Distinct nuclear and cytoplasmic assemblies and interactomes of the mammalian CTLH E3 ligase complex

Ubiquitination, the post-translational modification that involves the addition of ubiquitin (a 76 amino-acid protein) to a protein substrate, is essential for protein turnover or function (Komander and Rape, 2012). In this pathway, substrate specificity is mediated by E3 ubiquitin ligases. E3 ligases contain conserved protein interaction domains, which bind specific substrates, and the majority of them contain Really Interesting New Gene (RING) domains, which are essential for mediating ubiquitin transfer. Currently, there are over 600 RING E3 ligases that are predicted to target specific substrates for ubiquitination, but many of these substrates are currently unknown (Deshaies and Joazeiro, 2009).

RING E3 ligases can function as single proteins, dimers or multi-subunit protein complexes. An example of the latter is the C-terminal to LisH (CTLH) complex, also known as the mammalian glucose-induced degradation-deficient (GID) complex. This complex is composed of at least nine proteins: Ran-binding protein M (RanBPM, also called RanBP9), muskelin (encoded by MKLN1), WD repeat-containing protein 26 (WDR26), armadillo repeat-containing protein 8 (ARMC8) isoforms α and β, two hybrid-associated protein 1 with RanBPM (TWA1 or GID8), GID4, required for meiotic nuclear division 5A (RMND5A) and macrophage erythroblast attacher (MAEA) (Kobayashi et al., 2007; Maitland et al., 2019; Umeda et al., 2003). Additional members such as the RanBPM paralog RanBP10, RMND5A paralog RMND5B and protein yippee-like 5 (YPEL5) also associate with the complex, although whether they are constitutive complex members is not completely understood (Huffman et al., 2019; Lampert et al., 2018). Collectively, these subunits contain α-helical Lis1 homology (LisH) and CTLH domains, and several conserved protein interaction surfaces including WD40 repeats, armadillo repeats and kelch repeats, as well as discoidin and SPRY domains, which mediate a number of intracellular interactions (Francis et al., 2013; Kobayashi et al., 2007).

The CTLH complex has been shown to regulate the ubiquitination and proteasomal degradation of factors involved in cellular proliferation, growth and differentiation, including HBP1, c-Raf, β-catenin, lamin B1, AMPK and its own subunit muskelin (Lampert et al., 2018; Liu et al., 2020; Maitland et al., 2019; McTavish et al., 2019; Sato et al., 2020; Zhen et al., 2020). Additionally, the complex inhibits glycolysis by regulating non-degradative ubiquitination of PKM2 and LDHA (Maitland et al., 2021). The reported ubiquitination substrates, abundant protein interaction surfaces and highly conserved nature of the complex in higher-order organisms indicates that the CTLH complex is an important regulator of cellular homeostasis (Francis et al., 2013; Liu et al., 2020; Maitland et al., 2021; Pfirrmann et al., 2015; Santt et al., 2008).

Although it is recognized that most components of the CTLH complex are present in both nuclear and cytoplasmic compartments, there is still much uncertainty regarding the composition and function of the human CTLH complex in either compartment. Previous studies have shown that certain CTLH complex members mostly localize to the nucleus, whereas others are predominantly cytoplasmic, suggesting that the CTLH complex composition in vivo might differ between these two compartments (Francis et al., 2013; Kobayashi et al., 2007; Maitland et al., 2019; Sun et al., 2013; Umeda et al., 2003; Valiyaveettil et al., 2008). Other studies provide evidence that the CTLH complex shuttles between the cytoplasm and nucleus, potentially through nuclear localization signals present within the complex members RanBPM and WDR26, and exerts its function by binding to compartment-specific proteins (Boldt et al., 2016; Huttlin et al., 2017; Lampert et al., 2018; Malovannaya et al., 2011; Napoli et al., 2020; Salemi et al., 2015; Zhen et al., 2020).

Certain complex members also influence the stability of the complex. For example, knocking out individual CTLH complex members in mammalian cells changes the protein expression of other complex members and eliminating WDR26, but not muskelin, disrupts complex assembly (Maitland et al., 2019; Sherpa et al., 2021). Given the differing subcellular profile of complex subunits and the possibility of distinct complex assemblies, clarity is needed on the localization of endogenous CTLH complex members and the stability of compartment-specific CTLH complexes.

