 (ref ¹). T_H1 cells were thought to be responsible for delayed type hypersensitivity (DTH), activating macrophages via release of interferon (IFN)γ and enabling them to kill intracellular pathogens. T_H2 cells were considered the classical helper T cells providing help to B cells to generate class switched antibodies.

Figure 1.

Re-enforcement and destabilization of CD4⁺ T cell subsets. These four CD4⁺ subsets are induced in distinct conditions, but also can be reinforced or destabilized by other conditions, as discussed throughout the text.

Several attempts to add new subsets to this dichotomy were thwarted by the inability to identify consistent robust inducing conditions or transcriptional “signature”. As such, the formal status of the subsets previously known as T_H3 or TR1 appears uncertain. But in 2003, the requirement for interleukin-23 (IL-23) in IL-17-producing CD4⁺ T cells was recognized and a role for such cells, rather than T_H1 cells, was established in the experimental autoimmune encephalomyelitis (EAE) model². This data established a role for CD4⁺ T cells secreting IL-17 rather than IFNγ in diseases such as EAE or collagen-induced arthritis. Initially presumed to diverge from a common T_H1 precursor³ the IL-17 producing cells, named T_H17, were classified as a new subset on the basis of being independent of the transcription factors GATA-3 and T-bet^{4, 5}. The robust inducing conditions of IL-6 and TGFβ⁶ and the identification of RORγt and RORα as lineage defining transcription factors^{7, 8} finalized support of T_H17 as a separate subset.

The fourth major subset of CD4⁺ T cells are Treg cells⁹ characterized by expression of the transcription factor Foxp3. Tregs derived from the thymus are thought to be a stable subset. However, Tregs can be induced in the periphery from naïve CD4⁺ T cells by exposure to TGFβ. Similar to the stable thymus-derived Treg, inducible Treg (iTreg) express Foxp3, but may be less stable and share circuitry with T_H17 cells which also require TGFβ for their differentiation (reviewed in ¹⁰).

Subsets ‘in the making’

Follicular helper T cells (T_FH) residing in B cell follicles are essential for the generation of high-affinity isotype switched antibodies and B cell memory^11-14, a characteristic that originally defined CD4⁺ T cells as ‘helpers’. Although all CD4⁺ T cells migrate to follicular regions, T_FH cells are preferentially resident there by virtue of their continuous expression of the chemokine receptor CXCR5. CD4⁺ T cells expressing CXCR5 have the potential to secrete T_H1, T_H2 or T_H17 cytokines. Therefore, it is unclear whether T_FH are a distinct subset or rather a ‘chameleon’ state of other subsets that are imprinted by follicular location. T_FH produce high levels of IL-21, which acts in an autocrine manner together with IL-6 on their differentiation and expansion, a process that also depends on the Bcl-6 transcription factor^{11, 15-19}.

A population of IL-9 producing cells, derived from T_H2 cells by treatment with TGFβ, has also been described²⁰. IL-9 was once considered a T_H2 cytokine, but is now recognized as not being co-expressed with IL-4, IL-5 or IL-13. Although suggested to be produced by T_H17 cells or iTreg^21-23, IL-9 is not co-expressed with IL-17 or with IL-22, and is not expressed by nTregs or iTreg²⁰. Having been only examined in vitro, it is unclear whether IL-9 producers should be considered as a new subset, to be called T_H9, or whether expression of this cytokine reflects adaptation of T_H2 cells to a change in the microenvironment in the course of a response triggered by a pathogen or allergen.

More recently, human, but not mouse, T_H22 T cells (expressing IL-22 but not IL-17 or RORγt) were described^{24, 25}, and may represent a skin-homing subset responsible for skin inflammation such as psoriasis. These cells preferentially develop when cultured with plasmacytoid DC (pDC) which infiltrate psoriatic skin, but are independent of (and even inhibited by) IFNα,²⁴, making their link to skin inflammation still uncertain.

Flexibility of T cell subsets- plasticity or chaos?

