AIRE in context: leveraging chromatin plasticity to trigger ectopic gene expression

Summary

The emergence of antigen receptor diversity in clonotypic lymphocytes drove the evolution of a novel gene, Aire, that enabled the adaptive immune system to discriminate foreign invaders from self-constituents. AIRE functions in the epithelial cells of the thymus to express genes highly restricted to alternative cell lineages. This somatic plasticity facilitates the selection of a balanced repertoire of T cells that protects the host from harmful self-reactive clones, yet maintains a wide-range of affinities for virtually any foreign antigen. Here, we review the latest understanding of AIRE’s molecular actions with a focus on its interplay with chromatin. We argue that AIRE is a multi-valent chromatin effector that acts late in the transcription cycle to modulate the activity of previously poised non-coding regulatory elements of tissue-specific genes. We postulate a role for chromatin instability – caused in part by ATP-dependent chromatin remodeling – that variably sets the scope of the accessible landscape on which AIRE can act. We highlight AIRE’s intrinsic repressive function and its relevance in providing feedback control. We synthesize these recent advances into a putative model for the mechanistic modes by which AIRE triggers ectopic transcription for immune repertoire selection.

Role of self in the evolution of adaptive immunity

Self-reactivity is central to shaping the adaptive immune system. The immunological ‘big bang’¹ (as coined by Janeway), a retrotransposon invasion of an immunoglobulin-like gene, gave rise to RAG-mediated receptor diversity that generates specificities to self and foreign antigens.² Solving the problem of self/non-self discrimination then became a major catalyst in the evolution of the adaptive immune system.^3,4 What emerged was a clonal selection framework^5,6 based on the lymphocyte receptor’s affinity to self. The appearance of a new organ: the thymus,⁷ and a novel gene: Aire,^4,8 would facilitate the testing of lymphocyte clones for the capacity to recognize antigens in the ‘context’ of self (i.e. major histocompatibility complex), and against the capacity to harm self-constituents.⁹

Self-antigen diversity directs selection of a balanced T cell repertoire

Ancient Greeks, during sacrificial rites, found a pulp of white tissue above the heart and decided it must be the ‘soul’ of the young animal.¹⁰ The thymus, derived from the Greek word θυμος,¹⁰ is the site of T cell development and repertoire selection. The principal modes by which dangerous self-reactive clones are neutralized in the thymus are the triggering of cell death^11,12 or the functional deviation to the regulatory T cell (Treg) lineage^13,14 upon recognition of agonist T cell receptor (TCR) ligands. These functional outcomes in the thymus contrast those in the periphery where cognate ligands trigger T cell growth, proliferation and the secretion of cytokines. Thus, the self-ligands (peptide-MHC) presented in the thymus directly shape T cell repertoire selection. Insufficient scope of self-presentation would result in the escape of harmful self-reactive clones that could inflict autoimmunity. On the other hand, total purging of self-reactivity would effect “holes in the repertoire” (as suggested by Nossal) leaving the host vulnerable to infectious pathogens.¹⁵ Therefore, a delicate balance of self-presentation must be achieved in the thymus to select an immune repertoire with a wide range of affinities for any putative foreign invader, yet purged of the most dangerous autoreactive clones. To support this balance, the medullary epithelia of the thymus exert an extraordinary function: the ectopic expression of genes normally restricted to alternative lineages of specialized tissues (e.g. insulin).^4,16 This function dramatically increases the diversity of intrathymic self-antigens to represent essentially all parenchymal organs. The question of how exactly cells of a defined epithelial lineage express and display on their surface tissue-restricted antigens of wildly different cell types has riddled immunologists for decades.

Chromatin landscape and cell fate specification

Cell identity is generally thought to be a product of a forward-moving hierarchical process in development, defined by dynamic networks of lineage-specific transcription factors (TFs).^17–19 Chromatin, the template through which this process is exerted, is a macromolecular fiber of billions of DNA bases wrapped in millions of nucleosomes. Triggered by developmental and environmental cues, lineage-specific TFs orchestrate patterns of chromatin conformation and gene expression that define distinct cellular states. Chromatin effectors (e.g. histone chaperones,²⁰ DNA-²¹ or histone-modifying enzymes,²² and ATP-dependent chromatin remodelers²³) reinforce and stabilize these states by regulating intra- and inter-nucleosomal dynamics.^24,25 As cells differentiate during development, the chromatin barriers that define cell identity are progressively raised to gradually restrict multi-lineage potential.^19,26 This chromatin restriction provides stability to lineage commitment and serves as a barrier to cell fate changes^19,26 (Figure 1). Mutations that disrupt existing TF networks and/or nucleosome dynamics can lower these epigenetic barriers and alter developmental trajectories, classically exemplified by homeotic transformations in Drosophila.²⁷ Such mutations can also create unnatural stable states that allow sampling of gene regulatory programs usually restricted to other lineages or early developmental windows as observed in cancer.²⁶ Moreover, somatic cell fates can be deliberately reprogrammed via exogenous provision of lineage-defining TFs for both de-differentiation²⁸ (e.g. B lymphocyte to embryonic stem cell) and trans-differentiation²⁹ (e.g. hepatocyte to neuron). These perturbations indicate that the somatic chromatin landscape remains a malleable template that can be modified via intrinsic and extrinsic factors. Such somatic plasticity is likely crucial for organismal adaptation, as was the case for the thymus for jawed vertebrates 500 million years ago.⁷

Figure 1: Reshaping of the chromatin landscape allows mTECs to sample alternative lineage genes.

The chromatin landscape of a single cell is represented as a quasi-potential surface (i.e. chromatin barriers) in transcriptional phase space (horizontal plane). Red balls represent transcriptional machinery. Basins in the potential surfaces represent attractors to corresponding transcriptional programs. Transcriptional activity of a particular attractor is represented by the number of balls localized at the attractor. A: In typical epithelial cells, an ordered chromatin landscape shaped by epithelial-specific transcription factors (e.g. FOXN1) gives rise to a well-defined transcriptional program represented by the precise distribution of balls at the epithelial state. B: Induced pluripotent stem cell (iPSC) reprogramming via provision of OCT4, SOX2, KLF4, and c-MYC (OSKM) abolishes the epithelial state and creates a new chromatin landscape, giving rise to another ordered transcriptional program that mediates pluripotency. C, D: In mature medullary thymic epithelial cells (mTECs), the ordered epithelial chromatin state is disrupted stochastically (two examples shown). ATP-dependent chromatin remodelers (e.g. BAF) generate novel attractors and potentiate access to genomic loci normally highly restricted to alternative lineages (e.g. Insulin) without abolishing these cells’ epithelial identities. AIRE exploits this permissive landscape to actuate subsets of these divergent attractors, resulting in ectopic gene expression. This action is represented through an additional bucket of balls, illustrating AIRE’s ability to increase transcriptional output and diversity.

Plasticity of thymic epithelia

Medullary thymic epithelial cells (mTECs) fundamentally alter their chromatin landscapes to mediate the ectopic expression of nearly all coding genes in the genome. This process takes place at the terminal maturity of mTEC development, far downstream of the bipotent progenitor that gives rise to the cortical and medullary epithelia of the thymus.^30,31 Thus, mTECs are a remarkable example of somatic plasticity, driven by an endogenous differentiation program rather than exogenous perturbation. The specific features of ectopic gene expression in mTECs provide clues to the nature of this plasticity.