In this study, we evaluated the relative localization of CTLH complex members within the cytoplasm and nucleus of HeLa cells and investigated whether it changed in HeLa cells lacking specific CTLH complex members. We determined that loss of WDR26 severely impairs nuclear CTLH complex stability, revealing WDR26 as a critical determinant of CTLH complex nuclear function. CTLH complex formation was restored by treatment with a proteasome inhibitor, suggesting that the nuclear complex is protected from degradation by WDR26. We also identified high-confidence interacting proteins (HCIPs) in both cellular compartments by affinity-purification coupled to mass spectrometry (AP-MS) using endogenous CTLH complex member RanBPM as bait. Finally, we validated a subset of RanBPM HCIPs in vitro by co-immunoprecipitation and in situ by proximity ligation assay (PLA). Overall, we found that WDR26 is essential for nuclear CTLH complex formation and, by the identification of protein interactions with the CTLH complex that were specific to subcellular compartments, our findings implicated the CTLH complex in a number of novel mechanisms.

To study the CTLH complex in the cytoplasm and nucleus, we first devised a workflow to generate high-quality and reproducible nuclear and cytoplasmic fractions from HeLa cells (Fig. 1A). Equal quantities of cytoplasmic, nuclear and whole-cell extracts were analyzed by western blotting for the presence of endogenous CTLH complex members (Fig. 1B). To identify potential differences in the relative levels of nuclear versus cytoplasmic expression between CTLH complex members, we normalized the expression of three biological replicates to vinculin and SAP62 (or SF3A2) and converted the relative expression levels into ratios or proportions (Fig. 1C). These results revealed that all CTLH core members tested are found in both nuclear and cytoplasmic compartments. Interestingly, we noticed that some complex members displayed relatively higher abundance in the cytoplasm (muskelin, WDR26 and ARMC8ɑ) compared to the other complex members.

Previous sucrose density gradient experiments revealed that individual subunits of the CTLH complex migrate towards high molecular mass fractions (600–800 kDa), which exceed the predicted stoichiometric assembly in HEK293 and A549 mammalian cells (Kobayashi et al., 2007; Sherpa et al., 2021). Previous analyses also showed that in the absence of WDR26 and, to a lesser extent, muskelin, RanBPM migrated to lower molecular mass fractions, suggesting that removing WDR26 severely disrupts the higher molecular mass complex identified in wild-type (WT) cells (Sherpa et al., 2021). To investigate the impact of muskelin and WDR26 on the subcellular localization of CTLH complex members, we first performed western blot analyses using HeLa muskelin and WDR26 knockout (KO) nuclear and cytoplasmic protein fractions. The loss of muskelin, which does not significantly alter the stability of CTLH complex members in mammalian cells (Maitland et al., 2019), did not appreciably change the subcellular distribution of CTLH complex subunits; however, the loss of WDR26 resulted in a severe reduction of most CTLH subunit levels in the nuclear fraction without affecting subunit expression in the cytoplasm (Fig. 2A). In particular, the levels of muskelin, RanBPM, ARMC8β, MAEA and RMND5A were drastically reduced in the WDR26 KO nuclear extracts, but not in the cytoplasmic extracts. The loss of WDR26 resulted in a global decrease in expression of CTLH complex members in whole-cell extracts, indicating that the reduction of protein levels in the nucleus is likely due to a nuclear-specific loss of expression, rather than a change in localization to the cytoplasm (Fig. 2B). This effect was specific to WDR26, as re-expression of WDR26 in WDR26 KO cells restored CTLH complex subunit expression to near wild-type levels (Fig. 2C).

We then prepared 5–40% sucrose gradients to evaluate the effect of WDR26 loss on CTLH complex formation in whole-cell, cytoplasmic and nuclear WT HeLa extracts. In WT whole-cell extract (WCE), we observed the co-migration of CTLH complex subunits primarily in three fractions corresponding to over 670 kDa – as previously reported (Kobayashi et al., 2007; Sherpa et al., 2021) – with some complex members also being detected in lower molecular mass fractions (Fig. 3A). The WT nuclear fraction showed a similar migration pattern of CTLH subunits compared to that of WT WCE; however, the WT cytoplasmic fraction showed a different sedimentation behavior, with most subunits present in a wider range of fractions, which is indicative of lower molecular mass complexes (Fig. 3B,C).

Consistent with a recent study (Sherpa et al., 2021), most CTLH subunits shifted to lower molecular mass fractions in WDR26 KO WCE compared to WT WCE (Fig. 3A). Interestingly, in the WDR26 KO nuclear fraction, we observed a dramatic shift of most CTLH subunits to lower molecular mass fractions, suggesting disruption of CTLH complex formation (Fig. 3C). Surprisingly, the loss of WDR26 had little to no effect on the distribution of cytoplasmic CTLH complex subunits compared to WT, with only a slight change of ARMC8α to lower molecular fractions, suggesting that WDR26 does not significantly affect complex formation in the cytoplasm (Fig. 3B).

