Target-induced clustering activates Trim-Away of pathogens and proteins

TRIM21 assembles K63-linked Ub chains on itself upon target engagement

We first investigated whether target engagement is the trigger for TRIM21 enzymatic activity. We infected cells overexpressing His-tagged ubiquitin with adenovirus (AdV5) alone or in the presence of wild-type anti-AdV5 hexon antibody (9C12) or a mutant antibody that cannot bind TRIM21 [9C12(H433A)]³⁹. TRIM21 underwent polyubquitination in the presence of AdV5+9C12 antibody but not AdV5 alone. Polyubiquitination was lost with the 9C12(H433A) antibody, confirming that target engagement by TRIM21 is required (Extended Data Fig. 1a). In cells expressing His-tagged TRIM21 at endogenous levels and with endogenous ubiquitin, we observed a similar pattern of ubiquitination that was reversed with the deubiquitinase USP2 (Extended Data Fig. 1b-e).

To determine whether TRIM21 itself is activated, or if it merely becomes a target for another ubiquitin ligase, we introduced mutations into the TRIM21 RING domain (Extended Data Fig. 1f) that impair E2~Ub binding and reduce both Trim-Away and AdV5 neutralisation²⁰. Of these mutations, E12R and E13A substantially reduced polyubiquitination (Extended Data Fig. 1a,g). This indicates that TRIM21 is catalytically activated upon target engagement. To address which ubiquitin chain types TRIM21 modifies itself with upon target engagement in cells, we expressed different ubiquitin mutants. Expression of K48R, which prevents K48-chain formation, did not impact TRIM21 polyubiquitination. Consistent with this, no polyubiquitination was seen in cells expressing a ubiquitin mutant that forms K48-linked chains only. In contrast, K63R substantially reduced TRIM21 polyubiquitination whilst cells expressing only K63 maintained chain formation (Extended Data Fig. 1h). This suggests that TRIM21 builds K63 chains on itself upon target engagement in cells.

RING dimerisation is required for TRIM21 activation

To investigate whether TRIM21 RING dimerisation is required for function, we mutated two residues at the RING dimer interface (Fig. 1a) that are required for RING dimerisation³⁸. The M72E mutation abolished TRIM21's ability to degrade target protein using Trim-Away (Fig. 1b) and reduced TRIM21 autoubiquitination upon infection with AdV5+9C12 (Extended Data Fig. 2a-c). Previously, we showed that monomeric TRIM21 RING domain possesses catalytic activity but is constitutively inhibited by the B Box³⁸. However, comparing RING (T21R), RING-B Box (T21RB) or RING-linker-RING (T21RLR; two RINGs in a single polypeptide) constructs revealed that enforcing RING dimerisation significantly increases TRIM21 catalytic activity, as measured by polyubiquitin synthesis and ubiquitin discharge from Ube2N~Ub (Fig. 1c-e). This increase in activity of the RLR construct required the M10 dimerisation residue (Extended Data Fig. 2d). Thus, RING dimerisation both increases TRIM21 catalysis in vitro and is necessary for TRIM21 activity in cells.

TRIM21 is a multi-domain protein containing RING, B Box, coiled-coil and PRYSPRY domains (Extended Data Fig. 3a). Using SEC-MALS and isothermal calorimetry, we found that the coiled-coil (T21CC235; residues 129-235) from TRIM21 undergoes monomer-dimer exchange but forms no higher order multimers (Extended Data Fig. 3b,c). This result was confirmed by analytical ultracentrifugation using a T21RBCC construct containing residues 1-235 (Extended Data Fig. 3d). Small angle x-ray scattering (SAXS) data on T21CC235 and T21RBCC (Extended Data Fig. 3e-g) revealed that the TRIM21 dimer exists as an elongated antiparallel coiled-coil with the two copies of the RING domain held apart by at least 150 Å (Fig. 1f; Supplementary Note 1). This arrangement was maintained in full length TRIM21 bound to IgG Fc (Fig 1g; Extended Data Fig. 3h-j; Supplementary Note 1), suggesting that antibody binding does not induce large scale conformational changes in TRIM21 that could facilitate intramolecular RING dimerisation.

RING dimerisation is induced by recruitment of multiple TRIM21 molecules to a target

How can TRIM21 RING domains dimerise when they are held apart as monomers by an elongated coiled-coil? A potential mechanism could be dimerisation of RING domains between neighbouring TRIM21 dimers, for example as a result of TRIM21 clustering on the surface of an antibody-coated virion (Fig. 1h). This clustering mechanism predicts that recruitment of multiple TRIM21 molecules to the target is required for TRIM21 activation. We tested this prediction using a virus neutralisation assay in which the infectivity of adenoviral particles can be accurately quantified⁷. We used saturating concentrations of 9C12 antibody, in which ~ 200 antibodies bind per virion³⁹, but varied ratios of 9C12(WT) and 9C12(H433A) mutant, which lacks an intact TRIM21 binding site. Full neutralisation was abrogated when 9C12(H433A) reached >75%. For one natural log neutralization, ~12% of WT antibodies were required; by assuming WT and H433A antibodies were randomly distributed between viruses, we calculate that ~24 WT out of the 200 antibodies bound to AdV5 at saturation were required for neutralisation (Fig. 1i). Thus, efficient neutralisation only occurs when multiple TRIM21 molecules are recruited to the target.

