Zinc-mediated regulation of TMEM106B function in the endolysosomal pathway revealed through the systematic identification of TRIM2/3 ubiquitination substrates

doi:10.21203/rs.3.rs-5037938/v1

Download PDF

Article

Zinc-mediated regulation of TMEM106B function in the endolysosomal pathway revealed through the systematic identification of TRIM2/3 ubiquitination substrates

https://doi.org/10.21203/rs.3.rs-5037938/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The understanding of ubiquitin E3 ligase function hinges on thoroughly identifying their cellular targets, but the transient nature of signaling complexes leading to ubiquitination poses a significant challenge for detailed mechanistic studies. TRIM2 and TRIM3 are paralogous mammalian E3 ligases with particularly high expression in the brain, where they contribute to neuronal development and homeostasis. Here, we tailored recently developed ubiquitin-specific proximity labelling tools to identify substrates of TRIM2 and TRIM3 activity. We show that despite their high amino acid sequence identity, the ligases have distinct intracellular dynamics, binding partners, and ubiquitination substrates. Using biochemical and structural studies, we show that TRIM2 ubiquitinates the lysosomal protein TMEM106B at lysine residues located in the cytosolic N-terminal region. Substrate recognition involves a direct interaction between TRIM2 and a newly identified zinc-coordination motif in TMEM106B that mediates homodimerization and is required for lysosomal size regulation. We found that in addition to catalysis, the tripartite motif is involved in substrate recruitment, and we provide insights into the assembly of the ubiquitination complex. Deletion of TRIM2/TRIM3 in mouse embryonic stem-cell derived neurons impacted the extracellular matrix composition, likely through acting on the endolysosomal pathway. Our study thus contributes a catalogue of TRIM2 and TRIM3-associated effectors and supports a key role at the interface of vesicle trafficking and the cytoskeleton.

Biological sciences/Molecular biology/Proteomics/Protein–protein interaction networks

Biological sciences/Molecular biology/Post-translational modifications/Ubiquitylation

Biological sciences/Structural biology/NMR spectroscopy

Biological sciences/Structural biology/X-ray crystallography

Health sciences/Neurology/Neurological disorders/Neurodevelopmental disorders

The post-translational modification (PTM) of proteins with ubiquitin is one of the major cellular signals that maintain proteome homeostasis (“proteostasis”)^1–3. Imbalances in the levels, localizations, and folding states of a multitude of protein substrates can be countered by the concerted action of E1 activating enzymes, E2 conjugating enzymes, and E3 ligases in an ATP-dependent manner^4,5. As a result of this enzymatic cascade, the C-terminal glycine residue of ubiquitin is covalently linked to a particular target, most commonly through a lysine-mediated isopeptide bond^4,5. Mono- or poly-ubiquitin modification leads to various downstream effects, including protein trafficking and degradation through recruitment of ubiquitin binding entities that can specifically recognize the “ubiquitin code”^4–6.

The loss of proteostasis is a hallmark of diseases related to aging⁷. Many of the genes associated with neurodegeneration are involved in protein quality control, with E3 ligases being a particularly prominent group among these ^8,9. Because these enzymes catalyze the last and most specific step in the ubiquitination cascade, the identification of their substrates has been the focus of extensive work to understand their biological roles¹⁰. However, substrate-ligase interactions are weak and transient, making the systematic identification of these pairings challenging¹⁰. Recently, versatile proximity-based labelling methods specific for enriching ubiquitinated proteins have been developed^11–13. While still relying on the physical association of the ligase and the substrate, these tools additionally require ubiquitin to be covalently attached, and thereby address the difficulties of identifying the short-lived interactions that are commonplace in signalling cascades.

TRIM2 and TRIM3 are paralogous E3 ligases that are highly expressed in the brain and share close to 70% amino acid sequence identity (Extended Data Fig. 1). They belong to the tripartite motif (TRIM) family, defined by the presence of an N-terminal RING domain, one or two B-boxes, and a coiled-coil domain that mediates dimerization¹⁴. Animal models of TRIM2 deletion develop uncontrolled muscle movements (i.e., ataxia) in ageing mice, and peripheral neuropathy associated with the loss of Purkinje neurons of the cerebral cortex, where TRIM2 is particularly abundant¹⁵. The neuromuscular disorder Charcot-Marie-Tooth Disease is associated with TRIM2 mutations in humans and a dysregulation of endosomal trafficking in cell culture models^16–18. Deletion of TRIM3 impacts hippocampal plasticity and short-term memory in mice, as well as the regulation of dendritic spine morphology^19,20. These and many other observations point to TRIM2 and TRIM3 having crucial roles in neuronal health and differentiation^21–26.

Our understanding of the diverse functions of the TRIM family at the molecular level is still in its nascent phase. It is unclear to what extent the ubiquitination activity of the RING domain is necessary for their biological roles, what cofactors are involved in mediating activity, and how much functional overlap exists among highly paralogous members. For example, TRIM3 was shown to affect the motility of the kinesin motor protein KIF21B independently of the RING domain or protein degradation, suggesting that its non-catalytic functions are crucial²⁴. Several TRIM ligases, including the closely related TRIM32, have also been shown to act as SUMO ligases²⁷. TRIM3 can bind the SUMO conjugating enzyme UBE2I, and has relatively low ubiquitination activity in vitro, questioning the primary role of TRIM3 as a bona fide ubiquitin ligase^28,29. Nevertheless, ubiquitination activity has been demonstrated for both TRIM2 and TRIM3. TRIM2 can modify Bcl-2-interacting mediator of cell death (Bim) and neurofilament light chain (NF-L), while TRIM3 was shown to promote ubiquitination of guanylate kinase-associated protein (GKAP)^20,23,30. In addition, tandem-affinity ubiquitin binding entities (TUBEs) were used to identify γ-actin 1 (ACTG1) as a ubiquitination substrate of TRIM3 in hippocampal tissue¹⁹. Despite these findings, a general function for TRIM2 and TRIM3 as ubiquitin ligases has not been identified, their catalogue of substrates and associated cofactors is still largely unknown, and there is little mechanistic insight into how ubiquitination occurs.

The C-terminal regions in the TRIM family are thought to recruit substrates and are thus associated with target specificity^14,31. TRIM2 and TRIM3 contain Filamin and NHL (NCL-1, HT2A and Lin-41 homology) domains, and are hence grouped in the TRIM-NHL subfamily¹⁴. The two ligases exhibit small variations in protein sequence, and their overall structures as predicted by AlphaFold is not obviously different³². This raises the question of whether they recruit similar proteins through the Filamin and/or NHL domains, and if not, what is the basis for their substrate specificity. Previous work showed divergent oligomerization characteristics and E3 ligase activities of TRIM2 and TRIM3 in vitro, with potential implications for their cellular function²⁸. Specifically, the authors found that amino acids in the vicinity of the RING domain core, present in TRIM2 but not TRIM3, are responsible for promoting dimerization and catalytic activity. Whether these observations extend to their cellular behavior, however, has been unclear.

In this study, we tailored proximity-based labelling tools coupled to quantitative mass spectrometry (MS) to uncover previously unknown ubiquitination substrates and cofactors associated with TRIM2 and TRIM3 activity. We identified the lysosomal protein TMEM106B using this approach and validated that it is a bona fide TRIM2 target. Specifically, we show that TRIM2 ubiquitinates the N-terminal cytoplasmic region of TMEM106B in vitro, and that the two proteins interact directly. We find that substrate recruitment occurs through a region of TMEM106B involved in homodimerization that forms upon binding of a zinc metal ion, and we determine the structure of this previously unknown structural motif. Importantly, we demonstrate that zinc binding is key to the function of TMEM106B in lysosomal size regulation. These experiments are complemented with insight from mouse embryonic stem cell-derived neurons showing a crucial role of the ligases in maintaining the composition of the extracellular matrix (ECM), which we hypothesize is linked to their role in the endolysosomal pathway and vesicular trafficking.

Proteins affected by TRIM2/3 deletion in mESC-derived neurons are associated with extracellular matrix maintenance

TRIM2 and TRIM3 deletion in animal models is associated with neurodegeneration and memory impairment, respectively, but mice deficient in either of these proteins show relatively mild phenotypes early in life^19,23. To better understand their function as E3 ligases, we used the CRISPR-Cas9 system in mouse embryonic stem cells (mESCs) to guide the deletion of the coding exons of their catalytic RING domains. To reduce compensatory effects that could arise due to the high similarity of the two ligases (Extended Data Fig. 1), we targeted both genes simultaneously. We thus obtained two independent TRIM2/TRIM3 double knockout (DKO) cell lines, referred to as DKO1 and DKO2, which differentiated into glutamatergic neurons in the presence of retinoic acid according to published protocols (Fig. 1a)³³.

The WT and DKO cell lines differentiated without obvious phenotypic differences at the cellular level, as examined via wide-field light microscopy (Fig. 1a). Immunoblotting and quantitative MS confirmed the depletion of TRIM2/TRIM3 at day 12 of differentiation, when glutamatergic neurons are present³³ (Fig. 1b, c). TRIM3 levels were markedly reduced. Endogenous TRIM2 exhibits two proteoforms of distinct molecular weight (MW) corresponding to alternative splicing³⁴. The lower MW proteoform remained in the DKO cell lines, consistent with the MS results showing a mild reduction in TRIM2 levels (Fig. 1c).

We investigated the differences in protein expression comparing the WT and DKO neurons, focusing on those proteins which showed consistent changes in both DKO1 and DKO2 (see Methods) (Fig. 1c, d, Supplemental Table 1). Approximately 40 proteins were more highly expressed in WT neurons, enriched in ECM-associated proteins. These included membrane receptors, various collagen types (e.g., COL1A1 and COL3A1), adhesion molecules (POSTN, PCDH18, and PGM5), and metalloproteinases (MMP14) (Fig. 1c-e). Five proteins were more abundant in the DKO cells, including the lysosomal proteins MANBA and ATP6V0C. Thus, the combined effect of removing the catalytic activity of TRIM2 and TRIM3 is a loss in many of the proteins that shape the cellular periphery of neurons and is consistent with their biological roles in differentiation and homeostasis.

Distinct intracellular dynamics and ubiquitination substrates of TRIM2 and TRIM3

Next, we sought to disentangle the individual contributions of TRIM2 and TRIM3 ligase activity, and adapted a ubiquitin-specific proximity labelling approach described recently to identify ubiquitination substrates (Fig. 2a)¹³. In this method, an optimized AviTag present on ubiquitin (AviTag-Ub) allows specific BirA-mediated biotinylation of post-translationally modified substrates upon recruitment by BirA-tagged TRIM2 and TRIM3. Although both proteins have prominent roles in the brain, neurons do not readily take up exogenous DNA³⁵. HEK 293-T cells were thus used to allow for efficient co-transfection and protein expression of the AviTag-Ub and BirA-TRIM plasmids simultaneously (Fig. 2b and Extended Data Fig. 2a).

We first compared the expression levels of TRIM2 and TRIM3 under identical experimental conditions. We found that WT TRIM2 was expressed at low levels in the absence of proteasome inhibition, and protein levels decreased further upon co-transfection with AviTag-Ub (Fig. 2b). However, single point mutations of TRIM2 targeting the coordination of structural zinc atoms on the RING domain (C23S or C60S) led to its accumulation in cells without the need for proteasome inhibition (Fig. 2b). Autoubiquitination of TRIM2 in cells has been previously reported²³, but it was unclear whether this led to its degradation. Our results suggest that TRIM2 can be degraded in a proteasome-dependent manner aided by its own catalytic activity.

TRIM3, in contrast, exhibited small differences in expression levels between WT and catalytically inactive (“dead”) variants (C22S or C59S), and did not require proteasome inhibition for stable expression in cells. Its expression was also unaffected by co-transfection with AviTag-Ub (Fig. 2b and Extended Data Fig. 2a). Thus, despite their high sequence similarity, TRIM2 and TRIM3 exhibit distinct intracellular stabilities. Based on these results, we carried out the experiments either in the presence (TRIM2), or absence (TRIM3) of proteasome inhibition. Protein substrates were identified using MS based on tandem mass tag (TMT) labelling³⁶ after enrichment of the biotinylated substrates with streptavidin pulldowns (Fig. 2a).

