SPIN4 is a principal endogenous substrate of the E3 ubiquitin ligase DCAF16

Abstract

DCAF16 is a substrate recognition component of Cullin-RING E3 ubiquitin ligases that can be targeted by electrophilic PROTACs (PROteolysis Targeting Chimeras) to promote the nuclear-restricted degradation of proteins. The endogenous protein substates of DCAF16 remain unknown. In this study, we compared the protein content of DCAF16-wild type (WT) and DCAF16-knockout (KO) cells by untargeted mass spectrometry (MS)-based proteomics, identifying the Tudor domain-containing protein Spindlin-4 (SPIN4) as a protein that was substantially elevated in cells lacking DCAF16. Very few other proteomic changes were found in DCAF16-KO cells, pointing to a specific relationship between DCAF16 and SPIN4. Consistent with this hypothesis, we found that DCAF16 interacts with and ubiquitinates SPIN4, but not other related SPIN proteins, and identified a conserved lysine residue unique to SPIN4 that is involved in DCAF16 binding. Finally, we provide evidence that SPIN4 preferentially binds trimethylated histone H3K4 over other modified histone modifications. These results taken together, indicate that DCAF16 and SPIN4 form a dedicated E3 ligase-substrate complex that regulates the turnover and presumed functions of SPIN4 in human cells.

The ubiquitin-proteasome system (UPS) controls protein homeostasis by promoting protein ubiquitination and degradation. The covalent attachment of ubiquitin, a 76-amino acid protein, to lysine residues on another protein determines if that protein will be directed to the proteasome for destruction^{1, 2}. An enzyme cascade involving E1, E2, and E3 enzymes catalyzes the ubiquitination process³. E3 ligases mediate the final step of ubiquitin transfer and confer specificity by recognizing protein substrates. More than 600 E3 ligases are predicted to be expressed by human cells, with the RING E3 ligases constituting the largest subfamily⁴.

Multiple pharmacological approaches have been introduced to target the UPS, including direct inhibitors of the catalytic subunits of the proteasome, which have been approved as cancer therapies⁵, and small molecules that form ternary complexes with E3 ligases and protein substrates to promote the specific degradation of these substrates⁶. The latter category of chemical probes can be further divided into molecular glue compounds, such as the immunomodulatory IMiD drugs, which possess a monofunctional unit binding to both an E3 ligase and substrate protein⁷, and PROTACs (PROteolysis Targeting Chimeras), which are bifunctional compounds possessing separate units for simultaneously binding the E3 ligase and protein substrate⁸. We recently identified DCAF16, a poorly characterized Cullin-RING E3 ubiquitin ligase (CRL), as a target of electrophilic PROTACs that promote nuclear-restricted protein degradation⁹. We found that DCAF16 itself is localized to the nucleus⁹, offering a mechanistic explanation for the effects of the electrophilic PROTACs on nuclear, but not cytoplasmic protein degradation.

One question that emerges when developing and applying molecular glue or PROTAC compounds is the extent to which these agents may affect endogenous substrate-E3 ligase interactions in cells. We found that electrophilic PROTACs targeting DCAF16 were capable of promoting substrate degradation at low engagement stoichiometries with DCAF16 (10-40%), suggesting that these compounds, at pharmacologically relevant concentrations, may spare a substantial fraction of DCAF16 to perform physiological functions. Determining the endogenous protein substrates of DCAF16 would enrich our understanding of the functions of this E3 ligase and also point to pathways that may be perturbed by small molecules that more completely engage DCAF16. Here, we use untargeted mass spectrometry-based proteomics to compare the protein content of DCAF16-wild type (WT) and DCAF16-knockout (KO) human cells. Among > 5500 quantified proteins, we identified a single protein Spindlin-4 (SPIN4) that was substantially elevated in DCAF16-KO cells. We demonstrate that DCAF16 interacts with and ubiquitinates SPIN4, but not other related spindlin proteins. The functions of SPIN4 remain poorly understood, but the protein is predicted to bind methylated histones, and, consistent with this hypothesis, we found that SPIN4 interacts with trimethylated lysine-3 of histone H3 (H3K4me3) preferentially over a panel of other post-translational histone modifications.