Introducing the dimerisation mutants M72E or M10E significantly increased the amount of antibody required to initiate neutralisation and rendered TRIM21 unable to activate NFκB signalling (Fig. 1j,k). Furthermore, mutations N71D and R67A at the RING dimer:Ub interface (Extended Data Fig. 1f) also led to an increase in the critical threshold of antibodies required for neutralisation (Extended Data Fig. 2e-g). These mutations also reduced T21RLR polyubiquitination in vitro and rendered TRIM21 unable to activate NFκB signalling or degrade target using Trim-Away (Extended Data Fig. 2d,h,i). Altogether, these data indicate that catalytically-active RING dimers are formed upon recruitment of multiple TRIM21 molecules to the target.

Trim-Away requires recruitment of multiple TRIM21 to target

The above data support a model in which TRIM21 becomes catalytically-active when RING dimerisation is induced by the clustering of multiple TRIM21 molecules on an antibody-coated target (Fig. 1h). We tested this model by using an engineered target protein containing a repeated epitope sequence. To this end, we designed a series of monomeric EGFP (mEGFP) reporter constructs with four copies of the 10 amino acid c-myc tag at the N-terminus. We serially introduced the point mutations L4A and I5A, which ablate binding by the monoclonal anti-myc antibody 9E10⁴⁰. This antibody cannot induce antigen cross-linking owing to its atypical binding mode wherein both Fab domains contribute to antigen binding in a ‘clasping’ manner⁴¹. This resulted in a series of constructs with between zero and four anti-myc antibody binding sites per mEGFP (Fig. 2a). Upon electroporation of anti-myc antibody, we only observed efficient degradation of mEGFP with 4 active anti-myc binding sites (Fig. 2b–d). Degradation of the 4myc construct was antibody dose-dependent, whereas there was no loss of mEGFP containing only 3 anti-myc binding sites at any antibody concentration (Fig. 2e). Finally, to confirm that antibody binding takes place to the 2myc construct, but that a critical number of antibodies has not been reached, we co-electroporated anti-IgG polyclonal antibody alongside anti-myc and observed efficient degradation (Fig. 2f).

To confirm these results using a target protein without epitope tags, we mixed purified GFP together with either monoclonal or polyclonal anti-GFP antibodies and determined the molecular weight of the resulting protein complexes by mass photometry^42,43. In the absence of GFP, all the antibodies used gave a peak of approximately 150kDa (Fig. 2g–i, grey lines), consistent with the predicted size of a single IgG. For both a single monoclonal antibody (9F9.F9), and a mix of two monoclonals (7.1 + 13.1), the addition of GFP resulted in a peak shift to approximately 180-240 kDa, consistent with one or two GFP molecules (~27kDa) bound by just a single antibody (Fig. 2g,h, black lines). However, for the polyclonal antibody (ab6556), the addition of GFP resulted in the appearance of a peak at 330 kDa (Fig. 2i, black line), indicating two antibodies binding a single GFP molecule. We next tested which GFP:antibody complexes could trigger TRIM21 activation, and thus GFP degradation, inside cells. Only electroporation of polyclonal anti-GFP antibody (ab6556) resulted in efficient degradation of mEGFP (Fig. 2j). TRIM21 was co-depleted together with mEGFP when using the polyclonal antibody (Fig. 2j), consistent with TRIM21 activation only in this condition. All anti-GFP antibodies used were capable of binding GFP inside living cells (Extended Data Fig. 4). Altogether, these data suggest that protein degradation requires the recruitment of multiple TRIM21 molecules to the target.

