Antibodies mediate intracellular immunity through tripartite motif-containing 21 (TRIM21)

It is assumed that antibodies do not routinely enter the cytosol during viral infection. To test this, we preincubated adenovirus (a model virus that causes respiratory disease) with different antibodies and added the virions to cultured HeLa cells. Adenovirus was chosen because it is a nonenveloped virus, and its capsid is naturally exposed to serum antibody before cellular infection. This is in contrast to enveloped viruses, in which the capsid is protected from antibody. After 30 min of infection the cells were fixed, and a fluorescent anti-IgG antibody was added to detect antibody-coated virions. As can be seen in Fig. 1A, antibody-coated virions successfully enter cells. Similar results were obtained using different polyclonal antihexon antibodies and human serum IgG. Adenovirus enters the cell by binding to the coxsackie and adenovirus receptor and becoming endocytosed. We found that addition of antibody does not prevent this process and that antibody remains attached to virus after entry. To address whether antibody-coated virus is accessible to cytosolic TRIM21, we costained for TRIM21. As shown in Fig. 1A, TRIM21 is efficiently recruited to antibody-coated viral particles.

Next we tested the effect of TRIM21 recruitment to virions by quantifying the levels of adenovirus infection. We used a virus that carries a GFP gene so that infection efficiency could be determined by flow cytometry analysis. A standard viral neutralization assay was performed on HeLa cells pretreated with control siRNA, TRIM21 siRNA, IFN-α, or IFN-α and TRIM21 siRNA (Fig. 1B). To take account of toxicity and variable cell death between these different conditions, we measured the decrease in infection due to the addition of antibody. In the absence of antibody, adenovirus infected ≈50% of cells. The percentage of infected cells decreased rapidly with increasing antibody concentration, such that at 400 ng/mL antibody, infection was reduced by ≈60-fold (Fig. 1B). However, in cells depleted of TRIM21, addition of 400 ng/mL antibody had a minimal effect on infection (≈3-fold).

During an immune response, IFN-α activates the transcription of antiviral genes. TRIM21 is IFN-α regulated, and the modest levels of endogenous TRIM21 protein are substantially increased by IFN-α (Fig. 1C). Consistent with this regulation, preincubation of cells with IFN-α markedly increased the effect of antibody neutralization. IFN-α has pleiotropic effects, but without addition of antibody we observed little impact on adenovirus infection. To test whether IFN-α/antibody neutralization synergy is TRIM21 dependent, we specifically depleted the TRIM21 levels that are up-regulated by IFN-α, leading to >20-fold recovery of infectivity (Fig. 1B). In all experiments, antibody neutralization of viral infection directly correlated with TRIM21 levels (Fig. 1C). For example, cells expressing the most TRIM21 were almost 2 orders of magnitude more resistant to adenovirus infection than those expressing the least.

We confirmed the robustness of this phenotype by examining the effect of different siRNA sequences, cell types, and types of antibody. Different TRIM21 siRNAs with different target sequences reversed antibody neutralization of adenovirus infection by depleting TRIM21 levels (Fig. S1A). TRIM21 neutralization occurred in a range of cell lines including HeLa, HT1080, and TE671 (Fig. S1B). A stable TRIM21 knockdown line was established in each case using an shRNA vector based on the sequence of siRNA 2. In all cells, TRIM21 mediated antibody neutralization of adenovirus (Fig. S1B). Finally, we tested the effect on adenovirus neutralization of two different anti-Ad5 polyclonal antibodies (Abd Serotec and Millipore) and an anti-Ad5 hexon monoclonal. In every case, neutralization of adenovirus was enhanced by TRIM21 up-regulation and reversed by TRIM21 depletion (Fig. S1C).

Unlike entry neutralization, antibody effector mechanisms are reliant on interaction with the Fc domain. To test whether TRIM21 neutralization was Fc dependent, we treated IgG with pepsin to remove the Fc and generate Fab2 fragments. Fab2 fragments are bivalent, bind antigen with the same affinity as IgG, and can cross-link viral epitopes. However, we found that Fab2 fragments were not able to mediate efficient TRIM21 neutralization (Fig. 2A). Unlike with IgG, when using Fab2, IFN-α treatment or TRIM21 depletion no longer dramatically affected adenovirus infection. Thus in order for antibodies to mediate intracellular viral neutralization they must contain an Fc domain and TRIM21 must be present. These results reveal a unique effector mechanism in humoral immunity that extends protection into cells.

During the early stages of infection, in which innate immunity is critical, IgM rather than IgG antibodies dominate the antibody repertoire. We tested whether TRIM21 interacts with IgM and if so the importance of TRIM21 in IgM viral neutralization. To investigate TRIM21:IgM binding, we labeled the TRIM21 PRYSPRY domain with an Alexa 488 fluorophore and measured its fluorescence anisotropy upon titration of IgM (Fig. 2B). The resulting titration curve was fit (Materials and Methods) to give an affinity (K_D) of 16.8 ± 1.5 μM. This is weaker than the affinity with which TRIM21 binds IgG Fc [0.2 μM under physiological conditions (4)]. However, the affinity of TRIM21 to IgM is likely to be significantly higher in vivo because full-length TRIM21 is a multimer. Complement C1q, which binds IgM with nanomolar affinity, has undetectable affinity when measured as a monomer (6).

