TRIM21 is an IgG receptor that is structurally, thermodynamically, and kinetically conserved

Abstract

The newly identified tripartite motif (TRIM) family of proteins mediate innate immunity and other critical cellular functions. Here we show that TRIM21, which mediates the autoimmune diseases rheumatoid arthritis, systemic lupus erythematosus, and Sjögren's syndrome, is a previously undescribed IgG receptor with a binding mechanism unlike known mammalian Fcγ receptors. TRIM21 simultaneously targets conserved hot-spot residues on both Ig domains of the Fc fragment using a PRYSPRY domain with a preformed multisite interface. The binding sites on both TRIM21 and Fc are highly conserved to the extent that the proteins are functionally interchangeable through murine, canine, primate, and human species. Pre-steady-state analysis exposes mechanistic conservation at the level of individual residues, which make the same energetic and kinetic contributions to binding despite varying in sequence. Together, our results reveal that TRIM21 is a previously undescribed type of IgG receptor based on a non-Ig scaffold whose interaction at the fundamental level—structural, thermodynamic, and kinetic—is evolutionarily conserved.

The recently identified TRIM family comprises at least 70 proteins with diverse cellular roles from differentiation and development to innate immune function (1, 2). The antiviral and disease-associated TRIMs are particularly important in human health. TRIM5α is an antiviral protein that restricts the replication of lentiviruses such as HIV by an as-yet-unidentified mechanism (3, 4). Defects in TRIM18 cause the congenital disorder Opitz syndrome (5), and TRIM21 mediates multiple autoimmune diseases (6). Despite the importance of TRIM proteins, very little is known about their molecular, kinetic, and thermodynamic basis of function.

TRIM proteins share a conserved multidomain architecture of RING, B box, and coiled coil that may encode a common signal integration or mechanistic step. The specificity of many TRIMs appears to be determined by a C-terminal PRYSPRY domain. Domain-exchange experiments on TRIM5α have shown that whereas the tripartite domains can be swapped with those of other TRIMs, the PRYSPRY domain must be conserved for viral restriction (7). Furthermore, polymorphisms in TRIM20(pyrin) that lead to familial Mediterranean fever (FMF) (8) or in TRIM18 cause Optiz syndrome cluster within the PRYSPRY (9). The central importance of the PRYSPRY domain in TRIM function has stimulated interest in identifying its targets and understanding its mechanism of function. Homologues of the PRYSPRY domain are found in ≈2,000 proteins and 11 families in the human genome, yet its role here is also poorly understood (10).

TRIM21 (Ro52) is a major autoantigen in diseases including rheumatoid arthritis, systemic lupus erythematosus, and Sjögren's syndrome (11–13). Autoantibodies against TRIM21 are known to form pathogenic immune complexes and are used to diagnose disease and monitor disease progression (14–16). Recently we showed that, in addition to being targeted by autoantibodies, TRIM21 binds normal serum IgG through its PRYSPRY domain (17). However, this activity is incongruous with the fact that TRIM21 is a cytoplasmic protein and is neither secreted nor membrane-displayed (6). Moreover, the epitope on IgG bound by TRIM21 is different from that of known mammalian Fcγ receptors (18). Fc receptor function is largely determined by the mechanism by which it binds IgG. The neonatal Fc receptor binds at acidic but not basic pH to allow IgG recycling (19, 20), whereas Fcγ receptors utilize different binding mechanisms to enable different subtype specificities (21). We attempted to determine whether IgG binding is a specific function of TRIM21 or a cross-reaction that drives autoimmune disease. We hypothesized that if IgG binding is critical to TRIM21 function then it should be tightly mechanistically conserved across mammalian species. To address this, we examined the interaction of TRIM21 and IgG in a range of species using combined crystallographic and kinetic analysis. We find that not only is TRIM21:IgG binding conserved but also that structurally equivalent residues in TRIM21 orthologs make nearly identical contributions to component rate constants even when different in primary sequence. Our results show that TRIM21 is a highly conserved mammalian Fc receptor that is structurally and mechanistically distinct from previously identified Fc receptors.

Results

TRIM21 Is a Highly Specific IgG Receptor.

