Structural aspects of the MHC expression control system

Abstract

The major histocompatibility complex (MHC) spans innate and adaptive immunity by presenting antigenic peptides to CD4⁺ and CD8⁺ T cells. Multiple transcription factors form an enhanceosome complex on the MHC promoter and recruit transcriptional machinery to activate gene transcription. Immune signals such as interferon-γ (IFN-γ) control MHC level by up-regulating components of the enhanceosome complex. As MHC plays crucial roles in immune regulation, alterations in the MHC enhanceosome structure will alter the pace of rapid immune responses at the transcription level and lead to various diseases related to the immune system. In this review, we discuss the current understanding of the MHC enhanceosome, with a focus on the structures of MHC enhanceosome components and the molecular basis of MHC enhanceosome assembly.

Graphical abstract

Introduction:

The Major Histocompatibility Complex (MHC) is a group of molecules that activate the adaptive immunity by presenting antigens on the surface of cells for recognition by appropriate T cells. There are two classes of MHC molecules. Class I are expressed in all nucleated cells and present endogenous antigens to CD8⁺ T cells, while class II are specifically expressed in antigen-presenting cells and present exogenous antigens to CD4⁺ T cells (1). MHC class I and class II proteins are among the most important molecules in the immune system; their expression is tightly regulated according to immunological status. A remarkable feature of the transcription of MHC genes is the formation of an enhanceosome complex by multiple transcription factors and co-activators (2–5) (Fig. 1). The promoters of MHC class I and class II genes share four cis-regulatory elements, termed W/S, X1, X2, and Y-box, which recruit transcription factors including the regulatory factor binding to the X-box (RFX) heterodimeric complex, the cyclic AMP response element binding proteins (CREB), and the nuclear transcription factor Y (NF-Y) complex (4, 6). However, the master regulators recruited by MHC class I and class II promotors are different. The MHC class II trans-activator, CIITA, specifically controls the transcription of MHC class II genes, and the NLR family CARD domain containing 5 protein, NLRC5, was recently shown to regulate the expression of MHC class I genes (7–11) (Fig. 1). These master regulators activate gene transcription by recruiting chromatin remodelers (12), modifying the chromatin structure (13, 14), or directly recruiting general transcription factors to the transcription start site (15). The expression profiles of CIITA and NLRC5 determine the tissue-specific expression of MHC molecules. While the transcription factors and NLRC5 are ubiquitously expressed, CIITA is only constitutively expressed in antigen presenting cells, causing the cell type-specific expression of MHC class II molecules. Furthermore, as expression of CIITA can be induced by various immune signals (such as IFN-γ), it also connects MHC II expression levels with the immunological environment (16).

Figure 1.

The MHC enhanceosome. A) CIITA and NLRC5 are recruited by a similar set of transcription factors on the MHC promoters. The CIITA enhanceosome regulates MHC class II expression, while the NLRC5 enhanceosome regulates MHC class I expression. B) the consensus sequence in the MHC promoter region, adapted from Ludigs, K., et al., PLoS Genet, 2015 (10).

Abnormal expression of MHC leads to many diseases. One such example is Bare Lymphocyte Syndrome (BLS), in which severe human immunodeficiency results from genetic mutations that disrupt MHC enhanceosome assembly and thereby reduce T cell count (2). HIV, coronavirus, and other viruses also disrupt the formation of the MHC enhanceosome to facilitate immune evasion (17–20). Down regulation of CIITA or NLRC5 is also frequently observed in many solid and hematopoietic tumors (21–26). Given the critical roles of MHC molecules in immune regulation, understanding the molecular mechanism of MHC enhanceosome assembly can undoubtedly foster advances in the clinical treatment of multiple diseases through controlling the expression of MHC molecules. In this review, we will introduce the discovery and characterization of the MHC enhanceosome with a focus on the structure of its components. This will be facilitated by an analysis of the AlphaFold predicted structures of CIITA and NLRC5.