We genetically disrupted DCAF16 in multiple human cancer cell lines (MDA-MB-231, 22Rv1) using CRISPR/Cas9 methods¹⁰. To minimize the potential for clonal cell line artifacts, we generated populations of DCAF16-knockout (KO) cells and confirmed >95% DCAF16 disruption by genomic sequencing and mass spectrometry (MS)-based proteomics (Figure 1a, Figure S1a,b, and Figure S2a). We also verified that electrophilic PROTACs previously shown to degrade nuclear FKBP12⁹ in a DCAF16-dependent manner lost activity in DCAF16-KO cells (Figure S1c). We hypothesized that endogenous substrates for DCAF16 might show elevated expression in DCAF16-KO cells due to impairments in the degradation of these proteins. Toward this end, we compared DCAF16-WT and DCAF16-KO cells by untargeted, multiplexed (tandem mass tagging, or TMT) proteomics. We quantified > 5500 proteins each in MDA-MB-231 and 22Rv1 cells, and, in both cell lines, only Spindlin-4 (SPIN4) showed > 1.5-fold increases in DCAF16-KO cells (Figure 1a, Figure S2a, and Table S1).

Figure 1. Discovery of SPIN4 as an endogenous DCAF16 substrate.

a. Quantitative MS-based proteomics of DCAF16-WT and DCAF16-KO MDA-MB-231 cells. The y-axis and x-axis correspond to the average relative protein abundances (DCAF16-KO/DCAF16-WT) and coefficients of variation, respectively, from two experiments (n = 2 biologically independent experiments). b. Western blot of SPIN4 in DCAF16-WT and DCAF16-KO MDA-MB-231 cells. The result is representative of three biologically independent experiments. Bar graph (right) represents quantification of the SPIN4 protein content. Data are mean ± SEM (n = 3). Statistical significance was calculated with unpaired two-tailed Student’s t-tests comparing WT to KO cells. ****P < 0.0001. c. qPCR analysis of SPIN4 mRNA content in DCAF16-WT and DCAF16-KO MDA-MB-231 cells. The result is representative of six biologically independent experiments. Data are mean ± SEM (n = 6). d. Western blot of SPIN4 and NRF2 in DCAF16-WT and DCAF16-KO MDA-MB-231 cells treated with MLN4924 (1 μM, 8 h) or bortezomib (BTZ, 1 μM, 8 h). The result is representative of three biologically independent experiments. Bar graph (right) represents quantification of the SPIN4 protein content. Data are mean ± SEM (n = 3). Statistical significance was calculated with unpaired two-tailed Student’s t-tests comparing DCAF16-WT to DCAF16-KO, DCAF16-WT+MLN4924 or DCAF16-WT+ bortezomib cells. **P < 0.01.

We confirmed the increase in SPIN4 protein in DCAF16-KO cells by Western blotting (Figure 1b and Figure S2b), while quantitative polymerase chain reaction (qPCR) analysis showed that SPIN4 mRNA was unchanged (Figure 1c), indicating that a posttranscriptional mechanism was responsible for changes in SPIN4 protein in DCAF16-KO cells. We further found that both the proteasome inhibitor bortezomib and neddylation inhibitor MLN4924 increased the abundance of SPIN4 in DCAF16-WT cells to level that matched the quantity of this protein in DCAF16-KO cells (Figure 1d and Figure S2c). Bortezomib and MLN4924 did not affect SPIN4 abundance in DCAF16-KO cells (Figure 1d and Figure S2c). Since MLN4924 inhibits E3 ligases of the CRL class¹¹ that includes DCAF16, our findings suggested that SPIN4 may represent an endogenous substrate of DCAF16.

