A designed ubiquitin-binding protein selectively enriches and visualizes unanchored ubiquitin

Abstract

While protein ubiquitination has been extensively studied, the roles of unanchored ubiquitin and its chains remain less understood, largely due to a lack of specific, high-affinity tools for their enrichment and visualization. To address this, we used high-throughput affinity maturation and computational protein design to engineer novel proteins that selectively bind unanchored ubiquitin. We first enhanced the binding affinity of the zinc finger domain of histone deacetylase 6, a natural unanchored ubiquitin binder, by including its unstructured amino-terminal loop and introducing the V1091L mutation. Using RFdiffusion, we then designed 11 novel ubiquitin-binding proteins (UBiPs), among which UBiP10 showed specificity for unanchored ubiquitin. This protein enriched unanchored ubiquitin from cells and influenza A virions and served as a probe to determine its cellular levels, supporting a relative decrease upon proteasome inhibition. Imaging studies with UBiP10 further revealed a novel unanchored ubiquitin coat around aggresomes, illustrating the potential of UBiP10 as a unique tool to study unanchored ubiquitin in cellular regulation.

INTRODUCTION

Ubiquitin (Ub) is a ubiquitous 76-residue (8.6 kDa) protein that regulates many biological processes as a posttranslational modifier. The enzymatic process of attaching the C-terminal Gly residue of Ub to Lys residues on substrate proteins is called protein ubiquitination (1). When Ub is linked to another Ub molecule, it forms a Ub chain, which can be categorized on the basis of the position of the acceptor Lys, such as K48-linked or K63-linked chains, among others. In addition, Ub also exists in the cell as a monomer or polymeric chains, which are not attached to proteins and are then referred to as unanchored Ub (2, 3).

Whereas protein ubiquitination, which is primarily responsible for protein degradation via the Ub proteasome system (UPS), has been studied in exhaustive detail, unanchored Ub is much less studied. This lack of understanding is largely due to the limited tools available to distinguish unanchored Ub from its covalently bound forms. Emerging evidence suggests that unanchored Ub is critical in various biological pathways, such as in host-pathogen interactions and the life cycle of protein aggregates. For instance, during influenza A virus (IAV) infection, unanchored Ub within the virus particle recruits histone deacetylase 6 (HDAC6) and promotes formation of a complex with motor proteins dynein and myosin that facilitates virion uncoating (4). Once the viral RNA is released into the cytosol, K63-linked unanchored Ub chains activate the cellular innate immunity response by binding to the viral RNA sensor—melanoma differentiation–associated protein 5—subsequently triggering activation of the interferon regulatory factor 3 and promoting interferon expression (5). Moreover, unanchored Ub chains are implicated in proteasome-mediated disassembly of aggresomes (6), which are perinuclear aggregates of ubiquitinated, misfolded proteins that form when the UPS is impaired (7), and are associated with neurodegenerative diseases such as Alzheimer’s and Parkinson’s disease (8). However, direct evidence to show the presence of unanchored Ub in aggresomes is lacking.

Currently, the main tool for studying unanchored Ub is based on the zinc finger (ZnF) Ub-binding domain (BUZ domain) of the deubiquitinating enzyme isopeptidase T (IsoT; or USP5), which binds to unanchored Ub with a dissociation constant (K_d) of 2.8 μM (9). Two IsoT_BUZ-based methods have been developed to quantify unanchored Ub in cells: protein standard absolute quantification using mass spectrometry (MS) (10) and visualization of unanchored Ub fluctuations using the IsoT_BUZ-derived protein tUI fused to fluorophores (11). Both methods have their limitations, however. The low affinity of the IsoT_BUZ domain, compared to antibodies, makes the affinity purification of unanchored Ub less efficient, leading to potentially inaccurate quantification, and the tUI’s specificity for visualizing unanchored Ub in cells and for enriching it from cell lysates remains invalidated.

Thus, there is a clear need for more effective tools to capture and detect unanchored Ub. Recent advances in protein design, especially with deep learning algorithms such as RFdiffusion (12), have opened new avenues for engineering proteins with improved specificity and affinity. In this study, we aimed to develop a tool with enhanced affinity and specificity for targeting unanchored Ub, starting from the ZnF_UBP (hereafter ZnF) domain of HDAC6, which is a natural unanchored Ub–binding protein. We used multidimensional mutagenesis and affinity maturation of this ZnF domain to produce a mutant with notably improved binding affinity. We further engineered this mutant using RFdiffusion to combine it to de novo designed structures and other known Ub-binding domains. After in silico evaluation with AlphaFold2 (13), we designed 11 novel Ub-binding proteins (UBiPs), with UBiP10 emerging as the most selective for unanchored Ub. As an affinity purification tool, UBiP10 specifically enriched unanchored Ub from both cell and IAV lysates, even in the presence of abundant ubiquitinated proteins. We used UBiP10 in conjunction with immunofluorescence microscopy to determine and visualize the unanchored Ub level fluctuation in cells following treatment with either the proteasome inhibitor bortezomib or the E1 Ub-activating enzyme inhibitor TAK-243. In addition, UBiP10 enabled us to investigate the localization of unanchored Ub during aggresome formation elicited by proteasomal inhibition: We found that unanchored Ub accumulates around the cell nucleus during aggresome formation and forms a coat around the mature aggresome. Thus, UBiP10 is a novel and powerful tool for studying unanchored Ub in cellular regulation.

RESULTS

Maturation of HDAC6 ZnF affinity for Ub by multidimensional mutagenesis

Unanchored Ub is characterized by the presence of an unanchored C terminus (-LRLRGG) (Fig. 1A), a short, unstructured motif lacking the necessary epitopes for antibody production or de novo protein binder design. The HDAC6 ZnF domain is a natural unanchored Ub–binding motif. Crystal structure shows that the gatekeeper residues R1155/Y1156 and W1182 determine the specificity and affinity against unanchored Ub (14), respectively (fig. S1A). The structural simplicity and epitope limitations of unanchored Ub make it a very difficult target to develop specific antibodies; therefore, we chose the HDAC6 ZnF as a rational and effective starting point for novel tool development. We opted to begin by maturing the affinity of the HDAC6 ZnF domain (Fig. 1A) for Ub and subsequently adding additional Ub-binding motifs to develop a specialized tool for studying unanchored Ub. Early work demonstrated that a C-terminal protein fragment from murine (m) HDAC6, mHDAC6_(1007–1149), containing its ZnF domain and some upstream sequences, has an exceptionally high affinity for unanchored Ub with a K_d of ∼60 nM (15). To decide which construct was the best to start our optimizations, we first compared this mouse protein fragment to the equivalent fragment from the human (h) protein, hHDAC6_(1073–1215), as well as to a shorter core ZnF domain fragment that had been crystallized in complex with Ub, hHDAC6_(1108–1215)—hereafter, core ZnF domain (14)—and to the full-length protein, hHDAC6_(1–1215) (Fig. 1B). We measured the binding affinities of the purified proteins for Ub using isothermal titration calorimetry (ITC) and found that in good agreement with the earlier results, the mouse and human extended ZnF fragments have comparable affinities for Ub, with a K_d of (69.9 ± 4.6) nM (Fig. 1C) and (73.5 ± 3.3) nM (fig. S1B), respectively. Unexpectedly, the full-length protein and the core ZnF domain both had a much lower affinity, with K_d of (1.04 ± 0.20) μM (fig. S1c) and (5.16 ± 0.47) μM (Fig. 1D), respectively. This indicates that the flexible loop of ∼30 residues at the N terminus of the core ZnF domain substantially enhances Ub binding. To explore this further, we used locally installed AlphaFold2 to predict the structure of the extended mHDAC6_ZnF_(1007–1149) bound to Ub. The model showed the N-terminal flexible region preceding the first α helix wrapping around Ub (Fig. 1E), explaining the greater binding affinity of the extended ZnF domain.

Fig. 1. Affinity maturation of HDAC6 ZnF domain interaction with mono-Ub.

(A) Ub [white inside the dashed rectangle; Protein Data Bank (PDB):1UBQ] is aligned and superimposed on human HDAC6 ZnF (1108–1215) (orange; from PDB:3GV4). Structures are shown as surface. The green colored region indicates the Ub C terminus (-LRLRGG). (B) Constructs of human and mouse HDAC6. Orange and cyan rectangles represent the core ZnF domain. (C) Affinity determination by ITC between mHDAC6_(1007–1149) and mono-Ub. (D) Summary of different HDAC6 constructs’ affinity (K_d) for mono-Ub. aa, amino acid. hHDAC6_(1108–1215) affinity to mono-Ub was determined in our previous study (23). ITC plots for hHDAC6_(1–1215) and hHDAC6_(1073–1215) are displayed in fig. S1 (B and C). (E) AlphaFold2-predicted structure of mHDAC6_(1007–1149)-Ub complex. White, cyan, and red represent Ub, mHDAC6_ZnF_(1007–1149), and mHDAC6 ZnF N-terminal flexible loop region (1037 to 1044), respectively. (F and G) Heatmaps of fitness effects of single amino acid substitutions in mHDAC6 ZnF_(1007–1149) from BindingPCA (F) and AbundancePCA (G). The secondary structure of mHDAC6_ZnF_(1007–1149) is shown at the bottom of (G), and the flexible loop region (1037 to 1044) is colored red. Mutation W1116A is highlighted by an arrow in the “BindingPCA” heatmap. (H) Scatter plot comparing the abundance (x-axis value) and binding fitness (y-axis value) of single amino acid substitutions. V1091L and S1064H are circled as two mutations for further investigations. The diagonal dashed line (black) represents y = x. (I) Affinity determination by ITC between mHDAC6_ZnF_(1007–1149)_V1091L mutant and mono-Ub.

For affinity maturation, we selected the mHDAC6_ZnF_(1007–1149) fragment, which had the highest affinity for Ub. We used a multidimensional mutagenesis–based, high-throughput screening method (16), which used a protein fragment complementation assay (PCA) to evaluate the impact of mutations on protein abundance (AbundancePCA) and binding affinity (BindingPCA; fig. S1D). In BindingPCA, the interaction between the two proteins of interest is measured by fusing them to the two fragments (DH fragment and FR fragment) of a split dihydrofolate reductase (DHFR) enzyme, expressing the fused proteins in yeast cells and assaying for reconstituted DHFR function by measuring cell growth under selective conditions, which reflect the intracellular concentration of the protein complex. In AbundancePCA, only the ZnF domain is fused to one of the two DHFR fragments, whereas the other fragment is expressed at high (saturating) level by itself. In this case, functional DHFR is reconstituted through random encounters between the overexpressed fragment and the fragment fused to ZnF, and cell growth is proportional to the intracellular concentration of the ZnF domain. Both BindingPCA and AbundancePCA were performed by pooling plasmid libraries encoding various ZnF mutants (mutated at each position to every possible amino acid) and Ub and expressing them in yeast cells so that each cell expressed a single ZnF mutant–Ub combination. Cells expressing ZnF variants that strongly interacted with Ub were enriched using the BindingPCA method, whereas cells expressing ZnF variants that increased variant abundance were enriched using the AbundancePCA method. Enrichments were quantified by sequencing the plasmids extracted from the yeast cells after selective growth and calculating enrichment scores [log₂FC (fold change) as compared to wild-type ZnF fragment] by quantifying the plasmid amounts with deep sequencing.

We investigated the effect of every single mutation on both protein interaction and abundance and plotted the log₂FC of BindingPCA (Fig. 1F) and AbundancePCA (Fig. 1G) as heatmaps. The tryptophan residue 1116 (corresponding to W1182 in human HDAC6) has previously been shown to be essential for interaction with Ub (4); every amino acid change at this position disrupted the interaction, thus validating the specificity of the assay (Fig. 1F, column indicated by arrow). Mutations in the structured core regions of the ZnF (α helices and β sheets, indicated at the bottom of Fig. 1G) generally disrupted the binding (Fig. 1F). Notably, mutations in the loop region (residues 1037 to 1044; colored red at the bottom of Fig. 1G), which AlphaFold2 predicted to be important for binding Ub, also reduced the affinity (Fig. 1G). To validate these findings, we selected two mutations, Y1039R and A1040K, within this loop region, expressed and purified the mutated proteins, and measured their affinities for Ub. Both mutant proteins (fig. S1, E and F) had much lower affinities for Ub than the wild-type protein.