Bare Lymphocyte Syndrome and the discovery of the MHC enhanceosome complex

BLS, or MHC class II deficiency, was first discovered in the late 1970s as an autosomal recessive disease caused by the absence of MHC class II expression. Patients with this disease show severe immunodeficiency, recurrent infections, and frequently death in early childhood (27–30). Further studies found that the genetic defect associated with this disease did not segregate with the MHC gene itself, thus predicting a trans-acting regulatory factor unlinked to the MHC gene, that should control a function or a product necessary for the production of MHC class II molecules, and is affected in the BLS patients (31). This prediction spurred a series of exciting studies toward the identification of the MHC enhanceosome as the MHC expression control machinery (2, 3, 6, 32–36). The first gene identified in the MHC enhanceosome was CIITA, cloned by genetic complementation of an MHC class II mutant B lymphocyte line in 1993 (37). Later studies showed CIITA is expressed in a cell type-specific manner, and the pattern of its expression correlates with that of MHC class II molecules in numerous cell lines and tissues(16). Immune signals such as IFN-γ induce MHC class II expression through up-regulation of CIITA expression, and transfection of CIITA is sufficient to activate expression of MHC class II genes in class Il-negative cells (16).

Though CIITA is now confirmed as the master regulator for the expression of MHC class II genes, it does not directly bind the MHC promoter. Two years after the discovery of CIITA, genes encoding RFX5, RFXAP, and RFXANK were cloned from BLS patients (38). Further studies showed that these three RFX proteins form a trimeric transcription factor complex (the RFX complex), together recognizing the X1-box of the MHC promoter and physically interacting with CIITA. Further efforts found that CREB and NF-Y also exist in the same complex with RFX and CIITA (36, 39, 40). Below we will focus our discussion on the protein structures of the major transcription factors and co-activators in the MHC enhanceosome, including RFX, CREB, NF-Y, CIITA, and NLRC5 (Table 1).

Table 1:

protein components in the enhanceosome complexes

Protein	Function	Structural information
RFX5	Transcription factor. Forming RFX complex with RFXAP and RFXANK, bind to X1-box DNA motif.	PDB: 2KW3, (aa. 25–90, in complex with RFXAP)
RFXAP	Transcription co-factor. Forming RFX complex with RFX5 and RFXANK.	PDB: 2KW3, (aa. 214–272, in complex with RFX5)
RFXANK	Transcription co-factor. Forming RFX complex with RFXAP and RFXANK.	PDB: 3V30, (aa. 90–260, in complex with RFX5)
CREB	Transcription factor, bind to X2-box DNA motif.	PDB: 5ZKO, (bZIP domain)
NF-YA	Transcription factor. Forming NF-Y complex with NF-YB and NF-YC, bind to Y-box DNA motif.	PDB: 4AWL, (aa. 232–293, in complex with NF-YB, NF-YC, and Y-box DNA)
NF-YB	Transcription co-factor. Forming NF-Y complex with NF-YA and NF-YC.	PDB: 4AWL, (aa. 48–141, in complex with NF-YA, NF-YC, and Y-box DNA)
NF-YC	Transcription co-factor. Forming NF-Y complex with NF-YA and NF-YB.	PDB: 4AWL, (aa. 40–119, in complex with NF-YA, NF-YB, and Y-box DNA)
CIITA	Recruited by the transcription factors and activate MHC Class II transcription.	unknown
NLRC5	Recruited by the transcription factors and activate MHC Class I transcription.	unknown

MHC promoter and the transcription factors

The structure of the MHC promoter has been reviewed in detail elsewhere (4, 41, 42). Briefly, the W/S-, X1-, X2- and Y-boxes form the promoter-proximal regulatory modules of MHC class I and class II genes. These cis-acting elements recruit transcription factors in a sequence specific manner. Specifically, the X1-box recruits the RFX complex, the X2-box recruits CREB or a heterodimer formed by CREB and ATF1 (cyclic AMP-dependent activating transcription factor 1), and the Y-box recruits the NF-Y complex. It is still unknown if there is W/S-box specific transcription factors, as no transcription factors have been identified at this region (10, 43). While the sequence and spacing of X1-, X2- and Y-boxes are highly conserved, the W/S-box in the class I promoter diverges substantially from that in the class II promoter, which may confer the distinct transactivating specificities of NLRC5 and CIITA (10) (Fig. 1B).