A search of large-scale protein-protein interaction databases revealed that SPIN4 and DCAF16 have been found to associate with one another in affinity purification (AP)-MS proteomic experiments^{12, 13}. We therefore next sought to confirm that DCAF16 physically interacted with SPIN4 in cells. We generated HEK293T cells stably expressing FLAG epitope-tagged SPIN4 (Figure S3) and immunoprecipitated SPIN4 and associated proteins from these cells using anti-FLAG antibodies. To differentiate interacting proteins involved in proteasome-mediated degradation of SPIN4 from other constitutively interacting proteins, we treated cells with the proteasome inhibitor MG132, which has been shown to stabilize interactions between E3 ligases and client proteins¹⁴. Quantitative MS-based proteomic analysis comparing immunoprecipitated proteins from FLAG-SPIN4-expressing HEK293T cells treated with DMSO or MG132 identified DCAF16 and DDB1 as the only proteins showing substantial enrichment values (> 1.5-fold) in MG132-treated cells (Figure 2a and Table S2). DDB1 is an additional CRL component and serves as an adaptor between DCAF16 and Cullin 4A/B¹⁵. We did not detect Cullin 4A/B in our MS-based proteomic experiments. We also compared the SPIN4 interaction profile in FLAG-SPIN4-transduced MDA-MB-231 WT and DCAF16-KO cells, which revealed that DDB1 enrichment was dramatically decreased in DCAF16-KO cells (Figure 2b and Table S3). This finding suggests that the interaction between SPIN4 and DDB1 is mediated through DCAF16, as would be expected given the predicted role of DCAF16 as a substrate recognition component of CRLs. We next co-transfected FLAG-SPIN4 and HA-DCAF16 in HEK293T cells and found that HA-DCAF16, but not another control DCAF protein HA-DCAF2, interacted with and induced poly-ubiquitination of FLAG-SPIN4 (Figure 2c). Additionally, FLAG-SPIN4 showed substantially higher poly-ubiquitination in DCAF16-WT compared to DCAF16-KO cells (Figure 2d). We further found that the non-ubiquitinated FLAG-SPIN4 co-immunoprecipitated with HA-DCAF16 in HEK293T cells transiently expressing FLAG-SPIN4 and HA-DCAF16 (Figure S4), suggesting that ubiquitination of SPIN4 is not necessary for associating with DCAF16. Taken together, these results point to a specific interaction between SPIN4 and DCAF16 that leads to the DCAF16-dependent ubiquitination and degradation of SPIN4.

Figure 2. DCAF16 mediates poly-ubiquitination of SPIN4.

a. Quantitative MS-based proteomics showing MG132/DMSO ratio values of proteins identified in anti-FLAG affinity enrichment experiments from HEK293T cells stably expressing FLAG-SPIN4 treated with MG132 (10 μM, 2 h) or DMSO (2 h). The y-axis and x-axis correspond to the average relative log2 protein abundances (MG132/DMSO) and coefficients of variation, respectively, from two biologically independent experiments. b. Quantitative MS-based proteomics showing DCAF16-WT/DCAF16-KO ratio values of proteins identified in anti-FLAG affinity enrichment experiments from MDA-MB-231 WT and DCAF16 KO cells stably expressing FLAG-SPIN4 treated with MG132 (10 μM, 2 h). The y-axis and x-axis correspond to the average relative log2 protein abundances (DCAF16-WT/DCAF16-KO) and coefficients of variation, respectively, from two biologically independent experiments. c. DCAF16 interacted with and mediated polyubiquitination of SPIN4 in HEK293T cells. HEK293T cells were transiently transfected with FLAG-SPIN4 and HA-DCAF16 or HA-DCAF2 for 24 h and then treated with MG132 (10 μM, 2 h). The result is representative of three biologically independent experiments. Bar graph (right) represents quantification of the total polyubiquitination content after anti-FLAG affinity enrichment. Data are mean ± SEM (n = 3). Statistical significance was calculated with unpaired two-tailed Student’s t-tests comparing DCAF16 and SPIN4 co-transfected group to others. **P < 0.01. d. FLAG-SPIN4 shows greater polyubiquitination in DCAF16-WT MDA-MB-231 cells. DCAF16-WT and DCAF16-KO MDA-MB-231 cells stably expressing FLAG-SPIN4 were treated with MG132 (10 μM, 2 h). The result is representative of two biologically independent experiments.