In the case of TRIM2, we identified 29 proteins enriched in the presence of the WT variant relative to both negative controls (BirA-TRIM2^dead and BirA) (Fig. 2c, Supplemental Table 2). The most significantly enriched target was the myosin phosphatase regulatory subunit PPP1R12A (also called MYPT1), which had not been previously identified as a TRIM2 substrate. MYPT1 is highly expressed in many tissues and has a role in regulating mechanical forces and ECM assembly by affecting the actomyosin-microtubule crosstalk³⁷. Enrichment analyses found gene ontology (GO) terms such as ubiquitin-like protein ligase binding (Extended Data Fig. 2b), consistent with its role as an E3 ligase. More notably, we found terms related to multivesicular body transport, endocytosis, and autophagy (Fig. 2d), due to the presence of proteins such as HGS, STAM, STAM2, and RABEP2 (Fig. 2c). Genes related to the antiviral RIG-I-like receptor pathway also featured prominently (e.g., TRIM25, TRAF3, CYLD) (Fig. 2d and Supplemental Table 2). Despite conducting our experiments in HEK 293-T cells, we also found several proteins associated with neuronal homeostasis, such as TMEM106B, ITM2B, YARS, SLC20A1, and TRIB3 (Fig. 2c). In summary, TRIM2 can associate with a wide variety of cytosolic and membrane proteins. Its link to multivesicular bodies and the ESCRT-0 components STAM and STAM2 suggest that ubiquitination of substrates destined for the endolysosomal pathway is its primary function as a ubiquitin ligase.

In the case of TRIM3, we identified a smaller set of 15 candidate substrates with relatively low enrichment ratios (Fig. 2e). These included CAPRIN1 and RAD23B as the two most enriched. CAPRIN1 is a well-established RNA-binding protein that associates with stress and neuronal granules^38–42. TRIM3 has been previously implicated in the formation of messenger ribonucleoproteins (mRNP) in neurons^19,43, and the identification of CAPRIN1 is consistent with its capacity to associate with RNA-containing particles. RAD23B is a proteasome shuttling factor involved in DNA repair⁴⁴ and was shown to be responsible for liquid-liquid phase separation of the proteasome in the presence of increased levels of ubiquitinated proteins induced by hyperosmotic stress⁴⁵. We found several proteasomal subunits amongst our identified proteins, suggesting that TRIM3 can participate in the formation of protein assemblies composed of RAD23B, the proteasome, ubiquitinated factors, and “orphan” ribosomal subunits (Supplemental Table 2)⁴⁵. Previous work and this study also show relatively low ubiquitination activity of TRIM3, which could additionally impact substrate identification (see below). Nevertheless, we found several substrates or TRIM3-associated cofactors, including the metabolic enzymes PGAM1 and IMPDH2 involved in glycolysis and guanine biosynthesis, respectively; CKAP2 which affects the stability and nucleation of microtubules; and FKBP3 that has a role in immunoregulation^46–49.

To provide spatial context to the substrates identified above, we monitored the localization of GFP-tagged TRIM2 and TRIM3 using live confocal imaging (Fig. 2f, g). This approach avoided artifacts arising from fixation and antibody staining⁵⁰. In cells with relatively low expression, TRIM2 was found diffuse in the cytosol and along filaments. Cells with high TRIM2 levels, in contrast, exhibited irregular nuclei and blebbing, indicating that they may be apoptotic. In contrast, TRIM3 was often found concentrated as part of large cytosolic structures in intact cells, which could explain its association with proteins involved in phase separation (Fig. 2f, g). Recent findings in human osteosarcoma U2OS cells showed that TRIM2/3 can associate with microtubules through the NHL domain⁵¹. We observed that the GFP-TRIM2/3 signals found along filaments correlated also with microtubules in HEK 293-T cells (Extended Data Fig. 2c). Notably, several of the substrates identified above are known to associate with microtubules and/or centrosomes, including CKAP2, CEP85, and HMMR. Thus, the subcellular distribution of TRIM2 and TRIM3 is consistent with the types of proteins that they ubiquitinate and highlight their distinct intracellular behaviors.

We pooled the proteins differentially expressed in mESC-derived neurons and the substrate candidates identified in HEK 293-T cells to gain insights into commonly affected pathways through network analysis (Extended Data Fig. 3)⁵². Distinct sets of enriched proteins arose from the two types of experiments, with those involved in the endolysosomal pathway providing a possible link to remodelling of the ECM⁵³. By regulating protein shuttling in and out of the cell, we hypothesize that TRIM2/3 sustain key factors at the cellular periphery important for neuronal homeostasis.

TMEM106B is ubiquitinated by TRIM2 in vitro

Of the substrates identified through proximity labelling, we initially chose the top hit, MYPT1 (PPP1R12A), for further investigation. MYPT1 is the regulatory subunit of the protein phosphatase 1 (PP1) holoenzyme that tunes the phosphorylation state of myosin II to regulate apoptosis and membrane blebbing^54–56. We used immunoblotting to compare endogenous MYPT1 levels in HEK 293-T cells transfected with WT or catalytically dead variants of TRIM2 (Extended Data Fig. 4a). We observed no differences amongst our treatments, indicating that the association of TRIM2 and MYPT1 does not lead to substantial changes in protein levels, although it may alter its phosphatase activity. Consistent with this finding, MYPT1 was not differentially expressed in DKO mESC-derived neurons (Supplemental Table 1).

We next focused on TMEM106B, which like TRIM2, is highly expressed in Purkinje cells and has links to motor neuron degeneration and the endolysosomal pathway^57–62. TMEM106B contains a disordered N-terminal cytosolic region and a globular C-terminal domain projecting either into the lysosomal lumen or the cellular periphery, where it has been reported to act as an alternative receptor for SARS-CoV-2^{58,59,63–65} (Fig. 3a). We established in vitro ubiquitination assays to examine TRIM2 and TRIM3 enzymatic activities in a minimal system using purified components from recombinant expression in Escherichia coli (E. coli) and monitored ubiquitination upon addition of ATP with SDS-PAGE or immunoblotting (Extended Data Fig. 4b). TRIM2 was more catalytically active than TRIM3, as judged by the formation of high MW species corresponding to autoubiquitination, concomitant with the disappearance of unmodified TRIM2/3 protein bands (Extended Data Fig. 4b). This is consistent with previous reports²⁸, as well as the proximity labeling experiments shown above. Using MS, we identified the ubiquitination sites to occur mostly in the disordered linker (~ 50 amino acids) separating the FIL and NHL domains, with few modifications also occurring in the coiled-coil region (Supplemental Table 3). Given that TRIM2 is cytosolic, we then tested ubiquitination of the N-terminus of TMEM106B (residues 1–95) (Fig. 3b). Parallel to autoubiquitination of TRIM2, we observed the formation of species corresponding to mono- and di-ubiquitinated TMEM106B^1–95 over a range of substrate concentrations (Fig. 3b, c). Thus, TRIM2 catalyzes the modification of TMEM106B using the classical ubiquitin cascade components without the need for additional cellular cofactors. To a lesser extent, TRIM3 was also able to ubiquitinate TMEM106B^1–95 (Extended Data Fig. 4c).

The region of TMEM106B used in these assays contains three lysine residues at positions 3, 14, and 95, which could serve as potential acceptors for the modification. Using MS, we could assign both lysine 3 and 14 as ubiquitination sites based on the presence of the signature Gly-Gly dipeptide mass increase of ~ 114 Da that results after trypsin digestion (Fig. 3d, Supplemental Table 3).

A ubiquitin chain composed of four moieties is minimally required to promote proteasome-dependent substrate degradation⁶⁶. However, additional cellular factors absent in the above assay could extend ubiquitinated TMEM106B to promote its turnover. We thus tested whether TRIM2 had a role in TMEM106B degradation through the proteasome to clarify the role of its ubiquitination. The WT and TRIM2/3 DKO mESC-derived neurons did not exhibit consistent changes in TMEM106B protein levels as judged by immunoblotting (Extended Data Fig. 4d). In addition, transient co-transfections of TMEM106B and TRIM2 variants in HEK 293-T cells showed that the presence of TRIM2 did not alter TMEM106B expression, nor did proteosome inhibition stabilize its levels (Extended Data Fig. 4e). This is in line with previous observations that TMEM106B is likely turned over through the lysosomal pathway⁶⁷ and suggests that ubiquitination of TMEM106B by TRIM2 does not lead to proteasomal dependent degradation.

TRIM2 interacts directly with a zinc-coordinated homodimerization interface of TMEM106B

The C-terminal domains of TRIM proteins are thought to be involved in substrate recruitment, but this has only been demonstrated for a couple of PRYSPRY-containing members⁶⁸. We set out to investigate the interaction between TRIM2 and TMEM106B using NMR spectroscopy. The ¹H-¹⁵N HSQC spectrum of TMEM106B^1–95 indicated that the region is mostly disordered as previously shown⁶³. However, a few residues exhibited spectral dispersion consistent with some degree of ordering, which had not been described (Fig. 4a). Upon addition of full-length TRIM2 to ¹⁵N-labelled TMEM106B^1–95, we observed line broadening (i.e., signal loss) in several amide peaks, as would be expected if an interaction took place resulting in the slower tumbling of TMEM106B in solution (Fig. 4a, b). Notably, the largest changes in linewidth occurred in the most dispersed amide signals, indicating that TRIM2 recognizes a structured motif. To determine which domains were responsible for the interaction, we repeated this experiment with either the N-terminal RING, B-box, and coiled-coil region of TRIM2 (RBCC, residues 8-318), or the C-terminal FIL and NHL domains (residues 318–744). We found that both the RBCC and FIL-NHL domains caused line broadening and therefore contributed to binding (Fig. 4b, Extended Data Fig. 5a).

To identify the interaction interface, we assigned the backbone chemical shifts of TMEM106B^1–95 and quantified the signal loss in ¹H-¹⁵N-HSQC spectra as a function of residue number (Fig. 4b). The largest perturbations were centered around residue 60 in the C-terminal region of TMEM106B, which showed consistent changes in the presence of full-length TRIM2, as well as the RBCC and FIL-NHL regions (Fig. 4b). The AlphaFold model of TMEM106B predicts that cysteine residues 61 and 64 in this region could form a disulfide bond, given that the two sulfhydryl groups are in very close proximity (< 5 Å) and the peptide backbone forms a hairpin around them (Extended Data Fig. 5b) ³². This observation prompted us to test if the oxidation state of TMEM106B was important to its structure and recognition of TRIM2. However, upon addition of 10 mM of dithiothreitol (DTT, a reducing agent) to ¹⁵N-labelled TMEM106B^1–95, we observed no significant changes in the spectra, indicating that highly reducing conditions do not change its conformation (Extended Data Fig. 5c).

Previous findings indicated that TMEM106B could oligomerize, but it was unclear how and to what extent ⁶⁷. The predicted MW of monomeric TMEM106B^1–95 is 10.7 kDa. Using size exclusion chromatography coupled to multi-angle light scattering (SEC-MALS), we determined that at concentrations as low as ~ 0.3 mg/mL, the protein had a MW of 23.1 ± 0.4 kDa and is therefore dimeric (Fig. 4c). We used the AlphaFold-Multimer pipeline to predict the structure of homodimeric TMEM106B^{1–95 69}. The prediction suggested that residues C61 and C64 could participate in metal coordination, as the side chains arising from each of the two monomers formed a tetrahedral arrangement commonly observed in metal binding sites (Fig. 4d, Extended Data Fig. 5d). In addition, the chemical shifts of the ¹³C^β atoms of C61 and C64 were ~ 32 ppm, which is characteristic of cysteine residues participating in zinc-coordination (Extended Data Fig. 5e)^70,71.

To test whether metal coordination was important in promoting dimerization, we compared the ¹H-¹⁵N HSQC spectra of TMEM106B^1–95 in the absence and presence of excess ethylenediaminetetraacetic acid (EDTA, a metal chelator). Since metal ions can bind with very high affinity to protein coordination sites⁷², we increased the temperature to 45°C to promote metal ion release. Indeed, we observed that the presence of EDTA caused new amide peaks to appear in the centre of the spectrum, consistent with a monomeric form of TMEM106B (Fig. 4e). In the absence of EDTA, the protein did not exhibit changes in the spectrum after heating in the same manner, and thus the changes observed are not due to thermal denaturation of the dimer (Fig. 4e). In addition, SEC-MALS experiments detected a molecule with a MW of ~ 14 kDa only in the presence of EDTA (Extended Data Fig. 5f). Next, we introduced the point mutations C61S and C64S to prevent metal coordination. This variant was monomeric with a MW of 11.6 ± 0.2 kDa and is hereby referred to as TMEM106B^mono (Fig. 4c).

Given the large number of zinc-binding proteins that exist in the human proteome, estimated to be ~ 10%, and the prominent role of zinc in neuronal homeostasis, we hypothesized that the metal ion mediating TMEM106B dimerization could be zinc^73,74. We thus used a sample with a mixture of monomeric and dimeric forms of TMEM106B and compared the ¹H-¹⁵N HSQC upon supplementation with ZnCl₂. Indeed, we found that the amide peaks corresponding to monomeric TMEM106B disappeared, while peaks corresponding to the dimeric species became more pronounced upon addition of the metal ion (Fig. 4f). Therefore, the association of TMEM106B with zinc determines the monomer-dimer equilibrium.