RFX complex

As mentioned before, the RFX complex is a ternary complex formed by RFX5, RFXAP, and RFXANK proteins (Fig. 2A). RFX5 is the fifth member of the RFX family of DNA-binding proteins. Proteins in this family share a novel and highly characteristic DNA-binding domain (DBD) called the RFX motif (38, 44). RFX5^DBD mediates the sequence specific interaction between the RFX complex and the X1-box region in the MHC promoter. Full-length RFX5 has 616 amino acids that can be separated into five distinct regions: the oligomerization domain (OD), the DBD, the helical dimerization domain, the proline-rich region (P), and the transactivation domain (Fig. 2A). Insights into the DNA recognition mechanism of the RFX motif were provided by the crystal structure of RFX1^DBD, another RFX family member, in complex with its double stranded DNA substrate (45). Briefly, the RFX1^DBD consists of three α-helices (H), three β-strands (S), and three connecting loops (L), together forming a hydrophobic interface to mediate the protein-DNA interaction. Specifically, two copies of RFX1^DBD bind to DNA in a symmetric manner, and both strands of the DNA have direct contacts with RFX1^DBD. The sequence specificity is primarily determined by the interaction of K45 with the base of (T5’, C6, and A5), R58 with the base of G11, and R62 with the base of G15’ (Fig. 2B). The key residues in RFX1 that interact with DNA are all conserved in RFX5, and the AlphaFold predicted DBD of hRFX5 is highly similar with the crystal structure of RFX1^DBD, indicating RFX5^DBD may bind X1-box DNA in a similar manner (Fig. 2B).

Figure 2.

The transcription factors in the MHC enhanceosome. A) Domain organization of the components of the RFX complex. The blue colored domain within RFXAP form two a-helices, shown in panel C. Regions that mediate the intra-molecular interactions are indicated by the two-headed arrow; domains in grays are predicted disordered regions. B) Left, co-crystal structures of RFX1^DBD and X-box DNA. PDB: 1DP7. Residues on RFX1^DBD and bases of X-box DNA that mediate sequence specificity are labeled. Right, AlphaFold predicted structure of hRFX5 (AF-P48382-F1-model_v2, magenta) showed well agreement with the RFX1^DBD. The oligonucleotide used in crystallization and the X1-box region of HLA-RDA promoter are shown. C) NMR structure of RFX5 (25–90) and RFXAP (214–272). PDB: 2KW3. D) Crystal structure of RFX5 (167–183) and RFXANK (90–260), PDB: 3V30. E) Domain organization of CREB, with domains predicted as disordered regions in gray. F) Domain organization of proteins in NF-Y complex, with domains predicted as disordered regions in gray. G) Crystal structure of NF-Y in complex with Y-box DNA, PDB: 4AWL.

The dimerization and transactivation domains of RFX5 are also essential for MHC gene transcription, as a nonsense mutation that causes premature translational termination at residue 321 also leads to BLS(2).

RFX5/RFXAP/RFXANK interaction

RFXAP and RFXANK are identified together with RFX5 as co-activators that assemble on the X1-box DNA motif (2, 44, 46, 47). Despite their names, RFXAP and RFXANK are not members of the RFX family and do not contain the DNA binding motif. RFXAP is a 30 kDa protein containing a large disordered region at its N-terminus, and two α-helices at its C- terminus. RFXANK belongs to the Ankyrin repeat protein family (Fig. 2A). Though neither RFXAP nor RFXANK can directly bind DNA, mutations on RFXAP and RFXANK abolish the recruitment of RFX5 to the MHC promoter and lead to immunodeficiency (2). The structural mechanism of RFX complex formation is not fully understood. A DNA binding assay showed that pre-assembly of RFXAP and RFXANK with RFX5 is essential for RFX5/DNA interaction, suggesting they may induce a conformational change in RFX5 to expose the DBD (48). The difficuty in obtaining structures of the RFX complex is primarily caused by its disordered nature. NMR studies showed that two C-terminal helices of RFXAP (aa. 214–272) interact with a small N-terminal region of RFX5 (aa. 25–90) (Fig. 2C). A crystal structure illustrated the interaction between RFXANK and a peptide within the dimerization domain of RFX5, which forms a PxLPxL motif (¹⁶⁷KTLVSMPPLPGLDLKGS¹⁸³) (Fig. 2D) (49, 50). Further studies are needed to identify any additional interactions between RFX5, RFXAP, and RFXANK, to address how these interactions promote RFX5 trimeric complex formation, and to determine how they contribute to MHC enhanceosome assembly.