As the BindingPCA heatmap revealed no obvious mutations that might increase the binding affinity of ZnF for Ub, we searched for mutations that have a moderate or strong BindingPCA score despite reducing protein abundance, thus effectively having an increased affinity. To do so, we generated a scatter plot (Fig. 1H), plotting BindingPCA scores (y axis; Fig. 1F) against AbundancePCA scores (x axis; Fig. 1G), and focused on mutations among the furthest above the diagonal. This approach identified several mutations, including Ser to His at site 1064 (S1064H) and Val to Leu at site 1091 (V1091L) (Fig. 1H). We purified these two mutated proteins and tested their binding to Ub. Compared with the wild-type mZnF_(1007–1149), the V1091L mutation nearly doubled the affinity for Ub (K_d = 40.0 ± 3.8 nM) (Fig. 1I), whereas the S1064H and S1064H/V1091L mutations increased the affinity by only about 30% (fig. S1, G and H). Residue V1091 is far from the Ub C terminus (distance = 10 Å) (fig. S1I), suggesting that mutation of this residue affects Ub binding allosterically. On the basis of this, we selected this mZnF_(1007–1149)_V1091L variant as the template for the next round of optimization.

In silico de novo UBiP design and affinity determination

To further enhance the affinity of the mZnF_ (1007–1149)_V1091L mutant for unanchored Ub without compromising its selectivity, we designed a series of UBiPs by computationally fusing known Ub-binding domains to the mutant mZnF domain. Our protein design was guided by deep learning–based protein design tools RFdiffusion and ProteinMPNN (17), followed by an AlphaFold2 prediction of the designed UBiP in complex with Ub. Our strategy involved four steps: domain placement (step 1), backbone design (step 2), sequence sampling (step 3), and designability testing (step 4; Fig. 2A and details in fig. S2). In step 1, we superimposed known Ub-binding domains (i.e., the UBA domain from UQ1, the RUZ domain from Radex-5, and the UIM domain from VPS27 and RAP80) onto the AlphaFold2-predicted structure of the mZnF_(1007–1149)_V1091L–Ub complex. In step 2, we used RFdiffusion to inpaint the linker backbone (filling in missing regions with new structures without sequence generation) between the mZnF_(1007–1149)_V1091L and the Ub-binding domains added in step 1. These designed new structures were either unstructured linkers (such as in UBiP01 and UBiP10; Fig. 2B) or α helices or β sheets (such as in UBiP05; Fig. 2A). Depending on their spatial orientation, the new domains were fused to the nearest terminus of mZnF_(1007–1149) _V1091L; for example, UBA was fused to the N terminus, whereas RUZ and UIM were fused to the C terminus of mZnF (Fig. 2B). In step 3, the sequences of the linker backbones inpainted in step 2 were generated with ProteinMPNN. Last, in step 4, following the workflow described previously (18), we used AlphaFold2 to predict UBiPs-Ub complex structures based on the sequence obtained in step 3 and compared the predicted structures to the RFdiffusion designed model used as input template (the “initial guess” method). The predicted models were ranked and selected by the root mean square deviation (RMSD) value from the RFdiffusion designed models.

Fig. 2. In silico design of UBiPs and affinity determination with mono-Ub.

(A) Workflow for generating UBiPs based on the mZnF_V1091L mutant. mZnF_(1007–1149)_V1091L, RUZ domain of Radex-5, and designed new structures are colored as cyan, yellow, and blue, respectively. (B) Examples of UBiPs with N-terminally (UBiP01) and C-terminally (UBiP10) fused Ub-binding domains (UBA_UQ1, pink; UIM_RAP80, green; designed new structures, blue). (C) Affinity determination by ITC between UBiP05 or UBiP10 and mono-Ub. (D) AlphaFold2-predicted aligned error (PAE) plots for UBiP05 and UBiP10. The PAE increases from blue to red. aa, amino acids. (E and F) Affinity determination by ITC between UBiP10 mutants and mono-Ub. (E) is UBiP10_(L1091V), and (F) is UBiP10_(linker mut). Linker: RFdiffusion generated de novo sequence, shown as blue region in UBiP10 model in (B); mut: mutated sequence (VTPKTEPMTL to AAAAAAAAAA).

In summary, we built six different starting models and used a pipeline of RFdiffusion/ProteinMPNN/AlphaFold2 (locally installed or/and online via Colab) to generate the UBiP candidates (fig. S2). In total, we generated 10,081 UBiP candidates in silico designs using different domain fusion strategies and selected the 11 top candidates with the lowest RMSDs (in a range of 0.8 to 3.3 Å) and a predicted aligned error (PAE) in the range of 6 to 12 Å (table S1). These 11 candidates were expressed in Escherichia coli and purified via their N-terminal glutathione S-transferase (GST) tag (fig. S3A). Except for UBiP11, which failed to express, all other UBiPs were successfully purified. To assess their utilities as tools for detecting unanchored Ub, we measured the affinity of each UBiP for mono-Ub using ITC (table S2 and fig. S3B). The K_d values ranged from 8.55 to 885 nM. Notably, UBiP01, UBiP02, and UBiP03, all of which are ZnF domain N-terminal fusions, exhibited relatively low affinities compared to those with C-terminal fusions. The AlphaFold2-predicted complexes of UBiP01/02/03 together with Ub also have low interface predicted template modeling (ipTM) scores (<0.6) (table S1), indicating a weak protein interaction. Together, these observations underscore the importance of the ZnF domain’s N-terminal flexible region for Ub binding.

Among all the UBiPs, UBiP05 emerged as the best Ub binder with a K_d of (8.55 ± 3.28) nM, followed by UBiP10 with K_d values of (25.6 ± 3.9) nM (Fig. 2C and table S2). These results align well with AlphaFold2 predictions, where the UBiP05-Ub complex had a low PAE value and a high ipTM value, indicating the highest potential for enhanced binding (Fig. 2D). The designed model of UBiP05 contains a linker of three α helices between the ZnF and RUZ domains (fig. S3C). These α helices are predicted to bind to specific Ub regions (residues 33 to 35 and 55 to 67) and, together with the RUZ domain, wrap around Ub to form an extensive binding interface. In the case of UBiP10, a short linker connects the N terminus of the UIM domain (from RAP80) to the C terminus of ZnF (Fig. 2B).

To assess the necessity of previous mutagenesis and the de novo structural elements generated by RFdiffusion/ProteinMPNN for enhanced Ub recruitment, we designed and analyzed three UBiP10 variants: (i) UBiP10_(L1091V): This construct restores the valine residue at position 1091 in the mHDAC6 ZnF domain (Fig. 2E). (ii) UBiP10_(linker mut): All the residues in the RFdiffusion/ProteinMPNN-generated linker were mutated to alanine (Fig. 2F). (iii) UBiP10_(ΔUIM): The C-terminally fused UIM domain was deleted (fig. S3D). UBiP10_(ΔUIM) failed to express, suggesting that the C-terminal UIM domain contributes to protein stability (fig. S3E). UBiP10_(L1091V) and UBiP10_(linker mut) were purified (fig. S3F), and their affinities for mono-Ub were determined by ITC. Both mutants exhibited reduced binding affinities: (48.5 ± 7.6) nM for UBiP10_(L1091V) (Fig. 2E) and (98.0 ± 29.9) nM for UBiP10_(linker mut) (Fig. 2F), representing a ∼2 to 3-fold decrease compared to the original UBiP10. This confirms that both steps of our approach (mutation V1091L in the ZnF domain and addition of another Ub-binding domain via a designed linker) contribute to the increased affinity of UBiP10 for Ub. This approach, thus, yielded 10 UBiP candidates for experimental validation, among which, UBiP05 had the highest affinity for Ub, with a nearly 10-fold improvement in K_d when compared with the wild-type mZnF_(1007–1149). Given their promising binding characteristics, UBiP05 and UBiP10 were selected for further experiments.

Enrichment of unanchored Ub from cell lysates and influenza A virions

Although both UBiP05 and UBiP10 exhibit high affinity for Ub, their selectivity for unanchored Ub, rather than Ub covalently conjugated to a protein, was unclear. To investigate this, we used a pseudoubiquitinated protein—three Ub molecules fused to retinoic acid receptor alpha (3xUb-RARA)—which mimics a linearly ubiquitinated substrate with three Ub moieties linked to the initial Met residue of RARA’s N terminus, thereby representing anchored Ub. In this protein, every Ub C terminus (−LRGG) is engaged in a covalent linkage. We first prebound GST-tagged UBiP05 or UBiP10 to GST_trap beads separately, followed by the addition of purified 3xUb-RARA and/or mono-Ub to the prebound beads. After washes, we then assayed which protein (3xUb-RARA or mono-Ub) was bound to the beads by MS analysis after Lb^pro digestion or by immunoblotting.

As shown in Fig. 3A, digestion of mono-Ub by Lb^pro (19), which cleaves between R and GG in the Ub C-terminal tail (−LRLR│GG), yields a truncated form of Ub (minus GG, or –GG), with a mass of 8449.5 Da. In contrast, digestion of 3xUb-RARA produces two products: Ub (−GG) originating from the distal Ub and GG-Ub (−GG) (with a mass of 8563.5 Da) originating from the proximal Ub molecules. To minimize nonspecific binding to the GST_trap beads, we loaded minimal equal amounts of 3xUb-RARA and mono-Ub to the GST-UBiPs prebound beads. In the UBiP05-bound samples, both 8449.5 and 8563.5 Da mass peaks were detected by MS, indicating that UBiP05 binds to both unanchored and anchored Ub. In contrast, only the 8449.5-Da peak was observed in UBiP10-bound samples, indicating UBiP10’s selective interaction with unanchored Ub (Fig. 3B). For immunoblotting (Fig. 3C), higher amounts of 3xUb-RARA and mono-Ub were used. Although some purified 3xUb-RARA can bind to GST_trap beads via nonspecific interaction (Fig. 3C, pulldown, lane 1), after incubation with GST-UBiP05, more 3xUb-RARA was captured (compare pulldown lane 3 to lane 1), indicating a direct interaction between UBiP05 and anchored Ub. By contrast, we found no 3xUb-RARA binding to UBiP10 (lane 4). Moreover, when 3xUb-RARA and mono-Ub were added together to the prebound beads, UBiP10 selectively captured mono-Ub (lane 6), whereas UBiP05 bound to both proteins (lane 5). To investigate why UBiP10 exhibits greater selectivity than UBiP05, we used the latest AlphaFold3 to model the interactions between each protein and a truncated Ub variant lacking the C-terminal last six residues (–LRLRGG), referred to as Ub_(ΔC-ter). AlphaFold3 prediction shows that the ipTM score for the UBiP05–Ub_(ΔC-ter) complex is 0.81 (>0.8; fig. S4A), suggesting that UBiP05 can still recognize and bind Ub even in the absence of its C terminus (20). In contrast, the ipTM score for the UBiP10–Ub_(ΔC-ter) complex dropped notably to 0.25 (fig. S4B), indicating the absence of stable binding between UBiP10 and the truncated Ub.

Fig. 3. Selectivity of UBiP10 for unanchored Ub recruitment.