To establish whether IgG is a specific target of TRIM21 we explored binding in orthologs by immunoprecipitation of transiently transfected cells. Transiently expressed mouse TRIM21 immunoprecipitated mouse IgG in all constructs except where the C-terminal PRYSPRY domain was deleted (data not shown). Isothermal titration calorimetry (ITC) on recombinantly expressed mouse TRIM21 PRYSPRY domain revealed binding to IgG with an affinity of 500 nM and a stoichiometry of 2:1 (TRIM21:IgG) [supporting information (SI) Fig. S1]. These values are comparable to binding of human IgG by human TRIM21; however, the human interaction occurs with a more negative enthalpy, suggesting possible differences in interaction (17). To determine the mechanism of TRIM21:IgG interaction we solved high-resolution crystal structures of free mouse TRIM21 PRYSPRY to 1.3 Å and its complex with mouse Fc to 2.0 Å (Fig. 1 and Table S1). The free structure contains two copies in the asymmetric unit that superpose with an rmsd of 0.1 Å for all residues except for the six N-terminal residues. The complex structure contains a single PRYSPRY domain interacting with the C_H2 and C_H3 domains of one Fc heavy chain (Fig. 1). In both structures, the PRYSPRY domain has a twisted β-sheet architecture with six binding site loops arranged around a convex β-sheet (Fig. 1). The free and bound structures of TRIM21 superpose closely with an overall Cα rmsd of 0.37 Å and an rmsd for the six variable loops of 0.68 Å (Fig. 1). The binding site residues that directly contact Fc are found in their binding-competent rotamer conformations in the prebound state. This suggests that binding to Fc occurs without an induced-fit component.

Fig. 1.

Comparison of free and bound mouse TRIM21 structures. (Left) Superposition of free (wheat) with bound (orange) PRYSPRY domain. Binding-site loops and residues making important contributions are annotated. (Right) Superposition of TRIM21:Fc complex (orange and red, respectively) with one Fc heavy chain from mouse IgG structure 1IGT (24) (pink). A superposition of the free structure is shown in yellow.

There are two potential sources of conformational diversity in protein recognition—structural rearrangement upon binding (induced fit) and stabilization of a binding conformer from a mix of preexisting isomers (preequilibrium) (22, 23). The presence of a preequilibrium step has been shown to indicate cross-reactive or nonspecific binding (22). Structural similarity between free and complexed crystal structures does not rule out preequilibrium diversity, because the act of crystallization tends to stabilize a single state. To address the possibility that TRIM21:IgG interaction is a result of preequilibrium diversity we followed TRIM21:Fc interaction in the pre-steady state by monitoring changes in intrinsic tryptophan fluorescence. Experiments with increasing concentrations of Fc yielded a single exponential phase whose rate is directly proportional to ligand concentration and which fits to a bimolecular association with a rate constant of 3.16 × 10⁶ M⁻¹s⁻¹ and dissociation rate constant too slow to measure (Fig. 2A). To rule out multiphasic dissociation kinetics and determine the off-rate we mixed a 10-fold molar excess of bacterial protein A with preformed mouse PRYSPRY:Fc complex and measured the increase in fluorescence associated with accumulating free PRYSPRY. Dissociation occurred in a single step with a k_off of 1.35 s⁻¹, which, when combined with k_on, gives a kinetic dissociation constant of 437 nM, which is in close agreement with the equilibrium measurements by ITC (Fig. 2B). Thus, the kinetic and crystallographic data show that the binding site for IgG on TRIM21 is preformed.

Fig. 2.

Pre-steady-state kinetics of mouse TRIM21:Fc interaction. (A) Stopped-flow spectrofluorimetry was used to measure the rapid fluorescence quench upon mouse TRIM21–mouse IgG association (see Materials and Methods for details). The data (red circles) fit to a single exponential (solid black line). Linear regression of the linear increase in k_obs (Inset) yields an association rate constant (k_on) of 3.16 × 10⁶ M⁻¹s⁻¹. (B) Measurement of the dissociation rate constant (k_off) for mouse TRIM21–mouse IgG binding in a stopped-flow experiment by competition with excess protein A. This yields a k_off of 1.35 s⁻¹ resulting in a kinetic K_d of 437 nM, close to that observed by ITC.