CREB

The X2-box of the MHC promoter has sequence similarity to the cyclic AMP response element (CRE), and CRE binding protein (CREB) was identified as a co-migratee with X2 -box DNA in B cell nuclear extracts (41, 46, 51). Later, CREB was confirmed to be the X2-box binding protein through multiple assays: first, it cooperated with CIITA to activate MHC class II expression; second, crosslinking immunoprecipitation assay showed it bound to the class II promoter in vivo; and third, a dominant negative inhibitor of CREB, the inducible cAMP early repressor (ICER), was able to repress transcription from MHC class II promoters (34). CREB may assembles into the MHC enhanssosme in the form of a hetrodimer with ATF1 (36). Structurally, CREB contains a glutamine-rich 1 domain (Q1), a kinase inducible domain (KID), a glutamine-rich 2/constitutive activation domain (Q2/CAD), and a basic leucine zipper (bZIP) (52, 53). Co-immunoprecipitation showed that the bZIP domain of CREB is responsible for association with RFX5, while the bZIP and Q2/CAD domains are responsible for association with CIITA (53) (Fig. 2E).

NF-Y complex

NF-Y, also known as the CCAAT-binding factor (CBF), is a highly conserved and ubiquitously expressed heterotrimeric transcription factor. NF-Y is composed of three DNA-binding proteins, the NF-YA, NF-YB, and NF-YC subunits (Fig. 2F). A crystal structure of NF-Y bound to a 25 bp CCAAT oligonucleotide was solved in 2013, which showed that the histone-fold domains (HFDs) of NF-YB and NF-YC form a dimer that binds to the sugar-phosphate backbone of DNA, mimicking the nucleosome H2A/H2B-DNA assembly. NF-YA binds to NF-YB/NF-YC and inserts an α-helix deeply into the DNA minor groove, providing sequence-specific contacts to the CCAAT box (54) (Fig. 2G). NF-Y binds cooperatively with RFX to the MHC promoter, which is also essential for MHC enhanceosome assembly and MHC gene expression (41, 55).

Intrinsically disordered regions in the transcription factors

Recruitment of the transcription factors to the MHC promoter is the first step of enhanceosome assembly. CREB and NF-Y regulate many genes, but have only been observed cooperating with RFX on MHC promoters (6, 56) (Fig. 1). Strong cooperativity of these three transcription factors in MHC promoter binding was observed both in an in vitro binding assay (56) and in an in vivo co-IP experiment (6, 56), and the absence of any of them abolish the expression of MHC molecules, indicating they may function together as an intact machinery (4). While so far most structural studies of these transcription factors have focused on their DNA binding domains, they also contain multiple glutamine-, serine- and proline-rich intrinsically disordered regions (IDRs). Further studies are needed to address whether these IDRs may mediate the protein-protein and protein-DNA interactions (Fig. 2A, 2E, 2F).

CIITA, NLRC5 and the NLR protein family

CIITA is not only the first component identified in the MHC enhanceosome complex, but also a founding member of the nucleotide-binding domain (NBD) and leucine rich repeat (LRR) domain containing protein (NLR protein) family. NLRs distribute widely in plants and animals (57–59), and are involved in a wide range of processes. One of the most well-known subclasses of NLR proteins, the inflammasomal NLRs, form inflammasome complexes in response to pathogens and danger signals, and they initiate inflammation by activating caspase-1 (60). However, the involvement of CIITA and NLRC5 in MHC gene transcription is unique, as they are the only two NLRs known to regulate gene expression (Fig. 3A) (61, 62). Despite this functional diversity, the NLRs are structurally similar with each other. They share a three-domain architecture that comprise a varied N-terminal effector domain, a central NACHT domain, and a C-terminal LRR domain (58). Among them, NACHT is an NTPase domain which is featured in Walker A, Walker B, and five other motifs (58). NACHT and LRR domains mediate the functions of NLRs through ligand binding, protein-protein interaction and protein oligomerization (Fig. 3B).