(A) Scheme for Lb^pro enzyme–based Ub (anchored and unanchored) digestion and MS analysis (Ub clipping assay) (19). Lb^pro-treated mono-Ub and 3xUb-RARA produced peaks at 8449.5 Da [Ub(−GG)] and 8563.5 Da (proximal Ub retaining apparent wild-type mass through Gly-Gly attachment to Met1). (B) MS analysis of mono-Ub and 3xUb-RARA mixtures (input) and of proteins captured by GST-UBiP05 or GST-UBiP10. GST_trap beads saturated with GST-UBiP05 or GST-UBiP10 were incubated with mono-Ub and 3xUb-RARA (50 nM each), followed by Lb^pro digestion and MS. Input: 50%. (C) GST pulldown between GST-tagged UBiP05 or UBiP10 and 3xUb-RARA or/and mono-Ub. Captured proteins were visualized by immunoblotting (IB). (D) GST pulldown between GST-UBiP10 or GST (negative control) and linear, K48- and K63-linked unanchored di-Ub chains. (E) Affinity determination by ITC between UBiP10 and K48/K63-linked di-Ub chains. (F) GST pulldown between GST-UBiP05 or GST-UBiP10 and A549 cell lysates (equal protein input). MG132 treatment: 5 μM for 6 hours. Actin: loading control. (G) GST pulldown between GST, GST-hZnF_(1108–1215), or GST-UBiP10 and A549 lysates. (H) Comparison between UBiP10 and tUI using purified His-tagged proteins and Ni–nitrilotriacetic acid (NTA) pulldown from untreated A549 lysates. (I) GST pulldown between GST or GST-UBiP10 and lysates of 293T cells expressing deubiquitinase-resistant di-, tri-, or tetra-Ub chains carrying C-terminal mutations (-LRLRGG- to -LRAAAA-). (J) GST pulldown between GST-UBiP10 or GST and lysates from purified IAV (X31) virions disrupted with radioimmunoprecipitation assay (RIPA) buffer and sonication. NP, IAV nucleoprotein. (K) GST pulldown between GST-UBiP10 or GST and lysates from IAV-infected A549 cells [multiplicity of infection (MOI) = 50 for 6 hours] in the presence of 50 μM PR-619. (L) Affinity purification–mass spectrometry (AP-MS) analysis of UBiP10-interacting proteins in A549 lysates. GST served as negative control. Dashed lines indicate Log₂FC > 2 or < −2.

Because UBiP10 demonstrated superior binding to unanchored mono-Ub in vitro, we investigated whether it can also capture unanchored poly-Ub chains. We incubated purified linear, K48-, and K63-linked di-Ub chains with GST-UBiP10 prebound beads and analyzed the captured proteins by immunoblotting (Fig. 3D). As shown, GST-UBiP10 bound to all the three di-Ub chains with different linkages, indicating that UBiP10 can bind directly to unanchored Ub chains as well. Next, we determined the affinity of UBiP10 for purified K48-linked and K63-linked di-Ub chains (fig. S4C) by ITC (Fig. 3E) and found K_d values of (89.3 ± 47.4) nM for K48-linked and (35.3 ± 5.6) nM for K63-linked di-Ub chains. These affinities were lower than for unanchored mono-Ub, possibly due to the potential spatial clashes between the second Ub molecule in the chains and the C-terminal UIM domain of UBiP10 (fig. S4, D and E), especially for K48 di-Ub. Considering that the UIM domain from RAP80 is a K63-linked poly-Ub chain binding motif (21), it is plausible that UBiP10 binds to K63-linked di-Ub better than to K48-linked di-Ub. We next used AlphaFold3 to predict the structure of a UBiP10−linear di-Ub chain complex (as simulating a specific linkage with AlphaFold is still not possible; fig. S4F) and found that the ipTM value was 0.53, which is lower than for the UBiP10–mono-Ub complex, also indicating a weaker binding.

To further explore the specificity of UBiP10, we next evaluated its performance with cell lysates. We first treated A549 cells with or without the proteasome inhibitor MG132, prepared cell lysates, and incubated them with GST-UBiP05 or GST-UBiP10 prebound beads (Fig. 3F). By immunoblotting Ub in equal protein amount of cell lysate, we found that the characteristic smear of ubiquitinated proteins was increased by MG132 treatment, and as expected, these proteins were captured by GST-UBiP05, confirming that UBiP05 interacts with ubiquitinated proteins, in addition to unanchored mono-Ub. In contrast, GST-UBiP10 only bound to unanchored mono-Ub, independent of the ubiquitinated proteins increase under MG132 treatment. In addition, we compared UBiP10 and the hHDAC6_ZnF_(1108–1215) for their capacity to capture unanchored Ub from cell lysates. As shown in Fig. 3G, both proteins selectively pulled down unanchored mono-Ub, but UBiP10 captured substantially more mono-Ub than the natural ZnF domain. Last, we also compared our UBiP10 with the previously reported unanchored Ub binder tUI (11). Similar to UBiP05, purified tUI pulled down ubiquitinated proteins from the cell lysate. In contrast, under the same conditions, UBiP10 only pulled down the unanchored mono-Ub (Fig. 3H). AlphaFold3 prediction indicates that tUI may retain binding to Ub even when it lacks its C terminus; An ipTM score of 0.73 was obtained for the tUI–Ub_(ΔC-ter) complex (fig. S4G). Furthermore, we assessed whether UBiP10 interacts with Ub-like proteins (UBLs), including SUMO1, NEDD8, and ISG15. To this end, we performed coimmunoprecipitation (CoIP) in A549 cells stably expressing hemagglutinin (HA)–mScarlet or HA-mScarlet-UBiP10 using HA-based pulldown. As shown in fig. S4H, none of these UBLs were detected in the UBiP10 precipitates, indicating that UBiP10 does not appreciably bind SUMO1, NEDD8, or ISG15 under these conditions. Together, these findings confirm the high selectivity and capacity of UBiP10 to enrich unanchored Ub from cell lysates.

However, in the data presented above, only mono-Ub was detected from the cell lysate, while the characteristic Ub ladder—indicative of unanchored poly-Ub chains—was notably absent. We first supplied the cell lysates with purified K48-linked di-Ub chains and then incubated them with GST-UBiP10 prebound beads (fig. S4I). As expected, both K48-linked di-Ub chains and mono-Ub were captured by GST-UBiP10. Next, we overexpressed linear Ub chains in 293T cells. These poly-Ub chains (linear di-Ub, tri-Ub, and tetra-Ub) were designed with mutations to make them resistant to deubiquitinases: Except for the last distal Ub molecule, every proximal Ub has its C terminus: “-LRLRGG-” changed to “-LRAAAA-.” A clear Ub ladder was obtained following GST-UBiP10 pulldown (Fig. 3I), confirming the ability of UBiP10 to capture various unanchored poly-Ub chains, from dimer to nonamer from cell lysate.

In previous work, we demonstrated that IAV particles contain unanchored Ub (4), the presence of which is important for virus uncoating; therefore, we assessed the ability of UBiP10 to enrich unanchored Ub from purified IAV virus, which represents a natural sample as compared to overexpressed poly-Ub chains in cells. We used concentrated purified IAV viruses and first broke the viral capsids with radioimmunoprecipitation assay (RIPA) buffer, followed by sonication, after which the lysate was incubated with GST-UBiP10 prebound beads. As shown in Fig. 3J, GST-UBiP10 captured unanchored Ub, including Ub mono-, di-, tri-, and tetra-Ub chains in the IAV lysates. Given that K63-linked poly-Ub chains are up-regulated upon activation of the RIG-I signaling pathway during viral infection (22), we infected A549 cells with IAV [multiplicity of infection (MOI) = 50] or treated them with poly(I:C), a viral RNA mimic and RIG-I pathway activator. We then used UBiP10 to enrich endogenous K63-linked poly-Ub chains. As shown in Fig. 3K and fig. S4J, we successfully isolated di-, tri-, and even tetra-Ub species from the cell lysate. Last, we tested whether sequestering unanchored Ub with UBiP10 in cell affects IAV uncoating, a process dependent on unanchored Ub (4). A549 cells expressing HA–mScarlet or HA–mScarlet–UBiP10 were infected with IAV (MOI = 100). At 2 hours and 15 min postinfection, viral M1 protein distribution was analyzed as readout for uncoating (diffuse cytosolic staining indicates completion). UBiP10 expression markedly impaired M1 redistribution (fig. S4K), confirming that unanchored Ub is required for efficient IAV uncoating. In addition, we determined the cellular UBiP10-interacting proteins by affinity purification–mass spectrometry (AP-MS). Ub was identified as the most enriched protein by UBiP10 (Fig. 3L), and again, no UBLs were detected. In conclusion, UBiP10 is a highly specific unanchored Ub binder, which can efficiently capture unanchored Ub from biological samples.

Determination and visualization of unanchored Ub in cells with UBiP10

Given UBiP10’s high affinity and selectivity for unanchored Ub, we explored its potential as a tool for detecting unanchored Ub in fixed cells by immunofluorescence microscopy. We first incubated fixed A549 cells with excess GST-UBiP10 and washed the cells extensively. We then incubated the cells with a primary antibody against GST and a fluorescent secondary antibody to visualize the location of unanchored Ub (fig. S5A). To assess the method’s specificity, we also incubated fixed cells with an antibody against Ub as a positive control or with GST protein and GST antibody as a negative control. The cells incubated with GST-UBiP10 protein and GST antibody showed an even distribution of unanchored Ub throughout the cell (fig. S5B). By contrast, the Ub antibody stained the nucleus predominantly, indicating high levels of Ub, likely ubiquitinated histones, while the cells incubated with GST protein and GST antibody displayed only little background immunofluorescence. To confirm that the GST-UBiP10 immunofluorescent signal specifically reflects unanchored Ub, we treated cells with the proteasome inhibitor bortezomib, which induces formation of an aggresome, a large perinuclear structure filled with ubiquitinated proteins (7). Brightly stained aggresomes were detected by the Ub antibody; in contrast, staining with GST-UBiP10 did not reveal protein aggregates (fig. S5C), confirming that UBiP10 does not recognize ubiquitinated proteins. Thus, these data confirm that GST-UBiP10 can be used as a probe to detect unanchored Ub by immunofluorescence microscopy.

Next, we tested the ability of UBiP10 to visualize the variation of unanchored Ub level in cells treated with either the proteasome inhibitor bortezomib (to decrease unanchored Ub levels by inhibiting the degradation of ubiquitinated proteins) or with the E1 Ub-activating enzyme inhibitor TAK-243 (to increase unanchored Ub levels by stopping the ubiquitination cascade). To explore the applicability of UBiP10 in different tissues and species, we selected four cell lines: A549 (human lung epithelial cells), HeLa (human cervical epithelial cells), SW1353 (human chondrosarcoma cells), and MEFs (mouse embryonic fibroblasts). The intensity of the unanchored Ub signal, obtained by staining cells with GST-UBiP10 protein and GST antibody, significantly decreased in all cell lines treated with bortezomib and increased strongly in cells treated with TAK-243 (Fig. 4, A and B). We previously showed that IAV virions contain unanchored Ub, which plays an important role for virus uncoating (4, 23). As shown in Fig. 4C, GST-UBiP10 was also capable of visualizing the unanchored Ub inside IAV virions. Together, these data validate the effectiveness of UBiP10 as a probe for visualizing unanchored Ub by immunofluorescence microscopy.

Fig. 4. Determination and visualization of unanchored Ub in cells with UBiP10.

(A) Immunofluorescence microscopy analysis of unanchored Ub, visualized by GST-UBiP10 protein and GST antibody, in A549, HeLa, SW1353, and MEF cells under different treatments. Bortezomib treatment: 2 μM for 3 hours. TAK-243 treatment: 1 μM for 3 hours. Scale bars, 20 μm. (B) Quantification of unanchored Ub’s mean signal intensity in (A) (n = 3). Each dot represents a picture taken by 40× objective (with all the cells within the picture analyzed); for each condition, ∼200 cells were included in the quantification. One-way analysis of variance (ANOVA) test was used for statistical analysis. P values are listed, and P < 0.05 is considered as significant. Data are presented as mean ± SD. (C) Immunofluorescence of unanchored Ub in influenza virions. Red signal, viral HA protein; green signal, unanchored Ub (visualized by GST-UBiP10 protein and GST antibody incubation); yellow, merged signal. Scale bar, 1 μm. (D) Immunofluorescence of total Ub (obtained from GST-UBiP05 protein and GST antibody incubation) and unanchored Ub (obtained from GST-UBiP10 protein and GST antibody incubation) in dimethyl sulfoxide (DMSO)– and bortezomib (2 μM for 12 hours)–treated A549 cells. Green, actin signal. Nucleus was visualized by 4′,6-diamidino-2-phenylindole (DAPI) staining. Scale bars, 20 μm. Signal from GST group (obtained from GST protein and GST antibody incubation) represents the background intensity (n = 3). (E) Quantification of signal intensity of GST-UBiP05 (I_GST-UBiP05), GST-UBiP10 (I_GST-UBiP10), and GST (I_GST) in (D). Minimal 200 cells were analyzed for each condition. One dot represents one image taken by 40× objective from (D). One-way ANOVA test was used for statistical analysis. P values are listed, and P < 0.05 is considered as significant.