The accuracy of structural predictions is currently well established for ordered (i.e., folded) protein domains^32,69. However, flexible regions of proteins are not often characterized experimentally and exhibit poor prediction scores by the state-of-the-art tools available⁷⁵. We determined the NMR-derived structure of dimeric TMEM106B and compared it to the AlphaFold-Multimer predictions obtained above (Fig. 4g). To do this, we used a synthetic peptide spanning residues 54 to 92 supplemented with Zn^+ 2, corresponding to the most ordered residues that recapitulated the conformation observed in the longer TMEM106B^1–95 construct (Extended Data Fig. 5g). The structure shows that residues 57–71 directly involved in zinc coordination form a structure that is highly consistent with AlphaFold (Fig. 4g). Superposition of the highest ranking experimentally derived and predicted models resulted in a main-chain RMSD of 1.2 Å for those residues. Note that residues 69–71 and 58–60 adopt small β-strands appearing only in some of the predicted models. Around three quarters of the predicted models contain additional β-strands involving residues 80–83 that appear to further stabilize the dimer but are not experimentally validated. The NMR data shows that the remainder of TMEM106B (residues ~ 72–92) does not adopt stable secondary structure, exhibiting random coil chemical shifts (see BMRB deposition 52589). Altogether, our data show that both the N- and C-terminal domains of TRIM2 interact directly with the cytosolic region of TMEM106B through a newly identified zinc-dependent homodimerization interface involving cysteine coordinating residues C61 and C64.

Zinc binding by TMEM106B is important for recognition by TRIM2 and lysosomal regulation

We next tested whether zinc-mediated dimerization had an impact on TMEM106B function. TMEM106B^mono lacked the dispersed amide peaks observed in the WT variant, consistent with the loss of ordering of residues 54–74 (compare Fig. 4a and 5a). Upon addition of TRIM2^RBCC to TMEM106B^mono, much fewer amide peaks exhibited line broadening or chemical shift perturbation, and thus the interaction with TRIM2 is disrupted upon removal of the metal binding site (Fig. 5a, b). Specifically, the average ratio of signal intensities in the presence of TRIM2^RBCC relative to free TMEM106B^1–95 (I_RBCC/I₀) for well-resolved amide peaks was 0.67 ± 0.28 and 0.82 ± 0.15 in the case of WT and monomeric variants, respectively (Fig. 5b). Despite impaired TRIM2 association, the monomeric variant could be similarly ubiquitinated by full-length TRIM2 (Fig. 5c), indicating that TMEM106B dimerization is not required for ubiquitination in vitro.

TMEM106B regulates lysosomal acidification, size, and motility along microtubules⁷⁶. Notably, increased levels of TMEM106B result in enlarged and functionally defective lysosomes^67,77. We leveraged this phenotype to understand the role that ubiquitination and zinc-mediated dimerization might have, using live confocal imaging of HEK 293-T cells transfected with variants of full-length TMEM106B (Fig. 5d). The expression level of the variants was comparable and thus the mutations had no measurable impact on protein stability (Extended Data Fig. 6a). Overexpression of TMEM106B^WT caused an increase in lysosomal size as expected, in addition to perturbations in the intraluminal pH, as judged by the heterogeneous distribution of signal intensities arising from the pH-sensitive dye (LysoTracker) (Fig. 5d). The phenotype from the lysine deficient mutant (TMEM106B^K3A/K14A) was not distinct from WT. However, cells transfected with TMEM106B^mono (C61S/C64S) resembled the negative vector control, demonstrating that monomeric TMEM106B is not functional (Fig. 5d and Extended Data Fig. 6b). Using immunostaining, we confirmed the membrane localization of the overexpressed TMEM106B variants, which did not show obvious differences (Extended Data Fig. 6b). Thus, zinc-binding by TMEM106B is crucial to its function as a lysosomal regulator. The PTMs that occur at K3 and K14 did not affect the lysosomal size distribution or membrane localization, and thus the consequence of TMEM106B ubiquitination in cells is still enigmatic.

TRIM2 and TRIM3 in the context of neurobiology

Neurons are particularly susceptible to deficiencies in the long-distance transport of vesicles and associated cargo, and many of the genes linked to neurodegeneration regulate the endolysosomal pathway^78,79. In addition, remodelling of the ECM is highly dependent on lysosomal trafficking, as demonstrated during C. elegans larval development⁸⁰. The most striking phenotype we found in TRIM2/3 DKO neurons was the low abundance of many proteins comprising the ECM and outer plasma membrane (Fig. 1), in agreement with our hypothesis that TRIM2/3 may be important for the transport of proteins in and out of cells via the endolysosomal pathway. A few of our findings support this hypothesis. First, the most significantly enriched protein found using proximity labelling was MYPT1/PPP1R12A, a protein known to regulate the release of neurotransmitters from synaptic vesicles (Fig. 2)⁸¹. Second, a key component of the vacuolar H⁺-ATPase (V-ATPase) accumulated in TRIM2/3 DKO neurons (Fig. 1, Supplemental Table 1). The generation of proton gradients mediated by V-ATPases is crucial for synaptic vesicle exocytosis, and altered levels of ATP6V0C may impair acidification and adequate membrane fusion of intracellular vesicles^82,83. Furthermore, many of the proteins involved in synaptic vesicle trafficking overlap with the endolysosomal pathway⁸⁴, and we found TRIM2 to associate with endocytic components like SNF8, RABEP2, STAM, and HGS (Fig. 2, Supplemental Table 2). Among the proteins we identified through proximity labelling, we validated that the cytosolic domain of TMEM106B is ubiquitinated by TRIM2. TMEM106B regulates the formation, acidification, and trafficking of lysosomes along microtubules in neurons and other cell types^{57,60–62,85}. Importantly, a link between TMEM106B expression and alterations in the ECM pathway has been observed in microglia⁸⁶. TMEM106B was also shown to bind the V-ATPase accessory protein 1 (AP1), and to colocalize with the ESCRT-III component charged multivesicular body protein 2B (CHMP2B)^87,88. At the same time, TRIM2 can associate with microtubules and membrane phospholipids^51,89. Taken together, mounting evidence places both TRIM2 and TMEM106B at the interface of the microtubule cytoskeleton and the endolysosomal pathway through their functional or physical associations with many of the components that regulate vesicular trafficking, providing a plausible mechanism for their role in maintaining the ECM composition. Although we could not establish the direct role of ubiquitination in modulating TMEM106B activity, we showed that it does not affect the expression levels, localization, or function of TMEM106B as a lysosomal regulator in HEK 293-T cells. Ubiquitination may instead modulate interactions that occur through the N-terminal cytoplasmic region, or the kinetics of TMEM106B recycling in and out of cells.

Overlapping and divergent molecular roles of TRIM2 and TRIM3

It becomes apparent that the two TRIM paralogues have overlapping and divergent functions set by relatively few differences in amino acid sequence, highlighting how the expansion of the TRIM family of ligases has given rise to specific functional niches. We found that TRIM2 but not TRIM3 is readily degraded in cells in a proteasome-dependent manner (Fig. 2). In addition, we observed with live confocal microscopy that while both TRIM2 and TRIM3 colocalize with microtubules, TRIM3 accumulates in large intracellular structures (Fig. 2). Consistent with this, proximity labelling experiments link TRIM3 with proteins known to phase separate, while in the case of TRIM2, a wide variety of cellular targets with diverse localizations were identified as ubiquitination substrates (Fig. 2). Differences in protein concentrations resulting from distinct degradation kinetics, biophysical properties, PTMs, and the type of ubiquitin chain topology formed by these ligases⁹⁰ could all contribute to their distinct intracellular behaviour. TRIM2, for example, contains a greater number of lysine residues relative to TRIM3 (48 and 30, respectively), which may provide additional sites for ubiquitin modifications, influence solubility, and promote turnover (see also Supplemental Table 3). Nevertheless, MYPT1 and TMEM106B appeared enriched in both the TRIM2 and TRIM3 proximity labelling experiments, and the latter could be ubiquitinated by both ligases in vitro, indicating some degree of functional overlap (Extended Data Fig. 4, Supplementary Table 2).

Previous work demonstrated that the RING domain of TRIM3 is monomeric and somewhat inactive in vitro, but that its catalytic activity could be enhanced by promoting oligomerization²⁸. Here, we found the same trend in the relative activity of the two ligases in cells. The fact that few substrates were identified to be ubiquitinated by TRIM3 is consistent with its low ubiquitination efficiency relative to TRIM2 observed by us and others (see Fig. 2 and Extended Data Fig. 4)²⁸. As shown in the case of the TLR3, however, TRIM3 re-localization to early endosomes and its ubiquitination activity can be promoted upon treatment with the double-stranded RNA analog poly(I:C)⁹¹. Given that TRIM3 can also associate with enzymes involved in SUMOylation²⁹, future studies will be needed to establish how various cellular stimuli, including viral infection, influence its function; as well as the relative contribution of ubiquitination versus SUMOylation.

Many E3 ligases are associated with proteasomal dependent protein degradation. However, we did not find evidence that the presence of TRIM2 affected the levels of either MYPT1 or TMEM106B, despite clear evidence for their association. In addition, only a handful of proteins accumulated in TRIM2/3 DKO mESC-derived neurons. The only commonly enriched protein identified in the two cellular systems we studied was ITM2B, which appeared as a ubiquitination substrate of TRIM2 in HEK 293-T cells and was significantly depleted in the TRIM2/3 DKO neurons (Extended Data Fig. 3). These observations indicate that ubiquitination by TRIM2/3 does not necessarily lead to substrate degradation and may instead be stabilizing in some cases.

Mechanistic aspects of TMEM106B function and ubiquitination by TRIM2

RING-type E3 ligases exert their biological roles through the formation of a quaternary complex that brings the E2 conjugating enzyme primed for ubiquitin transfer close to the substrate. The ligase thus serves as an interaction platform that promotes catalysis⁹². We show here that the interaction between TRIM2 and TMEM106B is direct and mediated by both the tripartite motif and the C-terminal domains, revealing a role for substrate recognition of the RBCC region besides catalysis (Fig. 4). While mostly unstructured, TMEM106B contains a TRIM2-binding interface that is ordered around the metal-coordinating cysteine residues C61 and C64. The structure of the zinc-binding motif is characterized by two consecutive turns of the protein backbone that result in an S-shaped conformation (Fig. 4). The two TMEM106B monomers thus “sandwich” the metal ion, in a manner reminiscent of the zinc-hook found in Rad50⁹³. Unlike Rad50, however, TMEM106B lacks secondary structural elements that could further stabilize the formation of the zinc-binding site.

The dimerization of TMEM106B and its association with TRIM2 is disrupted upon mutation of the cysteine residues C61 and C64 (Fig. 5), and thus structural complementarity is important to the interaction between the E3 ligase and the substrate. Zinc binding and dimerization was also essential to the increase in lysosomal size caused by TMEM106B overexpression (Fig. 5), which raises the possibility that the availability of this metal ion has a regulatory role through control of the monomer-dimer equilibrium, especially in the context of neurons where zinc concentrations are tightly regulated⁹⁴. TMEM106B was previously shown to interact with paralogous TMEM106C⁷⁶. We observed that the cytosolic regions of both TMEM106A and TMEM106C contain cysteine residues that could participate in metal coordination, and thus heterodimerization of TMEM106A/B/C paralogues may occur through the same mechanism as homodimerization.

We determined the structure of a ternary complex consisting of the RING domain of TRIM2, the E2 conjugating enzyme UBE2D3 used in this work, and ubiquitin, to a resolution of ~ 2.4 Å using X-ray crystallography (Extended Data Fig. 7a). Relative to a previously described ternary TRIM2·E2·Ub complex containing UBE2D1 instead, the structure exhibits a very similar architecture, including the dimerization of the RING domain through N- and C-terminal α-helices (see PDB depositions 8AMS and 7ZJ3)²⁸. The structures show that a conserved arginine residue in RING-type ligases (R64 in TRIM2 and R54 in TRIM25) mediates contacts to both ubiquitin and the E2 to promote a “closed” conjugate conformation primed for catalysis⁹⁵. Nearby residues L73 in ubiquitin and Q92 in the E2 contribute to complex formation via stacking interactions and hydrogen bonding, respectively.

