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

Anthony H. Keeble, Zahra Khan, Alan Forster, and Leo C. James*

Division of and Nucleic Acid Chemistry, Laboratory of Molecular Biology, Medical Research Council, Hills Road, Cambridge CB2 2QH, United Kingdom

Edited by David R. Davies, National Institutes of Health, Bethesda, MD, and approved February 14, 2008 (received for review January 8, 2008) The newly identified tripartite motif (TRIM) family of membrane-displayed (6). Moreover, the epitope on IgG bound by mediate innate immunity and other critical cellular functions. Here TRIM21 is different from that of known mammalian Fc␥ receptors we show that TRIM21, which mediates the autoimmune diseases (18). Fc receptor function is largely determined by the mechanism rheumatoid arthritis, systemic erythematosus, and Sjo¨gren’s by which it binds IgG. The neonatal Fc receptor binds at acidic but syndrome, is a previously undescribed IgG receptor with a binding not basic pH to allow IgG recycling (19, 20), whereas Fc␥ receptors mechanism unlike known mammalian Fc␥ receptors. TRIM21 si- utilize different binding mechanisms to enable different subtype multaneously targets conserved hot-spot residues on both Ig specificities (21). We attempted to determine whether IgG binding domains of the Fc fragment using a PRYSPRY domain with a is a specific function of TRIM21 or a cross-reaction that drives preformed multisite interface. The binding sites on both TRIM21 autoimmune disease. We hypothesized that if IgG binding is critical and Fc are highly conserved to the extent that the proteins are to TRIM21 function then it should be tightly mechanistically functionally interchangeable through murine, canine, primate, and conserved across mammalian species. To address this, we examined human species. Pre-steady-state analysis exposes mechanistic con- the interaction of TRIM21 and IgG in a range of species using servation at the level of individual residues, which make the same combined crystallographic and kinetic analysis. We find that not energetic and kinetic contributions to binding despite varying in only is TRIM21:IgG binding conserved but also that structurally sequence. Together, our results reveal that TRIM21 is a previously equivalent residues in TRIM21 orthologs make nearly identical undescribed type of IgG receptor based on a non-Ig scaffold whose contributions to component rate constants even when different in interaction at the fundamental level—structural, thermodynamic, primary sequence. Our results show that TRIM21 is a highly and kinetic—is evolutionarily conserved. conserved mammalian Fc receptor that is structurally and mecha- nistically distinct from previously identified Fc receptors. PRYSPRY ͉ systemic lupus erythematosus ͉ TRIM5␣ ͉ Ro52 Results he recently identified TRIM family comprises at least 70 TRIM21 Is a Highly Specific IgG Receptor. To establish whether IgG is a specific target of TRIM21 we explored binding in orthologs by proteins with diverse cellular roles from differentiation and T immunoprecipitation of transiently transfected cells. Transiently development to innate immune function (1, 2). The antiviral and expressed mouse TRIM21 immunoprecipitated mouse IgG in all disease-associated TRIMs are particularly important in human constructs except where the C-terminal PRYSPRY domain was health. TRIM5␣ is an antiviral protein that restricts the replication deleted (data not shown). Isothermal titration calorimetry (ITC) on of lentiviruses such as HIV by an as-yet-unidentified mechanism (3, recombinantly expressed mouse TRIM21 PRYSPRY domain re- 4). Defects in TRIM18 cause the congenital disorder Opitz syn- vealed binding to IgG with an affinity of 500 nM and a stoichiom- drome (5), and TRIM21 mediates multiple autoimmune diseases etry of 2:1 (TRIM21:IgG) [supporting information (SI) Fig. S1]. (6). Despite the importance of TRIM proteins, very little is known These values are comparable to binding of human IgG by human about their molecular, kinetic, and thermodynamic basis of TRIM21; however, the human interaction occurs with a more function. negative enthalpy, suggesting possible differences in interaction TRIM proteins share a conserved multidomain architecture of (17). To determine the mechanism of TRIM21:IgG interaction we RING, B box, and that may encode a common signal solved high-resolution crystal structures of free mouse TRIM21 integration or mechanistic step. The specificity of many TRIMs PRYSPRY to 1.3 Å and its complex with mouse Fc to 2.0 Å (Fig. appears to be determined by a C-terminal PRYSPRY domain. ␣ 1 and Table S1). The free structure contains two copies in the Domain-exchange experiments on TRIM5 have shown that asymmetric unit that superpose with an rmsd of 0.1 Å for all whereas the tripartite domains can be swapped with those of other residues except for the six N-terminal residues. The complex TRIMs, the PRYSPRY domain must be conserved for viral structure contains a single PRYSPRY domain interacting with the restriction (7). Furthermore, polymorphisms in TRIM20(pyrin) CH2 and CH3 domains of one Fc heavy chain (Fig. 1). In both that lead to familial Mediterranean fever (FMF) (8) or in TRIM18 structures, the PRYSPRY domain has a twisted ␤-sheet architec- cause Optiz syndrome cluster within the PRYSPRY (9). The ture with six binding site loops arranged around a convex ␤-sheet central importance of the PRYSPRY domain in TRIM function (Fig. 1). The free and bound structures of TRIM21 superpose has stimulated interest in identifying its targets and understanding closely with an overall C␣ rmsd of 0.37 Å and an rmsd for the six its mechanism of function. Homologues of the PRYSPRY domain are found in Ϸ2,000 proteins and 11 families in the , yet its role here is also poorly understood (10). Author contributions: L.C.J. designed research; A.H.K., Z.K., A.F., and L.C.J. performed TRIM21 (Ro52) is a major autoantigen in diseases including research; L.C.J. analyzed data; and L.C.J. wrote the paper. rheumatoid arthritis, systemic lupus erythematosus, and Sjo¨gren’s The authors declare no conflict of interest. syndrome (11–13). Autoantibodies against TRIM21 are known to This article is a PNAS Direct Submission. form pathogenic immune complexes and are used to diagnose Data deposition: The atomic coordinates have been deposited in the , disease and monitor disease progression (14–16). Recently we www.pdb.org (PDB ID codes 2VOL and 2VOK). showed that, in addition to being targeted by autoantibodies, *To whom correspondence should be addressed. E-mail: [email protected]. TRIM21 binds normal serum IgG through its PRYSPRY domain This article contains supporting information online at www.pnas.org/cgi/data/0800159105/ BIOPHYSICS (17). However, this activity is incongruous with the fact that DCSupplemental. TRIM21 is a cytoplasmic protein and is neither secreted nor © 2008 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800159105 PNAS ͉ April 22, 2008 ͉ vol. 105 ͉ no. 16 ͉ 6045–6050 Downloaded by guest on September 27, 2021 VL3 C 2 TRIM21 VL1 H VL5