The nature of the target dictates TRIM21 activation

As Trim-Away of endogenous cellular proteins can be performed efficiently with monoclonal antibodies¹⁷, we hypothesised that the nature of the target itself could lead to the clustering of multiple TRIM21 molecules in close proximity and induce TRIM21 activation. To investigate this further, we compared the degradation of different GFP-fusion proteins (Fig. 3a–c) using an anti-GFP nanobody-Fc chimera (vhhGFP4-Fc)¹⁷, which binds a single epitope on GFP⁴⁴. We co-expressed vhhGFP4-Fc with either monomeric EGFP (mEGFP) or mEGFP fused to oligomeric scaffolds Histones (H2B-mEGFP) or Centrosomes (mEGFP-PACT). As with electroporation of monoclonal antibody (Fig. 2j), expression of vhhGFP4-Fc did not cause degradation of mEGFP (Fig. 3d,e). However, there was efficient degradation mEGFP when fused to the oligomeric scaffolds, as visualised by loss of nuclear GFP fluorescence for H2B-mEGFP (Fig. 3f,g) and loss of centrosomal puncta for mEGFP-PACT (Fig. 3h,i). Introducing the H433A mutation into the Fc region of vhhGFP4-Fc abolished degradation of mEGFP-PACT or H2B-mEGFP, confirming TRIM21 dependence (Fig. 3f-i). Expression of the vhhGFP4-Fc construct in cells where EGFP has been knocked-in to a single allele of Caveolin-1⁴⁵, which is highly oligomeric⁴⁶, led to similar TRIM21-dependent degradation (Fig. 3j-l). Altogether these data suggest that the nature of the target dictates TRIM21 activation during Trim-Away with single epitope binders.

We next asked if we could take advantage of target-induced TRIM21 activation to selectively degrade proteins that cause disease. Huntington’s disease is caused by expansion of the polyglutamine (polyQ) tract at the N-terminus of huntingtin (Htt) protein. Tracts <36Q do not cause disease, whereas tracts >39Q are certain to cause disease, with severity increasing with polyQ tract length⁴⁷. The monoclonal anti-polyglutamine antibody 3B5H10 binds to a 13Q epitope, which means that increasing numbers of 3B5H10 antibodies can bind as the Htt polyQ tract length increases⁴⁸. We hypothesised that the 3B5H10 antibody would recruit more TRIM21 molecules to long polyQ tracts than short, resulting in TRIM21 clustering and degradation of only disease-causing huntingtin (Fig. 4a). To test this, we expressed mEGFP-tagged huntingtin exon 1 with polyQ tracts increasing in 13Q increments up to 78Q and monitored GFP fluorescence following electroporation of 3B5H10 antibody. Strikingly, the degree of degradation caused by the 3B5H10 antibody correlated with polyQ tract length. Tracts >39Q were efficiently degraded, whereas tracts <39Q were spared from degradation, with tracts of exactly 39Q showing an intermediate level of degradation (Fig. 4b,c). Thus, target-induced TRIM21 activation could be exploited therapeutically for selective degradation of disease proteins without the need for selective binders.

Molecular clustering is sufficient to activate TRIM21

The above data suggest that TRIM21 activation is triggered by the clustering of multiple TRIM21 molecules in close proximity, either by multiple antibodies bound to the target or by the oligomeric nature of the target itself. To test if molecular clustering is sufficient to activate TRIM21, we developed an assay that would allow us to induce TRIM21 clustering independently of a target. To this end, we exploited the properties of engineered variants of the plant photoreceptor, Cryptochrome 2 (CRY2), which undergo clustering upon exposure to blue light^49-51. We generated constructs with CRY2 fused to TRIM21 (Fig. 5a) and compared their behaviour in cells. Exposure of cells to blue light resulted in the rapid formation of cytoplasmic puncta for both mRFP-CRY2 and mRFP-CRY2-TRIM21 constructs, consistent with CRY2 clustering. Strikingly, live microscopy revealed that these clusters rapidly disappeared in cells expressing mRFP-CRY2-TRIM21, but not in cells expressing mRFP-CRY2. The loss of CRY2-TRIM21 puncta was fully prevented by the addition of the proteasome inhibitor MG132, suggesting that this was due to proteasomal degradation (Fig 5b,c; Supplementary Video 1), which was confirmed by western blotting (Fig. 5d). Introducing mutations that abolish E2~Ub binding (E13A) and RING dimerisation (M10E/M72E) both prevented clustering-induced TRIM21 degradation (Fig. 5d). Together this shows that TRIM21 is normally catalytically inert, but that upon molecular clustering active RING dimers form and drive degradation.

Expanding the Trim-Away toolbox

We investigated if we could exploit clustering-induced TRIM21 activation to achieve optogenetic control of Trim-Away. To this end, we generated an optoTrim-Away construct containing TRIM21 fused to both the anti-GFP nanobody and light-controllable CRY2 domain (vhhGFP4-RFP-CRY2-TRIM21; Fig. 6a). We expressed this construct in cells together with the intended target of Trim-Away, GFP-tagged protein kinase C (GFP-aPKC), which behaves as a monomer in non-polarized cells⁵². At the onset of blue light stimulation, both proteins were expressed and distributed diffusely in the cytoplasm, suggesting that recruitment of TRIM21 to GFP-aPKC was insufficient to cause degradation. However, within minutes of blue light stimulation GFP-aPKC formed clusters together with vhhGFP4-RFP-CRY2-TRIM21 that were rapidly degraded (Fig. 6b,c). Degradation of GFP-aPKC was TRIM21-dependent, because co-clustering of aPKC with RFP-CRY2-vhhGFP4 in the absence of TRIM21 did not lead to degradation (Fig. 6d,e).