While the details of complex assembly leading to TMEM106B ubiquitination remain to be fully elucidated, we present a working model that summarizes our data and provides key insight into how this can happen (Fig. 6). The RING domain of TRIM2 engages with the ubiquitin machinery through an evolutionarily conserved interface to promote ubiquitination. The disordered cytosolic N-terminus of TMEM106B (residues ~ 1-53) provides accessible lysine residues where ubiquitination can occur. The ordered zinc motif (~ 54-74) is responsible for the interaction with TRIM2, while the C-terminus of cytosolic TMEM106B (~75-95) does not adopt secondary structure in vitro. While the RING domain promotes ubiquitination, the remainder of TRIM2 recruits TMEM106B in a manner that is not restricted to one individual region but involves both N- and C-terminal domains. Not shown in the schematic is the fact that TRIM2 dimerization can take place both through the RING and the coiled-coil domains. Thus, extensive oligomerization may occur as a function of TRIM2 concentration. Disordered linkers of ~ 20 and ~ 50 amino acids separate the RING and NHL domains, respectively, providing additional flexibility. Various lysine residues present in the latter region can also be modified due to autoubiquitination (Supplemental Table 3). These features highlight the complexity of ubiquitin signalling complexes that may include variable states arising from PTMs and oligomerization, and future research will be essential to better define the details of this dynamic assembly.

Generation of TRIM2/3 double-knockout cell lines

To generate a TRIM3 knockout line, regions encompassing coding exons 1 and 2 were targeted with the guide sequences specified in Extended Data Table 1. These were cloned into hSpCas9-GFP and -RFP plasmids, respectively, using the approach by Ran et al.⁹⁶. Two million mouse embryonic stem cells (mESCs, 129B-13) were transfected with 1.5 μg of each of the Cas9-containing plasmids using the Amaxa™ 4D-Nucleofector™ protocol according to manufacturer’s instructions (Lonza P3 kit, program CG-104 for mESC). Forty-eight hours post-transfection, fluorescence-activated cell sorting was performed to select monoclonal cell lines that were GFP and RFP-positive. These were sorted into 96-well plates, expanded in ES culture medium, and frozen. To screen for TRIM3 deletion, genomic DNA from ~ 1 million cells comprising a single colony was extracted using the Gentra Puregene Cell Kit (Qiagen) according to instructions and used for genotyping PCR and Sanger sequencing. Two clonal cell lines with successful TRIM3 deletion were used to further knock out TRIM2 using a similar approach targeting exons 1 and 2. The deletion of TRIM2 and TRIM3 proteins was further validated using immunoblotting and quantitative MS.

The chromosomal integrity of the clones generated was assessed via low-coverage whole-genome sequencing. Briefly, genomic DNA was extracted from ~ 1 million cells as above and the sample sonicated for 10 cycles (30 sec ON, 30 sec OFF) at 4°C, using a Bioruptor Pico sonication device (Diagenode) to obtain fragments of ~ 200 bp. The fragmentation pattern was assessed via agarose gel electrophoresis, and 0.5-1 μg of fragmented gDNA was used to prepare sequencing libraries using the NEBNext Ultra II DNA Library Preparation Kit (New England Biolabs), according to manufacturer’s instructions. Libraries were then quantified using the Qubit fluorometer (ThermoFisher Scientific) and the quality was assessed with a Bioanalyzer using the Agilent High Sensitivity DNA Kit (#5067-4626). Samples were then pooled and sequenced on an Illumina NextSeq-500 platform (75 bp single-end reads). The quality of the sequencing run was assessed using FastQC after removing adapter sequences with TrimGalore (https://github.com/FelixKrueger/TrimGalore). Sequencing reads were aligned to the mouse reference genome (GRCm38/mm10 assembly) using Bowtie 2⁹⁷ and uniquely mapped reads (MAPQ ≥ 30) were retained for the subsequent steps. A copy number variation analysis (bin size of 100 kb) was performed using a dedicated script (https:// github.com/tobiasrausch/coral) and the results are shown in Extended Data Fig. 7.

Mouse embryonic stem cell differentiation

The cells were differentiated into glutamatergic neurons following the procedure described by Bibel et al.³³ with minor modifications. Briefly, the cells were thawed on mouse embryonic fibroblasts (MEF)-coated T25 flasks (Corning) and maintained using KnockOut™ DMEM medium (Gibco, 10829018), including 15% fetal bovine serum (FBS, Gibco, 10270106), 1% NEAAs (Gibco, 11140035), 1% GlutaMAX (Gibco, 35050061), 1% Pen/Strep (Gibco, 15140122), 0.1 mM β-mercaptoethanol (BME, SigmaAldrich, M6250), and 20 ng/mL leukemia inhibitory factor for a week, in order to remove the feeder layer of fibroblasts before starting cell differentiation. At day 0 of the procedure, the cells were dissociated with TrypLE Express (ThermoFisher Scientific, 12605028) for 5 mins at 37 ºC, and ~ 4 million cells were seeded on non-adherent 10 cm dishes using 15 mL of CA medium containing high-glucose DMEM (Gibco, 41965039), 15% FBS, 1% NEAAs, 1% GlutaMAX, 1% Pen/Strep, 1% sodium pyruvate (Gibco, 11360070), and 0.15 mM BME. The resulting embryoid bodies were maintained for 8 days, with medium change every other day and addition of 5 µM retinoic acid (SigmaAldrich, R2625) from day 4 and onwards, to trigger differentiation towards the neuronal lineage. On day 8, the embryoid bodies were dissociated using trypsin and seeded on wells coated with 0.1 mg/ml poly-D-lysine (SigmaAldrich, P0899) and 2.5 µg/ml laminin (SigmaAldrich, L2020) at a density of 2 x 10⁵ cells per cm² using N-2 medium consisting of high-glucose DMEM, 1% Pen/Strep, 2% B-27 supplement without vitamin A (Gibco, 12587010) and 1% N-2 supplement (Gibco, 17502048). The cells were maintained for an additional 4 days and considered mature neurons ready for phenotypic characterization, corresponding to day 12 of differentiation.

Immunoblotting

The cleared lysate from cellular experiments or samples from in vitro assays (~ 2-50 µg) were run on a gradient denaturing SDS-PAGE gel (Mini-Protean TGX, BioRad) and transferred to a polyvinylidene difluoride membrane using the Trans-Blot Turbo transfer system according to manufacturer’s instructions (BioRad). The membrane was treated with a blocking buffer containing 5% milk powder and 0.1% Tween (Sigma) dissolved in 1x PBS (10 mM NaH₂PO₄, pH 7.4, 2.7 mM KCl, 1.5 mM KH₂PO₄, and 137 mM NaCl) and incubated at room temperature for 30 minutes while shaking. The membrane was then incubated with the primary antibodies listed below, diluted in blocking buffer for ~ 16 hours at 4 °C while mixing. The membranes were washed with 0.1% Tween/1x PBS three times for 5 minutes each at room temperature, and the secondary antibodies (goat anti-mouse-HRP, Abcam 97040 and goat anti-rabbit-HRP, Abcam 97051) diluted in blocking buffer 1:5000 and incubated for ~ 1 hour at room temperature. After similar washes, the membranes were treated with horseradish peroxidase substrate according to manufacturer’s instructions (Merck-Millipore Immobilon Western), and the chemiluminescence signal visualized using a BioRad ChemiDoc system. The following antibodies and dilutions were used in each case: anti-HA-tag antibody (Abcam 9110) 1:4000, anti-α-tubulin (Merck/Sigma-Aldrich T6199) 1:5000, anti-TRIM2 (Merck/Sigma-Aldrich SAB4200206) 1:2000, anti-TRIM3 (Bethyl Laboratories A301-209A) 1:1000, anti-TMEM106B (Biomol A303-439A) 1:2000-1:5000, anti-flag (Sigma A8592) 1:5000, anti-ubiquitin (Santa Cruz P4D1) 1:2000-1:4000, and anti-AviTag (Genscript A00674) 1:3000.

HEK 293-T cell culture

HEK 293-T cells (DSMZ - German Collection of Microorganisms and Cell Cultures, ACC: 635) were maintained in T-75 or T-25 cell culture flasks (Nunc EasYFlasks ThermoFisher, no. 156472, 156340) at 37 °C and 5 % CO₂, using DMEM + 1g/L D-glucose, L-glutamine, + pyruvate (ThermoFisher, 31885023) and supplemented with 10% heat-inactivated FBS, qualified from Brazil (ThermoFisher, 10270106), 1% PenStrep (ThermoFisher, 15140122), and 1% L-glutamine (ThermoFisher, A2916801). The cells were maintained and seeded according to the manufacturer’s instructions using trypsin-EDTA (ThermoFisher, 25200056). Quantification of cell numbers and viability was achieved using a TC20 automated cell counter (Bio-Rad Laboratories GmbH, DE). Where indicated, deubiquitinase (DUB) inhibition consisted of a treatment for 2.5 hours with a final concentration of 25 µM of PR-619 (Merck/SigmaAldrich #662141).

Mammalian expression plasmids

All plasmids used for mammalian transient transfections were prepared using endotoxin free (EndoFree) MaxiPrep kits (Qiagen no. 12362), and the primers were purchased from Sigma-Aldrich with standard desalting purification. To make GFP-expressing plasmids, the full-length genes encoding human TRIM2 and TRIM3 (NCBI ID: Q9C040 and O75382) were cloned into pcDNA3 containing mEGFP at the N-terminus with point mutations K206A and L221K ⁹⁸, using the restriction-free cloning method⁹⁹. A linker encoding 9 amino acids (GGSGGSGGG) separated the mEGFP and TRIM2/3 genes. The plasmids encoding HA-BirA-Rad18 and AviTag-Ubiquitin were kindly provided by Sagar Bhogaraju and have been described previously ¹³. The latter constitutes an optimized acceptor peptide with high selectivity named (−2)AP-Ub, in which two point mutations were introduced to change the acceptor peptide sequence from GLNDIFEAQKIEWHE to KGNDIFEAQKIEWHE ¹³. To generate HA-BirA-TRIM2/3 plasmids, similar restriction-free approaches were used. The catalytically dead variants contained the following point mutations introduced using site directed mutagenesis: C23S or C60S (TRIM2), and C22S or C59S (TRIM3)¹⁰⁰. We found no significant differences in the expression levels of these mutants, and thus used the C60S and C59S mutations in the proximity labelling approach as the catalytically inactive variants of TRIM2 and TRIM3, respectively (“dead”). Plasmids lacking the BirA gene, including HA-TRIM2^WT, HA-TRIM2^C60S, and HA-TRIM2^C23S/C60S were derived from those above and generated using standard cloning approaches to amplify the regions of interest. The gene for TMEM106B was obtained through Addgene (plasmid #179385, UniProt ID: Q9NUM4). Cloning of 3x flag-tagged TMEM106B^WT, TMEM106B^K3A/K14A and TMEM106B^mono (C61S/C64S) into pCDNA3.1 was accomplished using similar methods.

Transient co-transfections, lysis, and pull-down in substrate identification

Three million HEK 293-T cells were seeded in Falcon® 100 mm TC-treated cell culture dishes (Corning no. 353003) and maintained as described above. The following day (~ 16 hours), each dish was supplemented with a final concentration of 100 µM biotin (Sigma-Aldrich no. B4501) and shortly thereafter co-transfected with 5 µg of each of the BirA-containing and AviTag-Ub plasmids, using Fugene HD transfection reagent (Promega no. E2311) according to manufacturer’s instructions. Twenty-four hours post-transfection, the proteasome was inhibited with a final concentration of 0.5 µM Carfilzomib/PR-171 (Selleckchem no. S2853) for 5 hours in the case of TRIM2-treated cells and corresponding controls. For both TRIM2 and TRIM3 treatments, the cells were collected 29 hours post-transfection using plastic cell scrapers after briefly washing with ice-cold 1x PBS. The resulting cell suspension was centrifuged for 5 mins at 3, 000 r.c.f. and 4 °C, and the pellet frozen at - 80 °C until further processing.