The T_H1-T_H2 paradigm has been useful but also overused. It provided the needed framework to identify the cytokines, signaling molecules and transcription factors controlling important effector pathways²⁶. Overuse as a simplistic dichotomy was evident as well^{27, 28}, but within this framework came suggestions of flexibility in polarization both in mouse and human cells^29-33. The relative flexibility of early T_H1 cells compared to rapidly stabilized T_H2 cells is partly explained on the basis of early IL-12 receptor expression ³⁴, and by the fact that irreversible commitment of CD4⁺ T cells might be the end result of repeated stimulation with antigen in vitro or chronic disease in vivo³⁵.

The finding that irreversible fixation of T cell fate appears to require a certain number of divisions³⁶ established an early precedent for later considerations of chromatin modifications in gene regulation³⁷. This question seems particularly relevant to the origin of memory T cells from either committed effector cells or a branch-point before terminal commitment^{38, 39}. Mouse memory T cell populations are considered flexible with regard to cytokine production rather than irreversibly committed³¹, and this was later confirmed in human memory populations as well⁴⁰.

Several modes of plasticity of T cell subsets have recently been described (Fig. 1). First, inducible Treg (iTreg) and T_H17 cells (reviewed in ⁴¹) show effector differentiation plasticity. Thymus-derived Treg have stable Foxp3 expression, but iTreg generated in vitro are unstable (reviewed in ⁹). Runx transcription factors control the expression of Foxp3, allowing induction of iTreg and maintaining the Treg program in thymus derived Treg⁴². TGFβ induces both Foxp3 and RORγt in naïve T cells, but Foxp3 is dominant and antagonizes RORγt function unless IL-6 is present⁴³. Thus, an inflammatory environment tilts the balance between iTreg and T_H17 differentiation. The issue could be particularly important in the gut, where high production of TGFβ and retinoic acid by CD103-expressing dendritic cells supports favours transition from naïve CD4 cells to iTreg^44-46. Second, in vitro T_H17 differentiation shows a STAT4- and T-bet-dependent plasticity towards a T_H1 profile in mouse^47-49 and humans⁵⁰ (Fig. 1). Beyond this, even stably committed T_H2 cell can re-express IL-12Rβ2 and produce both IL-4 and IFNγ following a viral infection in vivo⁵¹. Both these transitions appear to depend on the IL-12 receptor as a central point of control. But certain transitions between subsets apparently do not take place; transition from T_H2 to T_H-17 or Treg, or T_H1 to Treg or naive T cell.

In trying to make sense of these observed or excluded transitions, the general notion of a potential function may be of value. For example, the relationships between CD4 subsets can be displayed in a manner analogous to the energy levels of an electron in the atom (Fig. 2). Stable energy states do not decay; they are not plastic (as if at an energy minimum). For the most part, transitions proceed from a higher energy (i.e., less stable) to lower energy (more stable) state, much like excited atomic states decay into less energetic modes. T_H1 cells can arise by differentiation of naïve T cells via IFNγ and IL-12 signaling, or by differentiation of T_H17 cells. The plasticity allowing T_H17 cells to convert to a T_H1 phenotype can thus be described as an allowed transition from ‘unstable’ to ‘stable’, or by higher to lower ‘energy’. Transitions are triggered by TCR stimulation provided in specific conditions. The primary antigen triggered state is the least stable and decays into one of the known subsets following T cell activation. Thus, what might initially appear as chaotic switching between various T cell phenotypes may actually reflect a predictable order of allowed transitions, based on some aspect of this relative stability.

Figure 2.

Transitions of CD4⁺ T cell subsets. Naïve T cells, when triggered, behave as an ‘unstable state’ that can transition into various other subsets, depending on the conditions as indicated. The diagram suggests a possible hierarchy of stability. Experimentally observed transitions are indicated, along with the required conditions and factors that appear to mediate each step. On the left, subsets that maintain IL-12 receptor remain responsive to IL-12, while on the right, T_H2 cells loose IL-12R expression. A hybrid state of T_H1+2 has been observed to be induced from T_H2 cells, in response to type I interferons, and IFN-γ and IL-12. Known interactions between transcription factors are indicated. Subsets whose status is still uncertain are not presented.