Our data suggests that intermolecular clustering activates TRIM21 by driving RING dimerisation. In support of this model, we found that just the TRIM21 RING domain alone was sufficient to induce proteasomal degradation upon molecular clustering (Fig. 6f). We exploited this finding to engineer a minimal degrader molecule that combines the catalytic activity of the TRIM21 RING together with a targeting domain in a single polypeptide chain (mini-Trim-Away; T21R-vhhGFP4; Fig. 6g). Strikingly, expression of T21R-vhhGFP4 caused rapid and stable proteasomal degradation of targets Caveolin-1-EGFP and H2B-mEGFP (Extended Data Fig. 5).

A key advantage of the original Trim-Away method is the direct delivery of protein reagents (whole antibodies) to the cell, which allows rapid target protein degradation even in primary cells that are sensitive to classical nucleotide delivery approaches¹⁷. We found that the T21R-vhhGFP4 protein expressed exceptionally well in bacteria and could be purified using a simple two-step protocol to give high yields and purity (Fig. 6h; Methods). Electroporation of this protein into Caveolin-1-EGFP knock-in cells resulted in dose-dependent degradation of Caveolin-1-EGFP beginning immediately after protein delivery (Fig. 6i). The small size of the T21R-vhh protein reagent (27kDa) makes it ideally suited to target proteins sequestered in the nucleus that cannot be accessed by whole antibodies¹⁷. To test this, we electroporated the protein into primary mouse embryonic fibroblasts (MEFs) isolated from H2B-GFP transgenic mice. Remarkably, H2B-GFP was degraded within 1h of T21R-vhhGFP4 protein electroporation (Fig. 6j). Protein delivery outperformed mRNA delivery in primary MEFs (Fig. 6j), likely because primary cells are sensitive to exogenous nucleotides, highlighting the advantage of the T21R-vhh protein reagent for Trim-Away in primary cells.

Mice

The experiments involving mice have been approved by the Medical Research Council Animal Welfare and Ethical Review Body and the UK Home Office under project licence number PCF3F9520. All mice were maintained in a specific pathogen-free environment according to UK Home Office regulations. H2B-GFP mice (CAG:H2B-EGFP) were obtained from The Jackson Laboratory (strain 006069).

Cell culture

HEK293T (ATCC) and NIH3T3-Caveolin-1-EGFP⁴⁵ were cultured in DMEM medium (Gibco; 31966021) supplemented with 10% Calf Serum and penicillin-streptomycin. RPE-1 cells (ATCC) were cultured in DMEM/F-12 medium (Gibco; 10565018) supplemented with 10% Calf Serum and penicillin-streptomycin. All cells were grown at 37°C in a 5% CO₂ humidified atmosphere and regularly checked to be mycoplasma-free. The sex of NIH3T3 cells is male. The sex of HEK293T and RPE-1 cells is female. Drosophila S2-DGRC cells (Drosophila Genomics Resource Centre) were cultured at 25°C in Schneider’s Insect medium (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% Calf serum. For primary mouse embryonic fibroblasts (MEFs) isolation, uteri isolated from 13.5-day-pregnant mice were washed with PBS. The head and visceral tissues were removed from isolated embryos. The remaining bodies were washed in fresh PBS, transferred into 3 ml of 0.25% trypsin, 1mM EDTA and minced using 2 scalpels. After 20 min incubation at 37°C cell suspension was mixed with 25ml of DMEM medium supplemented with 10% Calf Serum and strained through 40 μm cell strainer. Cells were collected by centrifugation (200 × g for 5 min at 4°C), resuspended in fresh medium and seeded in T75 flask (passage 1). In this study, we used MEFs within three passages to avoid replicative senescence.

Plasmids

A full list of plasmids used in this study and details of plasmid construction can be found in Supplementary Note 2. For stable expression of TRIM21 at endogenous levels, a native TRIM21 promoter-driven lentiviral vector (pT21MPP) was used⁵⁸. For in vitro mRNA transcription, constructs were cloned into pGEMHE⁵⁹, which contains UTR and polyA sequences for optimal mRNA stability and translation. For all TRIM21 constructs, sequences encoding amino acids 1-85 were used for the RING domain (T21R), amino acids 1-129 for RING-Box (T21RB), amino acids 132-235 for Coiled-Coil (T21CC235) and amino acids 1-235 for RING-Box-Coiled-Coil (T21RBCC).