To lyse, the cell pellet was thawed on ice for 15 mins and resuspended in 500 µL of ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 0.1% SDS, 1% Triton X-100, 500 mM NaCl, 1 mM EDTA, 1.5 mM MgCl₂, 0.5 mM TCEP, 10 mM N-Ethylmaleimide (NEM, Sigma-Aldrich no. E3876), 1 µg/mL RNAse A (Qiagen), 1 µg/mL DNAse I (Roche), and 1% Complete EDTA-free protease inhibitor cocktail tablet (Sigma-Aldrich no. 11873580001)), followed by a 20 minute incubation at 37 °C to allow DNA and RNA digestion. The cell suspension was then cooled on ice for 10 mins and lysed using a BioRuptor Plus sonicator (Diagenode) equilibrated at 4 °C, using 6 cycles of 30 seconds ON, 30 seconds OFF, with low power setting. The resulting lysate was centrifuged at 13, 000 r.c.f. for 10 mins at 4 °C, and the supernatant used for further analysis. The total protein concentration of each sample (soluble fraction) was determined using the Pierce BCA protein assay kit according to manufacturer’s instructions (ThermoFisher no. 23227) and normalized to the lowest protein concentration using lysis buffer. Twenty microliters of Pierce Streptavidin magnetic beads (ThermoFisher no. 88816) were incubated with 450 µL of each normalized cleared lysate and rotated at 4 °C for 2 hours to allow enrichment of biotinylated ubiquitin moieties. The beads were washed three times for 20 minutes each, with rotation at 4 °C, using 750 µL of wash buffer (50 mM Tris-HCl, pH 7.42, 0.1% SDS, 1% Triton X-100, 500 mM NaCl, 1 mM EDTA, 1.5 mM MgCl₂, 0.5 mM TCEP, 10 mM NEM, 10 mM of freshly prepared dithiothreitol (DTT), and 1% Complete EDTA-free protease inhibitor cocktail tablet). The bound proteins were eluted using 100 µL of elution buffer consisting of 100 mM Tris-HCl, pH 6.5, 2 % SDS, 200 mM DTT, 3 mM bromophenol blue, 2.15 M glycerol, 25 mM biotin, and boiled for 5 mins at 95 °C. Two biological and two technical replicates were conducted in each case (n=4), with proteasome inhibition in the case of TRIM2, and without in the case of TRIM3, with the exception of the TRIM3-catalytically dead variant, for which three replicates were used instead. The biological and technical replicates did not show differences in variability and were therefore treated equally during data analysis.

Quantitative mass spectrometry in substrate identification

Following elution from the streptavidin beads, the samples were further processed by reducing the disulfide bridges of cysteine-containing proteins with DTT (10 mM dissolved in 50 mM HEPES, pH 8.5 for 30 mins at 56 °C). The reduced cysteines were alkylated with 2-chloroacetamide (20 mM in 50 mM HEPES, pH 8.5) at room temperature and in the dark for 30 mins. The samples were then submitted to the SP3 protocol¹⁰¹ and trypsin was added in a ratio of 1 to 50 (enzyme to protein) for overnight digestion at 37 °C (sequencing grade, Promega no. V5111). The following day, the peptides were recovered using 50 mM HEPES (pH 8.5) by collecting the supernatant from the beads and combining it with a new wash with the same buffer. Peptides were labelled with the TMT10plex isobaric label reagent (ThermoFisher no. 90110) according to the manufacturer’s instructions¹⁰². Subsequently, the samples were combined for multiplexing and additional clean-up steps using an OASIS® HLB µElution Plate (Waters no. 186001828BA). Offline high pH reverse phase fractionation was carried out on an Agilent 1200 Infinity high-performance liquid chromatography system, equipped with a Gemini C18 column¹⁰³ (3 μm, 110 Å, 100 x 1.0 mm, from Phenomenex).

LC-MS/MS was carried out using an UltiMate 3000 RSLC nano LC system (Dionex) fitted with a trapping cartridge (µ-Precolumn C18 PepMap 100, 5µm, 300 µm i.d. x 5 mm, 100 Å) and an analytical column (nanoEase™ M/Z HSS T3 column 75 µm x 250 mm C18, 1.8 µm, 100 Å, Waters). Trapping was carried out with a constant flow of trapping solution (0.05% trifluoroacetic acid in water) at 30 µL/min into the trapping column for 6 mins. Subsequently, peptides were eluted while running solvent A (0.1% formic acid in water, 3% DMSO) with a constant flow of 0.3 µL/min, with increasing percentage of solvent B (0.1% formic acid in acetonitrile, 3% DMSO). The outlet of the analytical column was coupled directly to an Orbitrap Fusion™ Lumos™ Tribrid™ Mass Spectrometer (ThermoFisher) using the Nanospray Flex™ ion source in positive ion mode.

The peptides were introduced into the instrument via a Pico-Tip Emitter (360 µm OD x 20 µm ID; 10 µm tip, CoAnn Technologies) and an applied spray voltage of 2.4 kV. The capillary temperature was set to 275 °C. Full mass scans were acquired in the range of 375-1500 m/z in profile mode with a resolution of 120, 000. The filling time was set to a maximum of 50 ms with a limitation of 4x10⁵ ions. Data dependent acquisition was performed with the resolution set to 30000, with a fill time of 94 ms and a limitation of 1x10⁵ ions. A normalized collision energy of 38 was applied. MS² data was acquired in profile mode.

Mass spectrometry data analysis in substrate identification

IsobarQuant and Mascot (v2.2.07) were used to process the acquired data, which was searched against a Uniprot Homo sapiens (UP000005640) proteome database containing common contaminants and reversed sequences¹⁰⁴. The following modifications were included in the search parameters: carbamidomethyl (C) and TMT10 (K) as fixed modifications; as well as acetyl (N-term), oxidation (M), and TMT10 (N-term) as variable modifications. The following error tolerances were set: 10 ppm for MS¹ scans and 0.02 Da in the case of MS². In addition, trypsin was set as the protease with a maximum of two missed cleavages allowed, the minimum peptide length was set to seven amino acids, and at least two unique peptides were required for protein identification. The false discovery rate (fdr) at the peptide and protein level was set to 1%.

The raw output files of IsobarQuant (protein.txt) were processed using the R programming language¹⁰⁵. Only the proteins which were identified in two out of two technical replicate runs were kept. Raw TMT intensities ('signal_sum' columns) were first cleaned for batch effects using limma¹⁰⁶ and further normalized using variance stabilization normalization¹⁰⁷. Missing values were imputed with the knn method using the Msnbase package¹⁰⁸. Proteins were tested for differential expression using the limma package. The replicate information was added as a factor in the design matrix given as an argument to the ‘lmFit’ function of limma. In addition, imputed values were given a weight of 0.05 in the ‘lmFit’ function. A protein was annotated as a hit when the fdr was less than 5% and the fold-change was at least 100%, while protein candidates had a fdr less than 5% and a fold-change of at least 40%.

Proteins of interest (POI = TRUE, Supplemental Table 2) are highlighted in red in Fig. 2 and fulfilled the following criteria: i) an enrichment in the WT condition relative to catalytically dead of at least 40% with a fdr of 5% or less, ii) identical enrichment and fdr in the WT versus BirA-only comparison, and iii) they were not significantly different in the catalytically dead versus BirA-only conditions. This resulted in a set of 29 (TRIM2) and 15 (TRIM3) candidate proteins that were further considered for analysis.

Whole proteome profiling of mESC-derived neurons and data analysis

Quantitative MS analysis of the proteome corresponding to the three cell lines (WT, DKO1, and DKO2) was accomplished similarly as in the case of HEK 293-T cells (above) with the modifications that follow. The lysis buffer consisted of 50 mM Tris (pH 7.42), 150 mM NaCl, 1% Triton-X, 0.1 % SDS, and 0.5 mM TCEP, supplemented with 2.5 ng/mL of benzonase. The TMT6plex Isobaric Label Reagent (ThermoFisher) was used according to the manufacturer’s instructions. The Mus musculus database (UP000000589) was searched and TMT6 (K) or TMT6 (N-term) modifications were included in the search parameters. Only the proteins which were identified in two out of two replicates were further analyzed. A protein was annotated as a hit if the fdr was less than 5% and the fold-change was at least 50% (red circles in Fig. 1c). Proteins that were significantly affected in one of the two WT/DKO comparisons and exhibited a consistent fold change over 20% in the other, are highlighted with a black outline. Proteins considered hits in either of the WT/DKO comparisons and following a consistent trend in both are listed in Fig. 1d with the fold change quantification shown as a heat map (see also Supplemental Table 1).

Enrichment analysis

The gene symbols of protein candidates were used as input for the webserver program Enrichr¹⁰⁹. In cases of ambiguity in the protein identity (i.e., HSPA1B|HSPA1A), only the first annotation was retained.

Live confocal imaging and immunofluorescence

HEK 293-T cells cultured in glass-bottom dishes (Greiner Bio-One) were transfected one day after passaging with 0.5-2 µg of plasmid DNA containing GFP-TRIM2/TRIM3, using Fugene HD (Promega) according to manufacturer’s instructions for ~ 24 hours. The medium was changed to FluoroBright™ DMEM (Gibco, A1896701) the following day prior to imaging. To test microtubule co-localization, the cells were further incubated with 1 µM sirTubulin (Spirochrome) dissolved in FluoroBright™ DMEM for at least 30 mins before imaging. In the case of lysosomal staining after TMEM106B transfection, a similar procedure was applied for culturing and transfection. The lysosomal maker LysoRed (Abcam #112137) was used according to manufacturer’s instructions and incubated for ~ 45 mins 48 h post transfection. Rather than washing with HHBS, the cells were incubated with FluoroBright™ DMEM as above. To quantify the size distribution of lysosomes, 2D images collected at a resolution of 2048 x 2048 pixels (184 x 184 µm²) were processed in Fiji/ImageJ¹¹⁰. The Otsu threshold was applied automatically, and the images converted to binary with the watershed transformation to separate lysosomes in close proximity. The area of each of the resulting ~ 2500 particles was analyzed, excluding those smaller than 0.1 µm², larger than 100 µm², as well as those with a circularity lower than 0.8. An equal number of particles, selected in an unbiased manner from five images containing in total > 100 cells were quantified and plotted as a histogram for each condition. Particles with an area above 2.2 µm² were binned together to highlight the increase in size resulting from TMEM106B overexpression.

To perform fluorescent immunostaining of flag-tagged TMEM106B variants, similar procedures for culturing and transfection were applied. The following steps occurred at room temperature using 1x PBS as the buffer solution unless otherwise noted. The cells were fixed for 10 mins with 4% paraformaldehyde (Thermo #28908). The fixative was washed 3 times, and the cells permeabilized for 15 mins using 0.25 % Triton X-100 (Sigma-Aldrich #T8787). After similar washes, the samples were blocked with 1 % BSA (Sigma-Aldrich #A7906) and 22.5 mg/mL glycine (Sigma-Aldrich #50046) for 30 minutes. The cells were washed once, and incubated with the primary anti-flag antibody (Sigma-Aldrich #F1804) diluted 1:1000 in 1% BSA over ~ 16 h at 4 °C. The following day, the samples were washed three times, and the secondary antibody (Alexa-488 conjugate, Abcam #150113) was incubated for 1 h, diluted 1:1000 in 1% BSA and protected from light. The cells were counterstained with a mounting solution containing DAPI (ROTI®Mount FluorCare from Carl-Roth). In all cases, the samples were imaged using a Leica SP8 confocal microscope system with a 63x/1.4 numerical aperture oil objective using lasers at wavelengths of 405, 488, 552, and/or 638 nm.

Recombinant protein expression and purification

The genes encoding human TRIM2 and ubiquitin (NCBI gene IDs: Q9C040, P0CG48) were cloned into the expression plasmid pETM11 (generated at EMBL); while the genes encoding human TRIM3 and UBE2D3 (IDs: O75382, P61077) were cloned into pET28-MHL (Addgene, plasmid #26096). Human TMEM106B^1-95 and its variants (ID: Q9NUM4), as well as TRIM2^RBCC (residues 8-318) and TRIM2^RB (8-157) were cloned into pETM41 (EMBL), while TRIM2^FIL-NHL (residues 318-744) was cloned into pETM22 (EMBL), and mouse UBA1 (ID: Q02053) into pET28. The vectors produced encoded N-terminal poly-histidine (his₆) affinity tags used for purification with Ni-NTA Sepharose columns that could be removed by proteolytic cleavage with TEV or 3C proteases, save in the case of UBA1. pETM41 and pETM22 contained additional maltose-binding protein and thioredoxin solubility tags, respectively. Cloning involved standard restriction free methods⁹⁹. To create a stable E2~Ub conjugate, the point mutations C85S and S22A were introduced in UBE2D3^111,112. All plasmids were transformed into strain BL21 (λDE3) of Escherichia coli (E. coli) for recombinant expression. Protein production was induced with 0.2 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) over ~ 16 h at 20 °C, at optical densities (600 nm) of 0.6 to 1.5. The cells were grown in Lysogeny broth (LB) or Terrific broth (TB), while isotopically labeled proteins used for NMR spectroscopy were produced in M9 minimal media supplemented with 2 g/L ¹³C₆-D-glucose and/or 0.5 g/L ¹⁵NH₄Cl as sole carbon and nitrogen sources, respectively. Following expression, the cells were centrifuged at 3000 r.c.f. at 4 °C, and the pellet frozen at -20 °C until further processing. The cells were then thawed and resuspended in lysis buffer (20 mM Na₂HPO₄, 500 mM NaCl, 40 mM imidazole, pH 7.4), supplemented with 1 mg/mL lysozyme, 25 ng of benzonase, 0.5 mM TCEP, and 1x cOmplete™ Mini EDTA-free Protease Inhibitor Cocktail (Roche), followed by incubation at room temperature for ~ 30 mins. The cells were lysed using a homogenizer equilibrated at 4 °C, and the cleared supernatant applied to a Ni-NTA column (HisTrap HP, Cytvia) to allow binding. After washing with ~ 20 column volumes, the his-tagged proteins were eluted with 20 mM Na₂HPO₄ (pH 7.4), 500-1000 mM NaCl, and 250-500 mM imidazole. The appropriate fractions were pooled and dialyzed against 50 mM Tris (pH 7.5), 250 mM NaCl, and 1 mM DTT over ~ 16 h at 4 °C, with addition of 1-2 mg of his-tagged TEV or 3C proteases. A second Ni-NTA chromatography step was performed to separate the cleaved protein, and a final size-exclusion chromatography (SEC) step was included in all cases using HiLoad 16/600 Superdex 75 or 200 columns (Cytvia). In the case of proteins prepared for ubiquitination assays, the SEC buffer consisted of 50 mM HEPES (pH 7.5), and 150 mM NaCl. TRIM2 and TMEM106B preparations included 0.5 mM DTT and 10% glycerol additionally. For NMR spectroscopic studies, the SEC buffer consisted of 20 mM sodium phosphate (pH 6.5), 100 mM NaCl, and 0.5 mM TCEP.