Next, we leveraged the differential specificities of UBiP05 and UBiP10 to establish an imaging-based approach for monitoring unanchored Ub in cells. UBiP05 recognizes both conjugated and free Ub (Fig. 3F), whereas UBiP10 preferentially detects unanchored Ub. After background subtraction (signal of the GST control), the respective fluorescence signals provide complementary relative readouts of total and unanchored Ub pools. We therefore used their relative level as a proxy to compare unanchored Ub across conditions.

We first validated this strategy in vitro. In enzyme-linked immunosorbent assay (ELISA) assays with defined mixtures of mono- and anchored Ub (3 × Ub–RARA), the UBiP10 signal increased with the fraction of unanchored Ub (fig. S5D). A plate-based immunofluorescence titration, using the same detection setup as a microscopy experiment, showed that both probes scale approximately with Ub input after background subtraction (fig. S5E), albeit with ∼10 to 35% deviations from ideal linearity (fig. S5F). In cells treated with bortezomib, GST-UBiP05 recapitulated the increase in total Ub as the Ub antibody, whereas GST-UBiP10 did not show a corresponding rise (Fig. 4, D and E, and fig. S5G). Ratio-based analysis across ∼500 cells suggested a relative decrease in unanchored Ub upon proteasome inhibition (fig. S5H). Consistent with the in vitro characterization, we interpret these values as semiquantitative, supporting the use of UBiP05/UBiP10 to monitor relative changes in unanchored Ub. For precise quantification, enrichment of unanchored Ub using UBiP10 combined with MS and isotope-labeled standards would be required.

UBiP10 as a probe reveals an unanchored Ub coat surrounding aggresomes in fixed cells

Having demonstrated the feasibility of using UBiP10 as a probe to visualize unanchored Ub in cells, we used this tool to explore the spatial distribution of unanchored Ub during the aggresome life cycle. Although it is known that both the assembly and disassembly of aggresomes rely on unanchored Ub (6, 14), imaging data showing directly the presence of unanchored Ub within or near aggresomes have been lacking. We aimed to fill this gap using UBiP10 to visualize the localization of unanchored Ub at the aggresome site.

To investigate the distribution of unanchored Ub during aggresome formation, we treated A549 cells with bortezomib for up to 9 hours and fixed them at 3-hour intervals. Immunofluorescence staining with antibodies against Ub revealed an increasing number of cells containing Ub aggregates—preaggresome structures—near the nucleus over time (white arrows in “Ub Ab”; fig. S6A). Immunofluorescence staining with GST-UBiP10, by contrast, revealed a ring of unanchored Ub forming around the cell nucleus simultaneously with the formation of preaggresomes, with the number of cells displaying this perinuclear Ub ring increasing over time (white arrows in “GST-UBiP10 views 1 and 2”; fig. S6A). These findings suggest that the spatial distribution of unanchored Ub varies during aggresome formation.

To explore the presence of unanchored Ub in or around aggresomes, we treated A549 cells with bortezomib for 12 hours to induce fully grown aggresomes. Then, we fixed the cells, saturated them with GST-UBiP10 to bind all available unanchored Ub, and subsequently incubated them with the Ub antibody to visualize aggresomes and Ub-conjugated proteins (Fig. 5A). Bortezomib treatment effectively induced aggresomes (indicated by white arrows in the Ub antibody panel). To ensure complete binding of UBiPs to the unanchored Ub in fixed cells, we added 5 to 10 μg of purified UBiPs (corresponding to ∼740 to 1480 nM, which is >50 times higher than the K_d) to one well of the chamber slide (∼50,000 cells). Notably, we observed a “cloud” of unanchored Ub at the aggresome site and localized around the Ub antibody–stained aggresomes (see merged panels in Fig. 5A). In zoomed-in views, unanchored Ub was seen highly concentrated around the site of individual aggresome (see examples in Fig. 5B). Signal intensity profiles of these aggresome sites showed that the peak of unanchored Ub did not overlap with the peak of Ub signal in the aggresome itself (Fig. 5B, right). Instead, the unanchored Ub signal was enriched in the peripheral area surrounding the aggresome. Quantification of the unanchored Ub intensity at three distinct cellular locations (“aggresome,” “periaggresome,” and “remote area” as illustrated in fig. S6B) confirmed its enrichment selectively around the aggresome (Fig. 5C). In addition, we also examined the enrichment of unanchored Ub in differentiated SH-SY5Y cells, a widely used in vitro model for neurodegenerative and proteostasis-related disorders. As shown in fig. S7A, aggresome formation was observed in bortezomib-treated differentiated SH-SY5Y cells. Spatial analysis (fig. S7B) of the GST-UBiP10 signal revealed a clear enrichment of unanchored Ub in the periaggresomal region, consistent with our observations in A549 cells. Quantitative analysis (fig. S7C) of signal distribution confirmed that unanchored Ub is preferentially localized around, rather than within, the aggresome core. Thus, we propose that aggresomes have a previously unidentified outer coat structure that is rich in unanchored Ub.

Fig. 5. Identification of an unanchored Ub coat around the aggresome with UBiP10.

(A) Immunofluorescence microscopy of unanchored Ub (obtained by GST-UBiP10 protein and GST antibody incubation) and Ub (obtained by Ub antibody incubation) in aggresome. White arrows point to the aggresome sites. Aggresomes in white dashed rectangles were chosen for following analysis in (B). Scale bars, 20 μm. Bortezomib: 2 μM for 12 hours. (B) Zoomed-in views of two aggresomes in (A). Signal intensity distribution alongside the white dashed line (from left to right) was plotted by ImageJ. (C) Quantification of the unanchored Ub intensity at three different positions in bortezomib (2 μM for 12 hours)–treated A549 cells. In total, ∼50 aggresomes per condition were quantified. One dot represents one aggresome. Statistical analysis was performed with one-way ANOVA test; P values are indicated. P < 0.05 is considered as significant (n = 3). (D) IB for myosin 10 knockdown (KD) in A549 cells and HDAC6 in CRISPR-Cas9–generated stable HDAC6 KO A549 cells. (E) Immunofluorescence of Ub (obtained by Ub antibody) and unanchored Ub (obtained by GST-UBiP10 protein and GST antibody incubation) in cells from (D). Bortezomib (2 μM for 12 hours) was used to inducing aggresome formation. White rectangles: zoomed-in views of four aggresomes. Scale bar, 20 μm. (F) Signal intensity alongside the white dashed line [from the left to right in (E)] was analyzed by ImageJ. (G) Quantification of the unanchored Ub intensity at three different positions in bortezomib-treated, scrambled RNA (scRNA)–transfected, siRNA for myosin 10 (siMyo)–transfected, and HDAC6 KO A549 cells in (E) and fig. S6C. In total, ∼50 aggresomes per condition were quantified. Statistical analysis as in (C). n.s., not significant.

Previous studies have implicated HDAC6 and myosin 10 in the regulation of aggresome formation and disassembly (6, 14). Although in a different manner, both proteins are known to interact with Ub, and our previous work showed that myosin 10 also forms a coat around the aggresome (24). On the basis of this, we hypothesized that either myosin 10 or HDAC6 might be responsible for the localization of unanchored Ub around aggresomes. To test this hypothesis, we transiently depleted myosin 10 by small interfering RNA (siRNA) or used A549 HDAC6 knockout (KO) cells (Fig. 5D) and induced aggresome formation by bortezomib treatment for 12 hours. Unanchored Ub was then visualized by staining with purified GST-UBiP10, used similar to an antibody after cell fixation (Fig. 5, E and F). When myosin 10 was depleted, no coat of unanchored Ub was visible at the periphery of the aggresomes. By contrast, HDAC6 KO had no effect on the formation of the unanchored Ub coat surrounding the aggresome (Fig. 5, E and F), nor did treatment with scrambled siRNA (fig. S6C). Quantification of unanchored Ub intensity revealed that neither scrambled RNA (scRNA) transfection nor HDAC6 KO affected its enrichment around the aggresome, where the unanchored Ub signal remained notably higher than inside the aggresome. In contrast, knockdown of myosin 10 abolished this difference (Fig. 5G). These findings indicate that the mechanism by which unanchored Ub accumulates around the aggresome involves myosin 10, but not HDAC6. In summary, using UBiP10, we have shown a peripheral coat of unanchored Ub around the aggresome.

DISCUSSION

In this study, we set out to develop a novel tool to selectively enrich and detect unanchored Ub. Starting from the ZnF domain of HDAC6, which naturally binds unanchored Ub, we first used affinity maturation to enhance its affinity for Ub and identified a critical mutation that doubled the affinity. Using deep learning protein design tools, we then combined this mutant ZnF domain to other known Ub-binding domains to create a series of novel Ub binders. We demonstrated that one of the designed proteins, UBiP10, is specific for unanchored Ub, whereas another, UBiP05, recognizes both unanchored and protein-conjugated forms of Ub. We used these two proteins to estimate relative changes in the unanchored Ub fraction in cells in presence and absence of proteasome inhibitor. Moreover, using UBiP10, we identified a peripheral coat of unanchored Ub accumulating around mature aggresomes. This unanchored Ub coat formation is mediated by myosin 10. Thus, UBiP10 is a novel powerful tool to detect unanchored Ub, which opens the door for further exploration of the functional importance and dynamics of unanchored Ub in cells.

We demonstrated the importance of the flexible loop region preceding the structured HDAC6 ZnF domain for binding unanchored Ub: While the ZnF construct containing the loop has an affinity of ∼60 nM, in agreement with an early study (15), the core ZnF structured region, which had been crystallized together with Ub (14), has a ∼80× lower affinity. Using AlphaFold2 prediction, we found that this unstructured region, located just before the first α helix of the structured ZnF domain, may bind directly to Ub, suggesting that it represents a novel Ub-binding motif. The accessibility of this unstructured region appears to be tightly regulated, as the affinity of full-length hHDAC6 for Ub is lower, and similar to that of the core structured ZnF domain (∼1.0 μM). We hypothesize that in the full-length HDAC6, the flexible loop may be concealed, resulting in its lower affinity for Ub. It is possible that under certain conditions, this hidden region might become exposed, with a corresponding increase in HDAC6 affinity for Ub. In addition, during IAV infection, hHDAC6 is cleaved at site D1088 by caspase-3 (25, 26), releasing the ZnF domain together with its N-terminal flexible region. The released fragment is expected to have a high affinity for unanchored Ub, and because the ZnF domain of HDAC6 is crucial for uncoating IAV (4), this cleavage may impact IAV infection.

HDAC6 is an established marker for aggresomes, and an HDAC6-dynein motor complex transports misfolded proteins to the microtubule-organizing center, resulting in accumulation of ubiquitinated proteins within aggresomes (27). Although the ZnF domain of HDAC6 binds specifically to unanchored Ub, the mechanism by which HDAC6 is recruited to aggresomes remains unclear. One intriguing hypothesis proposes that during aggresome formation, the poly-glutamine–containing deubiquitinase ataxin-3 removes covalently bound Ub, converting it into unanchored Ub (14), which may then act as a docking site for HDAC6. However, no direct evidence was available to show the presence of unanchored Ub during aggresome formation. Here, using UBiP10, we unveiled the presence of unanchored Ub around the aggresome, providing evidence for unanchored Ub–based HDAC6 recruitment. Together with previous reports that myosin 10 and other cytoskeletal proteins, including vimentin and actin, form a cage around aggresomes (24, 28, 29), we propose a previously unidentifed model of the aggresome. In this model, the outer shell of aggresomes contains myosin 10 and unanchored Ub, while the inner core of aggresomes is filled with HDAC6 and anchored Ub.

To explain why myosin 10, rather than HDAC6, mediates the formation of the unanchored Ub coat, a consideration of their spatial distribution and functional roles may be useful. Myosin 10 forms together with actin a filamentous structure around the aggresome and interacts with K48-linked poly-Ub chains and ubiquitinated proteins (24). This suggests that myosin 10, as a motor protein, may transport anchored or/and unanchored Ub to the periaggresome, where deubiquitination by ataxin-3 (14) occurs and unanchored Ub then accumulates. This activity positions myosin 10 as a key mediator in forming the unanchored Ub coat. In contrast, HDAC6 binds to preexisting unanchored Ub to facilitate the maturation and stabilization of aggresomes rather than initiating coat formation. While the precise role of the unanchored Ub coat remains to be defined, it may facilitate aggresome formation by recruiting HDAC6 or, alternatively, contribute to aggresome disassembly in cooperation with myosin 10 and the deubiquitinase Poh1 (6). Understanding the role of these proteins in the assembly and disassembly of aggresomes may provide insights into the formation of pathological aggresome-like structures, which are hallmarks of many neurodegenerative diseases (30, 31).