To address the conservation of the binding site on IgG for TRIM21, we compared the mouse TRIM21:Fc complex with the only available structure of a free mouse IgG, IgG2a (Protein Data Bank ID code 1IGT) (24). Despite some sequence differences, the two Fc fragments superpose remarkably closely. Furthermore, the sequences of important contact residues are the same, in particular residues 433–435, which contact TRIM21 hot-spot residues. The superposition shows that remarkably little rearrangement of the Fc is required for TRIM21 binding (Fig. 1). TRIM21 binds at the hinge interface between the C_H2 and C_H3 domains and interacts with residues from both Ig domains (Table S2), including five of the six consensus residues shown by DeLano et al. (25) to be important for C_H2 and C_H3 binders: 252, 253, 254, 434, 435, and 436. The hinge region is potentially dynamic, particularly in the relative orientations of the Ig domains; however, the same elbow angle (104°) is observed in the free mouse IgG as the complexed mouse TRIM21:Fc structure. Specific structural epitopes bound by TRIM21 do not undergo conformational change upon binding. The hot-spot loop 430–435 is able to insert into the PRYSPRY binding site and pack the side chains of residues H433, N434, and H435 with a tetrad of hot-spot residues in TRIM21 without structural alteration. Residues 433–435 are highly conserved in human isotypes, with the exception of an IgG3 allotype that contains an H435R mutation. There is more diversity between mouse isotypes in this region, particularly between IgG2a and IgG2b, which are H433K and H435Y with respect to IgG2a (26). The effect of these differences is difficult to predict based on the structure alone; however, they are relatively conservative and might be accommodated in the binding site. In particular, because H435 makes stacking interactions with F449 and Y451, mutation to tyrosine is likely to be permissive. The observed structural rigidity of the TRIM21 binding site on Fc correlates with the observed single-step pre-steady-state binding kinetics in which Fc flexibility would be expected to manifest as an additional exponential phase. The absence of a second relaxation event also rules out allosteric binding effects. Isometric binding is confirmed both by the ITC data, which shows that binding occurs with a stoichiometry of 2:1, and by the structure in which binding of TRIM21 is symmetrical: the dyad axis between the two Fc heavy chains is crystallographic with a single TRIM21 molecule and heavy chain per asymmetric unit. Together, this suggests that the binding site for TRIM21 on IgG is also preformed and hence neither IgG nor TRIM21 needs to undergo structural rearrangement to interact.

TRIM21:IgG Interaction Is Highly Conserved Across Mammalian Species.

To investigate the kinetic and energetic basis for TRIM21:IgG interaction and its conservation across mammalian species we performed three types of experiments. In one experiment we determined how the free energy of the TRIM21:IgG complex is partitioned between contact residues in both the human and mouse interfaces. We find that individual TRIM21 residues in the mouse and human complexes contribute a comparable fraction of the free energy of complex formation (Figs. 3 and 4). This suggests that the contribution of individual residues is evolutionarily conserved.

Fig. 3.

Hot-spot, electrostatic potential, and hydrophobicity conservation in the TRIM21 PRYSPRY binding site. Human (Left) is compared with mouse (Right). The molecular surface of the binding site is shown and colored on a red scale for ΔΔG, on a green scale for hydrophobicity, and on a red to blue (negative to positive) scale for electrostatic potential.

Fig. 4.

Effect of site-directed mutagenesis on cognate versus noncognate species TRIM21:IgG interaction. When the effects of mouse TRIM21 mutations on binding mouse IgG (cognate; x axis) are plotted against the effect on binding human IgG (noncognate; y axis) there exists a striking linear correlation (R² = 0.98) suggesting that the same interactions are used to stabilize both complexes.

In a second experiment we compared the kinetics and thermodynamics of cognate and noncognate complexes to demonstrate that this conservation preserves the mechanism of binding. We find not only that TRIM21 and IgG are functionally interchangeable between species (mouse TRIM21 binds human IgG as well as mouse IgG) but that there is a striking linear correlation between the effects of mouse TRIM21 mutations on binding to IgG from human (noncognate) and mouse (cognate) (Fig. 4, Table 1, and Fig. S2). The results of both experiments also reveal that energetic conservation of TRIM21:IgG interaction occurs at the level of component rate constants—meaning that not only the thermodynamic but also the kinetic landscape of the binding site is conserved.

Table 1.