1. Single-cell heterogeneity: Each mTEC expresses only a subset of the total number of tissue-specific genes expressed by the entire mTEC compartment at any given time.^32–35 This heterogeneity could be the result of an evolutionary balance between generating self-antigen diversity and displaying the requisite density of a given ligand to induce the negative selection of its cognate autoreactive T cell.³⁶

2. Stereotypic stochasticity: The fraction of mTECs expressing a given tissue-specific gene is stereotypical, with different tissue-specific genes exhibiting distinct expression frequencies.^{32,33,35,37,38} Furthermore, tissue-specific genes appear to be coordinately co-expressed as discrete gene modules in single mTECs,^32,33,35 reflecting the actions of an ordered process. However, the genes within these modules are not related by tissue-specific function, lineage-specific co-regulation, developmental origin, or genomic location.^32,33,35 These observations, together with the significant cell-to-cell heterogeneity, suggest an underlying ordered yet stochastic mechanism. Perhaps this mechanism is akin to rolling weighted dice: the outcome of any particular roll is unpredictable, but the manner in which each die is weighted determines the probability distribution of the outcome.

3. Dynamics of ectopic gene expression: The breadth of tissue-specific genes expressed by each mTEC throughout its lifespan is greater than that observed at any given moment.^36,39 This was poignantly visualized by the heightened frequency of mTECs that had ever expressed a tissue-specific gene vs. those that are expressing the same tissue-specific gene at a given time.³⁹ mTECs likely shift between specific cohorts of co-expressed gene modules, as was described for clones of mTECs cultured ex vivo.³⁶ These dynamics increase the range of tissue-specific genes that a single mTEC can express and thereby increase the probability of encounter with cognate self-reactive T cells. Reciprocally, these dynamics decrease the number of mTECs required to encompass the total diversity of self-antigens presented in the thymic medulla.³⁶

4. Novel mode of gene activation: mTECs do not employ the same TFs necessary in peripheral counterparts. For example, PDX1 and MAFA that are indispensable for insulin expression in the pancreas are not necessary for insulin expression in the thymus.^38,40 Furthermore, tissue-specific gene activation in mTECs can often be monoallelic, a contrast to the largely biallelic transcription observed in peripheral tissues.³⁸ Likewise, tissue-specific transcripts in mTECs can often be found initiated at alternative start-sites with different splicing patterns than that in peripheral counterparts.³⁸ Thus, it appears mTECs do not adopt existing gene regulatory networks used in peripheral ‘home’ tissues to promote ectopic expression of tissue-specific genes. Rather, it seems mTECs have innovated novel modes of activating these tissue-specific loci.

Taken together, these features are not consistent with the view that mTECs create a mosaic of chromatin landscapes that mimic those of restricted fates in peripheral tissues. Instead, they indicate that mTECs establish novel permissive landscapes that variably activate clusters of genes unrelated by location, function, or tissue-of-origin. This would suggest that mTECs lower specific chromatin barriers to effect new states that allow the simultaneous sampling of transcriptional competencies from multiple alternative lineages (Figure 1). The genomic locations of the compromised chromatin barriers would differ between individual mTECs, causing the range of tissue-specific genes that each cell can express to vary correspondingly (Figure 1). The mode of this sampling does not resemble the acquisition of alternative TF network properties, as seen in cellular reprogramming. Rather, mTECs may exhibit a chromatin instability that is analogous to the genetic instability of cancer. As DNA repair defects differentially impact the mutation profiles of individual cancer clones, the mTEC differentiation program may stochastically disrupt chromatin barriers to impart ‘access’ to different cohorts of tissue-specific genes across individual mTECs. Thus, both forms of instability give rise to heterogeneous aberrant states. The precise mechanisms that drive this transcriptional plasticity in mTECs is largely unknown. However, the exploration of a rare monogenic autoimmune disease, autoimmune polyglandular syndrome type 1 (APS-1), revealed a central molecular determinant: AIRE.^41–43

AIRE, a lineage-specific transcription factor?

AIRE is a 54.5 kDa protein that controls thousands of genes encoding tissue-restricted antigens in mTECs.^32,44 Mutations in the AIRE gene cause devastating multi-organ autoimmunity featuring up to 20 different clinical manifestations including hypoparathyroidism, adrenal insufficiency, type-1 diabetes, gastritis, neuropathy, and ovarian failure.⁴⁵ Radiation bone marrow chimeras and thymic transplant studies using Aire-deficient mice attributed AIRE’s tolerance-inducing role to its function in mTECs.^43,46 Further, alterations in the cellular organization and composition of the thymic medulla in Aire-deficient mice indicated AIRE’s importance in mTEC differentiation.^47–49 The promoters of AIRE-induced tissue-specific genes maintained AIRE-dependent regulation even when randomly integrated into other genetic loci.^43,50–52 At first glance, these results suggest that AIRE recognizes tissue-specific cis-regulatory elements independent of their native chromatin context. These features – in addition to AIRE’s domain structure and punctate nuclear localization^46,51 – are reminiscent of the properties of lineage-specific transcription factors.

The hallmark function of transcription factors is the recruitment of enzymatic activity to distinct regions of the genome via sequence-specific DNA-binding activity. How AIRE targets the staggering number of disparate tissue-specific genes it regulates has been a long-standing conundrum. AIRE contains a DNA-binding domain termed SAND domain (SP100, AIRE, NucP41/P75, DEAF-1).⁵³ This domain’s functional importance is highlighted by the multitude of disease-causing missense mutations that result in impaired AIRE nuclear targeting.⁴⁵ However, AIRE’s SAND domain does not contain the critical α-helical KDWK binding module that has been reported to be necessary for its ability to recognize target sites.⁵³ Furthermore, AIRE’s SAND domain is not well-conserved across vertebrate species,⁸ and its interaction with DNA in vitro appears largely sequence-independent.⁵⁴ These observations bring to question whether AIRE should be classified as a TF. Indeed, it is unlikely that a sequence-specific TF could key the immense breadth of AIRE’s transcriptional footprint.

AIRE acts at promoter-distal regulatory elements

According to recent localization studies, AIRE directly binds to tens of thousands of genomic loci in mTECs, the large majority of which are distal to gene promoters and transcriptional start sites (TSSs).⁵⁵ This binding distribution suggests that AIRE exploits the immense non-coding regulatory content of the metazoan genome instead of focusing on the promoters of tissue-specific genes. Regulatory elements, e.g. enhancers, silencers, and insulators, provide precise spatiotemporal control of gene expression at distances that can span hundreds of kilobases.⁵⁶ This control stems from the combinatorial targeting of sequence-specific TFs and the activities they recruit.⁵⁶ Since AIRE is unlikely to act as a sequence-specific TF, perhaps AIRE directly cooperates with TFs via physical interface to target the specific loci it regulates. If this were the case, one would expect biases in sequence content of AIRE-bound genomic sites for target motifs of particular transcription factors. Indeed, bioinformatic analysis of AIRE-binding sites has revealed the enrichment of NF-κB motifs as the predominant feature of AIRE-targeted loci⁵⁷. Enrichment of other TFs, e.g. AP1, Fra1, BATF, were also observed, albeit at a lower significance than that for NF-κB.⁵⁷ The mechanistic implications of AIRE’s functional partnership with these TFs and their potential contribution to mTEC plasticity is discussed in multiple contexts below.