As we did not observe any change in subunit expression in the cytoplasm upon loss of WDR26, the decreased expression of CTLH complex subunits in the nucleus was likely not due to their relocalization to the cytoplasm, but rather to decreased stability and/or specific degradation in the nucleus. To test this possibility, we evaluated nuclear CTLH complex expression in WDR26 cells treated with the proteasome inhibitor MG132. The treatment restored expression of CTLH subunits muskelin, RanBPM and ARMC8, which confirms that in the absence of WDR26, subunits of the CTLH complex are targeted for degradation (Fig. 4A). To determine the effect the restored expression of these subunits had on CTLH complex formation, we further analyzed MG132-treated WDR26 KO nuclear extracts using sucrose gradients. Strikingly, MG132 treatment promoted the formation of the higher molecular mass complexes observed in the nuclear fractions in wild-type cells (Fig. 4B,C), suggesting that maintaining the expression of CTLH complex subunits in the absence of WDR26 is sufficient to restore higher-order complex formation.

Overall, our findings reveal that the CTLH complex adopts distinct molecular mass patterns, which likely reflect differential subunit assemblies in the cytoplasm compared to the nucleus. In addition, although all CTLH complex subunits are present in both the nucleus and the cytoplasm, we find that CTLH complex formation is differentially regulated by WDR26 in the nucleus compared to the cytoplasm of HeLa cells.

We previously demonstrated that we could efficiently immunoprecipitate the endogenous CTLH complex from mammalian cell extracts using a RanBPM antibody (Maitland et al., 2019, 2021). Therefore, we performed AP-MS using endogenous RanBPM as bait to identify interactors in the cytoplasmic and nuclear protein extracts. Immunoprecipitates were analyzed by liquid chromatography tandem mass spectrometry (LC-MS/MS). A complete overview of all identified proteins is provided in Table S2.

After filtering, 27 cytoplasmic and 155 nuclear HCIPs were identified to interact with RanBPM exclusively in their respective subcellular compartment, whereas 31 additional RanBPM HCIPs were identified in both compartments (Fig. 5A). As expected, all CTLH complex members were identified as HCIPs in both nuclear and cytoplasmic fractions, whereas the putative complex members YPEL5, RanBP10 and RMND5B were also identified in both compartments. We then performed Gene Ontology (GO) cellular component analysis with the cytoplasmic and nuclear HCIP list and determined that the proteins coincided with their expected localizations (Fig. 5B). Many statistically significant [false discovery rate (FDR)<0.05] GO components were also represented in both HCIP lists, but this is not surprising as many of the proteins that make up each GO component are nucleocytoplasmic in nature. Comparing our list to the BioGrid database, we found that 21 HCIPs were previously identified as interactors with other complex members (Fig. 5C).

Finally, we identified statistically enriched GO processes in our AP-MS dataset using STRING enrichment. Many of these GO processes – such as leukocyte mediated immunity, histone deacetylation and microtubule bundle formation – have previously been associated with CTLH complex function (Murrin and Talbot, 2007; Salemi et al., 2015, 2017). However, we found several other significantly enriched GO processes that have not previously been linked to CTLH complex function (Fig. 6). For example, we identified several factors implicated in senescence-associated heterochromatin foci assembly, long-chain fatty-acid import into peroxisomes and histone H4-K20 trimethylation.

To validate the AP-MS results, we performed co-immunoprecipitations (co-IPs) of nuclear and cytoplasmic fractions and analyzed immunoprecipitates by western blotting. Consistent with previous findings (Boldt et al., 2016; Huttlin et al., 2021; Kristensen et al., 2012; Lampert et al., 2018; Maitland et al., 2019, 2021; Wan et al., 2015), we found that RanBPM associated with all other CTLH complex members in whole-cell extracts (Fig. 7A). In addition, all complex members tested co-immunoprecipitated with RanBPM in both cytoplasmic and nuclear lysates.

We then validated selected putative interactors of RanBPM, such as the cytoplasmic protein tankyrase (TNKS1 or TNKS, and TNKS2, collectively referred to as TNKS1/2) and the nuclear proteins HDAC2, p14ARF (also known as CDKN2A) and macroH2A1 (Fig. 7B). As predicted by AP-MS, TNKS1/2 co-immunoprecipitated with RanBPM mostly in the cytoplasmic extracts, whereas HDAC2, p14ARF and macroH2A1 co-immunoprecipitated in the nuclear extracts. We repeated these experiments using extracts from HEK293 cells and determined that TNKS1/2 and macroH2A1 also associated with RanBPM in the cytoplasmic and nuclear fractions, respectively, confirming that these interactions are not cell-type dependent (Fig. S2).