VL4 Fig. 1. Comparison of free and bound mouse TRIM21 VL2 VL6 C 3 H structures. (Left) Superposition of free (wheat) with bound (orange) PRYSPRY domain. Binding-site loops and residues making important contributions are an- notated. (Right) Superposition of TRIM21:Fc complex (orange and red, respectively) with one Fc heavy chain Complexed Free Free Fc Free Complexed from mouse IgG structure 1IGT (24) (pink). A superpo- (2.0 Å) (1.6 Å) (1IGT) TR21 sition of the free structure is shown in yellow.

variable loops of 0.68 Å (Fig. 1). The binding site residues that site and pack the side chains of residues H433, N434, and H435 with directly contact Fc are found in their binding-competent rotamer a tetrad of hot-spot residues in TRIM21 without structural alter- conformations in the prebound state. This suggests that binding to ation. Residues 433–435 are highly conserved in human isotypes, Fc occurs without an induced-fit component. with the exception of an IgG3 allotype that contains an H435R There are two potential sources of conformational diversity in mutation. There is more diversity between mouse isotypes in this protein recognition—structural rearrangement upon binding (in- region, particularly between IgG2a and IgG2b, which are H433K duced fit) and stabilization of a binding conformer from a mix of and H435Y with respect to IgG2a (26). The effect of these preexisting isomers (preequilibrium) (22, 23). The presence of a differences is difficult to predict based on the structure alone; preequilibrium step has been shown to indicate cross-reactive or however, they are relatively conservative and might be accommo- nonspecific binding (22). Structural similarity between free and dated in the binding site. In particular, because H435 makes complexed crystal structures does not rule out preequilibrium stacking interactions with F449 and Y451, mutation to tyrosine is diversity, because the act of crystallization tends to stabilize a single likely to be permissive. The observed structural rigidity of the state. To address the possibility that TRIM21:IgG interaction is a TRIM21 binding site on Fc correlates with the observed single-step result of preequilibrium diversity we followed TRIM21:Fc interac- pre-steady-state binding kinetics in which Fc flexibility would be tion in the pre-steady state by monitoring changes in intrinsic expected to manifest as an additional exponential phase. The tryptophan fluorescence. Experiments with increasing concentra- absence of a second relaxation event also rules out allosteric binding tions of Fc yielded a single exponential phase whose rate is directly effects. Isometric binding is confirmed both by the ITC data, which proportional to ligand concentration and which fits to a bimolecular shows that binding occurs with a stoichiometry of 2:1, and by the association with a rate constant of 3.16 ϫ 106 MϪ1sϪ1 and disso- structure in which binding of TRIM21 is symmetrical: the dyad axis ciation rate constant too slow to measure (Fig. 2A). To rule out between the two Fc heavy chains is crystallographic with a single multiphasic dissociation kinetics and determine the off-rate we TRIM21 molecule and heavy chain per asymmetric unit. Together, mixed a 10-fold molar excess of bacterial protein A with preformed this suggests that the binding site for TRIM21 on IgG is also mouse PRYSPRY:Fc complex and measured the increase in preformed and hence neither IgG nor TRIM21 needs to undergo fluorescence associated with accumulating free PRYSPRY. Dis- structural rearrangement to interact. Ϫ1 sociation occurred in a single step with a koff of 1.35 s , which, when combined with kon, gives a kinetic dissociation constant of 437 TRIM21:IgG Interaction Is Highly Conserved Across Mammalian Spe- nM, which is in close agreement with the equilibrium measurements cies. To investigate the kinetic and energetic basis for TRIM21:IgG by ITC (Fig. 2B). Thus, the kinetic and crystallographic data show interaction and its conservation across mammalian species we that the binding site for IgG on TRIM21 is preformed. performed three types of experiments. In one experiment we To address the conservation of the binding site on IgG for determined how the free energy of the TRIM21:IgG complex is TRIM21, we compared the mouse TRIM21:Fc complex with the partitioned between contact residues in both the human and mouse only available structure of a free mouse IgG, IgG2a (Protein Data interfaces. We find that individual TRIM21 residues in the mouse Bank ID code 1IGT) (24). Despite some sequence differences, the and human complexes contribute a comparable fraction of the free two Fc fragments superpose remarkably closely. Furthermore, the energy of complex formation (Figs. 3 and 4). This suggests that the sequences of important contact residues are the same, in particular contribution of individual residues is evolutionarily conserved. residues 433–435, which contact TRIM21 hot-spot residues. The In a second experiment we compared the kinetics and thermo- superposition shows that remarkably little rearrangement of the Fc dynamics of cognate and noncognate complexes to demonstrate is required for TRIM21 binding (Fig. 1). TRIM21 binds at the hinge that this conservation preserves the mechanism of binding. We find interface between the CH2 and CH3 domains and interacts with not only that TRIM21 and IgG are functionally interchangeable residues from both Ig domains (Table S2), including five of the six between species (mouse TRIM21 binds human IgG as well as consensus residues shown by DeLano et al. (25) to be important for mouse IgG) but that there is a striking linear correlation between CH2 and CH3 binders: 252, 253, 254, 434, 435, and 436. The hinge the effects of mouse TRIM21 mutations on binding to IgG from region is potentially dynamic, particularly in the relative orienta- human (noncognate) and mouse (cognate) (Fig. 4, Table 1, and Fig. tions of the Ig domains; however, the same elbow angle (104°) is S2). The results of both experiments also reveal that energetic observed in the free mouse IgG as the complexed mouse conservation of TRIM21:IgG interaction occurs at the level of TRIM21:Fc structure. Specific structural epitopes bound by component rate constants—meaning that not only the thermody- TRIM21 do not undergo conformational change upon binding. The namic but also the kinetic landscape of the binding site is conserved. hot-spot loop 430–435 is able to insert into the PRYSPRY binding In a third experiment we determined how widely TRIM21:IgG