Figure 3.

CIITA and NLRC5 belong to the NLR protein family. A) Phylogenetic tree (unrooted) of human NLR proteins with CIITA and NLRC5 are highlighted in red and NLR proteins with solved structures highlighted in blue. B) Domain organization of NOD2, CIITA and NLRC5.

Although structures of CIITA and NLRC5 have not been reported, multiple structures of inflammasomal NLRs have been solved at atomic resolution that provided insights about NLR proteins (63–69). Recently, AlphaFold was developed as a computational approach capable of predicting protein structures from amino acid sequences to near experimental accuracy (70). More importantly, extensive biochemical studies have been performed to map the functional regions of CIITA (13, 71–77). Plotting the functional regions of CIITA to the AlphaFold predicted structure might provide clues suggesting how CIITA interacts with the enhanceosome components and how it regulates gene expression. Here we compare the AlphaFold predicted structures of CIITA and NLRC5 with the crystal structure of an inflammasomal NLR protein, NOD2 (PDB# 5IRM) (66), which is a close relative of CIITA on the phylogenetic tree (Fig. 3A, 4A). It should be noted that the inflammasomal NLRs adopt different conformations and oligomeric states in their inactive and active forms, which is beyond the scope of this review. We don’t know, however, whether CIITA and NLRC5 undergo conformational change when they assemble into the enhanceosome complex. It is also necessary to emphasize that we are only comparing CIITA/NLRC5 with NOD2 from a structural perspective; this does not imply functional similarity of these proteins (indeed, they regulate different immunological pathways). We will also discuss the major differences between the predicted CIITA and NLRC5 structures that may contribute to their target gene specificity.

Figure 4.

AlphaFold predicted structural model of CIITA and NLRC5. A) Crystal structure of NOD2(ΔCARD), domains are in the same color as Fig. 3B, with NBD-slate, HD1-green, WHD-magenta, HD2-orange, and LRR-cyan, specifically. B) AlphaFold predicted human CIITA structure (afCIITA), the colors represent the confidence score reported by AlphaFold. C) AlphaFold predicted human CIITA structure with individual domains colored as in A). D) Residues in afCIITA that mediate self-association. E)-I) Residues in afCIITA that interact with the transcription factors RFX5, RFXANK, CREB, NF-YB, and NF-YC, respectively. J) AlphaFold predicted human NLRC5 structure (afNLRC5), the colors represent the confidence score reported by AlphaFold. K) AlphaFold predicted human NLRC5 structure with individual domains colored as in A). L) Superimposition of afNLRC5 (blue) and afCIITA (red).

Crystal structure of NOD2

There are four inflammasomal NLR proteins with solved structures: NLRC4, NOD2, NLRP3, and NAIP (64, 66–68, 78). NOD2 is closest with CIITA at the sequence level (17.77% identity), and will be taken as a reference to dissect the structural domains of CIITA. NOD2, also known as NLRC2, has an N-terminal effector domain comprising two tandem caspase activation and recruitment domains (CARD), which activates caspases or other CARD containing proteins to initiate inflammation (66). Following CARD, NOD2 has the typical NACHT - LRR architecture that is shared across the NLR family (Fig. 2B). The crystal structure of Rabbit NOD2(ΔCARD) was solved at 3.3 Å resolution in 2016 (Fig. 4A) (66).

The AlphaFold predicted structure of CIITA

The AlphaFold predicted human CIITA structure is downloaded as AF-P33076-F1-model_v1 from the AlphaFold database (hereafter referred to as afCIITA). The per-residue confidence score (pLDDT) is high in most of the protein (>70 between 0–100), except in the N-terminal region. This is possibly due to the lack of structure templates and the highly disordered nature of the N-terminal domains (Fig. 4B).