Next we tested the effect of serum IgM on adenovirus infection. We found that pooled human serum IgM and TRIM21 operate synergistically to neutralize adenovirus infection (Fig. 2C). Furthermore, as with IgG, the neutralization of virus by IgM requires TRIM21. This suggests that TRIM21-mediated immunity could be effective in the early stages of an infection. Furthermore, it demonstrates that TRIM21 utilizes a broader usage of isotypes than classical Fc receptors, which are usually both isotype and subtype specific.

The previous experiments demonstrate that there is an intracellular immune response mediated by TRIM21 and antibodies that is capable of preventing viral infection. Next we examined the mechanism by which this intracellular neutralization occurs. We investigated the mechanism in three ways. First, we determined how TRIM21 targets antibody and the thermodynamics of interaction. Second, we examined what events subsequent to targeting are required for neutralization. Third, we asked how virus is neutralized.

TRIM21 is a multidomain protein consisting of RING, B Box, coiled-coil, and PRYSPRY domains. We tested the role of these domains in IgG binding using multiangle light scattering (MALS) and fluorescence titration spectroscopy. Analysis of the MALS data reveals that recombinant full-length TRIM21 forms a stable dimer and not a trimer as previously reported (7) (Fig. 3A). There are conflicting data on the oligomeric state of other TRIM proteins, which have been shown to be both trimers (8) and dimers (9). Our results support the view that a dimer is the preferred state for TRIM proteins. When mixed with IgG, TRIM21 forms a stoichiometric complex consisting of one antibody and one TRIM21 (Fig. 3A). Deletion of the RING domain alone resulted in a destabilized recombinant protein; however, deletion of both RING and B Box did not affect TRIM21 stability or its ability to dimerize. Fluorescence titration spectroscopy revealed that full-length TRIM21 and ΔRING-Box bound to IgG with a similar dissociation constant (K_D) of <1 nM (Fig. 3 B and C). Because the monomeric PRYSPRY domain binds with ≈150 nM affinity (4, 5), this indicates that the coiled-coil domain is required for TRIM21 dimerization and the simultaneous engagement of both IgG heavy chains. The sub-nM affinity of TRIM21 for IgG makes TRIM21 the highest-affinity antibody receptor in the human body. The evolution of such a high-affinity interaction explains how TRIM21 efficiently targets virus.

Next we looked at what happens to virus after TRIM21 is recruited and the role of the RING and B Box domains. Because RING domains often display E3 ubiquitin ligase activity, we hypothesized that TRIM21 may target bound virus for degradation via ubiquitination. Cells possess two pathways for degradation of ubiquitinated material: the proteasome and autophagy. To explore the role of these pathways in TRIM21 neutralization of virus we performed viral infection experiments in the presence of MG132 (a proteasome inhibitor) and 3-methyladenine (3-MA; an autophagy inhibitor). The autophagy inhibitor had no effect on infectivity; however, MG132 significantly reversed TRIM21 neutralization of infectivity (Fig. 3D). Titration experiments showed a direct correlation between increasing levels of MG132 and reduced neutralization (Fig. 3E). The ability of MG132 to reverse neutralization was dependent upon the presence of antibody and TRIM21. Moreover, addition of MG132 could not recover infection in cells depleted of TRIM21, showing that TRIM21 and proteasome function are essential components in the same neutralization pathway (Fig. 3F).

To determine whether ubiquitination is essential to target virus to the proteasome, we tested the ability of full-length TRIM21 and ΔRING-Box recombinant proteins to neutralize infection. We incubated protein with antibody-coated virions and allowed the virus to infect cells depleted of TRIM21. As can be seen in Fig. 4A, deletion of the RING and B Box domains prevents TRIM21 neutralization of virus. We attempted to repeat these experiments in cells overexpressing TRIM21; however, this led to loss of function. Loss of function could be partially reversed with IFN, suggesting that overexpression titrates essential cofactors rather than generates inactive protein. To confirm that the neutralization we observe with recombinant proteins correlates with ubiquitin ligase activity, we compared the ability of full-length and ΔRING-Box proteins to autoubiquitinate. Although deletion of the RING and B Box domains has no effect on IgG binding, it abolishes autoubiquitination (Fig. 4B). Thus both TRIM21 ubiquitination activity and proteasomal function are required for viral neutralization. However, although the in vitro assay indicates that TRIM21 is an active ligase, it cannot be assumed that it is autoubiquitination that drives proteasomal targeting in vivo.