6046 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800159105 Keeble et al. Downloaded by guest on September 27, 2021 Human Murine

L370 S370 S326

D355 Y328 D355 W381 W380 R324

W383 F450 W382 W299 W299 D452 Y451

∆∆G (kcal/mol) ≤ 0.5 0.51 - 1.0 3.0 ≥ 3.1

Kyte-Doolittle Hydrophobicity Score -2.0 -1.9 - 0.0 0.1 - 1.0 1.0 - 2.0

Electrostatics -10 0 +10 kbT/e c

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 ⌬⌬G, on a green scale for hydrophobicity, and on a red to blue (negative to positive) scale for electrostatic potential.

Fig. 2. Pre-steady-state kinetics of mouse TRIM21:Fc interaction. (A) tures reveals that the overall topology of the PRYSPRY binding Stopped-flow spectrofluorimetry was used to measure the rapid fluorescence site is highly conserved (rmsd of 1.5 Å for C␣ atoms) (Fig. 6). quench upon mouse TRIM21–mouse IgG association (see Materials and Meth- Furthermore, both the charge distribution and surface hydropho- ods for details). The data (red circles) fit to a single exponential (solid black bicity of the interface are closely maintained (Fig. 3). In particular, line). Linear regression of the linear increase in kobs (Inset) yields an association 6 Ϫ1 Ϫ1 a conserved ring of hydrophobic residues shields from solvent the rate constant (kon) of 3.16 ϫ 10 M s .(B) Measurement of the dissociation rate constant (koff) for mouse TRIM21–mouse IgG binding in a stopped-flow crucial interaction between D355 and Fc loop 430–435. The surface Ϫ1 experiment by competition with excess protein A. This yields a koff of 1.35 s around the D355 interaction has a conserved patch of negative resulting in a kinetic Kd of 437 nM, close to that observed by ITC. charge that is complementary to the histidine-containing positively charged loop 430–435. Conservation between mouse and human PRYSPRY extends to interaction is conserved across mammalian species. We find that the conformations of the six binding site loops and to the rotamers both human TRIM21 and mouse TRIM21 are able to bind to a of interacting side chains (Fig. 6). This structural homology is Ϸ ␮ wide range of species IgGs with Kd values of 10 M or better, with maintained despite differences in the composition of the binding the noncognate complexes in some cases (for example, human loops. Sequence differences occur in the two most important loops TRIM21–rhesus macaque IgG) binding as well as, if not better than, of the binding site—VL4 and VL6. The VL4 loop in human the cognate complex (Table 2). This suggests that the binding site TRIM21 contains two essential hot-spot residues—W380 and for TRIM21 on IgG is evolutionarily maintained across mammals. W382—that interact with the two histidine hot-spot residues within As in the previous two experiments, we find that the conservation Fc loop 430–435. The mouse VL4 loop contains both a deletion and of binding energetics occurs at the level of thermodynamics and a mutation with respect to human. The deletion is accommodated kinetics. This mechanistic conservation occurs despite sequence by shortening of the loop, which is tightly constrained at one end by variation in TRIM21 and IgG between species. ␤-zipper interactions and at the other by Y386, whose side chain forms a hydrophobic core with I384, I399, and I410. The loop Molecular Mechanism of TRIM21:IgG Interaction. Sequence identity shortening allows the position of hot-spot residues W380 and W382 between human and mouse TRIM21 is Ϸ70%, and there is to be maintained in mouse TRIM21. The effect on the energetics significant diversity in the PRYSPRY domain (Fig. 5). Human of binding (⌬⌬Gbinding) of mutation of these residues to alanine is anti-TRIM21 autoantibodies do not recognize endogenous sufficiently large (Ͼ3 kcal/mol) to effectively abolish binding, a TRIM21 in the cells of rodents or other nonprimate species. In feature conserved with human TRIM21 (Fig. 4 and Table 1). contrast, other human autoantibodies react with ortholog autoan- The VL6 loop is the most variable between TRIM proteins, and tigens from nonhuman species. To determine the structural basis in TRIM21 there are four differences between mouse and human. for TRIM21:IgG interaction we compared the mouse TRIM21:Fc These differences create substantially different loop conformations, BIOPHYSICS complex to its human counterpart. Superposition of the two struc- with a relative variation of Ͼ5 Å. Two of these differences, positions