TRIM21 knockout cell lines

HEK293T TRIM21 knockout cells were described previously³⁸. RPE-1 TRIM21 knockout cells were generated using the Alt-R CRISPR-Cas9 system from Integrated DNA technologies (IDT) with a custom-designed crRNA sequence (ATGCTCACAGGCTCCACGAA). Guide RNA in the form of crRNA-tracrRNA duplex was assembled with recombinant Cas9 protein (IDT #1081060) and electroporated into RPE-1 cells together with Alt-R Cas9 Electroporation Enhancer (IDT #1075915). Two days post-electroporation cells were plated one cell per well in 96 well plates and single cell clones screened by western blotting for TRIM21 protein. A single clone was chosen that contained no detectable TRIM21 protein and confirmed TRIM21 knockout phenotype in a Trim-Away assay.

Stable cell lines

Stable cell lines generated and used in this manuscript are detailed in Supplementary Note 2. HEK293T-mCherry-hTRIM21 cells were described previously¹⁷. Lentivirus particles were collected from HEK293T supernatant 3 days after co-transfection of lentiviral plasmid constructs together with HIV-1 GagPol expresser pcRV1 (a gift from Dr. Stuart Neil) and pMD2G, a gift from Didier Trono (Addgene plasmid #12259). Supernatant was filtered at 0.45 μm before storage at -80°C. For stable expression of TRIM21 at endogenous levels, HEK293T TRIM21 KO cells were transduced with lentivirus particles at multiplicity ~0.1 transducing units per cell and selected using puromycin at 2.5 μg/ml from 48 hours post-transduction⁵⁸. For stable expression of H2B-mEGFP-FKBP, RPE-1 cells were transduced at multiplicity ~0.1 transducing units per cell, sorted into single cells, and a single cell clone selected after screening for GFP expression by flow cytometry.

Electroporation

Electroporation was performed using the Neon® Transfection System (Thermo Fisher). Cells were washed with PBS and resuspended in Buffer R at a concentration of 8 × 10⁷ cells ml^-1. For each electroporation reaction 8 × 10⁵ cells (10.5 μl) were mixed with 2 μl of antibody (typically 0.5 mg/ml) or mRNA (typically 0.5 μM) or protein to be delivered. The mixture was taken up into a 10 μl Neon® Pipette Tip, electroporated at 1400V, 20 ms, 2 pulses and transferred to media without antibiotics.

Ubiquitin and TRIM21 pull-down assays

Approximately 4×10⁵ cells were transfected with 250 ng of pMT107-octa-6 His-ubiquitin or 200 ng of pcDNA3-2x-6His-ubiquitin. Media was changed 24h post-transfection. At 48h post-transfection, media was replaced with serum-free CO₂ independent medium (Gibco; 18045088) and the cells were chilled on ice for 30 min before infection with AdV5 or AdV5-9C12 complex for a further 30 min on ice. ΔE1ΔE3 AdV5-GFP (ViraQuest) was diluted 1:4 in PBS and mixed 1:1 with either 20 μg/ml of humanised 9C12 IgG3b or PBS and incubated for 1 hr at 20C to allow for complex formation. Following infection, cells were incubated at 37°C for 30 min, washed twice with PBS and collected by scraping. Cells pellets were collected by centrifugation and flash frozen in liquid nitrogen. For the denaturing His-Ubiquitin pulldowns, cell pellets were lysed with 500 μl of buffer A containing 6M Guanidinium chloride, 5 mM imidazole 100 mM sodium phosphate buffer pH 7.4, mixed by vortexing and sonicated in ultrasonic water bath for 2 min (10s on/off) with amplitude to 40% to shear the DNA. 20 μl of buffer equilibrated Ni-NTA bead slurry (50% bead) were added to each sample and incubated for at least 3 hrs on a rotating wheel at 4C. Beads were washed 2x with buffer A, 2x with buffer B (buffer A and buffer C mixed 1:3) and 2x with buffer C (20 mM Tris, 20 mM imidazole pH 6.8). Supernatant was removed using 30 G needle and beads boiled for 5 min in 10 μl of 2 X LDS + 100 mM DTT + 300 mM imidazole for SDS-PAGE analysis. For denaturing TRIM21-His pulldown, lysis buffer containing 4M Urea, 5 mM imidazole 50 mM Tris pH 7.4 was used and where indicated, the washed Ni-NTA beads were treated with USP2 (Boston Biochem; K-400) before immunoblotting.