Cytokines, costimulation, circuitry, clonality and chromatin

What could determine the relative stability between CD4⁺ subsets? Components that regulate T cell stability and plasticity can be divided into four categories; conditions, circuitry, clonality and chromatin. Conditions, both cytokines and costimulation, are the prime factors in differentiation, but also impact stability. Circuitry, meaning the network of interactions between transcription factors, impacts differentiation, stability and plasticity. The issue of T cell clonality has been largely ignored in recent discussions of plasticity, but as we will see, must be considered when examining T cell differentiation. Finally, chromatin modifications associated with active or repressed genes control the maintenance of the phenotype, and, as recognized more recently, its plasticity.

Antigen presenting cells drive CD4⁺ T cell differentiation through cytokines and costimulation⁵². Dendritic cells are important in the initial activation and development of CD4⁺ T cells⁵³, but additional types of innate cells influence CD4⁺ T cell subsets. For example, neutrophils promote the differentiation of the adaptive T_H17 response. Phagocytosis of apoptotic infected neutrophils by dendritic cells is a highly efficient stimulus for T_H17 differentiation⁵⁴. Because IL-17 itself plays an important role in promoting recruitment and differentiation of neutrophils ⁵⁵ it is clear that γδ TCR T cells that provide the first rapid IL-17 response not only precede differentiation of the adaptive T_H17 response, but are most likely instrumental in promoting their differentiation^56-58.

T_H1 responses on the other hand are helped on the way by natural killer (NK) cells providing the first source of IFNγ^{59, 60}. The crucial issue for T_H2 responses had always been the initial source of IL-4, which is important, albeit not indispensable, for their differentiation. Another innate cell type, the basophils, produces substantial amounts of IL-4 have been implicated in the induction of T_H2 responses^61-63. Other mediators, such as IL-25 and TSLP, coupled with suppression of IL-12 production by dendritic cells, can promote T_H2 differentiation even in the absence of IL-4 (reviewed in ⁶⁴). Beyond these major effects of cytokines derived from innate cells on T cell differentiation, it has been suggested recently that asymmetric cell divisions may contribute to heterogeneity in the expanding T cell population⁶⁵. How this process might be regulated or come into play with respect to flexibility of CD4⁺ T cells has not been addressed.

Circuit rules

Transcriptional circuitry can promote either stability or plasticity. First, circuits can stabilize phenotypes by being self-reinforcing. For example, the transcriptional circuitry in T_H1 and T_H2 cells underlies their relative stability compared to T_H17 cells and iTregs (Fig. 3). In T_H2 cells, the transcriptional autoactivation of GATA-3 provides a self-reinforcing feedback circuit⁶⁶. Likewise, T-bet induces its own expression, either directly⁶⁷ or indirectly⁶⁸. Second, transcriptional circuits can promote developmental divergence of subsets, and this divergent impulse acts to stabilize one phenotype over another. For example, the repression of IL-12 receptors during T_H2 differentiation makes these cells less responsive to IL-12, further promoting T_H2 development³⁴ Likewise, transcriptional interference between T-bet and GATA-3 proteins, where one factor neutralizes the transcriptional activities of the other, magnifies developmental imbalances and promotes divergence⁶⁹.

Figure 3.

Transcriptional circuits can stabilize or destabilize CD4⁺ T cell subsets. The relative stability of states is shown by a hypothetical energy function, with more stable states residing within low energy ‘wells’. The mutual antagonism between T-bet and GATA-3 activity makes the states that co-express the two factors unstable. The positive feedback loops known to reinforce the expression of these factors stabilize the states with high expression of one of the two factors.