To further confirm these associations in a cellular context, we investigated whether the interactions between RanBPM and selected prey proteins could be visualized in situ using PLA in HeLa cells (Söderberg et al., 2008). A microscopy-based technique also enabled us to determine whether the interactions occurred in the predicted cellular compartments. In addition to a negative control containing no primary antibodies, we performed the assay in previously generated and validated RanBPM shRNA HeLa cells, in which the expression of RanBPM is severely reduced (Maitland et al., 2019; McTavish et al., 2019). We observed that RanBPM and macroH2A1 associated exclusively in the nuclei of WT HeLa cells. In addition, we confirmed that RanBPM and TNKS1/2 primarily associated in the cytoplasm of WT HeLa cells (Fig. 7C). Overall, the PLA assay confirmed that RanBPM associates in situ with TNKS1/2 and macroH2A1 in the cytoplasm and nucleus of HeLa cells, respectively, further validating the AP-MS results.

All validations of the interactions were performed through experiments using a RanBPM antibody, as we have not identified (thus far) any other CTLH complex member antibody that can be used in immunoprecipitation or immunocytochemistry. Therefore, to validate that the interactions could be detected through immunoprecipitation with another CTLH complex member, we assessed TNKS1/2 association with WDR26 that was transfected in HeLa cells. We have previously reported that ectopically-expressed FLAG–WDR26 efficiently associates with endogenous CTLH complex members (Maitland et al., 2019). FLAG pulldown of transfected FLAG–WDR26 specifically co-immunoprecipitated endogenous TNKS1/2, confirming that TNKS1/2 associates with the CTLH complex (Fig. 7D).

In this study, we generated cytoplasmic and nuclear lysates to evaluate the distribution of endogenous CTLH complex subunits, determined differences in the stability between the compartment-specific complexes and identified compartment-specific CTLH complex member RanBPM HCIPs using AP-MS. First, our results establish that all endogenous CTLH complex members are found in both subcellular compartments, but that the CTLH complex exhibits different patterns in its molecular mass in the nuclear and cytoplasmic fractions, suggesting the existence of distinct complexes in the nucleus and the cytoplasm. Strikingly, we found that in the absence of the CTLH complex member WDR26, protein levels and overall complex formation decreased in the nucleus, whereas the cytoplasmic complex remained largely unaffected. This could be reversed through MG132 treatment, indicating that WDR26 protects the nuclear complex from degradation. Second, we identified over 200 high-confidence interactors that might implicate the CTLH complex in distinct compartment-specific pathways. GO process analysis revealed the potential involvement of the CTLH complex in diverse processes including chromatin assembly and ribonucleoprotein complex biogenesis. Finally, we confirmed the interactions between CTLH complex member RanBPM and several predicted interactors by co-IP, validated the histone macroH2A1 and TNKS1/2 as RanBPM-interacting proteins in the nucleus and cytoplasm, respectively, using PLA analyses, and determined that TNKS1/2 interacted with WDR26.

There have been conflicting findings regarding the localization of CTLH complex subunits. Although some subunits, such as RanBPM, were characterized as being nucleocytoplasmic, others have been reported to be localized predominantly in the cytoplasm (e.g. muskelin, WDR26) or the nucleus (e.g. MAEA, GID4) (Bala et al., 2006; Kobayashi et al., 2007; Maitland et al., 2019; Salemi et al., 2015; Valiyaveettil et al., 2008; Zhu et al., 2004). These studies mainly used transfected proteins, thus creating an artificial cellular environment in which overexpressed proteins might not all bind to the complex. In this study, we avoided these conditions by evaluating endogenous protein localization. The results obtained with our subcellular fractionation method show conclusively that all endogenous CTLH subunits are present in the nucleus and cytoplasm of mammalian cells.

Quantification of relative CTLH subunit abundance in the nuclear and cytoplasmic fractions revealed that, surprisingly, WDR26, muskelin and ARMC8α have a more cytoplasmic localization relative to the other complex members. This is, to our knowledge, the first report indicating that the ARMC8 ɑ and β subunits have different relative subcellular localizations. These proteins are alternatively spliced products from the ARMC8 gene (Kobayashi et al., 2007), with ARMC8β lacking the C-terminal end of the full-length construct. Recent studies have shown that each isoform can bind to different proteins, and that the common N-terminus is important for ARMC8 integration into the complex (Gul et al., 2019; Mohamed et al., 2021; Suzuki et al., 2008). Because of these properties, the CTLH complex might interact with certain proteins in an ARMC8ɑ- or ARMC8β-dependent manner. Recent findings have shown that GID4 associates with the CTLH complex by binding to ARMC8ɑ but not ARMC8β in vitro; in addition, there is evidence that the CTLH complex ubiquitinates ZMYND19 in a GID4-dependent manner (Mohamed et al., 2021). Our findings that ARMC8ɑ localization is more cytoplasmic than nuclear indicates that the cytoplasmic CTLH complex could recruit substrates through GID4 more so than the nuclear CTLH complex. It will be interesting to see whether this model is supported as more cytoplasmic complex-mediated ubiquitination substrates are identified.