Our UBiP10 is a versatile novel tool for studying unanchored Ub, which will complement the widely used tandem Ub-binding entities (fig. S7D) and enable the selective capture, quantification, and visualization of unanchored Ub in cells and other biological samples such as viral particles. Despite cell lysates containing more Ub than viral lysates, we can readily pull down unanchored Ub chains from the viral lysates but not from the cell lysates. This suggests a higher relative abundance of unanchored Ub chains within the viral particles, but the origin of this unanchored Ub is unclear. Fusing UBiP10 with a fluorescent label and expressing it in cells may allow to identify where this unanchored Ub originates. UBiP10 can also have broad applications in pathological contexts, such as neurodegenerative diseases, where it can help probe the role of unanchored Ub in Ub-positive protein aggregates, and in osteoarthritis, where an increased unanchored Ub level linked to reduced proteasome activity has been shown (32); UBiP10 may thus serve as a disease diagnostic tool. Additional uses include studying unanchored Ub dynamics in cancer, immune responses, and inflammatory diseases, offering insights into disease mechanisms and a potential diagnostic reagent.

MATERIALS AND METHODS

Cell culture

A549 [American Type Culture Collection (ATCC) #CRM-CCL-185], SW1353 (ATCC #HTB-94), MEF (ATCC #SCRC-1008), human embryonic kidney (HEK) 293T cells (ATCC #CRL-3216), HeLa cells (ATCC #CRM-CCL-2), and SH-SY5Y (ATCC #CRL-2266) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, #11-965-092) supplied with 10% fetal bovine serum (FBS; Gibco, #A5256701) and Penicillin-Streptomycin (Gibco, #15070063). Cells were maintained at 37°C with 5% CO₂ before and after the treatment. To differentiate SH-SY5Y, cells were incubated with all-trans retinoic acid (ATRA) (10 μM) for 3 to 5 days.

Antibodies and reagents

The antibodies and staining reagents used in Western blot and immunofluorescent experiments were as follows: Ub Monoclonal Antibody (Cytoskeleton, #AUB01, clone:P4D1), α-tubulin antibody (Merck, #T9026), acetyl–α-tubulin (Lys⁴⁰) monoclonal antibody (mAb; Thermo Fisher Scientific, #32-2700), myosin 10 antibody (CST, #3404), HDAC6 antibody (CST, #7558), HA-Tag (C29F4) mAb (CST, #3724), IAV X31 hemagglutinin (HA) antibody (33), GST antibody (CST, #2625; and Invitrogen, #13-6700), Actin Monoclonal Antibody (Sigma-Aldrich, #ACTN05, clone:C4), Goat anti-Mouse/Rabbit IgG (H + L) Superclonal Recombinant Secondary Antibody Alexa Fluor 488/568 (Invitrogen, #A28175/A11031), Acti-stain 488 phalloidin (Cytoskeleton, #PHDG1), and 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific, #D1306).

Protein expression and purification

Constructs of UBiPs and HDAC6 fragments were fused into pOPINJ vector and transformed into BL21(DE3) cells [New England Biolabs (NEB), #C2527H]. Successfully transformed bacteria were selected by ampicillin plate and seeded to 50 ml of LB medium (homemade) and cultured overnight at 37°C. A 10-ml overnight cultured medium was added to 1 liter of LB medium. For each purification, we used 6 liters of LB medium for culturing bacteria. After 3.5 hours culturing at 37°C on shaker (220 rpm), the optical density (OD) value of the medium was around 0.6, and isopropyl-β-d-thiogalactopyranoside (IPTG; Invitrogen, #AM9464) (final concentration = 1 mM) was added to the medium to induce protein expression. After IPTG addition, the bacteria medium was transferred to a shaker at 17°C and continued to be cultured for 18 hours. Bacteria were pelleted by centrifugation (room temperature, 4000 rpm, 15 min). Pellet from 1 liter of cultured bacteria was stored in one 50-ml Falcon tube (Thermo Fisher Scientific, #14-432-22) and stored at −80°C.

For GST-tagged protein purification, bacteria pellets were suspended in ice-cold phosphate-buffered saline (PBS) and sonicated (15-s on, 15-s off, 10 min in total, 35% amplitude) for lysing the cells. Lysates were centrifuged to remove debris (30,000 rpm, 45 min, 4°C) in a Beckman ultracentrifuge with a Ti 75 rotor. The supernatant was loaded to GSTrap HP Columns (Cytiva, #17528201), washed with 50 ml of PBS buffer (homemade), and eluted in 30 ml of elution buffer [50 mM tris (pH 8.5) and 10 mM reduced glutathione]. The flowthrough was collected and reloaded to PBS-washed GST_trap HP column again and eluted with 30 ml of elution buffer. These procedures were repeated three times. Eluted proteins were concentrated by an Amicon Ultra Centrifugal Filter (30 kDa) (Merck, #UFC9030), loaded onto a gel filtration system, and separated by a Superdex 200 Increase 10/300 GL column (GE, #GE28-9909-44) [buffer: 20 mM tris (pH 7.5), 150 mM NaCl, and 2 mM TCEP]. TCEP is short for tris(2-carboxyethyl)phosphine hydrochloride (Sigma-Aldrich, # C4706). Fractions were checked by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gel (NuPAGE Bis-Tris Mini Protein Gels, 4 to 12%, 1 mm, catalog number: NP0321BOX), and the fractions from main peak for GST-tagged proteins were pooled and concentrated to 5 to 10 mg/ml. Protein concentration was determined by NanoDrop (Thermo Fisher Scientific, NanoDrop 2000/2000c Spectrophotometers, catalog number: ND-2000). Concentrated proteins were flash frozen in liquid nitrogen.

Recombinant His-tagged tUI was cloned into the pET-28a (+) expression vector and transformed into E. coli BL21(DE3) cells. Protein expression was induced as previously described. Bacterial pellets were resuspended in HisTrap binding buffer [50 mM tris-HCl (pH 7.5), 500 mM NaCl, and 10 mM imidazole] and lysed by sonication on ice. Cell debris was removed by ultracentrifugation at 30,000 rpm for 45 min at 4°C. The clarified lysate was applied to a HisTrap HP column (Cytiva, #17524801) preequilibrated with binding buffer. Bound proteins were eluted using an imidazole gradient (10 to 500 mM) on an ÄKTA pure chromatography system. Fractions containing His-tUI were identified by SDS-PAGE, pooled, and dialyzed overnight at 4°C against low-salt buffer [50 mM tris-HCl (pH 7.5) and 100 mM NaCl]. The dialyzed sample was subsequently applied to a HiTrap Q HP anion exchange column (Cytiva, #17115401) and eluted with a linear NaCl gradient (100 mM to 1 M). Peak fractions containing His-tUI were pooled, concentrated to ∼500 μl using Amicon Ultra Centrifugal Filter (10 kDa) (Merck, #UFC9010), and subjected to size-exclusion chromatography on a Superdex 75 Increase 10/300 GL column (Cytiva, #29148721) equilibrated in 50 mM tris-HCl (pH 7.5), 150 mM NaCl, and 2 mM TCEP.

To generate K48- and K63-linked poly-Ub chains that were used for ITC with UBiP10, in vitro ubiquitination reactions were prepared in a 3-ml volume (can be scaled up) containing 50 mM tris-HCl (pH 8.0), 1 mM mono-Ub, 2 μM UBE1 (E1), 10 mM adenosine 5′-triphosphate (ATP), 10 mM MgCl₂, 1 mM dithiothreitol (DTT), and an ATP regeneration system [100 mM phosphocreatine, inorganic pyrophosphatase (6 U/ml), and creatine phosphokinase (6 U/ml)]. For K63-linked chains, 10 μM UEV1A and 10 μM UBCH13 (E2s) were included; for K48-linked chains, 20 μM E2-25K (E2) was used. Reactions were incubated at 37°C and monitored by SDS-PAGE and Coomassie staining until ∼90% of the mono-Ub substrate was consumed (5 to 6 hours). The reaction mixture was then diluted 20-fold in 50 mM ammonium acetate (pH 4.5) with 50 mM NaCl and subjected to cation exchange chromatography using a HiTrap Heparin HP column (Cytiva, #17040701), eluting with a 50 mM to 1 M NaCl gradient. Eluted fractions were pooled and further purified by Superdex 200 Increase 10/300 GL column (GE, #GE28-9909-44) in 20 mM tris-HCl (pH 7.5) to separate short and long chains. Linkage specificity was confirmed using K63- or K48-specific antibodies (CST, #5621; and Abcam, #ab140601, respectively), and concentrations were determined by absorbance at 280 nm. Purified chains were flash frozen in liquid nitrogen and stored at −80°C. Enzyme concentrations and incubation times were adjusted as needed to account for batch-to-batch variability and ensure optimal reaction efficiency. This method was adapted from the study of Li et al. (34).

ITC for affinity determination

Proteins (e.g., HDAC6 constructs, UBiPs, and Ub) were first dialyzed in the binding buffer [10 mM tris (pH 7.5) and 100 mM NaCl] for at least 3 hours at room temperature. Then, the protein concentrations were determined by NanoDrop (Thermo Fisher Scientific, NanoDrop 2000/2000c Spectrophotometers, catalog number: ND-2000). For Ub, which was added in the syringe of VP_ITC machine, the concentration was from 0.03 to 0.05 mM, and volume was 500 μl. For UBiPs, which were added in the cell of VP_ITC machine, the concentrations were varied from 0.001 to 0.005 mM, and volume was 2 ml. With varied concentrations, the K_d from different replicates fall into a narrow range. Protein solutions were degassed by centrifugation (14,000g for 10 min at 4°C) and placed on the bench at room temperature for 10 min. After loading the proteins to the VP_ITC system, the cell temperature should be set as 25°C.

Immunoblotting

Purified proteins were directly mixed with 4× LDS sample buffer (Thermo Fisher Scientific, #NP0007). Cell pellets were lysed in CoIP buffer [10 mM tris (pH 7.5), 100 mM NaCl, 0.5% NP-40, and 0.5 mM EDTA] on ice for 30 min. The lysate was then centrifuged at 14,000 rpm for 10 min at 4°C, and the supernatant was transferred to a new 1.5-ml tube. Protein concentration in the supernatant was measured using the Bradford assay. Equal amounts of protein were loaded onto NuPAGE 4 to 12% gradient Bis-Tris gels (200 V for 36 min), followed by membrane transfer using the Bio-Rad Trans-Blot Turbo transfer system with the 0.2-μm PVDF Transfer Packs (Bio-Rad, #1704156) using the “mixed molecular weight” program. After transfer, polyvinylidene difluoride (PVDF) membranes were rinsed with Milli-Q water for 3 min and then blocked in blocking buffer [5% skim milk, 20 mM tris (pH 7.5), and 100 mM NaCl] at room temperature for 1 hour. Primary antibodies were added to 5 ml of blocking buffer and incubated with the membranes overnight at 4°C. The membranes were washed three times with TBS-T and then incubated with the appropriate diluted secondary antibody (in 1× TBS-T with 5% skim milk) for 1 hour at room temperature with gentle rocking. The membranes were washed three more times with TBS-T. For detection, the membranes were treated with Amersham ECL Detection Reagents (Cytiva, #RPN3004) at room temperature for 3 min.

GST pulldown with 3xUb-RARA and mono-Ub

For a typical pull-down experiment, 20 μg of GST-tagged proteins was incubated with 10 μl of GST_trap beads (Proteintech, #sta) in a cold room for 30 min. The prebound beads were then centrifuged to remove the unbound proteins in the supernatant and washed twice with TBS [20 mM tris (pH 7.5) and 150 mM NaCl]. Subsequently, the beads were incubated with either 200 μg of (total protein amount) cell lysate or 10 μg of 3xUb-RARA or Ub chains in the cold room on a rotator (10 rpm) for 2 hours. The beads were washed three times with TBS. Captured proteins were either analyzed by immunoblot or by MS. For immunoblot, the bead samples were dissolved in 1× LDS sample buffer (Thermo Fisher Scientific, #NP0007) supplemented with 10 mM DTT and boiled for 5 min at 85°C before loading on the 4 to 12% Bis-Tris NuPAGE gels. Beads were spun down by centrifugation (×12,000g for 1 min).