By contrast, the transcriptional circuitry of iTregs and T_H-17 cells may dictate a hierarchy of instability relative to T_H1 and T_H2 cells. First, no functional evidence for transcriptional autoactivation for RORγt or Foxp3 has been established, although Foxp3 may play a role in maintenance of its expression in iTreg cells⁷⁰. Foxp3 can inhibit RORγt's transcriptional activity⁴³, but no reciprocal effect has been reported. At first, such unilateral inhibition of RORγt by Foxp3 would imply that iTreg should be more stable than T_H17 cells (Fig. 4a); in this scenario T_H17 cells would require continuous IL-6 input to maintain RORγt expression, at least without additional re-enforcement. However, RORγt induces IL-21, which acts by an autocrine pathway to maintain RORγt expression via STAT3 activation^71-73 (Fig. 4b); in this scenario, the inhibition of RORγt's transcriptional activity by Foxp3 produces only a transient stability for iTreg that is overcome by the self-reinforcement of RORγt through IL-21. It is not yet clear how the aryl hydrocarbon receptor (AhR) integrates into the transcriptional network of T_H-17 cells. Transduction of AhR deficient T cells with RORγt, AhR or both indicated that there does not seem to be cross talk between the two transcription factors; AhR drives the expression of IL-22, but only does so if both IL-6 and TGFβ are present⁷⁴ IL-23 appears responsible for amplification of IL-22 (ref ⁷⁵), but cannot restore IL-22 induction in AhR deficient T_H17 cells.

Figure 4.

Possible transcription factor interactions regulating intermediate CD4⁺ T cell transitions. (a) Hypothetical circuit in which Foxp3 inhibits the activity of RORγt, without a reciprocal inhibition or feedback from RORγt. This circuit might favor iTreg development over T_H17 development, making iTreg more stable. Without continuous stimulus by IL-6 to induce RORγt, Foxp3 might tend to repress T_H17 development. (b) This hypothetical T_H17 state is stabilized despite inhibition of RORγt by Foxp3 due to a known feedback loop, in which RORγt induces expression of IL-21, which acts in an autocrine manner to further induce STAT3 and stimulate RORγt expression. This feedback stabilizes the T_H17 state relative to iTreg cells. (c) The transcriptional inhibitor Bcl-6 inhibits its own transcription. Such a mechanism could destabilize the T_FH phenotype, consistent with the fact that in vitro cultures of T_FH phenotype cells have not been reported as stable, in contrast to the robust stability of T_H1 and T_H2 cells.

While the lineage status of T_FH cells is debated, their dependence on Bcl6 is clear^{16, 17, 76}, and this transcription factor is considered a master regulator for T_FH cells. It should be noted that after 10 years⁷⁷, T_FH cells have still not been stably derived in vitro, as it was readily achieved for the self-reinforcing T_H1 and T_H2 cells. Again, circuitry may be at work. It turns out that Bcl6 protein actually inhibits Bcl6 transcription^{78, 79}, the opposite pattern from the GATA-3 autoactivation seen in T_H2 cells (Fig. 4c). Such autoinactivation by Bcl6 could promote T_FH instability, much like the transcriptional interference between T-bet and GATA-3, or Foxp3 and RORγt⁴³, promotes developmental divergence.

Two other transitions can occur (Fig. 1). T_H17 can transit to T_H1 cells (at least in vitro), and T_H2 cells can acquire a stable, hybrid state of T_H1+2 in vivo in response to the combined actions of type I interferon, IFN-γ and IL-12 during viral infection⁸⁰. No known circuit explains the instability of T_H17 cells and their transition to T_H1 cells, but this transition requires the same factors already known to induce T_H1 differentiation. Conceivably, T-bet, like Foxp3, could directly bind and neutralize RORγt, or simply might inhibit RORγt transcription. Evidence that T-bet inhibits the level RORγt protein was recently provided, although a transcriptional basis for this was not revealed⁸¹. In the development of T_H1+2 cells, a key step is the re-expression of the IL-12 receptor through the actions of type I interferons and IFN-γ, although the transcription factors involved were not identified.