The SPIN family of proteins has five sequence-related members (SPIN1, 2A, 2B, 3, and 4) that all harbor three Tudor domains¹⁶. We wondered if other SPIN proteins might also interact with and serve as substrates for DCAF16. In our original proteomic experiments, SPIN1 signals were unchanged in DCAF16-WT and DCAF16-KO cells (KO/WT ratio value of protein abundance = 0.97; Table S1), which suggested that SPIN1 may not interact with DCAF16. Consistent with this conclusion, we compared interacting proteins for SPIN1 and SPIN4 using immunoprecipitation-MS-based proteomic experiments and found that DCAF16 and DDB1 interacted with SPIN4, but not SPIN1 (Figure 3a, Figure S5 and Table S4). We performed these proteomic experiments with and without MG132 treatment and, interestingly, found that DCAF16 and DDB1 interacted with SPIN4, but not SPIN1, under both conditions. No other proteins were found to preferentially bind SPIN4 over SPIN1. On the other hand, SPIN1 preferentially interacted with its established binding partners (e.g., C11orf84 (or SPINDOC), MORC4) and several additional nuclear proteins (e.g., BCLAF1, LBR, THRAP3) (Figure 3a and Table S4). Finally, few, if any proteins were identified that interacted with both SPIN1 and SPIN4 as determined by comparisons to mock-transfected cells (Table S4).

Figure 3. SPIN4, but not other SPIN family proteins, interacts with DCAF16.

a. Quantitative MS-based proteomics showing SPIN4/SPIN1 ratio values of proteins identified in anti-FLAG affinity enrichment experiments from HEK293T cells transiently expressing FLAG-SPIN1 or FLAG-SPIN4 treated with MG132 (10 μM, 2 h). The y-axis and x-axis correspond to coefficients of variation and the average relative log2 protein abundances (SPIN4/SPIN1), respectively, from two biologically independent experiments. b. FLAG-SPIN4, but not FLAG-SPIN1, FLAG-SPIN2A or FLAG-SPIN3, co-immunoprecipitated with HA-DCAF16. HEK293T cells were transiently transfected with FLAG-SPIN1/2A/3/4 and HA-DCAF16 for 24 h and then treated with MG132 (10 μM, 2 h). The result is representative of three biologically independent experiments. Bar graph (bottom) represents quantification of the HA-DCAF16 protein content after anti-FLAG affinity enrichment. Data are mean ± SEM (n = 3). Statistical significance was calculated with unpaired two-tailed Student’s t-tests comparing SPIN4 transfected group to others. ***P < 0.001. c. DCAF16 mediates polyubiquitination of SPIN4, but not SPIN1, in HEK293T cells. HEK293T cells were transiently transfected with FLAG-SPIN1 or FLAG-SPIN4 and HA-DCAF16 for 24 h and then treated with MG132 (10 μM, 2 h). The result is representative of three biologically independent experiments. Bar graph (bottom) represents quantification of the total polyubiquitination content after anti-FLAG affinity enrichment. Data are mean ± SEM (n = 3). Statistical significance was calculated with unpaired two-tailed Student’s t-tests comparing DCAF16 and SPIN4 co-transfected group to others. ***P < 0.001.

We did not identify SPIN2A/B and SPIN3 in our original proteomic analysis of DCAF16-WT and DCAF16-KO cells, but found that SPIN2A and SPIN3 did not interact with DCAF16 in co-expression and co-immunoprecipitation experiments in HEK293T cells (Figure 3b). Finally, we also found that DCAF16 does not ubiquitinate SPIN1 (Figure 3c). These results, taken together, indicate that SPIN4, but not other SPIN proteins, forms a dedicated complex with DCAF16/DDB1 that regulates turnover of SPIN4 in human cells.