Recent in vitro models have revealed that, similar to the yeast Gid complex, the CTLH complex can form a higher-order complex, which is mediated by WDR26 (Mohamed et al., 2021; Qiao et al., 2020; Sherpa et al., 2021). The sucrose gradient analyses of mammalian whole-cell extracts revealed that the loss of WDR26 shifted the endogenous CTLH complex to a lower molecular mass, suggesting that the higher-order assembly is disrupted in WDR26 knockout cells (Sherpa et al., 2021). Although our data corroborate these previous findings, we show here that this predominantly occurs in the nucleus, in which a larger proportion of CTLH complex members co-migrate in the high molecular mass fractions. Thus, WDR26 is necessary for the stabilization of the higher-order CTLH complex structure that is predominant in the nucleus. In addition to mediating the formation of higher-order complexes, WDR26 is part of a module that targets specific CTLH complex ubiquitination substrates, such as HBP1, for degradation (Mohamed et al., 2021; Sherpa et al., 2021). Similarly, its yeast homologue, Gid7, is required for Gid-mediated ubiquitination of specific oligomeric substrates (Qiao et al., 2020). Taken together with these studies, our data suggest that the higher-order CTLH complex might be used to primarily ubiquitinate nuclear substrates. In summary, our data show that there are compartment-specific compositions of the CTLH complex that likely have distinct targets and functions.

Our study also shows that the loss of WDR26, but not muskelin, leads to a severe decrease in protein levels of other CTLH complex members in the nuclei of HeLa cells. This is not due to altered translocation of complex members between the cytoplasm and the nucleus, as CTLH subunit levels are globally downregulated in whole-cell extracts. Our observation that treatment with the proteasome inhibitor MG132 restores expression of these subunits in the nucleus of WDR26 KO cells substantiates the idea that most CTLH subunits are destabilized and/or degraded in the absence of WDR26 in the nucleus. This indicates that the WDR26-dependent higher-order CTLH complex structure protects these nuclear complex members from degradative pathways.

One unexpected finding was that the higher-order complex can be restored upon MG132 treatment in WDR26 KO cells. The most likely explanation is that in these conditions, the CTLH complex member muskelin is able to compensate for the loss of WDR26 and promote multimerization of CTLH complex modules to form supramolecular complex(es). Muskelin was previously suggested to mediate multimerization of CTLH complex modules independently of WDR26, although its knockout did not have a major effect on high molecular mass fractions in sucrose gradients from whole-cell extracts (Sherpa et al., 2021).

In summary, our data suggest that the loss of WDR26 has minimal effects on the cytoplasmic CTLH complex, likely because muskelin can compensate for WDR26 in higher molecular mass complex formation; however, in the nucleus, WDR26 appears to have an essential, protective role, as its absence leads to degradation of muskelin and other CTLH complex members, thereby preventing CTLH complex assembly.

Unexpectedly, TWA1 and, to an extent, ARMC8ɑ nuclear sucrose gradient profiles were less affected by WDR26 loss compared to other complex members, as they were still present in higher molecular mass fractions. It is not known what proteins TWA1 and ARMC8ɑ associated with in these fractions, but their presence suggested that these two members might form alternate complexes that have yet to be identified.

Remarkably, ARMC8ɑ and ARMC8β are differentially affected by the loss of WDR26, with ARMC8β being severely reduced in the nucleus. It has been recently shown that ARMC8β could stabilize the tetrameric form of the CTLH complex (Mohamed et al., 2021); therefore, its disappearance from the nucleus in WDR26 KO cells could further contribute to the disaggregation of the complex in the nucleus. The reduced TWA1 levels in WDR26 KO nuclei might also explain the instability of other complex members, as TWA1 was previously shown to destabilize complex members in whole-cell mammalian extracts (Maitland et al., 2019). Overall, our data show that WDR26 is important for higher-order nuclear CTLH complex stability in vivo. Our data also suggest that different mechanisms are used to regulate the cytoplasmic and nuclear CTLH complex assemblies, which is indicative of the involvement of other unidentified factors in modulating CTLH complex formation.

The AP-MS analysis identified all CTLH complex members as nucleocytoplasmic RanBPM HCIPs; also present were the putative CTLH complex members RanBP10, RMND5B and YPEL5 (Boldt et al., 2016; Harada et al., 2008; Hosono et al., 2010; Lampert et al., 2018). RanBP10 is a paralog of RanBPM, with both proteins sharing the same functional domains. Interestingly, RanBP10 has a similar subcellular localization pattern as well as many common previously identified interactors of RanBPM, suggesting that both proteins functionally overlap (Palmieri et al., 2018). Similar to RanBP10, RMND5B is a relatively unstudied protein. RMND5B has been shown to interact with ARMC8 and WDR26 but not with RMND5A, suggesting a potential compensatory role (Lampert et al., 2018). Contrary to previous studies that examined RMND5B localization in Xenopus laevis cells (Pfirrmann et al., 2015), we show that RanBPM has the ability to interact with both RMND5A and RMND5B in both cellular compartments, suggesting that these two proteins are located in both compartments and that there might be species-specific differences.