AIRE acts on a previously poised chromatin landscape

Large-scale changes in chromatin accessibility can be detected during the differentiation of immature mTEC progenitors to mature, AIRE-expressing mTECs,⁵⁷ observed at tens of thousands of gene-distal regulatory elements neighboring tissue-specific genes.⁵⁷ Regulatory elements have high intrinsic affinity to histone octamers, creating a strong nucleosomal barrier.⁵⁸ Thus, chromatin accessibility at these sites is a function of TF competition with nucleosomes for cognate motifs.⁵⁶ Could AIRE mediate this accessibility via ‘pioneering’ activity (i.e. the direct binding of intranucleosomal target DNA⁵⁹) or via direct or indirect cooperativity⁶⁰ with other TFs, e.g. NF-κB? Alternatively, AIRE may have no role in establishing the initial chromatin accessibility. Instead, it may operate to recruit enzymatic activity (e.g. chromatin remodeling complexes or histone modifying enzymes) to augment pre-existing accessibility (e.g. via nucleosome eviction, histone variant exchange, or histone acetylation). Given the tissue-restricted nature of the genes AIRE regulates, the presumptive template for AIRE was thought to be inaccessible, sequestered heterochromatin. However, chromatin accessibility profiling of progenitor and mature mTECs in Aire-deficient mice indicated that the induction of chromatin accessibility at promoter-distal sites neighboring tissue-specific genes is AIRE-independent.⁵⁷ This result is consistent with the idea that AIRE amplifies rare but already ongoing transcription, rather than recruiting RNA polymerase II (Pol II) to initiate transcription.^{32,43,61–63} This result also provides more clarity to the stark differences in AIRE-regulated genes when AIRE is expressed in other cell types.^64–67 AIRE’s dependence on the pre-existing chromatin context is highlighted by the lack of expression of tissue-specific genes in cortical TECs (cTECs) genetically engineered to express AIRE.⁶⁶ When crossed to the Aire−/− background, the AIRE⁺ cTECs in these mice failed to rescue the negative selection and immune tolerance defects caused by Aire-deficiency in mTECs.⁶⁶ Thus, AIRE acts on a landscape that is poised prior to AIRE expression. This mode of gene regulation, where distinct factors are employed for chromatin accessibility/transcriptional initiation vs. transcriptional elongation, is a common theme in other systems including cellular stress response, immune response, hormone signaling, and pluripotency.⁶⁸ Perhaps AIRE’s dependence on a poised chromatin landscape provides distinct functionality for each cellular context in which it is expressed, e.g. in dendritic cells, B cells, and innate lymphoid cells (ILCs).^65–67

Determinants of mTEC chromatin accessibility landscapes

Since AIRE localization is contingent on the endogenous chromatin context, the induction of chromatin accessibility at tissue-specific loci appears to be a discrete ‘licensing’ step in mTEC maturation. The determinants that subtend this ‘licensing’ could key the specific range of genes that is activated by AIRE in each mTEC. These determinants could be TFs that directly target loci usually restricted to alternative lineages. Such actions could define the discrete clusters of co-expressed genes induced by AIRE.^32,33,35 These determinants could also be chromatin effectors that disrupt the stability of chromatin barriers that buttress epithelial identity. Such disruptions could be the basis for the epigenetic and transcriptional heterogeneity of mTECs.^{32,33,35,69,70} Therefore, identifying the determinants that impart the specific patterns of chromatin accessibility, and those that promote variance of these patterns is essential to understanding the nature of mTEC plasticity.

Quantitative analysis of TF features (e.g. TF motifs, ChIP-seq profiles, TF footprinting) within the differentially accessible sites between immature and mature mTECs revealed putative TFs that directly poise the mTEC chromatin landscape. The feature that was most enriched in mature mTEC vs. progenitor accessibility profiles was NF-κB motifs.⁵⁷ This enrichment was reinforced by differential TF footprinting at the same regions, which infers greater NF-κB binding in mature vs. progenitor mTECs.⁵⁷ Canonical and noncanonical NF-κB signaling are essential for mTEC lineage commitment and maturation to the AIRE-expressing state.^71–75 NF-κB signaling in mTECs is triggered by intercellular ‘cross-talk’ between tumor necrosis factor (TNF) receptors on the mTECs (i.e. CD40, receptor activator of NF-κB (RANK), lymphotoxin receptor-β (LTβR)), and their ligands produced by various developing lymphocytes (i.e. positively selected αβ T cells, γδ T cells, RORγt⁺ ILCs, and invariant natural killer T cells (iNKTs)).^71–75 NF-κB signaling precedes AIRE expression during mTEC differentiation. Further, the increase in accessibility at gene-distal sites co-localized by AIRE and NF-κB is AIRE-independent.⁵⁷ AIRE could be acting on NF-κB-poised regions to facilitate these developmental transitions, which would be consistent with AIRE’s role in mTEC terminal differentiation and medullary organization.^47–49 However, the abundance of NF-κB motifs at differentially accessible sites neighboring tissue-specific genes also implicates a direct role for NF-κB in poising tissue-specific loci for ectopic activation by AIRE.⁵⁷ Indeed, NF-κB signaling is critical to a wide range of disparate processes in many peripheral tissues. Beyond its well-characterized roles in cellular stress, immunity and inflammation, NF-κB signaling is also essential for the development of liver, lung, muscle, skin, limbs, as well as for hematopoiesis, insulin secretion, neurulation, hippocampal neurogenesis and neural function.^76,77

Comparative feature analysis of chromatin accessibility profiles between immature and mature mTECs also yielded significant enrichment for other TF motifs and footprints, including upstream stimulatory factor 1 (USF1), AP-1, members of the hepatocyte nuclear factor (HNF) family and ETS TF family.⁵⁷ Elucidating the individual and combinatorial roles of these and other factors in poising the mTEC chromatin landscape will bring further mechanistic insight into how mTECs promote ectopic gene expression.

ATP-dependent chromatin remodeling in mTEC plasticity

As cells differentiate along developmental trajectories, the chromatin landscape becomes progressively more restrictive to limit cell fate potential and facilitate lineage commitment.^19,26 This chromatin restriction is enacted in large measure by ATP-dependent chromatin remodeling complexes.^23,26,78 These multi-subunit macromolecular motors hydrolyze ATP to alter nucleosome composition, positioning, and occupancy.^23,78 Their role in maintaining cell identity is highlighted by the dramatic increase in reprogramming efficiency (fibroblast to induced pluripotent stem cells) after ablation of MBD3, a subunit of the nucleosome remodeling and deacetylation (NuRD) complex.⁷⁹ In another example, the Brahma-associated factor (BAF) ATP-dependent chromatin remodeling complex has an instructive role in promoting trans-differentiation of fibroblasts into neurons.⁸⁰ Thus, ATP-dependent chromatin remodeling is implicated in promoting changes to the chromatin landscape that set the permissive context for mTECs to activate tissue-specific genes. Indeed, two recent studies demonstrated that both BAF and NuRD complexes are important players in shaping the permissive chromatin accessibility landscape in mTECs.^57,81

The conditional deletion of the genes encoding the respective ATPase motors of BAF (Brg1cKO) and NuRD (Chd4cKO) in mTECs compromised the chromatin accessibility at different regions of the genome. The BAF perturbation severely impaired the induction of chromatin accessibility at promoter-distal sites near tissue-specific genes during mTEC maturation.⁵⁷ These are the same regions targeted by AIRE and replete with NF-κB motifs,⁵⁷ suggesting that BAF precedes AIRE to poise these sites in conjunction with NF-κB. In contrast, the Chd4cKO largely affected the accessibility at the promoter regions of FEZF2-regulated genes.¹³¹ FEZF2 is a transcription factor that promotes the ectopic expression of a distinct set of tissue-specific genes than those that are regulated by AIRE.⁸² The loss of chromatin accessibility at the respective target regions for BAF and NuRD perturbations critically diminished the ectopic expression of associated tissue-specific genes.^57,81 As a result, Brg1cKO and Chd4cKO mice exhibited systemic increases in frequencies of activated T cells and incurred lymphocytic infiltration at peripheral tissues, reflecting multi-organ autoimmunity.^57,81 Thus, the BAF and NuRD complexes are essential determinants of the mTEC chromatin landscape that poise tissue-specific loci for transcriptional activation by AIRE and FEZF2 respectively.