Considering that SPIN1 and SPIN4 share 75% sequence identity, we were initially surprised that each protein showed distinct interacting partners. The N-terminal ~50 amino acids represent the region of greatest sequence divergence between SPIN1 and SPIN4 (Figure S6). However, a chimeric protein where aa 1-49 of SPIN1 replaced aa 1-36 of SPIN4 (SPIN1-4, Figure 4a) still interacted with DCAF16 to a comparable extent as WT-SPIN4 (Figure 4b). We then replaced an internal region of SPIN4 (aa 37-96) with the corresponding region on SPIN1 (aa 50-109) and found that this chimeric protein (SPIN4-1-4) no longer interacted with DCAF16 (Figure 4b). We further compared the amino acid sequences within this region and noted three lysine residues in SPIN4 (K70, K81, and K93) that were not found in SPIN1 (corresponding to N83, F94, E106) (Figure 4a) or SPIN2A/B and SPIN3 (Figure S7). We wondered whether these amino acid differences within an otherwise highly conserved region of the SPIN proteins might contribute to DCAF16 interaction. Consistent with this hypothesis, we found that a K70N/K81F/K93E-triple mutant of SPIN4 (SPIN4 3K) no longer interacted with DCAF16 (Figure 4b). We then made individual mutations and found that K81F and K93E SPIN4 mutants maintained interactions with DCAF16, while the K70N SPIN4 mutant did not interact with DCAF16 (Figure 4c,d). In line with this, K70N SPIN4 was not ubiquitinated by DCAF16 (Figure S8). An unpublished structure of SPIN4 bound to a trimethylated H3K4 (H3K4me3) peptide has been deposited in the Protein Data Bank (PDB; 4UY4), which revealed that K70 is at a surface-exposed location on the opposite face of SPIN4 from the binding site for H3K4me3 (Figure S9). These data, taken together, designate K70 as a key residue involved in supporting SPIN4-DCAF16 interactions.

Figure 4. SPIN4 K70 is critical for interacting with DCAF16.

a. Schematic showing the chimeric proteins used in this study and the sequence alignment between SPIN1 (aa 50-109) and SPIN4 (aa 37-96). b. FLAG-SPIN4 and FLAG-SPIN1-4, but not FLAG-SPIN4-1-4 and FLAG-SPIN4 3K, co-immunoprecipitated with HA-DCAF16. HEK293T cells were transiently transfected with FLAG-SPIN4/1-4/4-1-4/3K and HA-DCAF16 for 24 h and then treated with MG132 (10 μM, 2 h). The result is representative of three biologically independent experiments. SPIN4 3K represent a triple lysine mutant (SPIN4 K70N/K81F/K93E). c. FLAG-SPIN4 K70N no longer interacts with HA-DCAF16. HEK293T cells were transiently transfected with FLAG-SPIN4 WT or mutants (3K, K70N, K81F, K93E) and HA-DCAF16 for 24 h and then treated with MG132 (10 μM, 2 h). The result is representative of three biologically independent experiments. Bar graph (bottom) represents quantification of the HA-DCAF16 protein content after anti-FLAG affinity enrichment. Data are mean ± SEM (n = 3). Statistical significance was calculated with unpaired two-tailed Student’s t-tests comparing SPIN4 transfected group to others. **P < 0.01.

All five members of SPIN family contain three Tudor methyl reader domains^17–19, and therefore SPIN4 is thought to bind methylated histones. As noted above, a structure has been solved of SPIN4 bound to a trimethylated H3K4 (H3K4me3) peptide (4UY4), providing support for this biochemical interaction. To better understand the specificity of SPIN4 interactions with modified histone sequences, we exposed FLAG-tagged SPIN4-transfected HEK293T cell lysates to a histone peptide microarray and quantified the relative signals of SPIN4 bound to each histone peptide. The strongest signals for SPIN4 were indeed associated with the H3K4me3 peptide among the 62 histone peptides on the microarray (Figure S10a). We also found that SPIN4 associated with a biotinylated H3K4me3 peptide, but not an unmodified H3K4 peptide (Figure S10b). Since H3K4me3 is thought to associate with the transcription start sites²⁰, we wondered if SPIN4 might regulate gene transcription and protein expression in cells. We generated SPIN4-KO MDA-MB-231 cells using CRISPR/Cas9 methods (Figure S11) and confirmed by MS-based proteomics > 90% reductions in SPIN4 in the KO population compared to SPIN4-WT cells (Figure S12). However, we did not observe any other proteins that showed substantial (> 1.5-fold) changes in expression (5476 total quantified proteins; Figure S12 and Table S5). These data indicate that, while SPIN4 can bind H3K4me3 in vitro, the genetic disruption of SPIN4 was insufficient to cause major constitutive changes in protein expression in human cells.