Key hot-spot residues are conserved between human and mouse TRIM21 cognate and noncognate complexes

Position mutated on TRIM21	Effect on M21–mouse IgG, kcal/mol	Effect on M21–human IgG, kcal/mol	Effect on H21–human IgG, kcal/mol	Effect on H21–mouse IgG, kcal/mol
299	1.00	0.9	0.7	≥1.5
355	≥3.2	≥2.8	≥5	≥1.5
370	0.26	0.08	1.1	0.57
381	≥3.2	≥2.8	4	≥1.5
383	≥3.2	≥2.8	4	≥1.5
452/1	1.27	1.03	0.6	≥1.5

In a third experiment we determined how widely TRIM21:IgG interaction is conserved across mammalian species. We find that both human TRIM21 and mouse TRIM21 are able to bind to a wide range of species IgGs with K_d values of ≈10 μM or better, with the noncognate complexes in some cases (for example, human TRIM21–rhesus macaque IgG) binding as well as, if not better than, the cognate complex (Table 2). This suggests that the binding site for TRIM21 on IgG is evolutionarily maintained across mammals. As in the previous two experiments, we find that the conservation of binding energetics occurs at the level of thermodynamics and kinetics. This mechanistic conservation occurs despite sequence variation in TRIM21 and IgG between species.

Table 2.

Binding kinetics of human and mouse TRIM21 to a range of mammalian IgGs

Complex	kon (×106), M−1·s−1	koff, s−1	Kd, M
H21–human IgG	3.5 ± 0.3	0.46 ± 0.01	1.3 ± 0.15 × 10−7
H21–canine IgG	1.54 ± 0.01	0.25 ± 0.05	1.6 ± 0.4 × 10−7
H21–guinea pig IgG	0.73 ± 0.02	10.43 ± 0.2	1.4 ± 0.07 × 10−5
H21–monkey IgG	2.7 ± 0.06	0.24 ± 0	8.9 ± 0.2 × 10−8
H21–mouse IgG	0.18 ± 0.01	1.3	7.2 × 10−6
H21–rat IgG	0.8 ± 0.02	1.33 ± 0.1	1.7 ± 0.02 × 10−6
M21–mouse IgG	3.16	1.35	4.3 × 10−7
M21–canine IgG	2.9 ± 0.1	0.5 ± 0.01	1.7 ± 0 × 10−7
M21–guinea pig IgG	2.9 ± 0.1	4.33 ± 0.02	1.5 ± 0 × 10−6
M21–human IgG	2.04	1.92	9.4 × 10−7
M21–monkey IgG	1.9 ± 0.1	1.1 ± 0.01	5.8 ± 0 × 10−7
M21–rat IgG	2.3 ± 0.1	2.77 ± 0.02	1.2 ± 0 × 10−6

Stopped-flow fluorescence was used to measure the association and dissociation kinetics of a range of mammalian IgGs to both human and mouse TRIM21 (see Materials and Methods for details).

Molecular Mechanism of TRIM21:IgG Interaction.

Sequence identity between human and mouse TRIM21 is ≈70%, and there is significant diversity in the PRYSPRY domain (Fig. 5). Human anti-TRIM21 autoantibodies do not recognize endogenous TRIM21 in the cells of rodents or other nonprimate species. In contrast, other human autoantibodies react with ortholog autoantigens from nonhuman species. To determine the structural basis for TRIM21:IgG interaction we compared the mouse TRIM21:Fc complex to its human counterpart. Superposition of the two structures reveals that the overall topology of the PRYSPRY binding site is highly conserved (rmsd of 1.5 Å for Cα atoms) (Fig. 6). Furthermore, both the charge distribution and surface hydrophobicity of the interface are closely maintained (Fig. 3). In particular, a conserved ring of hydrophobic residues shields from solvent the crucial interaction between D355 and Fc loop 430–435. The surface around the D355 interaction has a conserved patch of negative charge that is complementary to the histidine-containing positively charged loop 430–435.

Fig. 5.

Sequence alignment of mouse and human TRIM21. Identical residues are shaded gray. The positions of the variable binding loops are shaded purple.

Fig. 6.

Comparison of mouse and human TRIM21. (Left) Superposition of mouse (orange) and human (yellow) PRYSPRY domains. Where the sequences diverge the annotations are formatted so that the first residue is for human and the second is for mouse; e.g., D/S326 indicates an aspartic acid in human and a serine in mouse. (Right) Superposition of mouse (red and orange) and human (pink and yellow) complexes with Fc.