Keeble et al. PNAS ͉ April 22, 2008 ͉ vol. 105 ͉ no. 16 ͉ 6047 Downloaded by guest on September 27, 2021 Table 1. Key hot-spot residues are conserved between human and mouse TRIM21 cognate and noncognate complexes Position Effect on Effect on Effect on Effect on mutated M21–mouse IgG, M21–human IgG, H21–human IgG, H21–mouse IgG, on TRIM21 kcal/mol kcal/mol kcal/mol kcal/mol

299 1.00 0.9 0.7 Ն1.5 355 Ն3.2 Ն2.8 Ն5 Ն1.5 370 0.26 0.08 1.1 0.57 381 Ն3.2 Ն2.8 4 Ն1.5 383 Ն3.2 Ն2.8 4 Ն1.5 452/1 1.27 1.03 0.6 Ն1.5

451 and 452, are important Fc-contacting residues. Position 451 is interaction. However, it is not the case that the remaining an aspartic acid in human and a tyrosine in mouse, and these interaction residues are less important, as evidenced by the fact residues are located Ͼ4 Å apart in their respective binding sites (Fig. that the contribution of all interface residues is tightly evolu- 3). On the basis of the structure these residues would be expected tionarily conserved, both energetically and kinetically. This to contribute differently to the energetics of binding. However, reflects the biological context of protein:protein interactions mutation to alanine results in a similar ⌬⌬G of binding in both where the difference between function or no function depends species (Fig. 4 and Table 1). Furthermore, the alanine mutants on only a small change in binding. For instance, successful reveal that the aspartic acid in human and the tyrosine in mouse interaction between antigen and the B cell receptor requires a encode similar kon and koff rates. To confirm this finding we minimum affinity of 10 ␮M, a 1.5-fold drop in affinity resulting introduced the aspartic acid into mouse TRIM21. As predicted, the in no B cell triggering (28). Our data show that the contribution mutant maintained the same thermodynamics and kinetics of of structurally equivalent interaction residues in the TRIM21 binding (Fig. S2). binding site is maintained even when the primary sequence is different. This illustrates that mechanistic conservation does not Discussion require concomitant sequence conservation. TRIM21 Is a Previously Undescribed Mammalian Fc Receptor. We have found that TRIM21 is a previously undescribed type of widely TRIM21 in Autoimmune Disease. Our discovery that TRIM21 is a conserved mammalian Fc binding protein. TRIM21 interacts high-affinity receptor for IgG is highly significant for autoimmune with IgG specifically, displaying high affinity and a binding pathology. During autoimmunity, anti-TRIM21 autoantibodies are mechanism that is highly conserved. The structural character- generated against epitopes in the RING, B box, and coiled-coil istics of binding reflect an evolved rather than adventitious domains. Intriguingly, there are no reports of autoantibodies to interaction. This is in contrast to IgG complexes formed by epitopes in the PRYSPRY domain. This suggests that PRYSPRY autoantigens such as rheumatoid factor, in which binding occurs interaction with Fc takes place during autoimmunity and blocks Fab with poor complementarity and low affinity (27). Specific inter- targeting. We predict that anti-TRIM21 autoantibodies are likely to action of TRIM21 with IgG is therefore independent of its undergo ‘‘autoantibody bipolar bridging’’ or the simultaneous bind- targeting by autoantibodies during autoimmunity and does not ing of TRIM21 to both Fab and Fc domains. The resulting four itself explain pathogenic immune complex formation. TRIM21 binding sites per autoantibody combined with trimeric full-length is a previously undescribed mammalian Fc receptor, structurally TRIM21 would allow for multivalent binding and the creation of and thermodynamically distinct from known Fc receptors. All large highly cross-linked immune complexes that are potent Fc␥ previously identified mammalian Fc receptors are based on the receptor activators. Alternatively, TRIM21:autoantibody bridged Ig superfamily and related MHC class I-like scaffold. In complexes may be so large and cross-linked that binding to Fc␥ TRIM21, a C-terminal PRYSPRY domain with a ␤-sandwich- receptors and complement is blocked. In bacterial pathogenesis, like architecture mediates binding to IgG. Four PRYSPRY bipolar bridging of antibodies by superantigens such as gE-gI and hot-spot residues supply most of the binding energy for IgG protein A is presumed to block receptor binding, disconnecting