Lb^pro analysis of UBiP10’s specificity against 3xUb-RARA and mono-Ub

For Lb^Pro (a gift from D. Komander’s laboratory) digestion followed by MS analysis, samples from GST-UBiP10 and GST pull-down experiments were digested on-bead using Lb^Pro protease. GST_trap beads were first saturated with GST-UBiP05 or GST-UBiP10 proteins, and then a protein mixture of mono-Ub (50 μl, 50 nM) and 3xUb-RARA (50 μl, 50 nM) was incubated with the beads. After intensive washes, 50% of the bead volume was incubated with 10 μl of 10 μM Lb^Pro in digestion buffer [50 mM Hepes (pH 7.5) and 0.2 mM TCEP] at 37°C for 16 hours. After centrifugation to pellet the beads, 5 μl of the supernatant was mixed with 50 μl of a 1:1 (v/v) solution of Milli-Q water and acetonitrile (MeCN), both containing 0.1% formic acid. A 5-μl aliquot of this mixture was analyzed by liquid chromatography–MS using a Bruker compact 20260 QTOF mass spectrometer. Chromatographic separation was performed on an ACQUITY UPLC Protein BEH C4 column (300 Å, 1.7 μm, 2.1 mm by 100 mm; Waters, #186004496) at a flow rate of 0.4 ml/min with a 27-min gradient increasing MeCN from 10 to 95%. The mobile phases consisted of Milli-Q water with 0.1% formic acid (solvent A) and high-performance liquid chromatography–grade acetonitrile with 0.1% formic acid (solvent B). Electrospray ionization was used for ionization.

GST pulldown with linear, K48, and K63 di-Ub chains

GST and GST-UBiP10 (10 μg) were incubated with 10 μl of GST_trap beads (Proteintech, #sta) in a cold room for 30 min. The prebound beads were then centrifuged to remove the unbound proteins in the supernatant and washed twice by TBS. Linear (5 μg; LifeSensors, #SI-0102-0100), K48 (Enzo, #BML-UW9800) and K63 (Enzo, #BML-UW0730) di-Ub chains were added into either GST (negative control) or GST-UBiP10 prebound beads and incubated together on a roller in the cold room (4°C) for 30 min. After incubation, beads were washed three times with TBS-T [20 mM tris (pH 7.5), 150 mM NaCl, and 0.1% Triton X-100] and then three times with TBS [20 mM tris (pH 7.5) and 150 mM NaCl]. Supernatant was completely removed, and then 20 μl of 2× LDS sample loading buffer was added to the beads to elute all the proteins. Samples were boiled at 85°C for 10 min and loaded to NuPAGE Bis-Tris 4 to 12% gradient gel for following immunoblotting detection.

CoIP of HA-tagged proteins and UBLs

A549 cells stably expressing HA-mScarlet or HA-mScarlet-UBiP10 were seeded at 1.6 × 10⁶ cells per 10-cm dish and cultured overnight. Cells were harvested and lysed in CoIP buffer [10 mM tris-HCl (pH 7.5), 100 mM NaCl, 0.5% NP-40, and 0.5 mM EDTA] supplemented with protease inhibitors. Lysates were clarified by centrifugation, and protein concentration was determined before immunoprecipitation. Anti-HA magnetic beads (Thermo Fisher Scientific, #88836) were equilibrated in CoIP buffer and incubated with lysates (10 μl of bead slurry per 300 μg of total protein) overnight at 4°C with rotation. Beads were subsequently washed three times with TBS to remove nonspecifically bound proteins. Bound proteins were eluted by boiling in LDS sample buffer and analyzed by immunoblotting as described above.

MS analysis of UBiP10’s interactome

For the UBiP10’s interactome analysis in Fig. 3K, the samples from GST-UBiP10 and GST pulldown were digested on beads by trypsin (Thermo Fisher Scientific, #90305) and Lys-C (Thermo Fisher Scientific, #90051) enzyme. Proteins bound to beads were subjected to enzymatic digestion following a two-step protocol. First, 10 μl of Lys-C stock was diluted in 50 μl of digestion buffer. After completely removing the wash buffer from the beads, 6 μl of the Lys-C solution was added to the beads, briefly spun down (<1000g), and incubated at room temperature for 2 to 4 hours with shaking. Following incubation, 17 μl of 50 mM Hepes buffer (pH 8.5) was added, and the sample was spun down. Next, 1 μl of trypsin (0.2 μg/μl) was added, mixed by vortexing and centrifugation, and incubated overnight at 37°C. The next morning, another 1 μl of trypsin (0.2 μg/μl) was added to the reaction, which was then incubated for an additional 4 to 6 hours. After digestion, samples were spun down and either stored frozen or processed immediately for MS. For MS preparation, the digested peptide slurry was acidified with 1 μl of 20% trifluoroacetic acid, spun down, and sonicated for 2 min in an ultrasonic bath. After a final spin, magnetic beads were removed, and the supernatant was transferred to MS autosampler tubes to minimize peptide loss due to plastic adsorption. Samples were analyzed on Orbitrap Lumos (Thermo Fisher Scientific) with Vanquish Neo UHPLC in nano-flow LC mode, and an easy source with a 75 μm × 15 cm EasyC18 column. The samples were loaded on a C18 (0.3 × 5 mm) trap, and backward flush was used for the analysis. A MQ analysis was done searching DB human and E. coli.

His-tagged protein pulldown with Ni–nitrilotriacetic acid beads

Wild-type A549 cells were harvested from untreated cultures and lysed in CoIP buffer [10 mM tris-HCl (pH 7.5), 100 mM NaCl, 0.5% NP-40, and 0.5 mM EDTA]. Protein concentration was quantified using the Bradford assay, and 300 μg of total protein lysate was used per pull-down assay. Magnetic Ni–nitrilotriacetic acid (NTA) beads (Thermo Fisher Scientific, #78605) were prewashed three times with CoIP buffer by centrifugation at 500g for 30 s. A total of 15 μl of washed bead slurry was used per reaction. Purified His-HA-tUI or His-GST-UBiP10 (15 μg) was preincubated with the Ni-NTA beads to enable binding, followed by removal of the unbound proteins and the addition of A549 cell lysate. The mixture was incubated on a rotator in the cold room (4°C) overnight. Then, the beads were washed twice with CoIP buffer and subsequently washed twice with TBS buffer (10 mM tris-HCl (pH 7.5) and 100 mM NaCl] to reduce nonspecific interactions. Bound proteins were eluted by adding 20 μl of 1× LDS sample buffer (Invitrogen, #NP0007), followed by boiling at 85°C for 10 min. After centrifugation, the supernatant was analyzed by SDS-PAGE using a NuPAGE 4 to 12% Bis-Tris gel (Invitrogen, #NP0322BOX) for Western blotting.

GST pulldown between UBiP10 and overexpressed linear poly-Ub chains

On day 1, 1.6 × 10⁶ HEK293T cells were seeded in a 10-cm dish. On day 2, cells were transfected with 5 μg of plasmids encoding di-, tri-, or tetra-Ub constructs (designed by L.W. and synthesized by Twist Bioscience) using 10 μl of FuGENE HD transfection reagent (Promega, #E2311), following the manufacturer’s protocol. On day 5, cells were harvested in ice-cold PBS and pelleted by centrifugation at 500g for 10 min at 4°C. Cell pellets were stored at −20°C until lysis in CoIP buffer (composition as previously described). Total protein concentration was determined using the Bradford assay, and 500 μg of lysate was used for each GST pull-down assay. Pull-down experiments were performed as described in the previous section using 20 μg of GST or GST-UBiP10 and 10 μl of GST_trap beads per reaction.

IAV lysate preparation

IAV, strain X31, laboratory produced (titer: 1 × 10¹¹ per ml), was used for viral unanchored Ub pulldown by GST-UBiP10. A total of 50 μl of viral stock was used for lysis. A 500-μl RIPA buffer was added to 40 μl of viral stock, and the mixture was incubated overnight on a thermal shaker in the cold room (4°C for 800 rpm), followed by a sonication (in total 15 rounds; sonication program for each round was 15-s on with 15-s off). After this, the mixture was incubated with GST or GST-UBiP10 prebound GST_trap beads. Pull-down experiment was done as previously described in the “GST pulldown with linear, K48 and K63 di-Ub chains” section.

IAV uncoating assay

IAV infection was performed in infection medium consisting of DMEM supplemented with 50 mM Hepes and 0.2% bovine serum albumin (BSA; pH 6.8). For M1 uncoating assays, cells were maintained in the presence of 100 μM cycloheximide. Uncoating was assessed at 2.5 hours postinfection. A549 cells were seeded at 10,000 cells per well in 96-well optical-bottom plates (Greiner) 1 day before infection. IAV X31 was added at a MOI of 100, and plates were incubated on ice for 30 min to allow viral binding. Subsequently, cells were shifted to 37°C and incubated for 2 hours to permit viral entry and uncoating. Cells were then washed twice with PBS, fixed with 4% paraformaldehyde (PFA), and permeabilized with 0.5% Triton X-100 in PBS. After blocking with 5% BSA in PBS, cells were incubated overnight with anti-M1 antibody (HB64, in-house generated) diluted 1:100 in blocking buffer, followed by incubation with the appropriate fluorescent secondary antibody. Images were acquired using a Yokogawa CQ1 Benchtop High-Content Analysis System.

GST pulldown between UBiP10 and IAV-infected or poly(I:C)-treated A549 cells

A total of 60 μg of purified GST and GST-UBiP10 was preincubated with GST_trap beads for 30 min and then incubated with poly(I:C) (5 μg/ml; 4 hours)–treated or IAV-infected (strain X31, MOI = 50, 6 hours) A549 cell lysate (from two 10-cm dishes with 80% confluency) for 1 hour at cold room (4°C). The lysis buffer was supplied with 50 μM PR-619 to stop the degradation of unanchored poly-Ub chains. Enriched Ub signal was blotted with a specific antibody against K63-linked Ub chains (CST, #5621). Pull-down experiment was done as previously described in the “GST pulldown” section.

ELISA for visualizing the variation of unanchored Ub amount

Purified 3 × Ub–RARA and mono-Ub were prepared at 100 nM in PBS and combined in the following mixtures: (i) 60 μl of 3 × Ub–RARA, (ii) 40 μl of 3 × Ub–RARA + 20 μl of mono-Ub (2:1), (iii) 30 μl of 3 × Ub–RARA + 30 μl of mono-Ub (1:1), and (iv) 60 μl of mono-Ub. For blank control, only mono-Ub was used. Each mixture was immobilized on a Nunc MicroWell 96-Well Microplate (Thermo Fisher Scientific, #243656) and incubated overnight at 4°C. The next day, wells were washed six times with PBS and blocked with PBS-TB (PBS with 0.1% Tween-20 and 3% BSA) for 2.5 hours at room temperature, followed by five washes with PBS containing 0.1% Tween-20 (PBS-T) and another five with PBS. Wells were then incubated with 50 μl of 150 nM GST-UBiP05, GST-UBiP10, or GST (blank control) in PBS-TB for 1.5 hours at room temperature, followed by five PBS-T and five PBS washes. Subsequently, 60 μl of anti-GST antibody (CST, #2625; diluted 1:5000 in PBS-TB) was added to each well and incubated for 30 min at room temperature, followed by three washes with PBS-T and three with PBS. Detection was performed by adding 60 μl of TMB substrate (Thermo Fisher Scientific, #34028), and the reaction was stopped after 1 to 2 min, upon color development, by adding 25 μl of 1 M H₂SO₄. Absorbance was recorded at 450/620 nm using a microplate reader SpectraMax Plus (Molecular Devices).