How BAF and NuRD disrupt the epithelial-specific chromatin barriers to mediate access to tissue-specific loci is an open question. BAF remodeling activities are typically associated with nucleosome/histone eviction, nucleosome sliding, and histone variant exchange.^23,78 BAF can also directly antagonize the activity and localization of Polycomb group (PcG) proteins,^83,84 as well as recruit topoisomerases to resolve DNA catenation.⁸⁵ NuRD remodeling activities typically encompass spacing nucleosomes at ordered intervals, exposing promoters, and histone variant exchange.^23,78 NuRD complexes also consist of histone deacetylases (HDACs) and methyl-CpG-binding domains (MBDs) that stabilize nucleosomes and recruit the heterochromatin machinery respectively.^23,78 Hence, NuRD is commonly thought to function as a repressor, whereas BAF is typically thought to be an activator (although exceptions are not uncommon). BAF and NuRD remodeling complexes have highly specialized roles in a diverse range of cell types/tissues throughout mammalian development.²³ The breadth of lineage-specific functions that BAF and NuRD promote is facilitated by the differential incorporation of distinct subunits in the respective cell types in which they operate.²³ For example, cardiac fate and heart morphogenesis require switching of the BAF60A subunit for BAF60C.⁸⁶ Likewise, early post-implantation development requires NuRD complexes containing MBD3, which is mutually exclusive to those with MBD2.⁸⁷ These combinatorial assemblies of diverse complexes are thought to create binding interfaces for interaction with lineage-specific TFs and chromatin effectors, as well as to modify the complex’s enzymatic functions.^23,78 Identifying the subunit composition of BAF and NuRD complexes in mTECs and the TFs they interact with will be important investigative avenues, especially at the single cell level.

AIRE triggers RNA elongation

Transcription is pervasive across the non-coding genome,⁸⁸ where AIRE is largely localized.⁵⁵ Gene-distal regulatory elements (e.g. enhancers) are replete with core-promoter-like motifs that recruit general TFs (e.g. TFIID, TFIIB, TFIIH) and Pol II.^89,90 Shortly after transcriptional initiation and synthesis of short nascent enhancer RNA (eRNA), Pol II pauses, in part, due to the nonproductive conformation of the RNA-DNA hybrid in the Pol II catalytic site.^91,92 Negative elongation factors (e.g. SPT5, NELF) recognize and stabilize this paused state by preventing the reactivation of the Pol II catalytic site.^91,92 Pol II-pausing is widely accepted as a rate-limiting step in the transcription cycle at all active TSSs within enhancers^89,90 and promoters^93,94 alike. Accordingly, the efficiency of Pol II pause-release at enhancers is a critical determinant of transcriptional amplitude. Levels of eRNA at a given locus are highly correlated with enhancer activity at target promoters.⁹⁰

AIRE recruits P-TEFb via BRD4

AIRE likely imposes its function late in the transcription cycle at the elongation step, considering it targets accessible sites loaded with Pol II.^55,57 Indeed, AIRE interacts with the positive transcription elongation factor b (P-TEFb) through its well-conserved C-terminal domain.^8,95 The importance of this interaction is underscored by an APS-1 mutation causing a 25 amino acid truncation of AIRE’s carboxyl terminus. This mutation abolishes the AIRE:P-TEFb interface and results in multi-organ autoimmunity.⁹⁶ P-TEFb is a kinase that phosphorylates dozens of proteins, including negative elongation factors, Pol II, as well as factors mediating histone modifications and RNA processing.⁹⁷ P-TEFb’s pause-release function, however, is most directly attributed to its phosphorylation of SPT5.^98,99 This triggers the dissociation of NELF from Pol II, allowing the reactivation of its catalytic core and resumption of transcriptional elongation.^98,99 Thus, AIRE could be recruiting P-TEFb to tissue-specific loci to increase the efficiency of Pol II pause-release and augment productive eRNA synthesis. Indeed, forced recruitment of P-TEFb is sufficient to increase gene expression levels, suggesting that the levels of P-TEFb at a given locus can dictate its transcriptional amplitude.¹⁰⁰

P-TEFb is typically sequestered from chromatin via its interactions with an inhibitory complex that includes 7SK non-coding RNA and HEXIM proteins.¹⁰¹ This inactive complex can be recruited to specific sites in the genome, however, P-TEFb must be released from the inhibitory complex prior to activating its pause-release function.¹⁰² The bromodomain and extraterminal (BET) protein BRD4 is a potent factor that can dissociate P-TEFb from the 7SK complex by directly binding to P-TEFb through its C-terminal domain.¹⁰³ Moreover, BRD4:P-TEFb interactions cause allosteric changes that enhance P-TEFb’s kinase activity.^103,104 AIRE facilitates these interactions through the recruitment of BRD4, which recognizes acetylated lysines in AIRE’s caspase activation recruitment domain (CARD).¹⁰⁵ Mechanistically, BRD4 recruitment depends on the orchestrated actions of DNA-dependent protein kinase (DNA-PK) and the transcriptional co-activators p300 and CREB-binding protein (CBP) which respectively phosphorylate and acetylate AIRE’s CARD.^105,106 APS-1 missense mutations that impair these CARD post-translational modifications (but not AIRE’s punctate nuclear localization), highlight the functional importance of recruiting these elongation factors.¹⁰⁵ Thus, AIRE augments transcription at previously poised enhancers through the recruitment of P-TEFb and BRD4, mediating Pol II pause-release and increases the rate of eRNA synthesis.

AIRE recruits topoisomerases and elicits DNA-damage response

AIRE’s physical interaction with topoisomerases also contributes to its impact on transcriptional elongation. Nucleosome eviction and phasing by ATP-dependent chromatin remodelers generate negative supercoils at regulatory elements and promoters.¹⁰⁷ Moreover, Pol II progression along the DNA template during RNA synthesis causes torsional strain, generating positive supercoils ahead of Pol II and negative supercoils behind it (~2 supercoils for every 10 bp transcribed).¹⁰⁸ The accumulation of positive torsional strain and the formation of DNA-RNA hybrid R-loops by negative supercoiling directly impede transcriptional elongation at enhancers and gene bodies.^109,110 One of the major interacting partners of AIRE is DNA topoisomerase I (TOP1), which promotes transient single-strand DNA breaks to relieve the torsional stress caused by chromatin remodeling and Pol II.⁵⁵ In vivo treatment of mice with the TOP1 inhibitor topotecan specifically compromised the ectopic induction of a subset of AIRE-regulated genes and caused immune tolerance defects, highlighting the significance of TOP1 for AIRE’s function in mTECs.⁵⁵ The molecular mechanism underpinning TOP1’s influence has been proposed to be seeded in TOP1’s DNA nicking activity and the subsequent triggering of the DNA-damage response.⁵⁵ AIRE-recruited, TOP1-mediated resolution of torsional impediments to Pol II’s elongation activity is consistent with recent reports that TF-recruited TOP1 is crucial for Pol II pause-release and eRNA synthesis.^111,112 In addition, the strong overlap of AIRE and γ-H2AX localization⁵⁵ and AIRE’s reported interactions with DNA-damage repair proteins (e.g. poly(ADP-ribose) polymerase 1 (PARP-1), DNA-PK_cs, Ku70, Ku80)^55,113 are consistent with the connection between TOP1-triggered DNA-damage response and enhancer transcription.^110,114 Interestingly, shRNA-mediated knockdown of TOP1 in AIRE-expressing HEK293T cells resulted in a dramatic loss of AIRE’s interaction with BRD4 and P-TEFb,⁵⁵ supporting DNA-PK’s critical role in priming AIRE’s interface with transcriptional elongation factors.^103,104

AIRE’s interaction with TOP1 also results in the subsequent recruitment of type IIA topoisomerases, namely TOP2α and TOP2β.^55,113 TOP2 induces transient double-strand DNA breaks to resolve dynamic supercoiling predominantly at gene promoters.¹¹⁵ As a result, TOP2 activity recruits components of the non-homologous end joining (NHEJ) and homologous repair (HR) double-strand break machineries.¹¹⁵ There is emerging evidence that Pol II pause-release at promoters require TOP2’s resolution of torsional stress and its recruitment of DNA-damage repair proteins.^112,116 Indeed, in vivo treatment with the TOP2 inhibitor etoposide prevented the expression of a subset of AIRE-regulated genes in mTECs, increased the systemic frequency of self-reactive T cells, and caused lymphocytic infiltration in the eye and lung in the autoimmune-prone NOD background.⁵⁵ Thus, AIRE’s interactions with topoisomerases initiate a cascade of activities involving multi-protein complexes from transcriptional and DNA repair machineries that promote the release of paused Pol II.