Table 2. Binding kinetics of human and mouse TRIM21 D355A to a range of mammalian IgGs W381A W383A 6 kon (ϫ10 ), Ϫ1 Ϫ1 Ϫ1 Complex M ⅐s koff,s Kd,M Y452A W299A H21–human IgG 3.5 Ϯ 0.3 0.46 Ϯ 0.01 1.3 Ϯ 0.15 ϫ 10Ϫ7 S326D R324A H21–canine IgG 1.54 Ϯ 0.01 0.25 Ϯ 0.05 1.6 Ϯ 0.4 ϫ 10Ϫ7 S370A H21–guinea pig IgG 0.73 Ϯ 0.02 10.43 Ϯ 0.2 1.4 Ϯ 0.07 ϫ 10Ϫ5 Y452D H21–monkey IgG 2.7 Ϯ 0.06 0.24 Ϯ 0 8.9 Ϯ 0.2 ϫ 10Ϫ8 H21–mouse IgG 0.18 Ϯ 0.01 1.3 7.2 ϫ 10Ϫ6 H21–rat IgG 0.8 Ϯ 0.02 1.33 Ϯ 0.1 1.7 Ϯ 0.02 ϫ 10Ϫ6 S370L M21–mouse IgG 3.16 1.35 4.3 ϫ 10Ϫ7 M21–canine IgG 2.9 Ϯ 0.1 0.5 Ϯ 0.01 1.7 Ϯ 0 ϫ 10Ϫ7 M21–guinea pig IgG 2.9 Ϯ 0.1 4.33 Ϯ 0.02 1.5 Ϯ 0 ϫ 10Ϫ6 M21–human IgG 2.04 1.92 9.4 ϫ 10Ϫ7 M21–monkey IgG 1.9 Ϯ 0.1 1.1 Ϯ 0.01 5.8 Ϯ 0 ϫ 10Ϫ7 Fig. 4. Effect of site-directed mutagenesis on cognate versus noncognate M21–rat IgG 2.3 Ϯ 0.1 2.77 Ϯ 0.02 1.2 Ϯ 0 ϫ 10Ϫ6 species TRIM21:IgG interaction. When the effects of mouse TRIM21 mutations on binding mouse IgG (cognate; x axis) are plotted against the effect on Stopped-flow fluorescence was used to measure the association and dis- binding human IgG (noncognate; y axis) there exists a striking linear correla- sociation kinetics of a range of mammalian IgGs to both human and mouse tion (R2 ϭ 0.98) suggesting that the same interactions are used to stabilize TRIM21 (see Materials and Methods for details). both complexes.

6048 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800159105 Keeble et al. Downloaded by guest on September 27, 2021 VL1

Human 287 VHITLDPDTANPWLILSEDRRQVRLGDTQQSIPGNEERFDSYPMVLGAQHFHSGKHYWEV 347 Murine 287 VHITLDRNTANSWLI ISKDRRQVRMGDTHQNVSDNKERFSNYPMVLGAQRFSSGKMYWEV 347 VL2 VL3 VL4 VL5a VL5b Fig. 5. Sequence alignment of Human 348 DVTGKEAWDLGVCRDSVRRKGHFLLSSKSGFWT IWLWNKQKYEAGTYPQTPLHLQVPPCQ 407 mouse and human TRIM21. Identi- Murine 348 DVTQKEAWDLGVCRDSVQRKGQFSLSPENGFWT IWLWQDS - YEAGTSPQTTLHIQVPPCQ 406 VL6 cal residues are shaded gray. The Human 408 VGI FLDYEAGMVSFYNI TDHGSL I YSFSECAFTGPLRPFFSPGFNDGGKNTAPLTLCPLN 467 positions of the variable binding Murine 407 IGIFVDYEAGVVSFYNITDHGSLIYTFSECVFAGPLRPFFNVGFNYSGGNAAPLKLCPLK466 loops are shaded purple.