Our high-throughput screen identified a number of RanBPM interactors that had previously been shown to interact with other CTLH complex members. For example, previous studies showed that the CTLH complex interacts with and regulates ZMYND19 through its ubiquitination activity, whereas several CTLH complex members have been shown to interact with the DDX50 RNA helicase in other studies (Huttlin et al., 2021; Lampert et al., 2018; Mohamed et al., 2021). As this provides independent confirmation of the interaction between some of these factors and the CTLH complex, the RanBPM interactome identified in this study can be used as a resource to further explore the nature and outcome of these interactions.

Similar amounts of extracts from nuclear and cytoplasmic fractions were used for the AP-MS. Estimates indicate there are 2.3 times more cytoplasmic proteins compared to nuclear proteins in HeLa cells (Shaiken and Opekun, 2014); therefore, nuclear proteins were proportionally overrepresented. However, this study identified a far greater proportion of nuclear HCIPs (155) compared to cytoplasmic interactors (27) following MiST analysis, potentially suggesting the higher-order nuclear CTLH complex has a greater number of stable interactors.

The compartment-specific interactomes provide many clues to predict the specific localizations of the CTLH complex, particularly in the nucleus. For example, the nuclear HCIP list comprises many nucleolar proteins, which is consistent with previous findings. Recently, RanBPM was found to interact with the nucleolar protein nucleolin in mouse cells (Soliman et al., 2020), whereas a separate group identified the brain-specific nucleolar protein Centaurin ɑ-1 as a RanBPM interactor in human cells (Haase et al., 2008), further confirming that the CTLH complex is present in the nucleolus. Consistent with this localization, the most represented GO term in our dataset was the ribonucleoprotein complex biogenesis process. The proteins involved in this process regulate many cellular processes including gene and protein expression, as well as alternative splicing (Hogan et al., 2008). Interestingly, the CTLH complex was shown to regulate alternative splicing during male germ-cell development, suggesting a potential functional role of the interactions (Bao et al., 2014). Many of the proteins involved in ribonucleoprotein complex biogenesis, such as UTP18, RRP9, WDR12, WDR46, WDR74 and BOP1, contain Trp-Asp-40 (WD40) domains. These domains have been shown to dimerize, meaning that the CTLH complex member WDR26 might interact with these nuclear HCIPs directly (Zhang et al., 2019). The chromatin assembly GO process is also enriched in the nuclear HCIPs. Interestingly, the CTLH complex has been found on chromatin in past studies (Brunkhorst et al., 2005; Salemi et al., 2015). Although it is mechanistically still not understood, the interaction of the CTLH complex with chromatin could have important regulatory roles, as histone ubiquitination has been shown to maintain genome stability and regulate gene expression (Mattiroli and Penengo, 2021; Weake and Workman, 2008).

The compartment-specific co-immunoprecipitation of TNKS1/2, HDAC2, p14ARF and macroH2A1 with RanBPM validated the AP-MS results. In addition, PLA further confirmed proximal interaction in situ between cytoplasmic RanBPM and TNKS1/2. Tankyrase-1 (TNKS1) and tankyrase-2 (TNKS2) are functionally redundant poly(ADP ribose) (PAR) polymerases that regulate a number of cellular pathways, including Wnt/β-catenin signaling (Haikarainen et al., 2014; Lehtiö et al., 2013; Mariotti et al., 2017). Cytoplasmic TNKS1/2 promote Wnt/β-catenin signaling by PARylating axin, a key scaffold protein that inhibits Wnt signaling by maintaining low β-catenin protein levels (Huang et al., 2009). Individual members of the CTLH complex have also been shown to regulate Wnt signaling by interacting with axin and WNK proteins (Anvarian et al., 2016; Goto et al., 2016; Sato et al., 2020). Our finding that TNKS1/2 is associated with the CTLH complex suggests that the role of the CTLH complex in Wnt pathway regulation might be more intricate than previously assessed. Interestingly, and in support of our data, the CTLH complex was found to associate with TNKS1 and TNKS2 when the tankyrases were used as baits (Li et al., 2017).