AIRE’s role in enhancer-promoter contacts

AIRE-induced enhancer transcription could contribute to the ectopic expression of tissue-specific genes by several modes (Figure 2). Increased enhancer interactions with target promoters could be one of the major consequences of AIRE’s actions at gene-distal regulatory elements. Recent studies using live-cell imaging demonstrated that enhancer-promoter contacts are not as stable or long-lived as previously thought.¹¹⁷ Instead, they are transient, mediating ‘bursts’ of Pol II activity at target promoters.¹¹⁷ Hence, gene expression levels are more a product of the frequency, rather than the duration of enhancer-promoter contacts.¹¹⁷ Transcriptionally active enhancers exhibit greater rates of mobility, broader range of spatial territory sampled, and more frequent promoter contacts compared to inactive loci.^118,119 These properties are likely mediated by several factors including: (i) engagement of transcriptional co-activators (e.g. p300, CBP, MLL3, MLL4) with the actively transcribing Pol II;^120–123 (ii) enhanced catalytic activity of transcriptional co-activators by RNA-binding;¹²⁴ (iii) inhibition of repressive epigenetic regulators (e.g. Polycomb repressive complexes, DNA methyltransferases) via RNA-binding;^125–127 (iv) additional recruitment of TFs via low affinity recognition of RNAs;^128,129 (v) recruitment of topoisomerases and DNA-damage repair machineries.^{110,112,114,116} Altogether, AIRE-triggered RNA synthesis at enhancers could serve to generate a surfeit of favorable activity that increases transcriptional burst frequency at target promoters. This possibility is consistent with the recent observation that AIRE localizes to super-enhancers, i.e. clusters of lineage-defining enhancers harboring high local concentrations of TFs and co-activators (e.g. BRD4 and Mediator).¹³⁰

Figure 2: AIRE’s duality provides amplitude control of enhancer activity.

Molecular features of AIRE’s dual function at enhancers. Left: An enhancer is inaccessible, inactive, and unable to contact promoters at tissue-specific genes (TSGs) to release paused Pol II. Directed to the locus by transcription factors (TFs), the BAF chromatin remodeling complex mediates chromatin accessibility and transcriptional initiation. Right: AIRE (yellow oval) targets the accessible enhancer via interactions with TFs and unmodified H3 tails. AIRE recruits torsional strain-relieving DNA topoisomerases (e.g. TOP1), DNA-repair machinery (e.g. DNA-PK), elongation factors (e.g. P-TEFb, BRD4), transcriptional co-activators (e.g. CBP/p300), and other ancillary factors (blue ovals). The combined activity of these factors releases paused Pol II for RNA synthesis (eRNA). eRNAs are associated with enhancer mobility, promoter sampling, and promoter contacts, increasing transcriptional burst frequency at TSGs. AIRE simultaneously mediates nucleosome assembly and inter-nucleosomal interaction, antagonizing BAF-dependent accessibility, opposing enhancer activity, and preventing sustained ectopic gene expression.

AIRE is a lineage-specific chromatin effector

AIRE’s lack of DNA sequence-specificity, its target sites being accessible and poised prior to AIRE expression, and its connection to transcriptional elongation suggest that AIRE functions as a co-activator/effector protein rather than a lineage-specifying TF. Chromatin effectors facilitate the ability for TFs to activate transcription on chromatin templates but cannot target specific sites on their own.¹³¹ Chromatin effectors interact with a diverse range of TFs and tend to be associated with the most active regions of the genome.⁶⁰ Effector activities include those that compromise nucleosome stability¹³² (e.g. histone acetylation, histone variant exchange), nucleosome positioning¹³³ (e.g. nucleosome eviction, sliding, phasing via ATP-dependent chromatin remodeling), ‘crosstalk’ with promoters¹³⁴ (e.g. via Mediator complex), transcriptional initiation¹³⁵ (e.g. Pol II recruitment), and transcriptional elongation¹²⁹ (e.g. via PTEF-b). Since the steps preceding Pol II pause-release at tissue-specific loci are AIRE-independent (e.g. chromatin accessibility, pre-initiation complex formation, nascent transcription),^32,43,61,62 AIRE acts late in the transcription cycle as a chromatin effector to reinforce the actions of TFs and trigger transcriptional elongation.

AIRE is a reader of unmodified H3 tails

A common mode by which chromatin effectors engage target sites is through the recognition of specific covalent modifications on histone tails.¹³⁶ For example, bromodomains of BET proteins recognize histone acetylation,¹³⁷ and chromodomains of HP1 and Polycomb recognize methylated lysines on H3 (K9 and K27 respectively).¹³⁸ Likewise, AIRE’s effector function depends on the recognition of unmodified histone tails through its first plant homeodomain (PHD). PHD fingers are small structural folds (~50–80 amino acids) consisting of a two-strand anti-parallel β-sheet flanked by two zinc ions that coordinate the signature cross-brace topology of the Cys4-His-Cys3 motif.¹³⁹ The high prevalence of PHD fingers in chromatin effector complexes¹³⁹ was later explained by structural and biochemical studies showing PHD fingers recognize the methylation and acetylation states of histone tails.¹⁴⁰ AIRE contains two PHD fingers, however, only the N-terminal PHD finger (PHD1) has histone tail-binding activity.^54,141 AIRE’s PHD1 (but not PHD2) is highly conserved from zebrafish to humans, reinforcing the importance of this binding activity.⁸ AIRE recognizes the first 8 residues of the amino-terminus of H3 tails via a deep and extensive binding site in PHD1 that encompasses nearly a third of the structural fold.^142–144 The histone tail forms the third antiparallel β-strand, pairing with the existing two-strand β-sheet of AIRE’s PHD1.^142–144 The highly restrained binding pocket for the NH₃⁺-ARTK terminus on the histone tail allows AIRE to discriminate H3 from H2A, H2B, and H4 tails.^54,142 Strong electronegative properties of AIRE’s PHD1 (pI ~4.9) promote high-affinity interactions with H3 tail (~5–10 μM),^54,141 which is ~3–6-fold stronger than that of BHC80’s PHD finger to the same substrate.¹⁴⁵ Post-translational modifications (i.e. methylation, phosphorylation, acetylation) to H3R2, H3T3, or H3K4 block these electrostatic interactions and inhibit AIRE recognition.^54,141–144 For example, di- or tri-methylation at H3R2 or H3K4 abolishes AIRE binding, whereas mono-methylation severely impairs binding.^54,141–144 In contrast, modifications on H3R8 and beyond (e.g. H3K9me3, H3K27me3) have no impact on AIRE-recognition of H3.^54,141–144 These binding properties have important functional implications for how AIRE engages and targets the chromatin template.

AIRE’s histone-binding activity is essential for its function

The multi-organ autoimmunity caused by APS-1 missense mutations in AIRE’s PHD1 (e.g. C302Y, C311Y) underscores the functional importance of this domain.^45,146 Since these mutations disrupt zinc-coordination and unfold the PHD structure, the impact of an alanine substitution for the aspartic acid critical for recognition of H3K4 (i.e. D299A, D297 in human) was tested in vivo.¹⁴⁷ The structural integrity of AIRE’s PHD1 and the punctate nuclear localization were unaltered by this single amino acid substitution.^142,147,148 However, the majority of AIRE-regulated genes could not be induced by this histone-binding mutant in mTECs.¹⁴⁷ Moreover, mice harboring this histone-binding mutant exhibited immune tolerance defects that were identical in breadth and severity to the Aire−/− controls, highlighting the essential role of AIRE’s histone-binding activity for its transcriptional function and T cell repertoire selection.¹⁴⁷ Mechanistically, AIRE’s recognition of unmodified H3 tails appears crucial for its capacity to engage the chromatin template in vivo, given than the D299A histone-binding mutant disrupted H3 interactions and severely compromised AIRE’s localization to target sites in studies using transfected cell lines^54,141 (Figure 3). Taken together, these data reveal the following spatiotemporal constraints for AIRE’s interaction with chromatin:

Figure 3: AIRE’s CARD and PHD1 domains are critical for its dual function.

