Structural and biological features of FOXP3 dimerization relevant to regulatory T cell function

Structure of mFOXP3 zinc finger and leucine zipper (mFOXP3-ZL)

The crystal structure of the mouse FOXP3 (mFOXP3) domain containing the zinc finger and leucine zipper region (amino acids 196-276, designated as mFOXP3-ZL) was resolved to 2.1-Å resolution. The amino terminal region (V197-E209) corresponding to the zinc finger loop was poorly defined in the crystal structure. Based on the well determined part corresponding to the zinc finger helix region, a complete FOXP3 zinc finger was modeled using the five-finger structure (PDB code 2GLI) (Pavletich and Pabo, 1993) as a homologous template.

The structure defines amino acids V197-L223 as a zinc finger motif, and D224-K262 as the leucine zipper motif (Figure 1A). The zinc finger is immediately adjacent to the leucine zipper, connecting its α-helix directly to that of the leucine zipper, producing an extended single long helix. The zinc atom is coordinated by residues C198, C203, H216, H221 and in part by D220.

The zinc finger motif is not directly involved in dimerization; while the leucine zipper mediates inter-molecular interactions, consistent with biochemistry showing that FOXP3 may dimerize via this region. The coiled coil has a minimum size of only four heptad repeats (Figure 1A and 1B). Its α-helices wind around each other with an average pitch of 165 Å [calculated by using TWISTER (Strelkov and Burkhard, 2002)], in contrast to the 146 Å pitch observed for the prototypical coiled coil tropomyosin structure (Brown et al., 2005). The distance between the helical axes varies by less than 0.66 Å.

Homo-dimerization of FOXP3 via an unusual anti-parallel coiled coil

The mFOXP3-ZL homodimer features an unusual two-stranded anti-parallel α-helical coiled coil with a perfect 2-fold symmetry, resulting in two identical halves (K228-Q243 of subunit A paired with Q243-H258 of subunit B, versus Q243-H258 of subunit A paired with K228-Q243 of subunit B) (Figure 1A and S1). The two equivalent halves start respectively from the two distal ends of the elongated dimer and meet at the center core residues L241 and L245, where the 2-fold axis and the coiled coil super-helical axis intersect. Each half contains four pairs of core residues: C231-M255 (a-position), Q234-L252 (d-position), V238-E248 (a-position) and L241-L245 (d-position). Strikingly, two out of the four pairs contain polar amino acids, which is rarely observed in coiled coil core packing. In order to maintain the “knobs-into-holes” stacking pattern, the polar core residues Q234 and E248 turn the tips of their side-chains away from the core, creating a unique conformation that favors an inter-subunit hydrogen bonding essential for FOXP3 dimerization (Figure 1C and 1F).

The mFOXP3-ZL coiled coil is stabilized by a number of electrostatic interactions including 4 hydrogen bonds and 2 salt bridges. For example, the IPEX relevant residue K249 (equivalent to K250 of human FOXP3) and the breast cancer relevant residue E242 (equivalent to E243 of human FOXP3) form an inter-subunit salt bridge right on the opposite side of the Q234-E248 inter-subunit pairing (Figure 1D). The interface accessible surface area of mFOXP3 coiled coil is 966 Ų or 16% of total, which classifies the mFOXP3 homodimer as a transient one when compared with an average of 1462 Ų for all determined protein dimers (1515 Ų for homodimers and 1335 Ų for heterodimers), and 1005 Ų for all transient dimers (1153 Ų for strong ones and 693 Ų for weak ones)(Reynolds et al., 2009). The empty space on the dimeric interface is ~260 ų, corresponding to substantial packing holes found near to core residues L241 and L245 (Figure 1D and 1E). These features indicate that the FOXP3 coiled coil is rather flexible and the homo-dimeric association is dynamic.

Sequence alignment of the coiled coil motifs from the FOXP subfamily revealed an almost identical core, including the two polar residues at the d-position corresponding to mFOXP3 Q234 and E248 (Figure 1B). However, the molecular surface of the mFOXP3 homo-dimer shows a singularly high percentage of hydrophobic residues at b, c and g positions, which is unexpected for coiled coil structures. The three-dimensional arrangement reveals that these hydrophobic residues (specifically L222, A229, L232, L233, V237, L247, A254 and A257) form a stretch spanning the surface of the coiled coil (Figure 1B and S4).