antibody recognition from effector function (29). This type of block permits interaction inside the acidic environment of intracellular could have a number of serious deleterious effects on autoimmune vesicles but not the mildly basic pH of the cell. In rat FcRn this pH pathogenesis. Preventing complement binding and hence immune dependence results, in part, from interaction with two titratable complex recycling directly affects the persistence and deposition of histidine residues (H433 and H435) in the CH3 domain whose immune complex, which is a classic outcome of autoimmunity. deprotonation disrupts salt bridge formation (19, 32, 33). The role of these residues changes between species and isotypes and in TRIM21 Function. There are several hypotheses for the role of human and mouse IgG1 H433 appears to have no function (26). TRIM21:IgG interaction. One hypothesis is that TRIM21 mediates The interaction of pathogen superantigen HSV gE-gI with Fc is also clearance of apoptosed cells by opsonizing them in antibody and pH-dependent but in the opposite direction—binding taking place recruiting macrophage phagocytic function (17). This is supported at basic but not acidic pH—as a result of targeting the same two by the fact that TRIM21 migrates to the periphery of cells under- histidines, H433 and H435 (29, 34). TRIM21 also binds H433 and going apoptosis (30). Expression of TRIM21 is up-regulated by H435 (in fact these interactions are hot spots) but does not display IFN-␥ (6), which suggests that TRIM21-assisted clearance may be pH dependence, and its affinity is not reduced by protonation or specifically used during inflammation or tissue damage. A second deprotonation. The fact that TRIM21:IgG binding is not pH- hypothesis is that TRIM21 is involved in targeting unfolded IgG dependent suggests that it does not have an FcRn-like recycling made by B cells for proteasomal degradation (31). This provides a function. Instead of pH dependence, TRIM21 binding is salt- role for the tripartite motif domains, which are thought to have a dependent: increasing salt concentration from 20 mM to 200 mM conserved ubiquitination role in TRIM function, but does not results in a 5-fold drop in affinity. This behavior is a result of explain why TRIM21 is expressed at similar levels in other cell hydrophobic stacking interactions and hydrogen bonds with H433 types. and H435 that take place at the center of an interface surrounded Our data provide some clues to TRIM21 function and allow by a solvent-excluding seal of hydrophobic residues. The three Fc-binding mechanisms displayed by FcRn, gE-gI, and these hypotheses to be assessed. Binding of TRIM21 is entirely TRIM21 (acid dependence, basic dependence, and salt depen- symmetrical with a 2:1 stoichiometry, suggesting that it can be dence) produced by interaction with the same two histidine residues cross-linked by free antibody and that TRIM21 is unlikely to provide a remarkable example of how a binding mechanism can be function as a trigger receptor like the Fc␥ receptors. Because exquisitely tailored to give a desired cellular outcome. In the context TRIM21 is trimeric, it has the capacity to cross-link IgG and of IgG receptors, the utility of the two histidines, H433 and H435, stimulate B cells. The ability to cross-link the B cell receptor could provides one explanation for the observation by DeLano et al. (35) explain why TRIM21 is a dominant autoantigen and how TRIM21- that multiple binding solutions exist to this region of IgG. In a mediated autoimmunity is initiated, for example by stimulating T general context, the three binding mechanisms illustrate the ver- cell-independent B cell activation. satility of protein structure. We have previously commented that TRIM21 binds to all subclasses of IgG including IgG3 (Table S3). protein binding sites are intrinsically promiscuously active as result This suggests that TRIM21 interaction is not linked to a particular of their chemically heterogeneous nature (36). The present example effector function or arm of the immune response and would be illustrates how individual residues provide remarkable chemical consistent with a general role in intracellular IgG degradation. In and hence functional heterogeneity. contrast, Fc␥ receptors have specific patterns of IgG subclass specificity. The neonatal Fc receptor (FcRn) is the only mammalian Materials and Methods receptor known to bind IgG intracellularly. FcRn mediates the Protein Purification and Preparation. Mouse and human TRIM21 PRYSPRY trafficking of internalized IgG through a binding mechanism that domains were overexpressed and purified by using Ni2ϩ-NTA resin (Qiagen)

VL3 H/Q368 D/S326 TRIM21 VL5 L/S370 VL1 C 2 Y328 H D355 W381 F450 R324

VL4 W383 VL6 W299 D/Y452 Fig. 6. Comparison of mouse and human TRIM21. CH3 (Left) Superposition of mouse (orange) and human (yel- VL2 low) 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. BIOPHYSICS Mouse Human Mouse Human (Right) Superposition of mouse (red and orange) and human (pink and yellow) complexes with Fc.