In contrast to TNKS1/2, macroH2A1 associates with RanBPM exclusively in the nucleus. MacroH2A1 is a histone variant that is primarily deposited in transcriptionally repressed regions of the genome such as heterochromatin and heavily methylated DNA regions (Sun and Bernstein, 2019). MacroH2A1 is also alternatively spliced, thus generating two isoforms – macroH2A1.1 and macroH2A1.2 – which have many overlapping localizations on chromatin (Pehrson et al., 2014). Interestingly, previous studies have indicated that macroH2A1 ubiquitination modulates its transcriptional regulatory activity (Hernández-Muñoz et al., 2005; Kim et al., 2017). Many macroH2A1 ubiquitination sites have been identified using high-throughput methods, but only a small number of corresponding E3 ubiquitin ligases have been identified. Interestingly, macroH2A1 regulates genes that are essential for the proper development of mice, whereas development-related genes are transcriptionally deregulated in RanBPM knockdown mammalian cells (Atabakhsh et al., 2012; Changolkar et al., 2007; Nashun et al., 2010). These overlapping roles and the interaction between the CTLH complex and macroH2A1 support the possibility that the CTLH complex is a histone modifier, although this will require further investigation.

In summary, we demonstrate that the CTLH complex exists in both the nucleus and cytoplasm of HeLa cells, but has different compositions in these compartments. The nuclear complex structure and stability appear to depend on WDR26, whereas the cytoplasmic complex is largely unaffected by the loss of WDR26. The AP-MS analyses using endogenous cytoplasmic and nuclear CTLH complexes identified over 200 HCIPs, with a large majority (over 170) being compartment specific. Overall, our results suggest that the CTLH complex could be involved in many more processes than previously determined and might have distinct functions in the cytoplasm and the nucleus, which are performed by structurally distinct complexes.

HeLa and HEK293 cells were obtained from the American Type Culture Collection. RanBPM shRNA HeLa and HEK293 cells as well as muskelin HeLa KO cells have been previously described (Atabakhsh and Schild-Poulter, 2012; Maitland et al., 2019). Cells were cultured in high glucose Dulbecco's modified Eagle's medium (Wisent Bioproducts, St. Bruno, Quebec, Canada) supplemented with 10% fetal bovine serum (FBS), 5% glutamine and 5% sodium pyruvate (Wisent Bioproducts) at 37°C and 5% CO₂. Cells were regularly tested to ensure absence of mycoplasma contamination. For MG132 treatment, WT and WDR26 KO HeLa cells were treated with either DMSO (BioShop, Burlington, ON, Canada) or 10 μM MG132 (EMD-CalBiochem, San Diego, CA) for 24 h. Transfection with pCDNA-SBP-FLAG-WDR26 was performed as previously described (Maitland et al., 2019).

pCDNA-SBP-FLAG-WDR26 (described in Maitland et al., 2019) was a gift from Dr Songhai Chen, University of Iowa, Iowa City, IA. pSpCas9(BB)-2A-Puro V2.0 (PX459, Addgene plasmid #62988) was a gift from from Dr Joe Torchia, Western University, London, ON, Canada.

To generate WDR26 KO HeLa cells, single guide RNA (sgRNA) sequences were designed using the Benchling CRISPR tool (https://benchling.com). Top and bottom oligonucleotides with overhanging ends containing sgRNAs directed against WDR26 (5′-CACCGGGACCTGGCCCACGCCAAT-3′, 5′-AAACATTGGCGTGGGCCAGGTCCC-3′) were cloned into pSp-Cas9(BB)-2A-Puro V2.0 (PX459) digested with BpiI and then transfected into early passage cells using jetPRIME (Polyplus Transfection, Illkirch, France) according to the manufacturer's protocol. Forty-eight hours after transfection, cells were cultured under 0.3 µg ml⁻¹ puromycin selection for 7 days, followed by colony picking and expansion.

Preparation of whole-cell extracts was performed as previously described (Maitland et al., 2019). To prepare cytoplasmic and nuclear lysates, 3.6×10⁶ HeLa cells were seeded the night before. After overnight incubation, cells were scraped, washed in ice-cold PBS and lysed in cytoplasmic lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl) supplemented with 10 µg ml⁻¹ aprotinin (BioShop), 2 µg ml⁻¹ leupeptin (BioShop), 2.5 µg ml⁻¹ pepstatin (BioShop), 1 mM DTT, 2 mM NaF, 2 mM Na₃VO₄, 0.1 mM phenylmethylsulfonyl fluoride (PMSF) and 75 µg ml⁻¹ digitonin (Sigma-Aldrich). After a 10 min incubation on ice, samples were centrifuged at 2000 g for 1 min and the resulting supernatant was saved as the cytoplasmic lysate. The remaining pellet was resuspended in cytoplasmic lysis buffer and centrifuged as before (three times) to lyse any remaining cells. The resulting pellet was resuspended with nuclear lysis buffer (10 mM Tris-HCl pH 8.0, 100 mM NaCl, 0.1% sodium deoxycholate, 1 mM EDTA, 25% glycerol) supplemented with 10 µg ml⁻¹ aprotinin, 2 µg ml⁻¹ leupeptin, 2.5 µg ml⁻¹ pepstatin, 1 mM DTT, 2 mM NaF, 2 mM Na₃VO₄, 0.1 mM PMSF and 100 units ml⁻¹ benzonase nuclease (E1014−5KU, Sigma-Aldrich), and incubated on ice for 30 min. Samples were then centrifuged at 17,000 g for 20 min to collect the nuclear fraction. HEK293 cells were fractionated in the same way, but the cytoplasmic lysis buffer contained 50 µg ml⁻¹ digitonin. All protein lysates were quantified using Pierce 660 nm Protein Assay Reagent (Thermo Fisher Scientific, 22660).