Mutational analysis of interface residues and minimum region for FOXP3 homo-dimerization

Based on the mFOXP3 dimer structure, mutagenesis of full-length human FOXP3 was used to probe the homo-interacting contribution of individual residues (Figure 1F). Of note, several IPEX mutations including L242P, DelK250 and DelE251 (equivalent to mFOXP3 L241P, DelK249 and DelE250 respectively) are found in the coiled coil region, but with disparate three-dimensional positioning. In the mFOXP3 coiled coil structure, L241 is a core residue (d position) right on the dimeric interface; K249 is an interface-flanking residue (e position) that forms a hydrogen bond with E242 from the opposing subunit; yet E250 is a non-interface residue sandwiched by K249 and K251 (Figure 1C, 1D and 1E). Mutation or deletion of either L241 or K249 would directly affect dimerization, while deletion of E250 would alter the conformation of K249 and K251 to also affect dimerization.

We found that mutating L241 to a proline, or deletion of K249 disrupted FOXP3 homo-association (Figure 1F). By contrast, mutation of residues L223, K226 and A229, which are not located on the dimeric interface, did not disrupt FOXP3 homo-association. Mutation of another core residue (a position) V237 with an isoleucine structurally preserved the coiled coil packing and also did not disrupt FOXP3 homo-association.

Mutation of K251, a lysine residue involved in an interaction network essential for FOXP3 dimerization (Figure 1C), to arginine, which is found in other members of the FOXP subfamily (Figure 1B), only slightly reduced FOXP3 homo-association (Figure 3A). K249R and K251R double mutations had little effect on dimerization. However, mutations K251Q (mimicking acetylation) and K251D reduced FOXP3 homo-association respectively in these biochemical studies. We propose the reduction of FOXP3 homodimers occurs through modulation of the inter-subunit hydrogen bond formed between Q234 and E248 (Figure 1C, 1F and 3A). These observations indicate that the non-core residue K251, which is unique to the FOXP3 molecule, plays a role in regulating the stability of the FOXP3 homo-dimer.

Structure-based mutational analysis further revealed that a minimum of four haptad repeats (C₂₃₁LLQREV; VQSLEQQ; LELEKEK; LGAMQAH₂₅₈) is required for the FOXP3 dimer formation. Removal of the core residues (a or d position: C231 and M255) from the coiled coil terminal haptad repeats disrupted the FOXP3 dimer as shown by the FOXP3 truncation mutant a.a. 1–254; while the a.a. 1–258 truncation mutant can still form dimers, as does wild type FOXP3 (Figure 3A). Structure wise, the truncation mutant a.a. 1–254 would also lack the inter-subunit salt bridge formed by Q256 and R235, and would lose a pivotal core-packing residue M255. M255’s side-chain contributes to maintenance of the central conformational network formed by residues K251, Q234 and E248 (Figure 1C).

Lysine 251’s role in FOXP3 coiled coil mediated dimerization in Treg cells

Since K251 has been highlighted as a unique residue critical for FOXP3 homo-dimerization, we then expressed wild-type mFOXP3 and mutants of mFOXP3 lacking lysine 251 (delK251) in primary murine CD4+ T cells to verify its functional relevance (Figure 2A). Wild-type mFOXP3 repressed il2 gene expression by activated T cells (Figure 2B), and this was associated with direct occupancy of mFOXP3 and deacetylation of nucleosomes at the il2 promoter (Figure 2B and 2C). However, the delK251 mutant mFOXP3 failed to repress IL-2 production (Figure 2B), and completely failed to bind and remodel the chromatin at the endogenous il2 promoter (Figure 2B, 2C and 2D), indicating that this lysine 251 is required for proper DNA binding and transcriptional regulation in vivo. Furthermore, while CD4+ T cells transduced with wild-type mFOXP3 were able to suppress the proliferation of naïve T cells in an in vitro Treg assay (Figure 2E, black squares), CD4+ T cells expressing the delK251 mutant mFOXP3 exhibited a nearly complete loss of suppressive function (Figure 2E, gray circles). These results emphasize the functional significance of a single residue, K251, in FOXP3 homo-dimerization.