Clonality, plasticity and heterogeneity

Consideration of the plasticity or stability of CD4⁺ T cells is complicated by issues of clonality or heterogeneity. For example, early T_H2 cultures rapidly become refractory to reversal to a T_H1 phenotype³⁴; this correlated with the early loss of IL-12 receptor in T_H2 conditions. In contrast, early T_H1 populations remain reversible for longer³⁴. Importantly, this could either be due to maintenance of functional IL-4 signaling by cell in T_H1 conditions or simply from uncommitted cells present in a heterogeneous T_H1 population³⁴. The alternative explanations of flexibility versus heterogeneity need consideration whenever a population is described as showing ‘plasticity’. Indeed, in vitro polarization never yields 100% pure cells. Early T_H17 polarizations were in the range of 5 to 20%. Attributing a cytokine profile to such mixed populations on the basis of ELISA or RT-PCR data generates a considerable amount of ‘noise’.

Some of the variability in T_H17 polarization has been attributed to the presence or absence of natural AhR ligands in some culture media^{74, 82}. Iscove's modified Dulbecco medium (IMDM) promotes far better T_H17 polarization and IL-22 induction than RPMI and the addition of an AhR inhibitor to IMDM medium reduces polarization and abrogates IL-22 production to the level of that seen in cultures of AhR-deficient CD4⁺ T cells. Thus, the role of AhR in the induction of IL-22 highlights the great potential for variability in T_H17 cell generation in vitro. This fact directly impacts the studies of epigenetic modifications in T_H17 cells, as these have been so far carried out with in vitro generated T_H17 cells that lacked the full arsenal of signals⁸³.

The role of AhR in T_H17 development and effector function revealed an environmental effect on T_H17 generation ^84-86. Activation of AhR enhances the pathology of EAE by increasing T_H17 polarization and cytokine secretion and is indispensable for the induction of IL-22 in T_H17 cells as well as IL-17 producing γδ T cells^{57, 84} Because this transcription factor is activated by a wide range of ligands, including environmental pollutants as well as endogenous physiological ligands⁸⁷, there is substantial scope for external influences on the T_H17 program. Of particular interest, because of their wide availability, are ligands derived from tryptophan metabolism such as 6-formylindolo[3,2-b]carbazole (FICZ)⁸⁸. Remarkably, L-tryptophan is associated with the development of clinical disorders such as eosinophilia-myalgia-syndrome⁸⁹, but it remains to be determined if T_H17 cells and AhR ligands play any role in this syndrome.

Because studies of T cell subsets involve multiple rounds of cell division, distinguishing individual cell flexibility from heterogeneity should involve interrogation of responses of single cells, such as lineage tracing studies, or alternately, cloning studies. At least one study carried out clonal analysis to demonstrate the instructive mechanisms of GATA-3 in T_H2 commitment⁹⁰. However, no study describing plasticity of T_H17 or iTreg cells so far has taken the step to examine clonal populations. Lineage tracing using Cre-lox systems have not yet been applied to iTreg, T_H17 or the T_H1+2 patterns of flexibility. However, this approach will bring new difficulties, such as whether levels of Cre recombinase expressed by a heterologous locus accurately reflects the level of the endogenous gene in the committed state.

In vivo versus in vitro

In vitro plasticity or stability may not always reflect in vivo behaviour. In vitro differentiation conditions may differ from those in vivo, which may influence the degree of polarization and the nature of the cells generated. Indeed, some studies report that in vivo generated T_H17 cells are more stable than those derived in vitro⁹¹, while others suggest considerable plasticity in Tregs isolated ex vivo⁴⁷. Treg cells were shown to acquire the capacity to express either IL-17 or IFNγ (reviewed in ⁴⁷). More intriguingly, Treg seem to be able to adapt their suppressive mechanisms to particular effector programs. Thus, expression of T-bet in some Treg cells leads to upregulation of CXCR3 and correlates with their capacity to control T_H1 responses, while a T_H1 inducing environment promotes T-bet expression in Treg cells⁹² without inducing differentiation into effector cells. On the other hand, under highly inflammatory conditions that lead to host death, Treg cells not only express T-bet, but lose of their regulatory capacity and adopt an effector profile associated with production of IFNγ⁹³.