In vitro ubiquitination assays

Proteins stocks obtained as described above were diluted with reaction buffer (50 mM HEPES, 150 mM NaCl, pH 7.5) to the following final concentration ranges: 150-450 μM ubiquitin, 50-70 μM E2, 0.8-1.3 μM E1, 0.7-1 μM TRIM2 or TRIM3, and 5-80 μM TMEM106B^1-95. The reaction was started with the addition of a final concentration of 5 mM ATP-MgCl₂, followed by a 30-minute incubation at 37 °C. The reaction was quenched subsequently with a final concentration of 50 mM EDTA diluted in reaction buffer, and the resulting mixture analyzed by denaturing SDS-PAGE or immunoblotting.

Identification of ubiquitination sites

The assayed samples derived from in vitro experiments were applied to SDS-PAGE gels and visualized by Coomassie blue staining. The corresponding regions of interest were excised from the gel, cut into small pieces (~ 1 mm x 1 mm), and transferred to 0.5 ml Eppendorf tubes. In all following steps, each buffer was exchanged by two consecutive 15-minute incubations of the gel pieces with 200 µL of acetonitrile (ACN), followed by ACN removal. Proteins were reduced by the addition of 200 µL of aqueous solution consisting of 10 mM DTT and 100 mM ammonium bicarbonate (AmBiC, Sigma Aldrich, A6141), then incubated at 56°C for 30 min. Proteins were alkylated by the addition of 200 µL of aqueous solution of 55 mM chloroacetamide (CAA, Merck/Sigma-Aldrich, C0267), 100 mM AmBiC, and incubated for 20 min in the dark. A 0.1 µg/µL stock solution of trypsin (Promega, V511A) in trypsin resuspension buffer (Promega, V542A) was diluted with ice-cold 50 mM AmBiC aqueous solution to achieve a final concentration of 1 ng/µL. 50 µL thereof were added to gel pieces, which were incubated first for 30 min on ice and then overnight at 37 °C. Gel pieces were sonicated for 15 min, centrifuged briefly to collect them, and the supernatant was transferred into a glass vial for injection (VDS Optilab, 93908556). The remaining gel pieces were washed with 50 µL of an aqueous solution of 50% ACN and 1% formic acid and sonicated for 15 min. The combined supernatants were dried in a vacuum concentrator and reconstituted in 10 µL of an aqueous solution of 0.1% (v/v) formic acid.

Peptides were analyzed by LC-MS/MS using a similar strategy as described above. The following gradient program was used: from 4 to 8% solvent B in 6 min, 8 to 23% for a further 41 min, 23 to 38% in 5 mins, followed by an increase of B to 80% for 4 min, and a re-equilibration to 2% solvent B for 4 min. The Orbitrap Fusion Lumos was operated in positive ion mode with a spray voltage of 2.4 kV and capillary temperature of 275 °C. Full scan MS spectra with a mass range of 375–1200 m/z were acquired in profile mode using a resolution of 120, 000, a maximum injection time of 50 ms, and an AGC target of 2x10⁵. Precursors were isolated using the quadrupole with a window of 1.2 m/z. For fragmentation, HCD was used with a fixed collision energy of 34%. MS² spectra were acquired using the Orbitrap with a resolution of 15, 000. The maximum injection time was set to 54 ms and the AGC target to 2x10⁶.

The data acquired were analyzed using FragPipe version 21.1 and MSFragger 4.0¹¹³ using the E. coli protein sequence database (UniProt: UP000000625, 4402 entries, February 2022) that includes common contaminants, along with the sequences for human TRIM2/3 and TMEM106B. The following modifications were considered: carbamidomethyl (protein C-terminus, fixed), acetyl (protein N-terminus, variable), oxidation (methionine, variable), as well as Gly-Gly (lysine, variable). The mass error tolerance was set to 20 ppm for MS¹ as well as MS² spectra. A maximum of 3 missed cleavages were allowed. The minimum peptide length was seven amino acids. A false discovery rate below 0.01 was applied at the peptide and protein level.

NMR spectroscopy

NMR experiments were collected on Bruker Avance III 600, 700 MHz, and Bruker Avance III HD 1 GHz spectrometers equipped with a room temperature probe (700 MHz) or cryoprobes (600 and 1000 MHz). The experiments were collected at 298 K (25 °C) with samples containing 10% D₂O as a locking agent. The NMR sample buffer consisted of 20 mM sodium phosphate (pH 6.5), 100 mM NaCl, and 0.5 mM TCEP, supplemented with Zn⁺² (i.e., 100-800 µM of ZnCl₂ or ZnSO₄) in the case of samples derived from chemical synthesis. Isotopically ¹⁵N/¹³C-labelled samples of TMEM106B^1-95 were obtained as described above from recombinant expression in E. coli and concentrated to 100-600 µM using Amicon Ultra-15 3 kDa MWCO centrifugal filters (Sigma-Aldrich), calculated according to the predicted extinction coefficient of TMEM106B^1-95 at 280 nm. An unlabelled synthetic peptide spanning residues 54-92 (TMEM106B^54-92) was purchased from ProteoGenix SAS (Schiltigheim, France) in lyophilized form at 95% purity, and resuspended in NMR sample buffer to a concentration of 0.6 mM supplemented with fivefold excess of Zn⁺². Chemical shift assignments were obtained using either standard ¹H-¹³C-¹⁵N correlation experiments¹¹⁴ with apodization weighted sampling to reduce data collection times (Simon & Köstler, 2019) in the case of isotopically labelled TMEM106B^1-95, or homonuclear 2D ¹H,¹H-COSY, TOCSY and NOESY, along with ¹H-¹³C/¹⁵N-HSQC spectra at natural abundance in the case of TMEM106B^54-92. These have been deposited under BMRB accession codes 52589 and 34953, respectively. The raw data were processed and analyzed using NMRpipe¹¹⁵, NMRFAM-Sparky¹¹⁶, CARA (http://cara.nmr.ch), and NMRView¹¹⁷.

To test binding, samples containing equal quantities of ¹⁵N-labelled TMEM106B^1-95 were diluted either with buffer or TRIM2 variants prepared in the same conditions. The RBCC (residues 8-318) or FIL-NHL (318-744) regions of TRIM2 were added in a 5-fold molar excess, while full-length TRIM2 was added in a 1:1 ratio. Line broadening observed in ¹H-¹⁵N HSQC experiments resulting from binding was quantified from the height of well-resolved amide peaks, comparing the two types of samples. The errors were estimated from the signal-to-noise ratio in each spectrum. In the case of TMEM106B^mono, we did not assign the amide chemical shifts and the comparison with the WT variant was performed for all well-resolved peaks that could be quantified in both sample types. To test EDTA-mediated denaturation, ¹⁵N-labelled TMEM106B^1-95 was treated with a final concentration of 5 mM EDTA and heated to 45 °C for 2 h, before lowering the temperature to 25 °C for data collection. Zinc binding was assessed with a sample containing a mixture of monomeric and dimeric species, before and after supplementation with equimolar (100 µM) amounts of ZnCl₂. The addition of a final concentration of 10 mM DTT was used to observe changes in TMEM106B conformation under reducing conditions.

Structural ensemble calculations with CYANA

Both TMEM106B^1-95 and TMEM106B^54-92 gave rise to similar chemical shifts corresponding to the structured region (amino acids ~ 57-71). The remaining residues show random coil chemical shifts and only intra-residual and sequential NOEs. To reduce signal overlap resulting from the disordered regions, we thus focused on the shorter peptide. Distance restraints were obtained from 2D ¹H,¹H-NOESY spectra (mixing time 150 ms). The NOE cross-peaks were assigned automatically during structure calculation using the noeassign function of CYANA 3.98.15¹¹⁸, while those expected based on the Alphafold-Multimer model (i.e., intermolecular) were assigned manually to validate the prediction. Symmetry restraints were applied to the CA atoms of all residues in each of the two monomeric chains. The tetrahedral conformation of the Zn-(S^Cys)₄ cluster was maintained using Zn-S and S-S upper and lower distance restraints of 2.25-2.35 Å and 3.75–3.85 Å, respectively. Dihedral angles were obtained from TALOS+¹¹⁹. All restraints were used to calculate a structure ensemble using CYANA version 3.98.15¹¹⁸. The resulting 20 models with lowest energy (i.e., using the target function) have been deposited under PDB accession code 9GI8. While residues 57 to 71 superimpose very well with a main chain RMSD of 0.1 Å, the remainder of the peptide is unstructured. See also Extended Data Table 2 for the structural statistics.

Size exclusion chromatography – multi-angle light scattering (SEC-MALS)

TMEM106B^1-95 WT and C61S/C64S were prepared as described above for NMR experiments. A volume of 50 µL was injected onto a Superdex 200 Increase 5/150 GL gel-filtration column (Cytiva) on an Agilent 1260 Infinity II HPLC system (Agilent), at protein concentrations of 1 mg/mL (~ 0.3 mg/mL at the apex of the peak). The run was performed at room temperature and with a flow rate of 0.3 mL/min. The column was coupled to a MALS detector (MiniDAWN and Optilab, Wyatt Technology). Data were analyzed using the Astra 8.2.0 software (Wyatt Technology). The standard protein refractive index increment (dn/dc) value of 0.1850 was applied for molar mass determination. To determine metal-binding properties, a final concentration of 5 mM EDTA was added, and the sample heated to 45 °C for 2 h and cooled to room temperature before measurement. The negative controls consisted of samples that were heated in the absence of EDTA in the same manner or treated with EDTA only.

AlphaFold-Multimer

The amino acid sequence of human TMEM106B (residues 1-95) was used as input to predict a homodimer. The predictions of 25 structural models were performed with the AlphaFold2 (release version 2.3.2) multimer pipeline (model_preset=multimer) using the multimer_v3 parameter set and the UniRef30 database version 2023_02⁶⁹. The maximum template release date was set to the future. Default values were used for db_preset (full_dbs) and num_recycle (20), and all models were selected for Amber relaxation. Shown in Fig. 4d are the top 5 ranked relaxed models, after main chain alignment.