Conservation between mouse and human PRYSPRY extends to the conformations of the six binding site loops and to the rotamers of interacting side chains (Fig. 6). This structural homology is maintained despite differences in the composition of the binding loops. Sequence differences occur in the two most important loops of the binding site—VL4 and VL6. The VL4 loop in human TRIM21 contains two essential hot-spot residues—W380 and W382—that interact with the two histidine hot-spot residues within Fc loop 430–435. The mouse VL4 loop contains both a deletion and a mutation with respect to human. The deletion is accommodated by shortening of the loop, which is tightly constrained at one end by β-zipper interactions and at the other by Y386, whose side chain forms a hydrophobic core with I384, I399, and I410. The loop shortening allows the position of hot-spot residues W380 and W382 to be maintained in mouse TRIM21. The effect on the energetics of binding (ΔΔG_binding) of mutation of these residues to alanine is sufficiently large (>3 kcal/mol) to effectively abolish binding, a feature conserved with human TRIM21 (Fig. 4 and Table 1).

The VL6 loop is the most variable between TRIM proteins, and in TRIM21 there are four differences between mouse and human. These differences create substantially different loop conformations, with a relative variation of >5 Å. Two of these differences, positions 451 and 452, are important Fc-contacting residues. Position 451 is an aspartic acid in human and a tyrosine in mouse, and these residues are located >4 Å apart in their respective binding sites (Fig. 3). On the basis of the structure these residues would be expected to contribute differently to the energetics of binding. However, mutation to alanine results in a similar ΔΔG of binding in both species (Fig. 4 and Table 1). Furthermore, the alanine mutants reveal that the aspartic acid in human and the tyrosine in mouse encode similar k_on and k_off rates. To confirm this finding we introduced the aspartic acid into mouse TRIM21. As predicted, the mutant maintained the same thermodynamics and kinetics of binding (Fig. S2).

Discussion

TRIM21 Is a Previously Undescribed Mammalian Fc Receptor.

We have found that TRIM21 is a previously undescribed type of widely conserved mammalian Fc binding protein. TRIM21 interacts with IgG specifically, displaying high affinity and a binding mechanism that is highly conserved. The structural characteristics of binding reflect an evolved rather than adventitious interaction. This is in contrast to IgG complexes formed by autoantigens such as rheumatoid factor, in which binding occurs with poor complementarity and low affinity (27). Specific interaction of TRIM21 with IgG is therefore independent of its targeting by autoantibodies during autoimmunity and does not itself explain pathogenic immune complex formation. TRIM21 is a previously undescribed mammalian Fc receptor, structurally and thermodynamically distinct from known Fc receptors. All previously identified mammalian Fc receptors are based on the Ig superfamily and related MHC class I-like scaffold. In TRIM21, a C-terminal PRYSPRY domain with a β-sandwich-like architecture mediates binding to IgG. Four PRYSPRY hot-spot residues supply most of the binding energy for IgG interaction. However, it is not the case that the remaining interaction residues are less important, as evidenced by the fact that the contribution of all interface residues is tightly evolutionarily conserved, both energetically and kinetically. This reflects the biological context of protein:protein interactions where the difference between function or no function depends on only a small change in binding. For instance, successful interaction between antigen and the B cell receptor requires a minimum affinity of 10 μM, a 1.5-fold drop in affinity resulting in no B cell triggering (28). Our data show that the contribution of structurally equivalent interaction residues in the TRIM21 binding site is maintained even when the primary sequence is different. This illustrates that mechanistic conservation does not require concomitant sequence conservation.

TRIM21 in Autoimmune Disease.