Similarly, IRF4, which is essential for T_H2 effector differentiation, seems to be required by Treg cells to suppress T_H2 responses. Conditional ablation of Irf4 in Treg causes excessive T_H2 dominated responses⁹⁴. Furthermore, Treg cells depend on expression of STAT3, a transcription factor that is also essential for the differentiation of T_H17 cells, to suppress T_H17 responses⁹⁵. Thus, it appears that environmental factors that drive effector T cell subset differentiation also trigger the expression of transcription factors that can modulate Treg suppressive function, indicating a surprising adaptability in this T cell subset.

Chromatin and the maintenance or plasticity of phenotypes

Epigenetic modifications have recently been put forth as major determinants of the stability or plasticity of CD4⁺ T cell phenotypes^{41, 96}. This notion had its roots in the initial demonstration that cell division is required for CD4⁺ T cells to acquire differentiated cytokine expression patterns⁹⁷, followed by the demonstration that other epigenetic modifications, rather than a cell cycle requirement, were involved in active transcription of cytokine genes⁹⁸. Active or repressed states of transcription are now extensively correlated with specific modifications of histones, or chromatin marks. Such modifications have been suggested as responsible for the heritability of the differentiated states⁹⁶, in the case of active chromatin marks, and for the plasticity of a subset⁸³, in the case of “bivalent” marks, which posses features of both active and repressed chromatin. Formally, the causality of chromatin marks in phenotype stability or plasticity remains untested, since no study has directly manipulated chromatin states independently of the conditions or circuitry required to establish them^{99, 100}.

A current controversy is whether chromatin modifications influence heritability of gene expression in dividing cells, as compared with their role in non-dividing cells¹⁰¹. Assays that examine CD4⁺ T cell plasticity involve multiple rounds of cell division. During DNA replication, histones are thought to be removed from the nucleosome, and new histones inserted into nucleosomes on replicated DNA, presumably altering the locations of active or repressive histone marks^{99, 101}. DNA replication occurs from many sites simultaneously throughout the genome. Thus, for histone modifications to provide epigenetic control of phenotype would require a mechanism to restore the original pattern and prevent randomization of histone marks after cell division. However, whether histone modifications are in fact replicated during cell division is unknown^{99, 102}. A recent study showed that a fraction of H3-H4 histone tetramers could be divided and then associate with new H3-H4 histones after cell division¹⁰². Such a mechanism potentially could allow for a process to copy the histone marks from the older generation of H3-H4 tetramers onto the next generation of chimeric H3-H4 tetramers. However, no functional evidence for such a copy mechanism nor a mechanism to maintain nucleosomal positioning was provided.

A second concern discussed above is whether chromatin marks described for various subsets are representative of a uniform or heterogeneous population⁸³. This concern also applies to the claim that bivalent chromatin marks can explain the plasticity of a phenotype when measured in cell populations. Much of the recent support for epigenetic control of stability and plasticity arise from genome-wide application of the Chromatin immunoprecipitation (ChIP) technique that is used to interrogate population of cells for association of biochemical marks with regions of DNA. Thus, bivalent marks identified by ChIP may result from mixes of cells with active and repressive marks. Even beyond this, bivalent marks may represent one active locus and one inactive locus within a single cell. Indeed, such mono-allelic expression of cytokines clearly occurs during T cell differentiation¹⁰³ particularly during early or incomplete polarization¹⁰⁴. Thus, bivalent chromatin marks might indicate flexibility of individual gene loci within one cell, but might also result from heterogeneity or even from monoallelic expression of the locus within a single cell.