Acetylation of FOXP3 leucine zipper and its structural/functional consequences

Mass spectrometric studies of human FOXP3 (hFOXP3) revealed that lysine residues K250 and K252 (equivalent to mFOXP3 K249 and K251 respectively) are acetylated in vivo. Notably, hFOXP3 K250 (equivalent to mFOXP3 K249) is deleted in some IPEX patients; while K252 (equivalent to mFOXP3 K251) is represented by arginine in other members of the FOXP family.

As mentioned, mFOXP3 K249 and K251 mediate a network of interactions that are critical for dimerization. Specifically, the K249 side-chain forms an inter-subunit salt bridge with the side-chain of E242; while the side-chain of K251 helps to uniquely position or “hold” the side-chain of E248, a polar residue at the core (a position), so that it can form inter-subunit hydrogen bonding with the side-chain of Q234, also a polar residue at the core. The positive charges of both K249 and K251 are important for this dimer-stabilizing network. As predicted from the atomic structure, substitution of K249 and/or K251 with arginine, which maintains positive charges, would be permissive for dimerization of FOXP3 (Figure 3A). A K251D mutation that introduces negative charges disrupts FOXP3 homo-dimer. In addition, we found the K251D mutation also disrupts FOXP3-FOXP1 hetero-dimer formation (Figure 6). Acetylation of FOXP3 at K249 and/or K251 would neutralize the positive charges at these sites through which the FOXP3 dimer stability is modulated. Indeed, substitution of K251 with glutamine, which mimics acetylation, slightly reduces FOXP3 dimerization (Figure 3A).

Residues which guide Treg function

To explore how discrete lysines modify FOXP3 function, we transduced wild-type and mutant forms of mFOXP3 into conventional murine CD4+ T cells to gauge their suppressive functions. Mutation of lysines K249 and K251 to arginine exerted little effect on FOXP3-mediated Treg function, whereas substantial impairment of FOXP3 suppressive activity occurred with mutation to either aspartic acid (D) or glutamine (Q) (Figure 3B).

Supportive data were also observed using in vivo assays of the inhibitory effects of wild-type Tregs or T cells transduced with wild-type or mutant forms of mFOXP3 on the homeostatic proliferation of wild-type T cells. Transduction of T cells with either wild-type FOXP3 or mutant FOXP3 (K249/251R) led to comparable suppressive function to that of wild-type Tregs (p>0.05), whereas cells bearing the aspartic acid or glutamine mutations at these residues showed impaired Treg suppression (p<0.01) (Figure 3C). Since the lysines are post translationally modified, extrinsic signals appear to affect Treg cell function by charge alteration of the side-chains of K249 and K251.

Alteration of the FOXP3 acetylation pattern by discrete immune extrinsic signals

Various histone acetyltransferases could affect discrete lysine residues of FOXP3. p300 for example appears to target residues including but not limited to K250 and K252 of human FOXP3 (equivalent to mFOXP3 K249 and K251 respectively) (Figure 4). We initially used p300 and hFOXP3 co-transfected 293T cells in figure 4A. While lysines other than K250 or 252 may also be minimally acetylated, K250 and K252 are dominant but not the exclusive targets of the p300 HAT.

We then examined effects of extrinsic signals on this type of acetylation. CD3+ Jurkat T cells which contain several endogenous histone acetyltransferases including Tip60, CBP and p300 were transfected with wild type or the hFOXP3 K250/K252R double mutant constructs. The transfected Jurkat cells were then treated with TGF-β and/or IL-6. TGF-β treatment is known to increase the level of acetylated FOXP3 leading to enhanced chromatin binding of FOXP3 (Samanta et al., 2008) and this is also shown in figure 4(B and D). Strikingly, TGF-β treatment led to increased acetylation not only in wild type FOXP3, but also in K250 and K252 mutants such as K250/K252R double mutant cells, indicating that TGF-β treatment leads to FOXP3 acetylation at other lysine residues in addition to K250 and K252.

Therefore HATs other than p300 are likely to be induced by TGF-β and responsible for the acetylation of other sites. Interestingly, we found that iTreg inhibiting and Th17 polarizing signals, for example, TGF-β in combination with IL-6, limited acetylation of both wild type FOXP3 and the K250/K252R double mutant in transfected Jurkat cells (Figure 4C). Furthermore, deletion of K250, which would affect FOXP3 function as shown in figure 4, does not appear to alter the overall acetylation pattern induced by TGF-β. Taken together, these data indicate that even though FOXP3 acetylation is positively correlated with its activity as thought, TGF-β does not promote FOXP3 acetylation at sites K250 and K252, which are acetylated by p300. TGF-β promotes FOXP3a acetylation through residues other than K250 and K252, and acetylation of K250 and K252 is induced by extrinsic signals other than TGF-β.