Adenovirus neutralisation assay

HEK293T Cells were plated at a density of 5×10⁴ cells/well in 0.5 ml of media in 24-well plates and were allowed to attach overnight. ΔE1ΔE3 AdV5-GFP (Viraquest) was diluted 1:1250 in PBS and mixed 1:1 with 9C12 antibody at the indicated concentrations and incubated for 1 hour at 20°C to allow for complex formation. For estimating the stoichiometry of TRIM21 for neutralisation, 9C12(H433A) antibody was mixed at the indicated ratio with 9C12(WT) at 1 mg/ml before addition to AdV5-GFP particles. For the infection, 10 μl of the virus or virus-antibody complex was added to each well. Cells were harvested by trypsinisation at 16-20 hrs post infection and evaluated for GFP expression on a BD LSRFortessa cell analyser (BD Biosciences). The results were analysed using FlowJo software (FlowJo LLC) and relative infection was calculated using the method described previously⁷.

NFκB signalling assay

Approximately 3×10⁵ cells were transfected with 200 ng pGL4.32 NF-κB luciferase plasmid (Promega). Cells were incubated for six-hours at 37°C before reseeding at a density of 1 × 10⁴ per well in 96-well plates (Corning CellBIND; 3340) and allowed to attach overnight. ΔE1ΔE3 AdV5-GFP (ViraQuest) was diluted 1:3 in PBS and mixed 1:1 with either 20 μg/ml of humanised 9C12 IgG3b or PBS and incubated for 1 hr at 20°C to allow for complex formation. 5 μl of the virus or virus-antibody complex was added to each well in triplicates and allowed to incubate for 6 hrs before the cells were lysed with 50 μl/well of steadylite plus luciferase reagent (Perkin Elmer; 6066751). The luciferase activity was quantified using a BMG PHERAstar FS plate reader.

Light-induced CRY2 clustering

Drosophila S2 cells were transiently transfected with Hsp70-promotor-controlled optogenetic constructs, seeded on glass bottom-dishes (MatTek Corporation, Ashland, MA, USA) coated with poly-lysine and between 2-5 days post-transfection expression induced by incubation at 37°C for 40 min at least 5h prior to cell imaging. After induction, cells were kept in the dark and manipulated under a 593nm LED light (Super Bright LEDs, St. Louis, MO, USA). For proteasome inhibition, culture medium was supplemented with 20 μM MG132 at least 1h before live imaging. The 470nm filter of a Nikon Ti motorised inverted epifluorescence widefield microscope (Nikon, Japan) was used to induce CRY2 clustering, whereas the 555 nm filter was used to image RFP. For RPE-1 cells, following electroporation of mRNA, cells were incubated overnight. The following day, cells were shifted to CO₂-independent medium (Gibco; 18045088) supplemented with 10% Calf Serum and Glutamax (Gibco; 35050061) with either DMSO or 20 μM MG132 before being exposed to blue light (wavelength of 450-500 nm) from a Dark Reader transilluminator (Clare Chemical). Cells were harvested for immunoblot analysis after 3 hours of exposure to blue light. The washed cell pellets were lysed directly in 60 μl of 1x LDS sample buffer, sonicated in a water bath sonicator for 5 min (30s on/off, 4°C) and boiled at 95°C for 10 minutes before loading on to NuPAGE 4-12% Bis-Tris gels (Thermo Fisher).

Microscopy

Live imaging of mammalian cells was performed using the IncuCyte S3 live cell analysis system (Sartorius) housed within a 37°C, 5% CO₂ humidified incubator. Live imaging of Drosophila S2 cells was performed using a Nikon Ti motorised inverted epifluorescence widefield microscope (Nikon, Japan), equipped with a Lumencor SpectraX light engine and external emission filter wheel, and an Andor iXon888 camera (Andor, UK), with a PL APO LAMBDA 60X OIL/1,4 Oil WD 0,13mm objective, controlled by NIS elements 5.0 (Nikon, Japan) software. Z-stacks were acquired as 7 serial sections l.5μm apart. For fixed cells, images were acquired with a Leica SP8 microscope equipped with a 63x C-Apochromat 1.2 NA oil immersion objective.

Measurement of fluorescence in live cells

To quantify RFP and GFP degradation in S2 cells, images were captured just before blue light exposure and in 1-minute (Fig. 5b,c) or 2-minute (Fig. 6b-e) intervals from blue light exposure onwards. Using ImageJ/FIJI software⁶⁰, the mean pixel intensity was measured within regions of interest (ROIs) that were manually tracked to include the complete area of each individual cell in which clustering occurred. Fluorescence intensity was quantified in average-intensity Z-projections covering all imaged planes and the signal was also measured in 3 non-transfected cells in the same sample, which were used for background subtraction. Fluorescence intensity over time is shown as the ratio between each timepoint and the second timepoint after light-induction. To quantify GFP fluorescence in live mammalian cells, images were acquired and analysed using the IncuCyte live cell analysis system (Sartorius). Within the IncuCyte software, the integrated density (the product of the area and mean intensity) for GFP fluorescence was normalised to total cell area (phase) for each image.