Molecular functions of AIRE’s first plant homeodomain (PHD1) and caspase activation recruitment domain (CARD). Top: The PHD1 zinc fingers recognize unmodified H3 tails (i.e. H3R2me0, H3K4me0) and tether AIRE to chromatin, enabling recruitment of associated factors and promoting nucleosome assembly/stability. Individual AIRE molecules multimerize through CARD. The assembly of many AIRE molecules facilitates multivalent interactions with nucleosomes and augments the recruitment of transcription and elongation factors that mediate ectopic gene expression — and by consequence – immunological tolerance. Middle: Mutations in PHD1 do not affect AIRE multimerization but prevent chromatin targeting and nucleosome repositioning. Bottom: Mutations in CARD prevent AIRE multimerization and abrogate multivalent chromatin interaction, resulting in the failure to concentrate activating factors and repress chromatin accessibility. The failure of PHD1 and CARD mutants to concentrate activating factors at enhancers compromises ectopic gene expression, resulting in multi-organ autoimmunity (autoimmune polyglandular syndrome type 1 (APS-1)).

1. AIRE can tolerate repressive modifications (e.g. methylation at H3K9, H3K27) catalyzed by heterochromatin and Polycomb machineries at sequestered or silenced genomic regions. This tolerance is consistent with the reported deposition of H3K27me3 at AIRE-regulated genes and the putative interaction between AIRE and MBD1, a module that binds methylated DNA.^44,149 However, since AIRE largely acts on previously poised accessible loci, it is doubtful that compacted chromatin is a direct template for AIRE targeting.

2. AIRE is unlikely to target active enhancers with high densities of H3K4me1/2. Since eRNA synthesis precedes deposition of H3K4me1/2 by MLL3/4,^123,129 AIRE may focus its localization to regulatory elements that have lower rates of Pol II pause-release and have yet to recruit histone methyltransferase activity. Once AIRE induces enhancer transcription, there may be a dynamic competition between AIRE and MLL3/4 for H3 tail substrates. Approximately 85% of the solvent accessible surface of H3K4 becomes buried in a narrow binding groove of AIRE’s PHD1.¹⁴² This loss in solvent accessibility could restrict the engagement of H3K4 by other chromatin effectors to secure the unmodified state of H3. Such a function has been previously proposed for BHC80’s PHD finger securing the demethylated H3K4 state that was catalyzed by the associated lysine demethylase 1 (LSD-1).¹⁴⁵ The putative inhibition of H3K4 methylation by AIRE would not be deleterious to enhancer activity, as the essential role of MLL3/4 in enhancer function is not dependent on its catalytic activity.^150,151 The alternative outcome for the competition for H3K4 could be that MLL3/4 methylates H3K4 and evicts AIRE from the locus. MLL3/4 directly competes for H3 tail through its catalytic domains, which bind unmodified H3 at an affinity of ~4–6 uM (with cofactor S-adenosylmethionine bound).¹⁵² MLL3/4 have clusters of PHD fingers that recognize H4 tails, but none exhibit detectable binding to H3.^153,154 If AIRE is outcompeted and evicted due to H3K4 methylation, the initial AIRE-mediated recruitment of TOP1, BRD4 and P-TEFb could still elicit eRNA synthesis (albeit at lower levels). TOP1 interaction with the BRD4/P-TEFb-activated Pol II complex enhances the velocity of TOP1 nicking activity by at least 5-fold.¹¹² The resulting single-strand DNA breaks trigger the DNA-damage response, which recruits more TFs, TOP1, and elongation factors, creating positive feedback.^110,114 Thus, AIRE may only be required to initiate this cascade of events to ultimately amplify tissue-specific gene expression.

3. AIRE can target intronic regulatory elements at transcriptionally active gene bodies enriched with H3K36me3, given that H3K4 remains unmodified. SETD2 associates with actively elongating Pol II to deposit H3K36me3, which prevents spurious intragenic transcriptional initiation.¹⁵⁵ AIRE’s access to intronic enhancers is not inhibited by these processes.

4. AIRE is repelled by transcriptionally active promoters that have abundant deposition of H3K4me3.¹⁵⁶ Indeed, AIRE-regulated genes in mTECs are largely depleted of H3K4me3 and enriched for the repressive H3K27me3.^44,63 The lack of significant overlap between AIRE and H3K4me3 deposition further supports the role of AIRE’s histone-binding specificity in its targeting.^55,63 AIRE’s avoidance of H3K4me3 is intuitive, as AIRE has evolved to promote transcription at inactive tissue-specific loci. The lack of H3K4me3 at promoters of AIRE-induced genes could reflect several possibilities: (i) MLL1/2 is not recruited during AIRE-mediated transcription, (ii) AIRE outcompetes MLL1/2 for the unmodified H3 tail, or (iii) MLL1/2’s catalytic activity is not detectable due to the low frequency of cells expressing a given tissue-specific gene. In support of this third possibility, the promoters of AIRE-induced genes in HEK293 cells exhibited enrichment of H3K4me3 upon transcriptional activation by AIRE.⁶³ The eviction of AIRE due to MLL1/2’s methyltransferase activity could limit AIRE’s activating potential and restrain the transcriptional output at AIRE-regulated loci. Indeed, AIRE-regulated tissue-specific genes are expressed at low levels in mTECs compared to the levels observed in the respective peripheral tissues.^32,35,37,50

5. AIRE is unlikely to target bivalent promoters enriched for both H3K4me3 and H3K27me3 modifications. In embryonic stem cells, bivalent loci are highly accessible yet transcriptionally silent, resolving into an active or repressed state upon differentiation.¹⁵⁷ The prevalence of bivalency at promoters of mTECs is still unclear. There remains a possibility that AIRE could trigger transcription at bivalent genes indirectly via its activity at associated enhancers.

6. AIRE is unlikely to bind genomic regions with H3R2 methylation. Asymmetric dimethylation of H3R2 (H3R2me2a) largely functions to block the recruitment of chromatin effectors that recognize H3K4me3.^158,159 H3R2 is methylated by protein arginine methyltransferase 6 (PRMT6) at active promoters to repress transcription.^158,159 In contrast, H3R2me2a can be co-localized with active enhancers where it is reported to be associated with H3K4me1 and H3K27ac deposition.¹⁶⁰ H3R2me2a is also enriched at pericentromeric regions, whereas H3R2me1 is deposited at subtelomeric regions.¹⁶¹ Hence, these regions possibly further constrain the range of loci which AIRE can associate with.

7. AIRE cannot bind mitotic chromatin that is phosphorylated at H3T3 during prophase and maintained through the early stages of anaphase.¹⁶² The relevance of this inhibition is unclear as AIRE promotes ectopic gene expression after mature mTECs have exited the cell cycle.^30,31

The central role of AIRE’s histone-binding activity in its effector function suggests a major constraint in the chromatin landscape that AIRE is able to sample. This constraint contributes to the mode by which AIRE targets tissue-specific loci, but it is doubtful that this constraint is solely sufficient. Quantitative stoichiometric measurements by mass spectrometry estimated ~90% of the total chromatin fraction to be unmodified at H3K4.¹⁶³ The large majority of this fraction is expected to be unmodified at H3R2 as well.¹⁶³ The high abundance of unmodified H3K4 suggests AIRE’s recognition of unmodified H3 is likely a necessary but insufficient component of its targeting mechanism. Indeed, forced global demethylation of H3K4 by overexpression of JARID1B/KDM5B (H3K4-specific demethylase) did not significantly alter the range of genes regulated by AIRE.¹⁴⁷ Since AIRE targeting is strongly contingent on the chromatin landscape that is established prior to AIRE expression,^{32,43,57,61–63} the role of TFs and chromatin effectors that shape the chromatin accessibility landscape may be paramount in determining the genes regulated by AIRE.