Homo-clustering of FOXP3 dimers

The molecular surface of the FOXP3 coiled coil dimer is highly hydrophobic when judged by its unusual surface composition: 48% non-polar residues, in contrast with an average of 39% for general protein dimers (Reynolds et al., 2009). Consistently, crystal packing analysis revealed that the mFOXP3-ZL coiled coil dimer tends to form clusters by bundling up multiple α-helices from distinct dimers (Figure 5A). The extended hydrophobic stretch on the coiled coil surface facilitates this type of high order packing between dimers.

There are also a number of hydrophobic and electrostatic contacts between dimers (Figure 5A). The two neighboring α-helices from the two dimers pack symmetrically in an anti-parallel mode with residue A229 in the center. A large part of the hydrophobic stretch including residues L214, L222, K226, A229, L233, V237 and L247 were found on the dimer-dimer interface forming the cluster. In addition, residues E211, K215, Q218, E225, E236, S240, Q243, Q244, E248 and K251 contribute to electrostatic stabilization of the dimer-dimer packing in the cluster. The interface accessible surface area between the two adjacent dimers is ~800 Ų, which is comparable to that of the mFOXP3-ZL coiled coil dimer itself, representing a transient nature of the clustering interactions.

Dimer-based clustering of FOXP3 molecules might be an artifact caused by crystallization. We excluded this possibility by studying both mouse and human FOXP3’s clustering propensity in solution, by means of electron microscopy (EM) (Figure 5B) and also by cross-linking analysis using purified recombinant proteins (Figure 5C). While the FOXP3 sample appears as a single band on SDS-PAGE, cross-linking analyses revealed multiple levels of clustering with the dimer being a dominant species and small populations of species of higher order. Negative staining electron microscopy (Fowler and Erickson 1979) of maltose binding protein tagged human FOXP3 also revealed high ordered complexes of FOXP3 (Figure 5B arrow).

We had previously used gel filtration analyses followed by western blotting to show that full length FOXP3 was present as a mixture of monomers, dimers, tetramers and high-order oligomers (Li et al., 2007b). Jurkat cells stably transfected with human FOXP3 were treated with histone deacetylase inhibitors. Nuclear fractions were isolated and proteins fractionated by gel filtration (Figure S3). The FOXP3 complex was diminished in overall size from the 670 Kd range leading to a substantial fraction of the ensemble in the range ~158 kDa and less which tends to be smaller than the size of FOXP3 tetramer. HDACi treatment, via promoting FOXP3 acetylation, may lead to less homo-clustering of FOXP3, shifting the FOXP3 oligomerization status towards a dimer as well as loss of certain associated proteins.

Heterotypic interaction of FOXP3 with FOXP1

As shown by surface plasmon resonance, FOXP1 leucine zipper protein can form hetero-oligomers with FOXP3 leucine zipper protein (Figure 6A), consistent with our previous observation that FOXP1 co-localizes with FOXP3 at many sites within the nucleus of human CD4+CD25+ T cells. This hetero-association is disrupted by the IPEX mutation of FOXP3. Moreover, knockdown of endogenous FOXP1 in FOXP3-expressing Jurkat T cells relieves some FOXP3-mediated repression of IL-2 production (Li et al., 2007b).

Mutations (L241P, delK249 and K251D) that disrupt mFOXP3 homo-dimeric interactions also impaired FOXP3-FOXP1 hetero-association (Figure 6A and 6B). On the other hand, mutations that maintained the FOXP3 homo-association, including V238I on the dimer interface; L223D, K226R and A229C on the dimer surface, also preserved FOXP3-FOXP1 hetero-association (Figure 1F and 6B). Thus a similar interface is likely used to mediate both homo- and hetero- associations of FOXP3 and FOXP1. Furthermore, an arginine residue is found in FOXP1 at the position corresponding to mFOXP3 K251 that is important for the inter-subunit hydrogen bonding between Q234 and E248 of FOXP3 (Figure 1B and 1C). Replacement of lysine with arginine at 251 is predicted to constitutively maintain the electrostatic interactions with residues E248 and therefore would stabilize the FOXP1 homo-dimer.