Immunofluorescence-based plate assay and linear response analysis of UBiP05 and UBiP10

Purified mono-Ub was diluted in PBS to a starting concentration of 10 μg/ml and subjected to a series of six twofold serial dilutions, as indicated in fig. S5E. Plates were blocked with BSA overnight at 4°C before incubation. Subsequent incubation and washing steps were performed as described for the ELISA protocol above. GST (blank control and background subtraction), GST-tagged UBiP05, or UBiP10 probes were applied under identical conditions, followed by incubation with anti-GST primary antibody for 6 hours at room temperature. Fluorescent detection was achieved using Alexa Fluor 488–conjugated secondary antibodies (2 hours of incubation), replacing the horseradish peroxidase–based detection used in ELISA. After final washes, fluorescence signals were immediately measured using a Tecan Spark plate reader with appropriate excitation/emission settings for Alexa Fluor 488. For calibration experiments, mixtures of anchored Ub (3 × Ub–RARA) and mono-Ub were prepared at defined ratios (molar ratio of 1:1) while maintaining constant total protein concentration, as described for the ELISA setup. These samples were processed in parallel using the same staining and detection conditions.

Immunofluorescence of unanchored Ub inside IAV virions

Commercial X31 particles (Virapur; 1 × 10⁷ virions were used for each chamber) were prebound to chamber slide for 30 min at 4°C, washed with infection medium, and incubated with priming buffer [120 mM KCl and 20 mM NaCl (pH 5.8)] for 30 min at 37°C to increase the accessibility of inner core antigens. After fixation in 4% formaldehyde in PBS for 15 min, the particles were washed in PBS and permeabilized for 5 min in 0.5% Triton X-100/PBS. After blocking in blocking buffer [PBS, 10% goat serum, (Merck, #S26-100ML), and 0.1% Triton X-100] for 30 min, the particles were incubated with GST and GST-UBiP10 proteins (25 μg per well) for 2 hours at room temperature. Then, the chamber slide was washed five times with 0.1% Triton X-100/PBS. After that the chamber was incubated with anti-GST mouse antibody and anti-HA rabbit mAb overnight at 4°C. After washing five times in 0.1% Triton X-100/PBS, they were stained with anti-rabbit Alexa Flour 568 and anti-mouse Alexa Fluor 488 secondary antibodies for 1 hour at room temperature. The chamber slide was then mounted with ProLong Diamond Antifade Mountant (Thermo Fisher Scientific, #P36961) and covered with a coverslip.

Immunofluorescence and microscopy

A total of 0.05 million cells was seeded in a four-well chamber slide (Thermo Fisher Scientific, #154526) or eight-well chamber slide (Thermo Fisher Scientific, #154534; only for Fig. 4D). For SH-SY5Y cells, cells were incubated with ATRA (10 μM) for 3 to 5 days and then seeded into chamber slide. Following overnight culture, cells were treated or infected as depicted in the figures (Figs. 4 and 5 and figs. S5 to S7). Subsequently, the medium was aspirated, and cells were washed twice with ice-cold PBS. Cells were then fixed with 4% PFA for 10 min, permeabilized with 0.5% Triton X-100/PBS for 15 min, and blocked with blocking buffer (5% goat serum in 0.1% Triton X-100/PBS) for 1 hour. For GST-UBiPs staining (fig. S5A), cells were first incubated with 5 to 10 μg of GST-UBiPs or GST proteins for 6 to 12 hours at room temperature for each well on the chamber slide, followed by six washes with 0.1% Triton X-100/PBS on shaker. One wash lasted for 10 min. Then, primary antibodies (e.g., anti-GST antibodies) were applied to cells at a 1:150 dilution and incubated overnight at 4°C. The next day, cells were washed five times with 0.1% Triton X-100/PBS and incubated with fluorescent secondary antibodies in PBS or blocking buffer (1:250 dilution) for 1 hour at room temperature. DAPI (Thermo Fisher Scientific, #D1306) and Acti-stain 488 phalloidin (Cytoskeleton, #PHDG1) diluted in PBS were done according to the product instructions and added to cells for 20 min at room temperature. Following three washes with PBS, cells were sealed with ProLong Diamond Antifade Mountant (Thermo Fisher Scientific, #P36965). Slides were inspected with a spinning disk confocal scanning microscope (Yokogawa CSU-W1 with Dual Camera T2 and Nikon CSU-W1 SoRa). Signal intensity measurements were done with the ImageJ software.

Quantification of unanchored Ub intensity around aggresome

Image analysis in Fig. 5 (C and G) and fig. S7C was performed using the ImageJ software (version 1.54p) to segment and define regions of interest within cells. Initially, the Threshold function was applied to the Ub Ab channel to segment aggresomes, generating Mask 1 to represent aggresome sites. To define the periaggresomal region, Mask 1 was expanded using the Dilate function to create Mask 2, and then Mask 1 was subtracted from Mask 2 using the Image Calculator, resulting in Mask 3, which delineates the surrounding periaggresomal area. In parallel, cell boundaries were identified in the UBiP channel using the Threshold function to produce Mask 4. To define the remote membrane-proximal region, Mask 4 was eroded using the Erode function to generate Mask 5, and Mask 5 was then subtracted from Mask 4 using the Image Calculator, yielding Mask 6, which represents the area adjacent to the cell membrane.

Myosin 10 siRNA knockdown and A549 HDAC6 KO cell line establishment

For siRNA knockdown, 0.05 × 10⁶ A549 cells were seeded 1 day before transfection on each well of the chamber slide (Thermo Fisher Scientific, #154526), and a siRNA mixture was prepared as follows: A total of 6 μl of 10 nM siRNA was incubated with 15 μl of RNAiMAX reagent in 200 μl of DMEM medium (without FBS) at room temperature for 15 min. Then, 50 μl of mixture was added to each well of the chamber slide. Scramble siRNA is the AllStars Negative Control siRNA (GeneGlobe ID: SI03650318). The siRNA sequence for knocking down myosin 10 is UUCUGUACAAGUUUGGGUCCAAUUC.

For CRISPR-Cas9 HDAC6 KO A549 cells, the single guide RNA (sequence: GCCGGTTGAGGTCATAGTTG) was constructed in the plasmid px459 (Plasmid #62988), and the resulting construct was transfected into A549 cells by electroporation (4D-Nucleofector X Unit from Lonza, SF cell line kit, program: CM-130). The procedures exactly followed the published protocol (35). HDAC6 KO cells were selected with puromycin (2 μg/ml) or 3 days. HDAC6 KO was validated by immunoblotting and polymerase chain reaction (PCR) sequencing. Single cells were sorted by fluorescence-activated cell sorting and stored in liquid nitrogen.

BindingPCA and AbundancePCA

Media and buffer recipes for mutagenesis screening with yeast

LB medium: Luria Broth Base (25 g/liter; Invitrogen, Waltham, MA, USA). Autoclaved 20 min at 121°C.

LB-agar with 2× ampicillin (100 μg/ml): Luria Broth Base (25 g/liter; Invitrogen, Waltham, MA, USA), agar (7.5 g/liter), and MgSO₄ · H₂O (1.2 g/liter). Autoclaved 20 min at 121°C. Cool down to 45°C.Addition of ampicillin (100 mg/liter).

YPAD: bacto-peptone (20 g/liter), dextrose (20 g/liter), yeast extract (10 g/liter), and adenine (25 mg/liter). Filter sterilized (Millipore Express PLUS 0.22-μm PES, Merck, Darmstadt, Germany).

SC-ura: Yeast nitrogen base without amino acids (6.7 g/liter), glucose (20 g/liter), and complete supplement mixture dropout without uracil (0.77 g/liter). Filter sterilized (Millipore Express PLUS 0.22-μm PES, Merck, Darmstadt, Germany).

SC-ura/ade/met: Yeast nitrogen base without amino acids and folic acid (6.7 g/liter), glucose (20 g/liter), complete supplement mixture dropout without uracil (0.74 g/liter), and adenine and methionine. Filter sterilized (Millipore Express PLUS 0.22-μm PES, Merck, Darmstadt, Germany).

SORB: 1 M sorbitol, 100 mM LiOAc, 10 mM tris-HCl (pH 8.0), and 1 mM EDTA (pH 8.0). Filter sterilized (Millipore Express PLUS 0.22-μm PES, Merck, Darmstadt, Germany).

Plate mixture: 40% polyethylene glycol, molecular weight 3350; 100 mM LiOAc; 10 mM tris-HCL (pH 8.0); and 1 mM EDTA (pH 8.0). Filter sterilized (Millipore Express PLUS 0.22-μm PES, Merck, Darmstadt, Germany).

Recovery medium: YPAD + 0.5 M sorbitol. Filter sterilized (Millipore Express PLUS 0.22-μm PES, Merck, Darmstadt, Germany).

Competition medium: SC-ura/ade/met + methotrexate (200 μg/ml; BioShop Canada Inc., Canada) and 2% dimethyl sulfoxide (DMSO).

DTT buffer: 0.1 M EDTA-KOH (pH 7.5) and 10 mM DTT.

Zymolyase buffer: 20 mM K-phoshpate (pH 7.2), 1.2 M sorbitol, Zymolyase 20T (0.4 mg/ml; Amsbio, US Biological), and ribonuclease A (100 μg/ml).

ZnF library construction

A comprehensive mutational library was designed for the mHDAC6_ZnF domain_(1007–1149), incorporating every possible single amino acid substitution, including stop codons, resulting in 2860 variants. Because of the domain’s length of 143 amino acids, it was split into two parts for oligonucleotide synthesis. Two half-libraries covering the N- and C-terminal halves of the ZnF domain were created and ordered separately as 300–base pair (bp) oligonucleotide pools from Twist Bioscience (San Francisco, CA, USA).

Two cloning backbones, adapted from the BindingPCA backbone (36), were constructed to accommodate the two half-libraries. These backbones include a barcode landing pad for unique variant identification, flanked by SP1 and SP2 binding sites (Illumina, San Diego, CA) to enable easy PCR amplification of Illumina sequencing libraries. The barcode landing pad is followed by the BindingPCA prey cassette, comprising a CYC promoter, a triple Flag tag, the C-terminal half of the DHFR open reading frame (referred to as the FR fragment), a short linker (L4), and NheI and HindIII restriction sites for inserting the prey protein library in-frame. Downstream of the prey cassette is a bidirectional CYC terminator, followed by the DH cassette in the opposite direction. The DH cassette includes a CYC promoter, a triple Myc tag, the N-terminal half of the DHFR open reading frame (the DH fragment), a short linker (L3), and the Ub open reading frame, flanked by SpeI and BamHI restriction sites.

For the first sublibrary encoding the N-terminal half of the HDAC6 ZnF (amino acids 1 to 72), a gene fragment from Twist Bioscience was cloned into the backbone using KpnI-HF and HindIII-HF (NEB, Ipswitch, MA) following standard ligation protocols. The gene fragment contained a KpnI restriction site, an 18-bp placeholder barcode sequence, a BsaI site, the C-terminal ZnF wild-type sequence (amino acids 73 to 143), and a HindIII restriction site. Similarly, the second sublibrary encoding the C-terminal half of the ZnF domain (amino acids 73 to 143) was cloned into the backbone using NheI-HF and HindIII-HF. The gene fragment included an NheI restriction site, the N-terminal ZnF wild-type sequence (amino acids 1 to 72), and a HindIII restriction site. Sanger sequencing confirmed the accuracy of both plasmids, named pDL00473 and pDL00474 (plasmid sequence available on request), respectively.

The two mutagenized half-libraries were then amplified using PCR with primers oDL00763 and oDL01232. Forward primers contained 18-bp NNN-overhangs to introduce random barcodes, ensuring unique identification of each variant. PCR was carried out with Q5 polymerase, using 0.25 μM of each primer and 10 ng of oligonucleotide template in a 50-μl reaction volume. After 15 cycles of amplification, the products were gel purified using the QIAquick Gel Extraction Kit (QIAGEN, Hilden, Germany).

The amplified oligonucleotides from library 1 and pDL00473 were digested with KpnI-HF and BsaI-HFv2 and ligated using T4 ligase, while library 2 and pDL00474 were digested with NheI-HF and HindIII-HF and ligated similarly. Both ligation reactions were incubated overnight following a temperature-cycle ligation protocol to maximize efficiency. The resulting products were dialyzed to reduce salt concentration and transformed into NEB 10-beta Electrocompetent E. coli (NEB). After electroporation, the cells were recovered and plated on LB + 2× Amp plates to determine transformation efficiency. Approximately 46,000 and 43,000 transformants were harvested for library 1 and library 2, respectively, ensuring a barcode coverage of ∼30 barcodes per variant.