Keeble et al. PNAS ͉ April 22, 2008 ͉ vol. 105 ͉ no. 16 ͉ 6049 Downloaded by guest on September 27, 2021 followed by Superdex 75 gel filtration as previously described. Mouse serum IgG first-order rate constant, and Fe is the end-point fluorescence. The bimolecular (Equitech Bio) composed predominantly of IgG2 isotype was used in all biophys- association rate constant (kon) was determined by fitting the linear relation- ical and structural experiments. IgG2a is the most prevalent form of IgG2 and can ship between kobs and the increasing pseudofirst-order concentrations of IgG comprise Ϸ80% in serum (37). There are isotypic differences between IgG2a and to kobs ϭ kon(IgG binding sites) ϩ kreverse. The dissociation rate constant (koff) IgG2b, including at position 435, which is bound by TRIM21. It is not possible from was measured by chasing excess protein A into a preformed TRIM21:IgG our data to assess the effect of these differences on TRIM21 binding. The absence complex and fitting the resulting fluorescence enhancement to the fluores- of additional phases in the binding kinetics suggests that either the kinetic rate cence equation given above. Protein A binding to Fc occurs without a change constants are largely unaffected by these differences or binding is not observed in fluorescence as shown in Fig. S3. (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 Titration Calorimetry. Proteins were dialyzed into 20 mM potassium phosphate experiments with TRIM21 confirm that most protein is bound at concentrations (pH 8) and 125 mM NaCl overnight at 4°C, and experiments were carried out on a Microcal VP-ITC as described previously (17). over the Kd. 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 Crystallography. Mouse TRIM21-mouse Fc complex was purified by mixing the and subtypes were from Serotec and Athens Research and Technology. Pooled components in a 2:1 ratio and applying to a Superdex 200 gel filtration column. serum IgG composed of Ͼ90% IgG1 was used in all biophysical experiments. Free mouse TRIM21 was used without further purification. Crystals were ob- TRIM21 binding yielded single-phase kinetics to serum IgG, although we cannot served in many conditions, but the best were seen in 0.1 M Mes (pH 6.5) with 0.2 rule out the possibility that allotypes present in trace amounts bind with more M magnesium acetate and 20% PEG8000 for the free mouse TRIM21 and 0.1 M complex kinetics. 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 Radia- Site-Directed Mutagenesis. TRIM21 alanine mutants were produced by tion Facility beamline ID14-3, whereas data for the mouse complex were collected QuikChange mutagenesis (Stratagene) and purified as described above. 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. Rapid Reaction Kinetics. Experiments were carried out in 20 mM potassium The structure of human TRIM21 PRYSPRY [Protein Data Bank ID code 2IWG (17)] phosphate (pH 8) and 125 mM NaCl. Stopped-flow experiments were carried was used as a search model in PHASER. Structures were refined in REFMAC and ␲ out in an Applied Photophysics -Star stopped-flow spectrofluorimeter in 1:1 Coot. Structural figures were created by using Pymol (40) and Chimera (41). mixing mode and using a 320-nm cutoff filter essentially as described previ- ously (23). The resulting fluorescence quench that occurs upon association of ACKNOWLEDGMENTS. We thank beamline staff at the European Synchrotron TRIM21 with IgG was fitted to F ϭ⌬F exp(Ϫkobst) ϩ Fe, where F is the observed Radiation Facility for assistance. L.C.J., A.H.K., A.F., and Z.K. were supported by fluorescence, ⌬F is the fluorescence amplitude, kobs is the observed pseudo the Medical Research Council.