Our discovery that TRIM21 is a high-affinity receptor for IgG is highly significant for autoimmune pathology. During autoimmunity, anti-TRIM21 autoantibodies are generated against epitopes in the RING, B box, and coiled-coil domains. Intriguingly, there are no reports of autoantibodies to epitopes in the PRYSPRY domain. This suggests that PRYSPRY interaction with Fc takes place during autoimmunity and blocks Fab targeting. We predict that anti-TRIM21 autoantibodies are likely to undergo “autoantibody bipolar bridging” or the simultaneous binding of TRIM21 to both Fab and Fc domains. The resulting four binding sites per autoantibody combined with trimeric full-length TRIM21 would allow for multivalent binding and the creation of large highly cross-linked immune complexes that are potent Fcγ receptor activators. Alternatively, TRIM21:autoantibody bridged complexes may be so large and cross-linked that binding to Fcγ receptors and complement is blocked. In bacterial pathogenesis, bipolar bridging of antibodies by superantigens such as gE-gI and protein A is presumed to block receptor binding, disconnecting antibody recognition from effector function (29). This type of block could have a number of serious deleterious effects on autoimmune pathogenesis. Preventing complement binding and hence immune complex recycling directly affects the persistence and deposition of immune complex, which is a classic outcome of autoimmunity.

TRIM21 Function.

There are several hypotheses for the role of TRIM21:IgG interaction. One hypothesis is that TRIM21 mediates clearance of apoptosed cells by opsonizing them in antibody and recruiting macrophage phagocytic function (17). This is supported by the fact that TRIM21 migrates to the periphery of cells undergoing apoptosis (30). Expression of TRIM21 is up-regulated by IFN-γ (6), which suggests that TRIM21-assisted clearance may be specifically used during inflammation or tissue damage. A second hypothesis is that TRIM21 is involved in targeting unfolded IgG made by B cells for proteasomal degradation (31). This provides a role for the tripartite motif domains, which are thought to have a conserved ubiquitination role in TRIM function, but does not explain why TRIM21 is expressed at similar levels in other cell types.

Our data provide some clues to TRIM21 function and allow these hypotheses to be assessed. Binding of TRIM21 is entirely symmetrical with a 2:1 stoichiometry, suggesting that it can be cross-linked by free antibody and that TRIM21 is unlikely to function as a trigger receptor like the Fcγ receptors. Because TRIM21 is trimeric, it has the capacity to cross-link IgG and stimulate B cells. The ability to cross-link the B cell receptor could explain why TRIM21 is a dominant autoantigen and how TRIM21-mediated autoimmunity is initiated, for example by stimulating T cell-independent B cell activation.

TRIM21 binds to all subclasses of IgG including IgG3 (Table S3). This suggests that TRIM21 interaction is not linked to a particular effector function or arm of the immune response and would be consistent with a general role in intracellular IgG degradation. In contrast, Fcγ receptors have specific patterns of IgG subclass specificity. The neonatal Fc receptor (FcRn) is the only mammalian receptor known to bind IgG intracellularly. FcRn mediates the trafficking of internalized IgG through a binding mechanism that permits interaction inside the acidic environment of intracellular vesicles but not the mildly basic pH of the cell. In rat FcRn this pH dependence results, in part, from interaction with two titratable histidine residues (H433 and H435) in the C_H3 domain whose deprotonation disrupts salt bridge formation (19, 32, 33). The role of these residues changes between species and isotypes and in human and mouse IgG1 H433 appears to have no function (26). The interaction of pathogen superantigen HSV gE-gI with Fc is also pH-dependent but in the opposite direction—binding taking place at basic but not acidic pH—as a result of targeting the same two histidines, H433 and H435 (29, 34). TRIM21 also binds H433 and H435 (in fact these interactions are hot spots) but does not display pH dependence, and its affinity is not reduced by protonation or deprotonation. The fact that TRIM21:IgG binding is not pH-dependent suggests that it does not have an FcRn-like recycling function. Instead of pH dependence, TRIM21 binding is salt-dependent: increasing salt concentration from 20 mM to 200 mM results in a 5-fold drop in affinity. This behavior is a result of hydrophobic stacking interactions and hydrogen bonds with H433 and H435 that take place at the center of an interface surrounded by a solvent-excluding seal of hydrophobic residues.

The three Fc-binding mechanisms displayed by FcRn, gE-gI, and TRIM21 (acid dependence, basic dependence, and salt dependence) produced by interaction with the same two histidine residues provide a remarkable example of how a binding mechanism can be exquisitely tailored to give a desired cellular outcome. In the context of IgG receptors, the utility of the two histidines, H433 and H435, provides one explanation for the observation by DeLano et al. (35) that multiple binding solutions exist to this region of IgG. In a general context, the three binding mechanisms illustrate the versatility of protein structure. We have previously commented that protein binding sites are intrinsically promiscuously active as result of their chemically heterogeneous nature (36). The present example illustrates how individual residues provide remarkable chemical and hence functional heterogeneity.