Immunofluorescence

Cells were fixed for 30 min at 37°C in 100 mM HEPES (pH 7; titrated with KOH), 50 mM EGTA (pH 7; titrated with KOH), 2% formaldehyde (methanol free) and 0.2% Triton X-100 and incubated in blocking buffer (PBS, 3% BSA and 0.1% Triton X-100) for 1h at room temp prior to antibody incubations. Antibodies and dilutions (in blocking buffer) used for immunofluorescence (IF) are detailed in Supplementary Note 2. DNA was stained with 5 mg ml⁻¹ Hoechst 33342 (Molecular Probes H3570). To measure mEGFP fluorescence, Image J software was used to segment images in the mEGFP channel and measure mean mEGFP intensity within the segmented region. H2B-mEGFP-positive nuclei and mEGFP-PACT foci, as well as total nuclei (Hoechst), were manually counted using the Image J cell counter tool.

Immunoblotting

Samples were run on NuPAGE 4-12% Bis-Tris gels (Thermo Fisher) and transferred onto nitrocellulose membrane. Membranes were incubated in blocking buffer (PBS, 0.1% Tween20, 5% milk) for 1h at room temp prior to incubation with antibodies. Antibodies and dilutions (in blocking buffer) used for immunoblotting (IB) are detailed in Supplementary Note 2. HRP-coupled antibodies were detected by enhanced chemiluminescence (Amersham, GE Healthcare) and X-ray films. IRDye-coupled antibodies were detected using LI-COR Odyssey CLx imaging system.

Protein Purification

Ubiquitin, Ube1, Ube2N(K92R), T21R (amino acids 1-85) and T21RB (amino acids 1-129) were produced as described previously^19,20,38. The T21RLR and T21RBCC (amino acids 1-235) proteins were produced using the same protocol as T21RB³⁸. To generate full length TRIM21 protein (T21FL), MBP-tagged TRIM21 (pFastBac-MBP-TRIM21) was expressed in SF9 insect cells in Lonza Insect-XPRESS according to standard protocols (Lonza). Cleared cell lysates were prepared by sonication of cell pellets in 50 mM Tris pH 8, 200 mM NaCl, 1 mM DTT, 10 μM ZnCl₂, 15% (vol/vol) BugBuster (Novagen) and cOmplete protease inhibitors (Roche), followed by centrifugation 16,000 × g for 30 min. Lysates were loaded onto Amylose resin and the protein eluted with 10 mM Maltose. MBP was cleaved from the eluted protein with TEV protease, and the protein was passed through a HiLoad 16/600 Superdex 200 PG size exclusion column using an ÄKTA pure purification system (GE Healthcare). Peak fractions were pooled and concentrated using Amicon Ultra-4 centrifugal filter units (Millipore). Proteins MBP-T21CC235 (containing MBP fused to TRIM21 residues 129-235) and T21CC235 (containing TRIM21 residues 129-235) were expressed in Escherichia coli C41(DE3) cells in 2TY medium and lysed and purified by affinity (NiNTA- or Amylose-resin, as appropriate) and size exclusion chromatography (SEC) as above, except the MBP tag was retained on MBP-T21CC235.

To generate the TRIM21-nanobody fusion protein, pOPTH-T21R-vhhGFP4 was transformed into E. coli C41(DE3) and TRIM21-nanobody protein expressed using autoinduction media (ZYP-5052) for 20 hours at 30°C. Cleared cell lysates were prepared by sonication of cell pellets in 50 mM Tris pH 8, 300 mM NaCl, 2 mM DTT, 10 mM Imidazole pH 8, 15% (vol/vol) BugBuster (Novagen) and cOmplete protease inhibitors (Roche), followed by centrifugation 16,000 × g for 30 min. Lysates were loaded onto Ni-NTA agarose (Qiagen), and protein eluted with 400 mM Imidazole. Eluted protein was passed through a HiLoad 10/300 Superdex 75 GL size exclusion column using an ÄKTA pure purification system (GE Healthcare). Peak fractions were pooled and concentrated using Amicon Ultra-4 centrifugal filter units (Millipore). For electroporation into cells, proteins were dialysed in PBS for 2 × 2 hours using Slide-A-Lyzer™ Dialysis Cassettes (Thermo Fisher). Protein aliquots were frozen and stored at -80°C until use.