X-ray crystallography

We reconstituted a stable, non-hydrolyzable E2~Ub conjugate, utilizing human proteins obtained from recombinant expression in E. coli according to established approaches^111,112. Briefly, the E2~Ub conjugate was produced over ~ 24 hours at 37 °C after addition of 5 mM ATP-MgCl₂ to a solution containing 200 µM of UBE2D3^S22A/C85S, 400 µM of his-tagged ubiquitin, and 2 µM UBA1, in a buffer consisting of 50 mM HEPES (pH 7.5) and 150 mM NaCl. The reaction products were applied onto a Ni-NTA column and eluted as above to remove any unreacted E2. Subsequently, the (his)₆ tag was cleaved with TEV protease as described previously, and a second Ni-NTA step applied to remove all tagged proteins. The flow-through fraction was concentrated and applied onto a HiLoad 16/60 Superdex 75 gel filtration column (Cytvia) equilibrated with 50 mM HEPES (pH 7.5), 150 mM NaCl, and 1 mM TCEP. TRIM2^RB (residues 8-157) was purified as described above and mixed in a 1:1.5 molar ratio with the conjugate to give a final protein concentration of ~ 28 mg/mL. The complex was further mixed in a 1:1 volume ratio with a crystallization solution consisting of 0.1 M bis-tris propane (pH 7.5), 0.2 M sodium formate, and 20% (w/v) polyethylene glycol 3350 (PACT screen, Molecular dimensions). Crystals grew at 20 °C using sitting drop vapour diffusion, and they were cryo-protected with 25% glycerol before flash-freezing and storage in liquid nitrogen. Residues corresponding to the B-box (~ 115-157) did not appear in the electron density. The data were acquired at the ESRF beamline MASSIF-3¹²⁰ at 100 K using single wavelength X-ray diffraction and processed using XDS for data reduction and scaling¹²¹, Phenix for phasing and refinement¹²², and Coot for model building¹²³. Phasing was performed using molecular replacement with the TRIM2 RING domain structure (PDB: 8A38), along with ubiquitin (PDB: 1UBQ) as search models. The structure has been deposited under accession number 8AMS and the structure statistics are presented in Extended Data Table 3.

Accession codes

The NMR-derived structure of TMEM106B^54-92 has the PDB accession number 9GI8. The X-ray derived structure of the ternary TRIM2·E2·Ub complex has been deposited at the PDB with accession number 8AMS. The chemical shifts of TMEM106B^1-95 and TMEM106B^54-92 can be found at the BMRB with codes 52589 and 34953, respectively. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE¹²⁴ partner repository with the dataset identifier PXD055297.

Acknowledgements

We are grateful to the Proteomics, Advanced Light Microscopy, and Protein Expression and Purification core facilities at EMBL, as well as HPC resources through EMBL IT services for access to scientific instrumentation, plasmids, reagents, infrastructure, and training. We thank the European Synchrotron Radiation Facility (ESRF) and beamline scientist Montserrat Soler Lopez for access to MASSIF-3 and technical assistance. We thank Sandra Augsten, Pravin Jagtap, and Bernd Simon for advice on cell culture experiments, X-ray crystallography, and NMR spectroscopy, respectively; as well as Stefan Terjung and Jean-Karim Hériché for help with confocal microscopy. We thank members of the Hennig and Mahamid groups for fruitful discussions. C.P.-B. was supported by the EMBL Interdisciplinary Postdoc (EI₃POD) Program fellowship under Marie Sklodowska-Curie Actions COFUND (grant no. 664726). This study was supported by funding from the Deutsche Forschungsgemeinschaft (DFG) via an Emmy-Noether Fellowship (HE 7291) to J.H..

Author contributions

C.P.-B. and J.H. carried out the study conceptualization and design. C.P.-B. performed, analyzed, and interpreted all experiments; prepared figures and wrote the manuscript with feedback from all co-authors. F.S. analyzed the quantitative MS data and assisted with experimental design and figure preparation. K.S. and J.H. determined the NMR-derived structure of TMEM106B^54-92. I.K. optimized the purification of the TRIM2·E2·Ub complex and set up crystallization trials with assistance from B.M. and advice from C.P.-B.. N.F-N. maintained, differentiated, and assisted with establishing mESC lines with advice from K-M.N.. M.R., J.J.S., and P.H. assisted with experimental design, processed samples, collected LC-MS data, and assisted with data analysis, with guidance from M.S.. K.L. assisted with experimental design and collected and analyzed the SEC-MALS data. J.G. performed autoubiquitination assays with advice from C.P.-B.. I.L. led the approach to create TRIM2/3 DKO mESC lines, with help from W.S.. M.T. and N.F-N. aided with the genome integrity check of said lines. T.H. established the AlphaFold-Multimer pipelines for efficient computation and ran the predictions. S.B. provided the BirA-containing plasmids and advice on the method. J.M., K-M.N., and M.S. assisted with project guidance and supervision, and J.H. directed the study.

Competing interests

The authors declare no competing interests.

List of supplemental tables

Supplemental Table 1: Quantitative proteomics of mESC-derived neurons

Supplemental Table 2: Ubiquitination substrates identified using proximity labeling

Supplemental Table 3: Ubiquitination sites of TRIM2, TRIM3, and TMEM106B found in vitro