AIRE is a chromatin repressor

AIRE’s actions as a chromatin effector are expected to promote chromatin accessibility given that: (i) AIRE-induced transcriptional elongation at enhancers and promoters recruits and modifies the activity of co-activators, topoisomerases, and DNA-damage repair proteins that contribute to the chromatin accessibility at target sites;^{110,112,114,116,120–129} and (ii) the strong bias for transcriptional induction (vs. repression) in the thousands of genes regulated by AIRE.^32,44 Accordingly, one would expect a decrease in chromatin accessibility at AIRE-targeted sites in an Aire-deficient setting. Contrary to expectations however, ATAC-seq profiles of mTECs from Aire-deficient mice showed a surprising increase in chromatin accessibility at sites directly targeted by AIRE compared to those from AIRE-sufficient controls.⁵⁷ These are the same sites that neighbor tissue-specific genes and exhibit large BAF-dependent increases in accessibility during differentiation of mTECs.⁵⁷ Furthermore, NF-κB motifs are the most highly enriched feature at these AIRE-targeted, BAF-induced, AIRE-repressed regions of the genome.⁵⁷ These data indicate that AIRE represses the chromatin accessibility that was previously induced by BAF and NF-κB during mTEC differentiation (Figures 2–4). Indeed, the quantitative difference in chromatin accessibility at tissue-specific loci between immature and mature mTECs in Aire−/− mice was much greater than that observed in WT littermates.⁵⁷ This result is the diametric opposite of the difference in transcription of tissue-specific genes between immature and mature mTECs in Aire−/− vs. WT mice.⁵⁷ Altogether, AIRE appears to impose a form of negative feedback on the transcriptional activity at target sites by modulating chromatin accessibility.

Figure 4: A speculative model of ectopic gene expression dynamics.

Top: mRNA levels of two tissue-specific genes (TSGs) through stages in mTEC transcription cycle. Bottom: Molecular events at two enhancers that individually target two distinct TSGs. BAF cooperates with TFs to promote chromatin accessibility at enhancers of moderate (TSG1) and weak (TSG2) strengths (indicated by dials and number of TFs engaged with loci). Insufficient recruitment of elongation factors (EFs) causes Pol II-pausing. Recruitment of AIRE at TSG1 enhancer mediates Pol II pause-release and augments enhancer activity. Subsequent chromatin repression by AIRE evicts TFs and co-activators, antagonizes BAF activity, and hinders Pol II re-loading, eventually silencing transcription. Newly released TFs target next preferential sites (at TSG2 enhancer), allowing sufficient chromatin accessibility, recruitment of AIRE, and recruitment of EFs at TSG2 enhancer for Pol II pause-release and enhancer activation.

AIRE’s repressive influence on chromatin accessibility could reflect the direct intrinsic actions of AIRE, as well as indirect consequences of transcriptional elongation. Pol II is in dynamic competition with nucleosomes, such that stable Pol II occupancy caused by its pausing creates a barrier for nucleosome assembly at enhancers and promoters.^90,164,165 Therefore, what appears to be AIRE’s repression of chromatin accessibility could be in part due to AIRE-mediated Pol II pause-release which would yield nucleosome assembly at target sites. AIRE might also compromise Pol II pausing, analogous to the disruption of the negative elongation factor, NELF. Ablation of NELF results in local increases of nucleosome occupancy, reducing the overall transcriptional output.^90,164,165 Such a scenario would be consistent with the lower levels of gene expression at tissue-specific loci in mTECs vs. those in respective peripheral tissues.^32,35,37,50

AIRE’s repressive influence on chromatin accessibility may also be due to the direct repressive actions of AIRE. Multiple lines of evidence indicate that AIRE has an inherent capacity to inhibit chromatin accessibility:

1. Inducible recruitment of AIRE to an accessible site upstream of the promoter of an AIRE-regulated tissue-specific gene (i.e. Oct4) resulted in the rapid loss of accessibility (~10 min) that was entirely uncoupled to transcription (Pol II was not recruited and the locus remained inactive).⁵⁷

2. Recruitment of AIRE to an active enhancer of a lineage-restricted gene (i.e. β5t) in a TEC line using an orthogonal recruitment system (via dCas9 vs. chemical-induced proximity) yielded the same result: the loss of chromatin accessibility.⁵⁷

3. Inducible recruitment of AIRE to a locus undergoing active transcription reduced the amplitude of transcription.⁵⁷

4. This direct and rapid repression of accessibility and transcription by AIRE was abolished by APS-1 mutations that compromised AIRE’s histone-binding activity (via PHD1) or multimerizing capacity (via CARD), but not by an APS-1 mutation in its SAND domain⁵⁷ (Figure 3).

5. Recombinantly expressed AIRE inhibited the BAF-catalyzed accessibility of intranucleosomal restriction sites in a dose-dependent manner (A. Koh, J. Antao, C. Benoist, D. Mathis, R. Kingston, unpublished results).

AIRE’s intrinsic repressive function is not mutually exclusive to the aforementioned indirect effects of Pol II pause-release. In fact, it is plausible that AIRE works in synergy with the negative supercoils generated behind the catalytically active Pol II to promote and stabilize nucleosome formation.¹⁶⁶ This repressive influence is likely to impact multiple steps in the transcription cycle (e.g. efficiency of BAF chromatin remodeling, recognition of cognate motifs by TFs, re-loading of Pol II, rate of eRNA synthesis, magnitude of co-activator recruitment). Thus, AIRE’s repressive function could dampen the transcriptional activity at enhancers, ultimately reducing the transcriptional burst frequency at target promoters of tissue-specific genes.^118,119 AIRE’s direct repression of chromatin accessibility could be the major factor causing the low expression levels of tissue-specific genes in mTECs.^32,35,37,50

The duality of AIRE’s activating and repressive functions is teleologically intuitive. A critical balance must be achieved to both promote immune tolerance and prevent overexpression of tissue-specific proteins like insulin, calcitonin, and blood coagulation factors, which can have catastrophic consequences.^167–169 Furthermore, mTECs exhibit disruption of chromatin barriers, the induction of DNA damage, and the ectopic expression of genes normally restricted to alternative lineages – traits which are all hallmarks of cancer.^26,170 Perhaps AIRE’s repressive function (in addition to the postmitotic state and short lifespan of mature mTECs^30,31) evolved to suppress the oncogenic potential of mTEC plasticity.

AIRE’s repressive capacity is not limitless, however. Recruitment of AIRE to the active Oct4 locus in embryonic stem cells did not impact the chromatin accessibility or the transcriptional amplitude of the locus (A. Koh, G. Crabtree, unpublished results).⁵⁷ In contrast, recruitment of heterochromatin protein 1α (HP1α) to the same region repressed chromatin accessibility and transcription within days.¹⁷¹ Hence, a critical density of activating TFs and chromatin effectors can nullify AIRE’s repressive function. This is consistent with AIRE’s localization to super-enhancers and its positive influence on chromatin accessibility at these dense clusters of enhancers.⁵⁵

AIRE acts as a multi-valent scaffold

AIRE does not operate as a monomer, but rather as a multi-valent chromatin effector that exhibits homotypic multimerization.^172–174 This multimerization is mediated by AIRE’s N-terminal caspase activation recruitment domain (CARD). APS-1 mutations in AIRE’s CARD impair its capacity to multimerize, form punctate nuclear localization, maintain protein stability, activate transcription, and repress chromatin^{57,148,172–174} (Figure 3). The number of AIRE monomers that interact to form a functional unit in mTECs is unclear. Recent biochemical studies substituting AIRE’s CARD for variable tandem repeats of FK506-binding proteins (FKBPs) required the tetramerization of this chimeric AIRE to partially rescue AIRE’s transcriptional function.¹⁷⁴ This study provokes curiosity about what the results would have been if the FKBP repeats had the acetylated interface for BRD4 recruitment.¹⁰⁵ Homo-oligomerization of AIRE creates multi-valent interfaces to increase the stoichiometric levels of interacting proteins tethered to the complex. This multi-valency is also expected to augment AIRE’s affinity to its interacting partners, consistent with the requirement of CARD for binding its functional allies (e.g. BRD4, HIPK2^105,175). Accordingly, AIRE exists as a large complex, estimated to be ~650 kDa – 2 MDa, over an order of magnitude larger than the AIRE monomer (55 kDa).^55,172 AIRE multiplexes itself to create a scaffold that bridges enzymatic activities to its substrates via chromatin targeting (Figure 3).