Coiled coil mediated dimer is a regulatory structure important for FOXP3 function

Our studies revealed substantial structural instability and irregularities of the FOXP3 leucine zipper. For example, a high proportion of polar residues were found at the core packing (Figure 1A), and sizable holes were formed on the dimeric interface passing through the hydrophobic core (Figure 1D and 1E). Moreover, statistical analyses of accessible surface and empty space of the dimeric interface indicated transient and dynamic features of the FOXP3 homo-dimer.

The intact leucine zipper is required for DNA binding through the FOXP3 forkhead domain (Koh et al., 2009; Li et al., 2007b). Our current study explains at the atomic level how disruption of homo-dimerization abrogates FOXP3 function in vitro and in vivo. Based on its anti-parallel conformation, the FOXP3 coiled coil also plays a role in coordinating DNA binding with co-regulator recruitment (Figure S5). Since the FOXP3 coiled coil is not tightly assembled, we envision it to exist in a flexible conformation able to undergo tightening to form a stable dimer for high-affinity DNA binding upon differential posttranslational modifications, or loosening to break apart for hetero-association with FOXP1. Thus the plasticity of the FOXP3 coiled coil could provide a regulatory basis, through which the oligomerization status and stability of FOXP3 is controlled in a cell context “signaling” dependent manner.

Atomic explanation of IPEX disease-causing mutations on FOXP3 structure

Due to the 2-fold symmetrical conformation of the FOXP3 coiled coil, mutation of any single residue will affect two sites of the dimer, which therefore will enhance the biological consequence of the mutational changes caused by diseases. The structure of mFOXP3-ZL revealed that deletion of a single residue in the coiled coil would break the heptad-based symmetrical packing, resulting in mismatch of residues on the dimeric interface (Figure 1A). A network of interactions involving Q234, E242, E248, K249 and K251 has been identified to stabilize FOXP3 homo-dimerization (Figure 1C, 1D, 7A and 7B). Notably K251 of α-helix 1 forms an intra-subunit hydrogen bond with E248, holding E248 to further form an inter-subunit hydrogen bond with Q234 of α-helix 2 (Note that α-helices 1 and 2 form the mFOXP3 coiled coil). In addition, K249 of α-helix 1 directly forms an inter-subunit salt bridge with E242 of α-helix 2. These observations readily explain several reported deletions found in an IPEX patient (human FOXP3 DelK250 and DelE251, equivalent to mFOXP3 K249 and E250) (Chatila et al., 2000; Wildin et al., 2002). Deletions of these key residues may either directly break the E242-K249 inter-subunit salt bridge, or cause a positional shift and strain or break the Q234-E248 inter-subunit hydrogen bond through conformational “releasing” of E248 (Figure 7A and 7B). As a result, these deletion mutants will disrupt the homo-dimerization of FOXP3 (Figure 1F) and lead to impaired ability to repress transcription (Figure 2 and 3) (Li et al., 2007b; Ziegler, 2006).

Human FOXP3 L242P (equivalent to mFOXP3 L241P) is a novel mutation found in an IPEX patient (Gambineri et al., 2008; Zuo et al., 2007). It is worth noting that L241 of mFOXP3 is located at the core position of the coiled coil center, where the 2-fold axis passes along (Figure 1E and 7C). Variation at this site is expected to directly alter the coiled coil conformation and stability, and therefore affect the regulatory function of the FOXP3 homo-dimer. In the wild type mFOXP3 coiled coil, there is a hole near L241 and L245, indicative of a flexible nature of the homo-dimer. Substitution of L241 with a proline may not only distort the helical conformation, but also make the hole near L241 and L245 even larger. The smaller side-chain of proline would leave more empty space in the core packing, generating a groove (Figure 7D). One consequence of this packing is that water molecules may pass through the dimer interface causing instability (Figure 1F).