For library 2, the wild-type C-terminal ZnF sequence was added via digesting pDL00474 with NheI-HF and BsaI-HFv2, followed by cloning the 217-bp fragment into the digested library 2. This step was performed using high-efficiency ligation, and approximately 37 million transformants were harvested and purified with the NucleoBond PC 500 kit (Macherey-Nagel, Düren, Germany).

To link the randomly generated barcodes to their corresponding ZnF variants, an amplicon library was generated. A two-step PCR process was used to add Illumina adapter sequences. In the first round, the SP2 primer binding site was added using 1 ng of plasmid library template, amplifying 487-bp fragments containing the barcodes and ZnF variant sequences with the KAPA HiFi HotStart DNA Polymerase Kit (Roche Sequencing Solutions Inc.). After ExoSAP-IT treatment and purification, the second PCR round incorporated the Illumina adapter using NEBNext Multiplex Oligos for Illumina.

The final PCR products were pooled and gel purified, and their concentrations were determined by a Qubit 4 fluorometer (Invitrogen, #Q33238). The two libraries were then pooled at equimolar ratios and further purified using AMPure XP beads (Beckman Coulter, Brea, CA). Quality control was performed via Bioanalyzer (Agilent Technologies, Santa Clara, CA) and quantitative PCR (qPCR) using KAPA Library Quantification Standards. Last, the library was sequenced on an Illumina MiSeq System (600 cycle paired-end run), with 25% PhiX phage library spiked in to manage the low complexity of the sequence pool.

Barcodes and variants were associated as described in (37). In the end, a total of 41,716 barcodes that could be matched to an expected variant sequence was recovered.

To finalize the library, the missing part of the backbone containing the downstream region of the barcode landing pad, the CYC promoter for the prey cassette, and the FR fragment with 3× Flag tag and linker had to be cloned into the plasmid libraries using SalI-HF and NheI-HF (NEB, Ipswitch, MA). High-efficiency cloning with electroporation was performed as described above. For library 1, 12 million transformants were harvested, and for library 2, 37 million transformants were harvested. Both were purified using a NucleoBond PC 500 kit (Macherey-Nagel, Düren, Germany) and then pooled at equimolar ratios to create the final full library.

To measure the abundance of the individual ZnF variants, the bait cassette of the entire library was replaced by an alternative cassette expressing the DH tag only with 3× Myc tag and the linker still attached under the control of a strong yeast promoter, GPDp. The abundance bait cassette was cloned in with SpeI-HF and BsmI, and a high-efficiency cloning with electroporation was performed as described above. A total of 232 million transformants was harvested and purified using a NucleoBond PC 500 kit (Macherey-Nagel, Düren, Germany).

Yeast large-scale transformation

All transformations and screens were performed in six replicates. For each replicate, 5 ml of YPAD was inoculated with a single colony of BY4742 and grown to saturation overnight. In the morning, a preculture of 175 ml of YPAD was inoculated at optical density at 600 nm (OD₆₀₀) 0.3 and grown for ∼4 hours until an OD₆₀₀ of 1.2 to 1.4 was reached. Cells were then harvested for 5 min at ×3000g, washed once with 50 ml of H₂O and once with 50 ml of SORB, and lastly resuspended in 7 ml of SORB and incubated for 30 min at room temperature on a rotating wheel. Preboiled salmon sperm DNA (175 μl; Agilent Technologies, Santa Clara, CA), 1 μg of plasmid library, and 35 ml of plate mixture were added to the cells that were then incubated for another 30 min at room temperature on a rotating wheel. After the addition of 3.5 ml of DMSO, the cells were heat shocked at 42°C for 20 min with occasional inversion of the tube. They were then centrifuged at 3000 rpm for 5 min, the supernatant was removed, and the cells were resuspended in 50 ml of recovery medium and incubated at 30°C for an hour. Afterward, they were spun down again, the recovery medium was removed, and all cells were used to inoculate 350 ml of SC-ura medium. They were incubated at 30°C and 200 rpm for 49 hours until they reached saturation.

ddPCA screen

For a second round of transformant selection, 200 ml of SC-ura-ade-met was inoculated at an OD₆₀₀ of 0.1 from the saturated SC-ura culture and grown at 30°C and 200 rpm for ∼12 hours. As soon as the cells reached an OD₆₀₀ of 1.2, they were used to inoculate 200 ml of competition medium at OD₆₀₀ of 0.05 and put back at 30°C. The remaining culture was harvested as input sample. When the competition culture reached an OD₆₀₀ of 1.6, the cells were harvested as output culture. Both input and output cultures were washed twice with H₂O and then frozen at −20°C until further use.

Yeast plasmid extraction

The harvested yeast cells were thawed and then spheroplasted by first resuspending each sample in 4 ml of DTT buffer and incubating at 30°C for 15 min at 200 rpm, spinning them at ×2500g for 5 min, and then resuspending them in 4 ml of Zymolyase buffer and incubating them for ∼2 hours. The cells were harvested again at ×2500g for 5 min and resuspended in 1.6 ml of homemade buffer P1. Next, 1.6 ml of homemade buffer P2 was added, the tube was inverted several times and incubated at room temperature for 10 min. Subsequently, 1.6 ml of commercial buffer N3 (QIAGEN, Düren, Germany) was added, and the tubes were inverted again. The cells were then transferred to 5-ml reaction tubes (Eppendorf, Hamburg, Germany) and spun at max. Speed in a tabletop centrifuge for 20 min. The supernatant was recovered and purified over a Miniprep column (QIAGEN, Düren, Germany) according to the manufacturer’s instructions but washing the column twice with buffer PE. Plasmid enrichment in the eluates was determined by qPCR.

Sequencing library preparation

Because the backbone contained SP1 and SP2 primer binding sites (Illumina, San Diego, CA) flanking the barcode landing pad, a single PCR was sufficient to generate the sequencing library. For each sample, 200 million plasmid molecules were used as the template in a PCR, which was performed with Q5 polymerase (NEB, Ipswitch, MA) following the manufacturer’s protocol. NEBNext Multiplex Oligos for Illumina (NEB) were used for amplification. The PCR was run for 18 cycles with an annealing temperature of 63°C. The resulting 421-bp amplicons were purified using a 1% agarose gel and the QIAquick Gel Extraction Kit (QIAGEN, Düren, Germany). Their concentrations were measured using a Qubit 4 fluorometer (Invitrogen, #Q33238), and the samples were pooled at equimolar ratios. The pooled library was further purified using AMPure XP beads (Beckman Coulter, Brea, CA, USA) at a 1:1 bead-to-sample ratio, following the manufacturer’s instructions. Quality control was performed with a Bioanalyzer (Agilent Technologies, Santa Clara, CA), which confirmed the presence of single peaks at the expected size. In addition, a qPCR was conducted using the KAPA Library Quantification Standards with Primer (Roche Sequencing Solutions Inc., Pleasanton, CA) for further validation. The final library was sequenced on an Illumina NovaSeq Sequencing System (Illumina, San Diego, CA) with an SP flow cell. To address the low complexity of the library, 25% PhiX phage library was spiked in.

Data analysis

Read counts were preprocessed, and log₂FC was calculated with mutscan (38). The data were visualized using R.

Oligonucleotides

Primer: oDL00763: CCGTACGTTGCAGTCAG; oDL01232: TTCAGGAGCATCTGCTAGC; oDL01110: ACACTCTTTCCCTACACGACGCTCTTCCGATCT; oDL00833: TTCAGACGTGTGCTCTTCCGATCTAGCGTGACATAACTAATAAGC.

UBiP design strategies

Partial starting model generation

The six starting models for RFdiffusion were generated as follows: For UBiP01 to UBiP03 (starting model 1), the Protein Data Bank (PDB) structure of UBA_UQ1 (PDB ID: 2JY6) was merged with the AlphaFold2-predicted mHDAC6_(1026–1149)–Ub complex. For UBiP04 (starting model 2), the UIM domain from Vps27 (PDB ID: 1Q0W) and UBA_UQ1 (PDB ID: 2JY6) were merged with the AlphaFold2-predicted mHDAC6_(1026–1149)–Ub complex. For UBiP05 to UBiP06 (starting model 3), the RUZ domain from PDB structure 2FIF was merged with the AlphaFold2-predicted mHDAC6_(1026–1149)–Ub complex. For UBiP07 to UBiP09 (starting model 4), the RUZ domain from PDB structure 2FIF was merged with the AlphaFold2-predicted mHDAC6_(1007–1149)–Ub complex. In the case of UBiP09, which was designed by Colab, the C-terminal RUZ domain was removed. For UBiP10 (starting model 5), the PDB structure of RAP80 (PDB ID: 3A1Q) was merged with the AlphaFold2-predicted mHDAC6_(1007–1149)–Ub complex. For UBiP11 (starting model 6), the PDB structure of RPN13 (PDB ID: 5YMY) was merged with the AlphaFold2-predicted mHDAC6_(1007–1149)–Ub complex.

Linker backbone design

Although a local pipeline (RF-diffusion/MPNN/AlphaFold2) was established at FMI, we also tried some UBiP designs with the online Colab from Google (https://colab.research.google.com/github/sokrypton/ColabDesign/blob/main/rf/examples/diffusion.ipynb). The local pipeline generated 1000 candidates for each design, and Colab generated several additional candidates (fig. S2)

To connect the HDAC6 ZnF domain to the various Ub-binding domains, linkers/backbones were designed using RFdiffusion. In the RFdiffusion inference script’s contigmap, residues in the input models were fixed, and variable linker lengths were defined:

1) UBiP01 to UBiP03: A linker of 3 to 10 residues between mZnF and UBA_UQ1 was generated. UBiP01 and UBiP02 were generated using a local pipeline, while UBiP03 was generated using Colab.

2) UBiP04: The model was manually generated.

3) UBiP05 and UBiP06: A linker of 10 to 50 residues was generated, with specific Ub (chain A) hotspot residues defined at positions (A6, A16, A18, A25, A29, and A63).

4) UBiP07 and UBiP08: The same contig as UBiP05 and UBiP06 was used, but the N-terminal flexible region was extended by 20 amino acids.

5) UBiP09: A linker of 10 to 50 residues was generated at the C terminus of mZnF_(1007–1149) before RUZ without specifying hotspots, and the RUZ domain was removed.

6) UBiP10: A 5- to 30-residue linker was generated between mZnF_(1007–1149) and the UIM domain of RAP80 without defining hotspots.

7) UBiP11: A 5- to 30-residue linker was generated between mZnF_(1007–1149) and the RPN13 domain without defining hotspots.

For UBiP04 to UBiP11, the initial 10 to 20 amino acids at N terminus were deleted for final construct synthesis to increase the structure stability by reducing the unstructured region while still including the key Ub-binding region (1037 to 1044). UBiP1 to UBiP11 were selected on the basis of the lowest RMSD values within each design round. We used RFdiffusion to generate 5025 different models, and after sequence generation with ProteinMPNN, we obtained 10,081 candidates. Details of the generation of each UBiP are presented in fig. S2.

Sequence sampling and designability test

ProteinMPNN was used to predict sequences for the designed linkers. A custom wrapper script was used to automatically generate inputs and run the “designability_test.py” script from the ColabDesign GitHub repository (https://github.com/sokrypton/ColabDesign/blob/d193e44f77a8f382cdd8fdb49c565c17e61fa13a/colabdesign/rf/designability_test.py) for all RFdiffusion backbone designs. The Colab design pipeline runs ProteinMPNN, followed by structure prediction with Alphafold2 using the input model as template. Contigs were specified in which all residues except for the designed linker segments were masked. For local pipeline, two sequences were generated for each RFdiffusion input model, and Alphafold2 predictions were run with “multimer,” “num_recycles 3,” and “initial_guess” flags. For each linker backbone design experiment, designability’ results were sorted by “RMSD” (with respect to the input model). The ProteinMPNN score is the average over residues that were designed negative log probability of sampled amino acids. For Colab, we followed the instructions at website: https://colab.research.google.com/github/sokrypton/ColabDesign/blob/v1.1.1/rf/examples/diffusion.ipynb.

Statistical analysis

Data were plotted using Prism v.10.2.3 (GraphPad) and presented as the means of multiple experiments, with error bars showing the ±SD. The significance of results was determined using ordinary one-way analysis of variance (ANOVA), as indicated in figure legends.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Tables S1 and S2

Uncropped Western blots