Previous efforts to characterize the functional domains of CIITA have identified an acidic domain (AD) in the N-terminus (79). Unlike the N-terminal effector domain in other NLR family members, which often forms an oligomer, there is currently no evidence suggesting the AD in CIITA can also oligomerize. Instead, the AD contains a significant level of acidic amino acids among residues 30–160 and defines CIITA as a transactivator (13, 14, 37); the transcription activation function of this domain was tested via fusion to a DNA-binding sequence associated with a reporter gene in both yeast and human cells (32, 80). Progressive C-terminal deletions of the AD in one of these fusion constructs indicated residues 1–125 as the minimal CIITA region required for full activity (80).

Most NLRs have the NACHT domain follow immediately after the N-terminal effector domain. However, CIITA has approximately around 160 residues inserted between AD and NACHT with high proline, serine, and threonine content. This has been named the PST domain (37). Deletion of the N- or C-terminal half of PST had no effect, however, deletion of the entire region eliminated the transcription activation of MHC induced by CIITA (81). It is still unknown how PST is involved in MHC expression, because neither the transactivation activity of the acidic domain, nor the MHC class II specificity of CIITA require the PST domain (32, 80). The crystal structure of NOD2 does not contain the N-terminal effector domain (Fig. 4A), while afCIITA has the AD of CIITA predicted as a four-helical bundle and the PST as an unstructured loop (Fig. 4B, 4C).

Starting from the NACHT domain, afCIITA shows a fold that is highly similar to NOD2, with some unique features that will be discussed in detail below. The NACHT module in NOD2 (and all other inflammasomal NLRs that have solved structures) comprises an NBD, a helical domain (HD1), and a winged-helix domain (WHD) (63) (Fig. 3B, 4A). Structure alignment defines CIITA residues 336–568 as the NBD, residues 569–634 as the HD1, residues 635–726 as the WHD, and 727–830 as the HD2 (Fig. 3B, 4C). Functional studies showed CIITA has a unique nucleotide binding property in the NACHT module. Instead of ATP or ADP that is present in other NLRs, residues 420–561 of CIITA are involved in the interaction with a GTP molecule. CIITA showed one-third of the GTP-binding activity of a GTPase-defective H-Ras in a GTP exchange assay (81). There are three GTP-binding motifs identified in CIITA-NACHT: G1, G3, and G4 (71, 81). G1 (⁴²⁰GKAGQGKS⁴²⁷) is a phosphate-binding motif (GXXXXGK[S/T]), G3 (⁴⁶¹DAYG⁴⁶⁴) is a magnesium-binding motif (DXXG), and G4 (⁵⁵⁸SKAD⁵⁶¹) is a guanosine-binding motif ([N/T/S] KXD) (71, 81). Mutations of G1, G3, or G4 caused a significant decrease in GTP binding and CIITA activity, though reduced expression of the G3 mutant protein was observed (71, 81). However, CIITA does not display detectable GTPase activity (71). The work that has been done in the interest of elucidating the role of GTP binding in the function of CIITA suggests that GTP is intrinsically bound to CIITA to allow proper structural confirmation (71). Additionally, GTP binding ability seems to affect the cellular localization of CIITA, as CIITA with mutations affecting GTP binding failed to properly localize to the nucleus (71).

In NOD2, the HD2 caps the N-terminal side of the LRR domain via the formation of extensive hydrophobic interactions (63). AfCIITA HD2 does not have direct interaction with LRR. Instead, it is separated from LRR by an insertion domain formed by residues 831–957 (Fig. 4C). The insertion domain adopts an S-H-S-H-S-H-H-S-H structure (S: β-strand; H: α- helices). It is not known whether the HD2 and insertion domains have their specific functions other than maintaining the structure of CIITA.

LRR domains in toll-like receptors (TLRs) and inflammasomal NLRs mediate protein-protein interactions that involve pathogen recognition and receptor-coreceptor complex formation (64, 65, 67, 82, 83). Residues 958–1130 of CIITA form a classical LRR with the horseshoe shape. Co-immunoprecipitation experiments showed residues 939–1130 interacting with 336–702 (within the NACHT domain) (74, 79). These two regions do not show intramolecular interactions in afCIITA, which may indicate that they are interacting with other CIITA molecules and mediating CIITA oligomerization (Fig. 4D).