Materials and Methods

Protein Purification and Preparation.

Mouse and human TRIM21 PRYSPRY domains were overexpressed and purified by using Ni²⁺-NTA resin (Qiagen) followed by Superdex 75 gel filtration as previously described. Mouse serum IgG (Equitech Bio) composed predominantly of IgG2 isotype was used in all biophysical and structural experiments. IgG2a is the most prevalent form of IgG2 and can comprise ≈80% in serum (37). There are isotypic differences between IgG2a and IgG2b, including at position 435, which is bound by TRIM21. It is not possible from our data to assess the effect of these differences on TRIM21 binding. The absence of additional phases in the binding kinetics suggests that either the kinetic rate constants are largely unaffected by these differences or binding is not observed (because of poor affinity or poor signal). It is likely, however, that nonbinding species represent only a small proportion of the serum IgG because titration experiments with TRIM21 confirm that most protein is bound at concentrations over the K_d. Mouse Fc was prepared from whole IgG by papain cleavage and then purified on protein A Sepharose (Amersham Pharmacia Biosciences) followed by Superdex S200 gel filtration (Amersham Pharmacia Biosciences) (38). Human IgG and subtypes were from Serotec and Athens Research and Technology. Pooled serum IgG composed of >90% IgG1 was used in all biophysical experiments. TRIM21 binding yielded single-phase kinetics to serum IgG, although we cannot rule out the possibility that allotypes present in trace amounts bind with more complex kinetics.

Site-Directed Mutagenesis.

TRIM21 alanine mutants were produced by QuikChange mutagenesis (Stratagene) and purified as described above.

Rapid Reaction Kinetics.

Experiments were carried out in 20 mM potassium phosphate (pH 8) and 125 mM NaCl. Stopped-flow experiments were carried out in an Applied Photophysics π-Star stopped-flow spectrofluorimeter in 1:1 mixing mode and using a 320-nm cutoff filter essentially as described previously (23). The resulting fluorescence quench that occurs upon association of TRIM21 with IgG was fitted to F = ΔF exp(−k_obst) + F_e, where F is the observed fluorescence, ΔF is the fluorescence amplitude, k_obs is the observed pseudo first-order rate constant, and F_e is the end-point fluorescence. The bimolecular association rate constant (k_on) was determined by fitting the linear relationship between k_obs and the increasing pseudofirst-order concentrations of IgG to k_obs = k_on(IgG binding sites) + k_reverse. The dissociation rate constant (k_off) was measured by chasing excess protein A into a preformed TRIM21:IgG complex and fitting the resulting fluorescence enhancement to the fluorescence equation given above. Protein A binding to Fc occurs without a change in fluorescence as shown in Fig. S3.

Titration Calorimetry.

Proteins were dialyzed into 20 mM potassium phosphate (pH 8) and 125 mM NaCl overnight at 4°C, and experiments were carried out on a Microcal VP-ITC as described previously (17).

Crystallography.

Mouse TRIM21-mouse Fc complex was purified by mixing the components in a 2:1 ratio and applying to a Superdex 200 gel filtration column. Free mouse TRIM21 was used without further purification. Crystals were observed in many conditions, but the best were seen in 0.1 M Mes (pH 6.5) with 0.2 M magnesium acetate and 20% PEG8000 for the free mouse TRIM21 and 0.1 M sodium citrate (pH 5.5) with 0.1 M magnesium chloride, 0.1 M sodium chloride, and 26% PEG400 for the mouse TRIM21–mouse Fc. Data for the free mouse TRIM21 PRYSPRY structure were collected at the European Synchrotron Radiation Facility beamline ID14-3, whereas data for the mouse complex were collected at beamline ID23-1. Crystallographic analysis was performed by using programs from the CCP4 suite (39). All data were indexed in MOSFLM and scaled in SCALA. The structure of human TRIM21 PRYSPRY [Protein Data Bank ID code 2IWG (17)] was used as a search model in PHASER. Structures were refined in REFMAC and Coot. Structural figures were created by using Pymol (40) and Chimera (41).