AIRE’s multimerized state forges multi-valency for nucleosomes through its histone-binding PHD finger. If the AIRE complex is a tetramer, it would be able to engage up to 4 nucleosomes simultaneously. A tetrameric complex could also strengthen the affinity for unmodified H3 tails to nanomolar levels since the monomer PHD1 alone exhibits K_D ~5–10 μM.^54,141 The engagement of multiple nucleosomes at once can play crucial roles in AIRE’s activating (via associating with target sites) and repressive (via nucleosomal compaction) functions (Figure 3). The importance of AIRE’s multi-valent engagement of nucleosomes is underscored by the dominant-negative effects of APS-1 mutations to AIRE’s histone-binding PHD finger.¹⁴⁸ The pairing of histone-binding mutants with WT AIRE could also compromise AIRE’s affinity to target sites, ultimately impairing AIRE’s capacity to bridge transcription elongation factors with paused Pol II. On the other hand, the disruption to poly-nucleosome engagement would also impair AIRE’s capacity to repress chromatin accessibility, diminishing the magnitude of the overall transcriptional defect. The ameliorative effect of this latter impairment is consistent with the milder autoimmune manifestations of patients with heterozygous AIRE PHD1 mutations. This autoimmunity presents later in age, has reduced penetrance, and displays milder phenotypes compared to the disease in patients with homozygous PHD1 mutations.¹⁴⁶

AIRE localizes as discrete nuclear bodies, often immediately adjacent to nuclear speckles.^176,177 It is yet unclear whether the AIRE complexes in these nuclear bodies have engaged their target sites or are associated with their interacting partners in mTECs. APS-1 mutations in AIRE’s CARD dissociated these punctate AIRE nuclear bodies into a diffuse distribution, indicating the central importance of AIRE’s multimerization for AIRE localization and function.^{148,172–174,176} It is doubtful that these AIRE nuclear bodies represent depots of unengaged AIRE, as the dosage of AIRE in mTECs is limiting, demonstrated by its haploinsufficiency.^51,52,148 Many of AIRE’s interacting partners can promote the formation of macromolecular condensates via liquid-liquid phase separation¹⁷⁸. TFs,¹⁷⁹ Pol II,¹⁸⁰ BRD4,¹⁸¹ and DNA-damage repair proteins¹⁸² have been hypothesized to form phase-separated molecular assemblies at enhancers and promoters to limit diffusion and augment effective molecular collisions to mediate their respective biological processes.¹⁸³ It is thus plausible that through its multimerization, AIRE functions as a nucleation site to concentrate associated enzymes with their substrates at target genomic regions. Feedback control might be imposed by AIRE’s interaction with unmodified nucleosomes, which also have been shown to undergo phase separation.¹⁸⁴

Instead of multimerized AIRE serving as a nucleation site, an alternative hypothesis has been proposed whereby AIRE would translocate its target regions to nuclear territories replete with factors that can mediate transcriptional elongation.¹⁸⁵ Such a scenario is consistent with AIRE’s proximity to nuclear speckles. Indeed, recent live-cell imaging studies demonstrated that the spatial proximity of genomic loci to nuclear speckles directly correlates with their transcriptional amplitude.¹⁸⁶ While nuclear speckles contain an abundance of RNA-splicing factors,¹⁸⁷ the model genes used in these studies did not have introns, indicating that the results are not necessarily due to splicing effects. Facets of the ‘nucleation’ and ‘translocation’ models are not mutually exclusive, and perhaps mTECs employ a combination of their features to promote the ectopic expression of tissue-specific genes.

A Speculative Model

Much remains to be learned about the modes by which mTECs establish a permissive chromatin landscape and the mechanisms underlying transcriptional triggering by AIRE. Immense strides have been made, however, since the initial discovery of AIRE nearly 25 years ago. Based on the findings of the past two decades, we propose a hypothetical working model for AIRE’s multi-faceted modes of action to facilitate ectopic gene expression in mTECs (Figure 4).

We postulate that during the differentiation of mTEC progenitors, chromatin barriers to alternate cell fates are lowered in part through the actions of ATP-dependent chromatin remodeling (e.g. BAF). We propose this disruption of chromatin homeostasis has a stochastic component, whereby the exact regions of tissue-specific loci that become accessible vary between individual mTECs. This stochasticity is constrained by 3 targeting properties of BAF: (i) its ~8 DNA-binding domains; (ii) its ~12 histone-binding modules (e.g. chromodomains, bromodomains, PHD fingers) with specificities to different modifications on various histone tails; and (iii) its physical association with TFs with distinct target motifs (e.g. NF-κB).^23,188 These regions made newly accessible by BAF could define the scope of the chromatin landscape on which AIRE can act in each cell.

Each mTEC can express different cohorts of AIRE-regulated genes during its lifespan,^36,39 suggesting that at any given time, AIRE acts on only a fraction of the total accessible loci that it can target. A hierarchy might exist for AIRE’s spatiotemporal targeting that could be determined by several factors. The loci that AIRE preferentially targets first could be a product of their regulatory syntax (e.g. number, binding affinity, spacing, order, orientation of consensus TF motifs⁶⁰) that influence the local concentration of TFs and magnitude of chromatin accessibility. Another major factor could be the methylation state of the H3 tail, where AIRE would be strongly repelled by any methylation at H3K4 or H3R2. Once AIRE engages its target sites, it would initiate the cascade of events leading to Pol II pause-release, eRNA synthesis, and enhancer-promoter contacts, culminating in transcriptional bursts at cognate promoters. Concomitantly, AIRE’s multi-valent engagement of nucleosomes through its histone-binding PHD finger progressively represses chromatin accessibility and generates a barrier for the next round of Pol II loading. Previously accessible TF motifs would then be occluded by nucleosomes, causing the translocation of TFs to engage accessible target motifs at other sites in the genome. There at these secondary sites, the AIRE-mediated transcription cycle repeats (Figure 4).

The limiting levels of AIRE in mTECs would also be a factor in regulating these transcriptional dynamics. The NF-κB-responsive regulatory element required for Aire expression in mTECs^189,190 is strongly targeted by AIRE and its chromatin accessibility is dependent on BAF (A. Koh, G. Crabtree, unpublished results).⁵⁵ AIRE’s repressive influence on chromatin accessibility at this critical enhancer could cause the attenuation, oscillation, or extinguishment of Aire expression. Thus, the different stoichiometric levels of AIRE during the lifespan of mTECs could alter the number of loci that AIRE has engaged at any given time. This negative feedback could also contribute to the transition of mTECs to the ‘post-AIRE’ stage, potentially including the differentiation of the recently discovered tuft-like mTECs.^34,191

AIRE – and more broadly, the capacity of the immune system to distinguish self vs. foreign constituents – has fascinated immunologists for decades. With the continuing innovation of technologies, especially at single cell resolution, the remaining enigmas are soon to be solved with new surprises awaiting us.