Interaction between CIITA and the transcription factors

CIITA interacts simultaneously with the RFX complex, CREB, and the NF-Y complex in the MHC enhanceosome (14, 148, 150). Regions on CIITA that bind these transcription factors were mapped by co-immunoprecipitation, which indicated that CIITA residues 336–612 binds RFX5, 1–335 binds RFXANK, 1–735 binds CREB, 518–612 binds NF-YB, and 218–335 binds NF-YC (150) (Fig. 4E–I). Further studies especially structural approaches are needed to map the specific residues in these regions that directly mediate protein-protein interactions.

Post-translational modification of CIITA

Post-translational modification of CIITA is a crucial regulatory point, and multiple post-translational modifications including acetylation, phosphorylation, and ubiquitination are identified. Acetylation of K141 and K144 of CIITA by CREB and acetyltransferase PCAF may enhance CIITA nuclear accumulation and transactivation of MHC class II genes (84). CIITA is phosphorylated at multiple sites with apparently different outcomes. Phosphorylation of serine 280 may increase the cellular level of CIITA (85, 86), while phosphorylation of serines 286, 288, 293, 834 and 1050 were shown to downregulate CIITA function (87, 88). CIITA at the MHC class II promoter is also shown to bee ubiquitinated, which enhances the stability and transcriptional function of CIITA (86, 89).

The AlphaFold predicted structure of NLRC5

The function of NLRC5 was controversial for a long time but more recent studies have supported its role as the master regulator of MHC class I gene expression (7–10). The AlphaFold predicted human NLRC5 structure, AF-Q86WI3-F1-model_v1 or afNLRC5, showed similar overall fold with NOD2 and afCIITA, with a significantly longer LRR domain (Fig. 4J, K, L).

The N-terminal effector domain of NLRC5 adopts a CARD-like fold, though it does not bind and activate caspases (90). A domain swap showed the CARD of NLRC5 can substitute for the N-terminal domain of CIITA, confirmed that it confers transactivation activity instead of caspase activation (90, 91). After the N-terminal effector domain, a linker containing residues 93–131 leads directly to the NACHT domain of afNLRC5, which is significantly shorter than the afCIITA PST domain (Fig. 4K).

AfNLRC5-NACHT also displayed similar organization of NBD (131–367), HD1(368–425), WHD (426–530), and HD2 (531–635). Further investigation is needed to evaluate whether NLRC5 has nucleotide binding potential similar to the previously discussed GTP binding potential of CIITA. The HD2 of afNLRC5 forms extensive interactions with WHD and directly caps the LRR domain, which spans from residue 636 to 1847. This is the longest LRR among NLRs, making NLRC5 the biggest protein in this family. Though afNLRC5 does not have an entire insertion domain as observed in afCIITA, there are several insertion loops in its LRR formed by residues 816–843, 944–976, and 1079–1107. They do not disrupt the horseshoe-like LRR structure but may afford unique docking sites in protein-protein interaction. It will be interesting to test whether these insertions are functionally important for NLRC5 (Fig. 4K).

Future directions and challenges

The MHC enhanceosome formation is essential for MHC expression, and no bypass or alternative pathways can compensate efficiently for its absence. Despite significant efforts have been invested to elucidate the functions of the transcription factors and master regulators, our understanding of the structural basis of MHC enhanceosome assembly is still limited by the lack of atomic models. Major challenges include the aggregation-prone nature of NLR proteins and difficulties in reconstituting the complex in vitro. AlphaFold predicted CIITA and NLRC5 structure models provided information to interpret previous functional studies, but also raised new questions. Further studies are needed to understand how CIITA and NLRC5 recognize the transcription factors, and to determine what excludes NLRC5 from the MHC class II promoter locus. The fundamental biological and clinical importance of the MHC makes the MHC enhanceosome a key determinant of the functional integrity of the immune system, and we believe a deeper understanding of the MHC enhanceosome holds the promise of targeting this process for therapeutic benefit through modulating immune function.

• MHC enhanceosome is essential for MHC gene expression

• Structural and biochemical studies provided insights into the MHC enhanceosome assembly

• AlphaFold predicted structures of CIITA and NLRC5 advanced our understanding of these two NLR proteins in gene transcription control