1. Nisole S, Stoye JP, Saib A (2005) TRIM family proteins: Retroviral restriction and antiviral 22. James LC, Roversi P, Tawfik DS (2003) Antibody multispecificity mediated by confor- defence. Nat Rev Microbiol 3:799–808. mational diversity. Science 299:1362–1367. 2. Meroni G, Diez-Roux G (2005) TRIM/RBCC, a novel class of ‘single protein RING finger’ 23. James LC, Tawfik DS (2005) Structure and kinetics of a transient antibody binding E3 ligases. BioEssays 27:1147–1157. intermediate reveal a kinetic discrimination mechanism in antigen recognition. Proc 3. Stremlau M, et al. (2004) The cytoplasmic body component TRIM5alpha restricts HIV-1 Natl Acad Sci USA 102:12730–12735. infection in Old World monkeys. Nature 427:848–853. 24. Harris LJ, Larson SB, Hasel KW, McPherson A (1997) Refined structure of an intact IgG2a 4. Towers GJ (2007) The control of viral infection by tripartite motif proteins and cyclo- monoclonal antibody. Biochemistry 36:1581–1597. philin A. Retrovirology 4:40–50. 25. DeLano WL, Ultsch MH, de Vos AM, Wells JA (2000) Convergent solutions to binding at 5. Trockenbacher A, et al. (2001) MID1, mutated in Opitz syndrome, encodes an ubiquitin a protein-protein interface. Science 287:1279–1283. ligase that targets phosphatase 2A for degradation. Nat Genet 29:287–294. 26. Kim JK, et al. (1999) Mapping the site on human IgG for binding of the MHC class 6. Rhodes DA, et al. (2002) The 52 000 MW Ro/SS-A autoantigen in Sjogren’s syndrome/ I-related receptor, FcRn. Eur J Immunol 29:2819–2825. systemic lupus erythematosus (Ro52) is an interferon-gamma inducible tripartite motif 27. Corper AL, et al. (1997) Structure of human IgM rheumatoid factor Fab bound to its protein associated with membrane proximal structures. Immunology 106:246–256. autoantigen IgG Fc reveals a novel topology of antibody-antigen interaction. Nat 7. Li X, et al. (2006) Functional replacement of the RING, B-box 2, and coiled-coil domains Struct Biol 4:374–381. of tripartite motif 5alpha (TRIM5alpha) by heterologous TRIM domains. J Virol 28. Batista FD, Neuberger MS (1998) Affinity dependence of the B cell response to antigen: 80:6198–6206. A threshold, a ceiling, and the importance of off-rate. Immunity 8:751–759. 8. The International FMF Consortium (1997) Ancient missense mutations in a new member 29. Sprague ER, Wang C, Baker D, Bjorkman PJ (2006) Crystal structure of the HSV-1 Fc receptor of the RoRet family are likely to cause familial Mediterranean fever. Cell 90:797–807. bound to Fc reveals a mechanism for antibody bipolar bridging. PLoS Biol 4:e148. 9. Ferrentino R, et al. (2007) MID1 mutation screening in a large cohort of Opitz G/BBB 30. Casciola-Rosen LA, Anhalt G, Rosen A (1994) Autoantigens targeted in systemic lupus syndrome patients: Twenty-nine novel mutations identified. Hum Mutat 28:206–207. erythematosus are clustered in two populations of surface structures on apoptotic 10. Rhodes DA, de Bono B, Trowsdale J (2005) Relationship between SPRY and B30.2 protein keratinocytes. J Exp Med 179:1317–1330. domains. Evolution of a component of immune defence? Immunology 116:411–417. 31. Takahata M, et al. (2007) Ro52 functionally interacts with IgG1 and regulates its quality 11. Ben-Chetrit E, Chan EK, Sullivan KF, Tan EM (1988) A 52-kD protein is a novel compo- control via the ERAD system. Mol Immunol 45:2045–2054. nent of the SS-A/Ro antigenic particle. J Exp Med 167:1560–1571. 32. Raghavan M, Bonagura VR, Morrison SL, Bjorkman PJ (1995) Analysis of the pH 12. Ben-Chetrit E, Fox RI, Tan EM (1990) Dissociation of immune responses to the SS-A (Ro) dependence of the neonatal Fc receptor/ interaction using antibody 52-kd and 60-kd polypeptides in systemic lupus erythematosus and Sjogren’s syn- and receptor variants. Biochemistry 34:14649–14657. drome. Arthritis Rheum 33:349–355. 33. Martin WL, West AP, Jr, Gan L, Bjorkman PJ (2001) Crystal structure at 2.8 A of an 13. Moutsopoulos HM, et al. (1985) Anti-Ro(SSA) positive rheumatoid arthritis (RA): A clinicoserological group of patients with high incidence of D-penicillamine side effects. FcRn/heterodimeric Fc complex: Mechanism of pH-dependent binding. Mol Cell 7:867– Ann Rheum Dis 44:215–219. 877. 14. Frank MB, et al. (1993) The mapping of the human 52-kD Ro/SSA autoantigen gene to 34. Sprague ER, Martin WL, Bjorkman PJ (2004) pH dependence and stoichiometry of human 11, and its polymorphisms. Am J Hum Genet 52:183–191. binding to the Fc region of IgG by the herpes simplex Fc receptor gE-gI. J Biol Chem 15. McCauliffe DP, et al. (1997) Recombinant 52 kDa Ro(SSA) ELISA detects autoantibodies 279:14184–14193. in Sjogren’s syndrome sera that go undetected by conventional serologic assays. 35. DeLano WL, Ultsch MH, de Vos AM, Wells JA (2000) Convergent solutions to binding at J Rheumatol 24:860–866. a protein-protein interface. Science 287:1279–1283. 16. Salomonsson S, et al. (2005) Ro/SSA autoantibodies directly bind cardiomyocytes, disturb 36. James LC, Tawfik DS (2001) Catalytic and binding poly-reactivities shared by two calcium homeostasis, and mediate congenital heart block. J Exp Med 201:11–17. unrelated proteins: The potential role of promiscuity in enzyme evolution. Protein Sci 17. James LC, et al. (2007) Structural basis for PRYSPRY-mediated tripartite motif (TRIM) 10:2600–2607. protein function. Proc Natl Acad Sci USA 104:6200–6205. 37. Jones RHV, Rademacher TW, Williams PJ (1996) Bias in murine IgG isotype immobil- 18. Woof JM, Burton DR (2004) Human antibody-Fc receptor interactions illuminated by isation implications for IgG glycoform analysis ELISA procedures. J Immunol Methods crystal structures. Nat Rev Immunol 4:89–99. 197:109–120. 19. Burmeister WP, Huber AH, Bjorkman PJ (1994) Crystal structure of the complex of rat 38. James LC, et al. (1999) 1.9 A structure of the therapeutic antibody CAMPATH-1H fab in neonatal Fc receptor with Fc. Nature 372:379–383. complex with a synthetic peptide antigen. J Mol Biol 289:293–301. 20. Prabhat P, et al. (2007) Elucidation of intracellular recycling pathways leading to 39. CCP4 (1994) Collaborative Computational Project 4. Acta Crystallogr D 50:760–763. exocytosis of the Fc receptor, FcRn, by using multifocal plane microscopy. Proc Natl 40. DeLano WL (2002) The PyMOL Molecular Graphics System (DeLano Scientific, San Acad Sci USA 104:5889–5894. Carlos, CA). 21. Nimmerjahn F, Ravetch JV (2006) Fc gamma receptors: Old friends and new family 41. Pettersen EF, et al. (2004) UCSF Chimera—A visualization system for exploratory members. Immunity 24:19–28. research and analysis. J Comput Chem 25:1605–1612.

6050 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800159105 Keeble et al. Downloaded by guest on September 27, 2021