Kaushik S, Cuervo AM (2015) Proteostasis and aging. Nat Med 21:1406–1415
Franić D, Zubčić K, Boban M (2021) Nuclear ubiquitin-proteasome pathways in proteostasis maintenance. Biomolecules 11:1–16
Bett JS (2016) Proteostasis regulation by the ubiquitin system. Essays Biochem 60:143–151
Pickart CM (2001) Mechanisms underlying ubiquitination. Annu Rev Biochem 70:503–533
Swatek KN, Komander D (2016) Ubiquitin modifications. Cell Res 26:399–422
Komander D, Rape M (2012) The ubiquitin code. Annu Rev Biochem 81:203–229
López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G (2023) Hallmarks of aging: An expanding universe. Cell 186:243–278
Liu N, Lin MM, Wang Y (2023) The Emerging Roles of E3 Ligases and DUBs in Neurodegenerative Diseases. Mol Neurobiol 60:247–263
Gan L, Cookson MR, Petrucelli L, La Spada (2018) A. R. Converging pathways in neurodegeneration, from genetics to mechanisms. Nat Neurosci 21:1300–1309
Iconomou M, Saunders DN (2016) Systematic approaches to identify E3 ligase substrates. Biochem J 473:4083–4101
Barroso-Gomila O et al (2023) BioE3 identifies specific substrates of ubiquitin E3 ligases. Nat Commun 14
Huang H-T et al (2024) Ubiquitin-specific proximity labeling for the identification of E3 ligase substrates. Nat Chem Biol 2024:1–10. 10.1038/s41589-024-01590-9
Mukhopadhyay U et al (2024) A ubiquitin-specific, proximity-based labeling approach for the identification of ubiquitin ligase substrates. Sci Adv 10:3000
Williams FP, Haubrich K, Perez-Borrajero C, Hennig J (2019) Emerging RNA-binding roles in the TRIM family of ubiquitin ligases. Biol Chem. 10.1515/hsz-2019-0158
Li JJ et al (2020) A recessive Trim2 mutation causes an axonal neuropathy in mice. Neurobiol Dis 140
Ylikallio E et al (2013) Deficiency of the E3 ubiquitin ligase TRIM2 in early-onset axonal neuropathy. Hum Mol Genet 22:2975–2983
BasuRay S, Mukherjee S, Romero EG, Seaman MNJ (2013) Wandinger-Ness, A. Rab7 mutants associated with Charcot-Marie-Tooth disease cause delayed growth factor receptor transport and altered endosomal and nuclear signaling. J Biol Chem 288:1135–1149
Lee SM, Chin LS, Li L (2012) Charcot-Marie-Tooth disease-linked protein SIMPLE functions with the ESCRT machinery in endosomal trafficking. J Cell Biol 199:799–816
Schreiber J et al (2015) Ubiquitin ligase TRIM3 controls hippocampal plasticity and learning by regulating synaptic γ-actin levels. J Cell Biol 211:569–586
Hung AY, Sung CC, Brito IL, Sheng M (2010) Degradation of postsynaptic scaffold GKAP and regulation of dendritic spine morphology by the TRIM3 ubiquitin ligase in rat hippocampal neurons. PLoS ONE 5
Khazaei MR et al (2011) The E3-ubiquitin ligase TRIM2 regulates neuronal polarization. J Neurochem 117:29–37
Lokapally A, Neuhaus H, Herfurth J, Hollemann T (2020) Interplay of TRIM2 E3 Ubiquitin Ligase and ALIX/ESCRT Complex: Control of Developmental Plasticity During Early Neurogenesis. Cells 9
Balastik M et al (2008) Deficiency in ubiquitin ligase TRIM2 causes accumulation of neurofilament light chain and neurodegeneration. Proc Natl Acad Sci U S A 105:12016–12021
Labonté D et al (2013) TRIM3 Regulates the Motility of the Kinesin Motor Protein KIF21B. PLoS ONE 8
Cheung CC et al (2010) Identification of BERP (brain-expressed RING finger protein) as a p53 target gene that modulates seizure susceptibility through interacting with GABAA receptors. Proc Natl Acad Sci U S A 107:11883–11888
Fang C et al (2017) MicroRNA-181c Ameliorates Cognitive Impairment Induced by Chronic Cerebral Hypoperfusion in Rats. Mol Neurobiol 54:8370–8385
Chu Y, Yang X (2011) SUMO E3 ligase activity of TRIM proteins. Oncogene 30:1108–1116
Esposito D, Dudley-Fraser J, Garza-Garcia A, Rittinger K (2022) Divergent self-association properties of paralogous proteins TRIM2 and TRIM3 regulate their E3 ligase activity. Nat Commun 13
Ye R et al (2021) Tripartite motif-containing 3 (TRIM3) enhances ER signaling and confers tamoxifen resistance in breast cancer. Oncogenesis 10:60
Thompson S et al (2011) Identification of a novel Bcl-2-interacting mediator of cell death (Bim) E3 ligase, tripartite motif-containing protein 2 (TRIM2), and its role in rapid ischemic tolerance-induced neuroprotection. J Biol Chem 286:19331–19339
Wang HT, Hur S (2021) Substrate recognition by TRIM and TRIM-like proteins in innate immunity. Semin Cell Dev Biol 111:76–85
Jumper J et al (2021) Highly accurate protein structure prediction with AlphaFold. Nature 596:583–589
Bibel M, Richter J, Lacroix E, Barde YA (2007) Generation of a defined and uniform population of CNS progenitors and neurons from mouse embryonic stem cells. Nat Protoc 2:1034–1043
Zhou A et al (2016) Lipopolysaccharide treatment induces genome-wide pre-mRNA splicing pattern changes in mouse bone marrow stromal stem cells. BMC Genomics 17 Suppl 7
Karra D, Dahm R (2010) Transfection techniques for neuronal cells. J Neurosci 30:6171–6177
Dayon L et al (2008) Relative quantification of proteins in human cerebrospinal fluids by MS/MS using 6-plex isobaric tags. Anal Chem 80:2921–2931
Joo EE, Yamada KM (2014) MYPT1 regulates contractility and microtubule acetylation to modulate integrin adhesions and matrix assembly. Nat Commun 5
Solomon S et al (2007) Distinct Structural Features ofCaprin-1 Mediate Its Interaction with G3BP-1 and Its Induction of Phosphorylation of Eukaryotic Translation InitiationFactor 2 α, Entry to Cytoplasmic Stress Granules, and Selective Interaction with a Subset of mRNAs. Mol Cell Biol 27:2324–2342
Kedersha N et al (2016) G3BP-Caprin1-USP10 complexes mediate stress granule condensation and associate with 40S subunits. J Cell Biol 212:845–860
Nakayama K et al (2017) RNG105/caprin1, an RNA granule protein for dendritic mRNA localization, is essential for long-term memory formation. Elife 6
Shiina N, Yamaguchi K, Tokunaga M (2010) RNG105 deficiency impairs the dendritic localization of mRNAs for Na +/K + ATPase subunit isoforms and leads to the degeneration of neuronal networks. J Neurosci 30:12816–12830
Schulte T et al (2023) Caprin-1 binding to the critical stress granule protein G3BP1 is influenced by pH. Open Biol 13
Kanai Y, Dohmae N, Hirokawa N (2004) Kinesin transports RNA: Isolation and characterization of an RNA-transporting granule. Neuron 43:513–525
Schauber C et al (1998) Rad23 links DNA repair to the ubiquitin/proteasome pathway. Nature 391:715–718
Yasuda S et al (2020) Stress- and ubiquitylation-dependent phase separation of the proteasome. Nature 578:296–300
Toriyama K et al (2020) T cell-specific deletion of Pgam1 reveals a critical role for glycolysis in T cell responses. Commun Biol 3
Johnson MC, Kollman JM (2020) Cryo-EM structures demonstrate human IMPDH2 filament assembly tunes allosteric regulation. Elife 9
McAlear TS, Bechstedt S (2022) The mitotic spindle protein CKAP2 potently increases formation and stability of microtubules. Elife 11
Yang X et al (2021) Fkbp3 induces human immunodeficiency virus type 1 latency by recruiting histone deacetylase 1/2 to the viral long terminal repeat. mBio 12
Yoshida SR, Maity BK, Chong S (2023) Visualizing Protein Localizations in Fixed Cells: Caveats and the Underlying Mechanisms. J Phys Chem B 127:4165
Glover HL et al (2023) Microtubule association of TRIM3 revealed by differential extraction proteomics. J Cell Sci. 10.1242/JCS.261522
Szklarczyk D et al (2023) The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res 51:D638–D646
Dankovich TM et al (2021) Extracellular matrix remodeling through endocytosis and resurfacing of Tenascin-R. Nature Communications 2021 12:1 12, 1–23
Clark K, Langeslag M, Figdor CG, van Leeuwen FN (2007) Myosin II and mechanotransduction: a balancing act. Trends Cell Biol 17:178–186
Qiao YN et al (2014) Myosin phosphatase target subunit 1 (MYPT1) regulates the contraction and relaxation of vascular smooth muscle and maintains blood pressure. J Biol Chem 289:22512–22523
Mills JC, Stone NL, Erhardt J, Pittman RN (1998) Apoptotic membrane blebbing is regulated by myosin light chain phosphorylation. J Cell Biol 140:627–636
Schwenk BM et al (2014) The FTLD risk factor TMEM106B and MAP6 control dendritic trafficking of lysosomes. EMBO J 33:450–467
Chang A et al (2022) Homotypic fibrillization of TMEM106B across diverse neurodegenerative diseases. Cell 185:1346–1355e15
Schweighauser M et al (2022) Age-dependent formation of TMEM106B amyloid filaments in human brains. Nature 605:310–314
Feng T et al (2022) TMEM106B deficiency impairs cerebellar myelination and synaptic integrity with Purkinje cell loss. Acta Neuropathol Commun 10
Rademakers R et al (2021) Loss of Tmem106b leads to cerebellum Purkinje cell death and motor deficits. Brain Pathol 31
Stroobants S, D’Hooge R, Damme M (2021) Aged Tmem106b knockout mice display gait deficits in coincidence with Purkinje cell loss and only limited signs of non-motor dysfunction. Brain Pathol 31:223–238
Kang J, Lim L, Song J (2018) TMEM106B, a risk factor for FTLD and aging, has an intrinsically disordered cytoplasmic domain. PLoS ONE 13
Baggen J et al (2023) TMEM106B is a receptor mediating ACE2-independent SARS-CoV-2 cell entry. Cell 186:3427–3442e22
Baggen J et al (2021) Genome-wide CRISPR screening identifies TMEM106B as a proviral host factor for SARS-CoV-2. Nat Genet 53:435–444
Thrower JS, Hoffman L, Rechsteiner M, Pickart CM (2000) Recognition of the polyubiquitin proteolytic signal. EMBO J 19:94–102
Brady OA, Zheng Y, Murphy K, Huang M, Hu F (2013) The frontotemporal lobar degeneration risk factor, TMEM106B, regulates lysosomal morphology and function. Hum Mol Genet 22:685–695
Wang HT, Hur S (2021) Substrate recognition by TRIM and TRIM-like proteins in innate immunity. Semin Cell Dev Biol 111:76–85
Evans R et al (2022) Protein complex prediction with AlphaFold-Multimer. bioRxiv 2021.10.04.463034 10.1101/2021.10.04.463034
Martin OA, Villegas ME, Vila JA, Scheraga HA (2010) Analysis of 13Cα and 1313C β chemical shifts of cysteine and cystine residues in proteins: A quantum chemical approach. J Biomol NMR 46:217–225
Kornhaber GJ, Snyder D, Moseley HNB, Montelione GT (2006) Identification of zinc-ligated cysteine residues based on 13Calpha and 13Cbeta chemical shift data. J Biomol NMR 34:259–269
Kluska K, Adamczyk J, Krężel A (2018) Metal binding properties, stability and reactivity of zinc fingers. Coord Chem Rev 367:18–64
Andreini C, Banci L, Bertini I, Rosato A (2006) Counting the zinc-proteins encoded in the human genome. J Proteome Res 5:196–201
Frederickson CJ, Suh SW, Silva D, Frederickson CJ, Thompson RB (2000) Importance of Zinc in the Central Nervous System: The Zinc-Containing Neuron. J Nutr 130:1471S–1483S
Terwilliger TC et al (2023) AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination. Nature Methods 21, 110–116 (2023)
Stagi M, Klein ZA, Gould TJ, Bewersdorf J, Strittmatter SM (2014) Lysosome size, motility and stress response regulated by fronto-temporal dementia modifier TMEM106B. Mol Cell Neurosci 61:226–240
Busch JI et al (2016) Increased expression of the frontotemporal dementia risk factor TMEM106B causes C9orf72-dependent alterations in lysosomes. Hum Mol Genet 25:2681–2697
Morfini GA et al (2009) Axonal transport defects in neurodegenerative diseases. J Neurosci 29:12776–12786
Grochowska KM, Andres-Alonso M, Karpova A, Kreutz MR (2022) The needs of a synapse—How local organelles serve synaptic proteostasis. EMBO J 41
Miao R, Li M, Zhang Q, Yang C, Wang X (2020) An ECM-to-Nucleus Signaling Pathway Activates Lysosomes for C. elegans Larval Development. Dev Cell 52:21–37e5
Kiss A, Erdődi F, Lontay B (2019) Myosin phosphatase: Unexpected functions of a long-known enzyme. Biochim et Biophys Acta (BBA) - Mol Cell Res 1866:2–15
Poëa-Guyon S et al (2013) The V-ATPase membrane domain is a sensor of granular pH that controls the exocytotic machinery. J Cell Biol 203:283–298
Bodzęta A, Kahms M, Klingauf J (2017) The Presynaptic v-ATPase Reversibly Disassembles and Thereby Modulates Exocytosis but Is Not Part of the Fusion Machinery. Cell Rep 20:1348–1359
Ivanova D, Cousin MA (2022) Synaptic Vesicle Recycling and the Endolysosomal System: A Reappraisal of Form and Function. Front Synaptic Neurosci 14:826098
Lüningschrör P et al (2020) The FTLD Risk Factor TMEM106B Regulates the Transport of Lysosomes at the Axon Initial Segment of Motoneurons. Cell Rep 30:3506–3519e6
Dominguez SL et al (2023) TMEM106B reduction does not rescue GRN deficiency in iPSC-derived human microglia and mouse models. iScience 26:108362
Klein ZA et al (2017) Loss of TMEM106B Ameliorates Lysosomal and Frontotemporal Dementia-Related Phenotypes in Progranulin-Deficient Mice. Neuron 95:281–296e6
Jun MH et al (2015) TMEM106B, a frontotemporal lobar dementia (FTLD) modifier, associates with FTD-3-linked CHMP2B, a complex of ESCRT-III. Mol Brain 8
Tsujita K et al (2010) Proteome of Acidic Phospholipid-binding Proteins: SPATIAL AND TEMPORAL REGULATION OF CORONIN 1A BY PHOSPHOINOSITIDES. J Biol Chem 285:6781–6789
Galagedera SKK et al (2023) Polyubiquitin ligand-induced phase transitions are optimized by spacing between ubiquitin units. Proc Natl Acad Sci U S A 120
Li WW et al (2020) Ubiquitination of TLR3 by TRIM3 signals its ESCRTmediated trafficking to the endolysosomes for innate antiviral response. Proc Natl Acad Sci U S A 117:23707–23716
Yang Q, Zhao J, Chen D, Wang Y (2021) E3 ubiquitin ligases: styles, structures and functions. Mol Biomed 2
Hopfner KP et al (2002) The Rad50 zinc-hook is a structure joining Mre11 complexes in DNA recombination and repair. Nature 2002 418:6897 418, 562–566
Shannon P et al (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13:2498–2504
Branigan E, Carlos Penedo J, Hay RT (2020) Ubiquitin transfer by a RING E3 ligase occurs from a closed E2 ~ ubiquitin conformation. Nature Communications 2020 11:1 11, 1–11
Ran FA et al (2013) Genome engineering using the CRISPR-Cas9 system. Nat Protoc 8:2281–2308
Langmead B, Salzberg SL (2012) Fast gapped-read alignment with Bowtie 2. Nat Methods 9:357–359
Zacharias DA, Violin JD, Newton AC, Tsien RY (2002) Partitioning of lipid-modified monomeric GFPs into membrane microdomains of live cells. Science 296:913–916
Bond SR, Naus CC (2012) RF-Cloning.org: An online tool for the design of restriction-free cloning projects. Nucleic Acids Res 40
Reikofski J, Tao BY (1992) Polymerase chain reaction (PCR) techniques for site-directed mutagenesis. Biotechnol Adv 10:535–547
Hughes CS et al (2019) Single-pot, solid-phase-enhanced sample preparation for proteomics experiments. Nat Protoc 14:68–85
Werner T et al (2014) Ion coalescence of neutron encoded TMT 10-plex reporter ions. Anal Chem 86:3594–3601
Reichel M et al (2016) In planta determination of the mRNA-binding proteome of arabidopsis etiolated seedlings. Plant Cell 28:2435–2452
Franken H et al (2015) Thermal proteome profiling for unbiased identification of direct and indirect drug targets using multiplexed quantitative mass spectrometry. Nat Protoc 10:1567–1593
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing Preprint at (2019)
Ritchie ME et al (2015) Limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43:e47
Huber W, Von Heydebreck A, Sültmann H, Poustka A, Vingron M (2002) Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics 18
Gatto L, Lilley KS (2012) Msnbase-an R/Bioconductor package for isobaric tagged mass spectrometry data visualization, processing and quantitation. Bioinformatics 28:288–289
Kuleshov MV et al (2016) Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res 44:W90–W97
Schindelin J et al (2012) Fiji: an open-source platform for biological-image analysis. Nature Methods 2012 9:7 9, 676–682
Sakata E et al (2010) Crystal Structure of UbcH5b∼Ubiquitin Intermediate: Insight into the Formation of the Self-Assembled E2∼Ub Conjugates. Structure 18:138–147
Brzovic PS, Lissounov A, Christensen DE, Hoyt DW, Klevit R (2006) E. A UbcH5/Ubiquitin Noncovalent Complex Is Required for Processive BRCA1-Directed Ubiquitination. Mol Cell 21:873–880
Kong AT, Leprevost FV, Avtonomov DM, Mellacheruvu D, Nesvizhskii A (2017) I. MSFragger: ultrafast and comprehensive peptide identification in mass spectrometry–based proteomics. Nature Methods 2017 14:5 14, 513–520
Sattler M, Schleucher J, Griesinger C (1999) Heteronuclear multidimensional NMR experiments for the structure determination of proteins in solution employing pulsed field gradients. Progress in Nuclear Magnetic Resonance Spectroscopy vol. 34 Preprint at https://doi.org/10.1016/S0079-6565(98)00025-9
Delaglio F et al (1995) NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J Biomol NMR 6:277–293
Lee W, Tonelli M, Markley JL (2015) NMRFAM-SPARKY: enhanced software for biomolecular NMR spectroscopy. Bioinformatics 31:1325–1327
Johnson BA (2004) Using NMRView to visualize and analyze the NMR spectra of macromolecules. Methods Mol Biol 278:313–352
Güntert P, Buchner L (2015) Combined automated NOE assignment and structure calculation with CYANA. J Biomol NMR 62:453–471
Shen Y, Delaglio F, Cornilescu G, Bax A (2009) TALOS+: a hybrid method for predicting protein backbone torsion angles from NMR chemical shifts. J Biomol NMR 44:213–223
Von Stetten D et al (2020) ID30A-3 (MASSIF-3) - a beamline for macromolecular crystallography at the ESRF with a small intense beam. J Synchrotron Radiat 27:844–851
Kabsch WXDS (2010) Acta Crystallogr D Biol Crystallogr 66:125–132
Liebschner D et al (2019) Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr D Struct Biol 75:861–877
Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot. Acta Crystallogr D Biol Crystallogr 66:486–501
Perez-Riverol Y et al (2022) The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidences. Nucleic Acids Res 50:D543–D552

There is NO Competing Interest.

ExData20240903PerezBorrajero.pdf
TableS1MSneurons20240828.xlsx
Table S1
TableS2MSproximity20240828.xlsx
Table S2
TableS3Ubiquitinationsites20240828.xlsx
Table S3
StructurefilesPerezBorrajero.zip
Structural files

Download PDF

Version 1

posted

You are reading this latest preprint version

Zinc-mediated regulation of TMEM106B function in the endolysosomal pathway revealed through the systematic identification of TRIM2/3 ubiquitination substrates

Status:

Version 1

Abstract

Figures

Introduction

Results

Proteins affected by TRIM2/3 deletion in mESC-derived neurons are associated with extracellular matrix maintenance

Distinct intracellular dynamics and ubiquitination substrates of TRIM2 and TRIM3

TRIM2 interacts directly with a zinc-coordinated homodimerization interface of TMEM106B

Zinc binding by TMEM106B is important for recognition by TRIM2 and lysosomal regulation

Discussion

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1