Modulation of FOXP3 structure and function by acetylation

Mass spectrometry and biochemical studies identified and confirmed human FOXP3 K250 and K252, i.e., K249 and K251 of mFOXP3, as acetylation sites in the coiled coil. Our current structure indicates that acetylation sites K251 and K249 are electrostatically involved in the interface network critical for FOXP3 dimerization (Figure 7). p300 appears to target these sites (K250 and K252 in human FOXP3) to acetylate FOXP3. Mutational analysis of these sites (K249-251D/Q/R) indicates that acetylation appears to be important for regulating the FOXP3 dimer formation and, suppressive activity. Recently von Loosdregt et al. suggested that p300 could promote the acetylation level of wild type human FOXP3 (van Loosdregt et al., 2010).

Acetylation of K249 and/or K251 appears to modify the same set of dimeric interactions affected by IPEX mutations. As discussed above, the key inter-subunit hydrogen bond between Q234 and E248 is dependent on the unique conformation of the E248 side-chain that is held by the side-chain of K251 (Figure 7A and 7E). Acetylation of K251 will neutralize the positive charges of its side-chain and, therefore, decrease its interaction with E248. In turn, the release of the E248 side-chain will break the inter-subunit hydrogen pairing between Q234 and E248, further releasing the side-chain of Q234 (Figure 7B and 7F). As a result, the steric tension caused by E248-L241 and Q234-M255 may further relax/destabilize the FOXP3 homo-dimerization (Figure 3A and 7F). Similarly, acetylation of K249 will directly decrease its inter-subunit pairing with E242, as observed for the IPEX mutant DelK249 (Figure 7G and 7H). Taken together, acetylation of K251 and/or K249 may impose structural and functional modifications on the FOXP3 molecules similar to the IPEX mutations, but in a regulated manner.

Considering the irregularity and instability of the FOXP3 coiled coil and the potential acetylation sites, we speculate that acetylation/deacetylation of the coiled coil regulatory structure possibly by TIP60 and HDAC7/HDAC9 may reversibly shift the FOXP3 molecule ensemble from non-stable homo-dimers to stable homo-dimers and hetero-dimers. From the in vitro experiment (Figure 3A), we observed that the flexibility of FOXP3 to have acetylation at K251 is optimal for the formation of dimer. When K251 is replaced by an acetylation-mimicking residue K251Q or the charged K251R, formation of dimer is reduced. A positively charged residue, however, is preferred at K249 for dimer formation, as no effect is observed for the K249R mutant.

Acetylation of FOXP3 by discrete signals may have contrasting effects on different lysines. FOXP3 acetylation has been shown to increase FOXP3 in vivo protein levels by lessening degradation (van Loosdregt et al., 2010). We found that acetylation at FOXP3 K249/251 may down-regulate its suppressive activity by causing dimer relaxation. However, TGF-β does not promote FOXP3 acetylation at K249 and K251 that are dominantly targeted by p300 (Figure 4). These data indicate that lysine modifications of K249 and K251 lead to “mechanistic adjustments” of the dimeric complex and then increase the stability of FOXP3 proteins but are not relevant to the type of the functional change induced by TGF-β and the acetylation pattern that cytokine stimulates. In fact other lysine residues undergo specific acetylation encouraged by TGF-β, and we propose that those modifications alone or in combination with other post translational modifications are responsible for increased affinity to chromatin and enhanced suppressive behavior.

Regulation of FOXP3 complex through homo-clustering and heterotypic interactions

Our crystallographic study identified a hydrophobic stretch on the molecular surface of the FOXP3 dimer that may promote homo-clustering (Figure S4). Further solution studies confirmed that FOXP3 may indeed form high molecular weight clusters (Li et al., 2007b)(Figure 5B and 5C). FOXP3 homo-clustering is diminished by mutations of the residues on the homo-dimer interface that disrupts homo-dimerization. Homo-clustering of human FOXP3 itself can be modified by HDAC inhibitor treatment (Figure S3), suggesting a physiological role in FOXP3 mediated suppression of higher order complex formation.

FOXP3 can interact with FOXP1 through the leucine zipper domain. Specifically, FOXP3 forms hetero-dimers with FOXP1 likely via an interface similar to that involved in FOXP3 homo-dimer formation (Figure 3). FOXP1-FOXP3 hetero-associations may compete with FOXP3 homo-dimerization and/or homo-clustering and the dynamic balance of discrete forms of FOXP3 complexes may directly affect its repressor activity.