Ubiquitin discharge assay

Ube2N(K92R) was loaded with ubiquitin by mixing 40 μM of the E2 with 1 μM Ube1, 0.37 mM Ub and 3 mM ATP in 50 mM HEPES pH 7.5, 150 mM NaCl, 20 mM MgCl₂, and incubating the reaction at 37°C for 30 min. The reaction was transferred to ice and used immediately. To observe E3 mediated discharge of ubiquitin, 2 μM ubiquitin-loaded E2 was mixed with 2.5 μM Ube2V2 and 1.5 μM TRIM21 protein constructs in 50 mM HEPES pH 7.5, 150 mM NaCl, 20 mM MgCl₂, 50 mM L-lysine. Samples were taken at timepoints post-mixing and reactions stopped by adding LDS sample buffer at 4°C. The samples were boiled, resolved by SDS-PAGE and observed by immunoblot using rabbit anti-Ube2N (Bio-Rad, AHP974, 1:1000).

Ubiquitin chain formation assay

Reactions were carried out in 50 mM Tris at pH 8, 2.5 mM MgCl2, 0.5 mM DTT with 0.2 mM Ub, 2 mM ATP, 1 μM Ube1, 0.5 μM Ube2N, 0.5 μM Ube2V2 and 1.5 μM TRIM21 protein constructs. The reaction was started upon incubation at 37°C for the time points indicated in the figures. The reaction was stopped via the addition of LDS sample buffer at 4°C, followed by boiling at 90°C for 2 min. The reactions were resolved by SDS-PAGE and ubiquitin detected by immunoblot using anti-ubiquitin-HRP (P4D1; Santa Cruz Biotechnology sc8017-HRP, 1:1000).

Isothermal Titration Calorimetry (ITC)

ITC dilution experiments were performed using an iTC200 titration calorimeter (Malvern Panalytical). T21CC235 was carefully dialysed into 20mM TRIS, pH 8, 100mM NaCl buffer at diluted to a concentration of 200uM. This was titrated into the same dialysis buffer in the ITC cell using twenty 2 μL injections. The resultant endothermic heat effects were integrated using Origin software supplied with the calorimeter.

SEC-MALS

Size exclusion chromatography coupled to multi angle light scattering (SEC-MALS) measurements were performed using a Wyatt Heleos II light scattering instrument coupled to a Wyatt Optilab rEX online refractive index detector (Wyatt Technology). Samples of T21CC235 (100uL) were resolved on a Superdex S-200 10/300 analytical gel filtration column (GE Healthcare) running at 0.5 ml/min in 20 mM Tris pH 8, 250mM NaCl buffer before passing through the light scattering and refractive index detectors in a standard SEC-MALS format. Protein concentration was determined from the excess differential refractive index based on 0.186 RI increment for 1 g/ml protein solution. The concentration and the observed scattered intensity at each point in chromatogram were used to calculate the molar mass from the intercept of the Debye plot using Zimm’s model as implemented in Wyatt’s ASTRA software.

Analytical ultracentrifugation (AUC)

T21RBCC at approximately 0.5 mg/mL in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM DTT was loaded into 12 mm 2-sector cells, placed in an An50Ti rotor and centrifugated 50,000 rpm at 10 °C using an Optima XL-I analytical ultracentrifuge (Beckman). The data were analysed in SEDFIT 16.1c⁶¹ using a c(s) distribution model. The partial-specific volumes (v-bar), density and viscosity of the buffer were calculated using Sednterp (Dr Thomas Laue, University of New Hampshire). Data were plotted with the program GUSSI⁶².

Small Angle X-ray Scattering (SAXS)

SAXS data (I(q) vs q, where q = (4πsinθ)/λ, where λ = 1.24 Å (10 kEV), 2θ = scattering angle), were collected at the beam line P12 of the EMBL at the Petra-III storage ring (DESY, Hamburg). Samples were collected at 277 Kelvin using an EMBL/ESRF new generation automated sample changer (1.5 mm sample capillary). Samples were measured while flowing through the capillary with scattering intensities recorded using 50 ms exposure times (20 frames total) on a 2D photon counting Pilatus 2M pixel X-ray detector (Dectris), maintaining a fixed sample to detector distance of 3.1 m (q-range 0.008-0.47 Å^-1). For processing details, see Supplementary Note 1.

Mass Photometry

GFP:antibody mixtures were prepared in 300 mM NaCl, 50 mM Tris pH 8 and 1 mM DTT and incubated on ice prior to loading onto the Refeyn One^MP instrument (Refyn Ltd.) according to manufacturer’s instructions. Data were processed and analysed using Discover MP v1.2.4 (Refeyn Ltd.). Events were then normalised with a calibration curve and the resulting data exported and processed in GraphPad Prism.

Statistical Analysis

Average (mean), standard deviation (s.d.), standard error of the mean (s.e.m) and statistical significance based on Student’s t-test (two-tailed) and one- or two-way ANOVA were calculated in Microsoft Excel or Graphpad Prism. Significance are represented with labels ns (not significant, P>0.05), * (P≤0.05), ** (P≤0.01), *** (P≤0.001), **** (P≤0.0001).