Although E3 ubiquitin ligases are known to autoubiquitinate, it is the transfer of ubiquitin to substrate that is thought to be important for proteasomal targeting. However, we hypothesized that proteasomal targeting via autoubiquitination would allow TRIM21 to neutralize any virus and prevent evolution of viral mutants that escape ubiquitin conjugation. Consistent with this mechanism, although we found that TRIM21 efficiently forms ubiquitin chains on itself, we found no detectable ubiquitination of either IgG or virus in our in vitro ubiquitination assay (Fig. 4C). This suggests that recruitment to the proteasome is not dependent upon direct ubiquitination of either the antibody or virus but rather autoubiquitination of TRIM21 or a recruited cellular cofactor. Proteasomal recruitment through TRIM21 autoubiquitination is consistent with the extremely high affinity with which TRIM21 has evolved to bind antibody. If TRIM21 were transferring ubiquitin to antibody or virus through normal enzymatic turnover, then a high affinity would translate as a highly inefficient K_M. However, we cannot rule out that a capsid protein other than hexon is ubiquitinated or that although undetected in vitro, ubiquitination of hexon or antibody occurs in vivo. Finally, to demonstrate that the intracellular TRIM21-associated viral particle is ubiquitinated, we examined infected cells by confocal microscopy and stained for ubiquitin. As can be seen in Fig. 4D, virions colocalized with TRIM21 are also positive for ubiquitin.

To determine what happens to virus after TRIM21-mediated targeting to the proteasome, we performed a fate-of-capsid time-course experiment. In this experiment we infected cells with virus preincubated with antibody and compared the levels of hexon protein (viral capsid) in infected HeLa cells with those in cells depleted of TRIM21. By 2 h after infection there was markedly less hexon in HeLa compared with TRIM21-depleted cells (Fig. 5). This indicates both that TRIM21 mediates degradation of virus and that it is a rapid process. Addition of MG132 slowed the decline in hexon levels, confirming that virus is being physically degraded in a proteasome-dependent manner. Because proteasomal targeting by TRIM21 requires virus to be antibody-bound, we also looked at the antibody levels in infected cells. We found that destruction of virus is paralleled by disposal of antibody (Fig. 5). In contrast to hexon and antibody, we saw little change in TRIM21 levels after 2 h. Treatment with MG132 did not increase TRIM21 levels, suggesting either that TRIM21 is not degraded by the proteasome or that ubiquitination and degradation are coupled to IgG binding. In the latter case, only a fraction of the total pool of cellular TRIM21 would be degraded. We also cannot rule out the possibility of a feedback loop that maintains TRIM21 levels.

The combination of antibody targeting and TRIM21 autoubiquitination implies that no direct viral interactions are required for neutralization. To test this, we transfected cells with streptavidin latex beads coated in antistreptavidin antibody. We found that TRIM21 is efficiently recruited to the antibody-bound beads (Fig. 6). Furthermore, TRIM21-associated beads are positive for ubiquitin. This result suggests that direct interaction with an invading viral particle may not be required to recruit TRIM21 and that virus may not be the target of ubiquitination. We believe that TRIM21-mediated neutralization may be effective against other nonenveloped viruses.

HEK293T, HeLa, TE671, and HT1080 were maintained in standard conditions. 293F cells (Invitrogen) were grown in serum-free Freestyle medium (Invitrogen). Adenovirus Ad5-GFP (20) was grown in transcomplementation cell line 293F for 72 h and purified to 10⁸ to 10⁹ IU/mL. Transient siRNA knockdowns were performed in six-well plates with 150 pmol of oligo per 1 × 10⁵ cells.

HeLa cells in six-well plates were infected with 5 × 10⁴ infectious units (IU) AdV5-GFP for 30 min and then incubated for 48 h before fixing. Goat antiadenovirus polyclonal antibody was used in all virus neutralization assay experiments unless otherwise stated (AB1056; Millipore).

HeLa cells were seeded onto coverslips in 24-well plates and infected essentially as described above, then fixed and stained with appropriate primary and secondary antibodies. Confocal images were taken using a Zeiss 63× lens on a Jena LSM 710 microscope (Carl Zeiss MicroImaging).

Full-length and ΔRING-Box recombinant TRIM21 was expressed as maltose binding protein-fusion proteins in Escherichia coli and purified using amylose resin and size-exclusion chromatography. Fluorescence experiments were performed by titrating IgG using a Cary Eclipse fluorescence spectrophotometer.

The PRYSPRY domain of TRIM21 was expressed and purified as previously described (4, 5). The protein was labeled with Alexa Fluor 488 5-SDP ester (Invitrogen) and its polarized fluorescence measured at 530 nm using a Cary Eclipse fluorescence spectrophotometer. IgM (Athens Research and Technology) was titrated into 50 nM PRYSPRY and the change in anisotropy averaged over 5 s.

Size-exclusion chromatography (SEC) MALS was performed using a Wyatt Heleos II 18-angle light-scattering instrument coupled to a Wyatt Optilab rEX online refractive index detector. Samples were prepared as described above and passed through the light-scattering and refractive index detectors in a standard SEC MALS format.

In vitro assays were carried out largely as previously described (21). Reaction mixtures were incubated at 37 °C for 1 h then visualized by Western blot for TRIM21 (1:500, sc-25351; Santa Cruz Biotechnology), Ad5 hexon (donkey anti-goat IgG HRP 1:5,000 sc-2056; Santa Cruz Biotechnology), or ubiquitin (1:1,000, FK-2; Enzo Life Sciences) as indicated.