Previously we reported that the FOXP3 coiled coil may associate with HEAT-SHOCK 70-KD PROTEIN 5 (HSPA5/BiP/GRP78), as well as with the linker histone H1.5 (Li et al., 2007c). Lately Su and colleagues found that histone H1.5 and FOXP3 interact with each other to cooperatively repress IL-2 transcription in human T cells (Mackey-Cushman et al., 2011). IPEX mutations found in the leucine zipper domain that disrupted FOXP3 homo-dimerization abolished FOXP3 binding to histone H1.5, indicating that hetero-association between H1.5 and FOXP3 is dependent on the coiled coil mediated homo-dimerization of FOXP3.

Summary

We have resolved certain of the atomic features of homo- and hetero- association of FOXP3. Clearly it is the homomeric FOXP3 dimer assembled with other co-regulators and in combination with discrete post translational modifications of discrete FOXP3 residues, that ultimately alters the conformation and activity of the FOXP3 protein complex to modulate Treg function. Appreciation of the structural features of the FOXP3 coiled coil explains the actions of many of the mutations involved in human diseases. FOXP3 are acetylated at several lysines, K8, K382, and K389 (Tao et al 2007, Samanta et al 2008) and K249 and K251 (this work), K262 and K267 (unpublished data). A recent study on acetylation sites of FOXP3, have found a role for K31 and K267 in the regulation of the suppressive activity of regulatory T cells (Kwon et al., 2012). A novel feature of the Kwon et al study that extends our thesis is that K31 and K262 may be targets of Sirtuin deacetylation. It is also worth noting that discret acetylation sites could be differentially modified by different HATs depending on distinct stimulations and cellular context, which may result in differential control of the FOXP3 molecules. Our current work has focused on the acetylation of only K249 and K251, and described an atomic explanation for FOXP3 acetylation at K249/251, which alters the stability of the dimeric complex of FOXP3 proteins.

Protein expression and purification

Proteins used in this study were all expressed in Escherichia coli and purified by Strep Tactin affinity column followed by gel filtration. All the mutants used for immunoprecipitation were made based on the full length FOXP3.

Crystallographic studies

Crystals were obtained by mixing equal volumes of sample solution (20mM Hepes, pH 7.3, 1mM TCEP, 0.5%Chaps and 100mM NaCl) and well buffer (100 mM NaAC pH 4.5 and ~1.8M MgSO4) for hanging drop. Crystals were cro-protected by adding 30% glycerol. Native and multi-wavelength anomalous dispersion (Selenium) data were collected at BL17U beam line, Shanghai Synchrotron Radiation Facility (Table S1). Structure determination and modeling were carried out with standard procedures as previously published (Brown et al., 2005). The atomic coordinates and structure factors have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code 3HSX).

Cell culture and immunoprecipitation

293T cells transfected with HA-FOXP3 or FLAG-FOXP3 were prepared for immunoprecipitation with anti-FLAG antibody, followed by Western Blot analysis using anti-HA antibody. The expression levels of the exogeneous FOXP3 proteins were also examined by Western Blot using anti-HA or anti-FLAG antibody.

Spectroscopy-LC-MS

10 μL of the IP pulldown solution was diluted with the running buffer of a SDS-PAGE gel (10% Tris-HEPES-SDS gel) for separation. The gel bands of interest were identified and isolated for trypsin digestion. After MS data collection, protein identification was achieved through computer aided searching of the Swiss-Prot human protein database using SEQUEST.

Biacore analysis

Binding experiments were performed with the surface plasmon resonance-based biosensor instrument Biacore 2000 (Biacore), at 25°C, essentially as described before (Berezov et al., 2002). Immobilization of FOXP3 on the sensor surface was performed following the standard amine coupling procedure according to manufacturer’s instructions.

suppression assay

In vitro assays of Treg function were performed using MACS-purified CFSE-labeled CD4+ CD25− T cells from normal C57BL/6 mice as effector cells, and normal CD4+ CD25+ Tregs or Foxp3-transduced cells as regulatory cells, with flow cytometric assessment of CFSE dilution at 72 h and data analysis using Flowjo 4.2 software (TreeStar) (Tao et al., 2007). The ability of Tregs to suppress the homeostatic proliferation of Thy1.1+ B6 CD4+CD25− T cells was assessed by the co-injection, i.v., of 10⁶ cells of each type into B6/RAG2−/− mice (Tao et al. 2007).