Besides histone modifications, DNA modifications may provide a means to control the stability of differentiated phenotypes⁹⁶. Cytosine methylation at CpG is regulated by DNA (cytosine-5-)-methyltransferase 1 (DNMT1); patterns of methylation can persist throughout multiple rounds of cell divisions by the action of this maintenance methylase on hemimethylated CpG sites⁹⁶. As such, the methylation status of gene regulatory elements could provide epigenetic control that is stable in the setting of cellular division. However, current technologies have not allowed genome-wide description of DNA methylation patterns, making large scale correlations unavailable.

Where the marks are

Another issue complicates the simple interpretation that ChIP studies identify the chromatin marks associated with regions of DNA. Typically, data from ChIP experiments are interpreted as indicating direct interactions of an identified chromatin regions with the modified histone or sequence-specific DNA binding factors, a presumption based on a linear model chromosomal DNA. However, the initial experimental steps in ChIP protocols¹⁰⁵ are the same as for the Chromatin Conformation Capture (3C) technique^{106, 107}. In both techniques, the nucleus is chemically fixed to induce cross-linking^{105, 108, 109}, followed by sonic-degradation of the DNA into small fragments that remain attached to the proteins that are about to be immunoprecipitated. In ChIP and 3C, both protein-protein and protein-DNA interactions are captured by chemical fixation, but 3C further ‘captures’ the proximity of elements caused by DNA looping by introducing a DNA ligation step later on. In discussing results based on ChIP, typically only the protein-DNA cross-linking is emphasized. But importantly, even without this DNA ligation step, indirectly associated DNA has already been immunoprecipitated by ChIP, and so will be identified as a ‘binding site’ by ChIP as it is in 3C. In short, ChIP identifies both direct and indirect DNA regions associated with a histone mark or transcription factor. For example, our interpretation that BATF binds to both the IL-17 promoter and an intergenic enhancer might simply be due to an interaction with one of these regions, if the promoter and enhancer are brought into close proximity by DNA looping in vivo¹¹⁰. This caveat applies also when two transcription factors, such as STAT3 and IRF4, are examined by ChIP, with data suggesting that factors must bind at non-canonical sites¹¹¹. It is possible in such instances that looping is again at work, with each factor binding to distant canonical sites that are brought into proximity by looping mediated by protein-protein interactions. Such complexities have largely been overlooked as global surveys of factor binding sites have supplanted older techniques such as EMSA, DNA foot printing and site-directed mutagenesis⁸³. Clearly, it is important that speculative interpretations derived from ChIP studies be validated with such additional techniques.

Concluding remarks

Flexibility or stability in T helper subsets can be represented as a series of transitions from less stable to more stable states. Transcriptional circuitry that reinforces or destabilizes expression of specific factors may contribute to the relative stability and to the plasticity of CD4⁺ T cell subsets. Reinforcement can be direct and cell autonomous, or act in a paracrine manner through the actions of cytokines. Transcription factors that reinforce their own expression, or interfere with the expression or activity of alternative factors, produce stable states that behave as ‘attractors’.

To what extent recent findings of plasticity truly reflect individual CD4⁺ T cell flexibility or simple population-based heterogeneity is still unresolved. Heterogeneity can occur at the level of individual cells within a population, and at the level of individual gene alleles within a single cell. Furthermore, plasticity and stability are evaluated in the context of cells undergoing many rounds of cell division. The actions of the transcription factors, which persist through cell division, should not be ignored as a mechanism to maintain cellular phenotype, since mechanisms that would re-establish the pattern of specific histone modifications to their original nucleosomal location after cell division have not been identified. In contrast, patterns of cytosine methylation can be replicated through the maintenance methylase DNMT1. For these reasons, the emerging paradigm that histone modifications are a major determinant of heritability of differentiated CD4 subsets remains an intriguing, but untested hypothesis. As additional CD4⁺ ‘subsets’ emerge, it will be important to elucidate their transcriptional regulation to understand whether they represent novel ‘lineages’ or alternative ‘pathways’ of cellular activation.