With more than 600 predicted E3 ligases in the human proteome, an important objective is to match these components of the proteostasis network with their endogenous protein substrates. Some E3 ligases appear to have very restricted, even single substrate relationships (e.g., KEAP1-NRF2^21,22, MDM2-p53^23,24, VHL-HIF1/2a^{25, 26} etc.), while others may be more promiscuous in their substrate scope. In this study, we have used an array of proteomic approaches to identify and characterize SPIN4 as an interaction partner and endogenous substrate of DCAF16. Interestingly, a previous study²⁷ showed that SPIN4 is the only SPIN family member that was regulated in its cellular abundance by CRLs, and our findings that other SPINs do not interact with DCAF16 are consistent with this past work. Conversely, we also found that the protein interaction partners of SPIN1 do not associate with SPIN4. These results suggest that each SPIN family member, despite sharing high sequence homology with one another, may engage in different and largely non-overlapping protein complexes. We made some progress in deciphering the features of SPIN4 that impart binding to DCAF16, including the identification of a key lysine residue K70 that is unique to SPIN4 and required for this protein-protein interaction.

Why SPIN4 would be part of a protein complex responsible for its constitutive ubiquitination and degradation is unclear, but, for other substrate-E3 ligase pairs, such as NRF2-KEAP1^21,22 and HIF1/2a-VHL^{25, 26}, this type of relationship enables exquisite post-translational control over the cellular concentrations of transcriptional regulatory proteins that, in turn, promote stimulus-responsive gene expression networks. Our initial attempts to relate the H3K4me3-binding activity of SPIN4 to effects on gene and protein expression, however, were unsuccessful, as our proteomic experiments did not detect constitutive changes in protein abundances in SPIN4-KO cells. It is possible that SPIN4 regulates gene and protein expression in specific, as-of-yet unidentified cellular states or in response to discrete signals; alternatively, SPIN4 may serve a different function altogether in cells that is unrelated to regulating gene and protein expression. Moreover, given the similarity across the SPIN family proteins, it is also possible that potential alterations in protein expression in SPIN4-KO cells are counteracted by compensatory functions of other SPIN family proteins. Future studies where multiple, or the entire family of SPIN proteins are disrupted may shed light on this possibility.

We have shown previously that DCAF16 harbors reactive cysteines that can be engaged by electrophilic PROTACs to promote targeted protein degradation⁹. Whether the same region of DCAF16 is involved in mediating the activity of electrophilic PROTACs and interactions with SPIN4 remains unclear. If so, then these events could compete with one another, and, for instance, high stoichiometric engagement of the relevant cysteines in DCAF16 by electrophilic compounds might provide a way to pharmacologically suppress SPIN4 interactions and turnover. Tudor domains are methyl-lysine/arginine reader domains that have been found in ~40 proteins, including five SPIN family proteins²⁸. The Tudor domain’s aromatic cage pocket recognizes methyl-lysine or methyl-arginine through cation-π and π-stacking interactions²⁸. Other strategies to pharmacologically target SPIN4 could involve developing small molecules that directly bind the Tudor domain to disrupt methylated histone interactions, as has been demonstrated for SPIN1^29,30.

We conclude with some speculation regarding whether DCAF16 may represent another example of an E3 ligase with a highly restricted substrate scope. In two different human cancer cell lines, untargeted MS-based proteomics identified SPIN4 as the only protein substantially affected by loss of DCAF16. That DCAF16 also represents the main interacting protein for SPIN4 in human cells further supports a potentially dedicated relationship. Alternatively, additional endogenous substrates for DCAF16 may be identified as more cell types are examined or as cells are studied under different stress conditions. We finally note that, even though the cellular functions of SPIN4 remain mysterious, the SPIN4 protein is highly conserved across mammalian species (> 95% sequence identity between human and mouse), and the SPIN4 gene has a strong EXAC pLI score (0.72; https://gnomad.broadinstitute.org/), suggesting natural selective pressure against the emergence of loss-of-function alleles. We anticipate that our findings will provide a biochemical framework to support deeper investigations into the functions of SPIN4 and DCAF16 in human health and disease.