Subcellular fractionation was performed as described above, but glycerol was omitted from the lysis buffers. Whole-cell extracts were prepared using whole-cell sucrose gradient buffer (10 mM Tris-HCl pH 8.0, 100 mM NaCl, 0.1% sodium deoxycholate, 1 mM EDTA) supplemented with 10 µg ml⁻¹ aprotinin, 2 µg ml⁻¹ leupeptin, 2.5 µg ml⁻¹ pepstatin, 1 mM DTT, 2 mM NaF, 2 mM Na₃VO₄, 75 µg ml⁻¹ digitonin, 100 units ml⁻¹ benzonase nuclease and 0.1 mM PMSF. The cytoplasmic extract was adjusted to contain 100 mM NaCl, 0.1% sodium deoxycholate and 1 mM EDTA. 300 µg of cytoplasmic, nuclear or whole-cell extracts were loaded onto a 5–40% sucrose gradient (w/v), made with whole-cell sucrose gradient buffer the night before. Gradients were centrifuged in an SW41Ti rotor (Beckman Coulter) at 158,043 g for 16 h at 4°C. Fractions were collected and 10% trichloroacetic acid and acetone were used to precipitate the proteins from each fraction. The protein pellets were resuspended in SDS loading buffer and boiled at 95°C for 10 min prior to SDS-PAGE. Gradients containing Gel Filtration Standard (Bio-Rad, 1511901) were prepared in parallel, loaded on 4% SDS-PAGE and stained with Coomassie Blue to determine the sedimentation pattern using a 670 kDa protein standard.

Prepared samples were resolved using 8%, 10% or 12% SDS-PAGE. Gels were transferred onto polyvinylidene fluoride membranes, blocked in 5% milk and hybridized with the following primary antibodies: anti-ARMC8 (1:500; E-1, sc-365307, Santa Cruz Biotechnology); anti-MAEA (1:200; AF7288, R&D Systems); anti-muskelin (1:500; C-12, sc-398956, Santa Cruz Biotechnology); anti-RanBPM (1:4000; 5 M, 71-001, Bioacademia); anti-TWA1 (1:200; NBP1-32596, Novus Biologicals); anti-vinculin (1:10,000; E1E9V, Cell Signaling Technology); anti-WDR26 (1:4000; ab85962, Abcam – this antibody also gives rise to non-specific products as shown in Fig. S1); anti-RMND5A (1:2000; custom-made antibody from Yenzym Antibodies); anti-SAP62 (1:1000; A-3, sc-390444, Santa Cruz Biotechnology); tankyrase-1/2 (1:1000; E-10, sc-365897, Santa Cruz Biotechnology); anti-HDAC2 (1:500; 5113, Cell Signaling Technology); anti-macroH2A1 (1:1000; ab37264, Abcam); anti-p14ARF (1:1000; ARF 4C6/4, sc-53392, Santa Cruz Biotechnology); anti-GAPDH (1:5000; 6C5, sc-32233, Santa Cruz Biotechnology); anti-actin (1:10,000; A5441, Sigma-Aldrich); and anti-lamin A/C (1:1000; E1, sc-376248). The following secondary antibodies were used: peroxidase-conjugated AffiniPure goat anti-mouse IgG (H+L) (1:5000; 115-035-003, Cedarlane); goat anti-rabbit IgG (H+L)-HRP conjugate (1:5000; 1706515, Bio-Rad); and donkey anti-sheep IgG HRP affinity-purified pAb (1:5000; HAF016, Cedarlane). Blots were developed using Clarity Western ECL Substrate (Bio-Rad) and imaged using ChemiDoc MP (Bio-Rad). ImageLab (Bio-Rad) was used to quantify band intensity.

Pre-conjugation of Dynabeads Protein G (10004D, Invitrogen) and 25 µg of mouse IgG (sc-2025, Santa Cruz Biotechnology) or RanBPM (F-1, sc-271727, Santa Cruz Biotechnology) antibody was performed as previously described (Maitland et al., 2021). Prior to immunoprecipitation, cytoplasmic and whole-cell extracts were adjusted to 10% glycerol, 0.25% NP-40 and 1 mM EDTA. The crosslinked beads were then incubated with extracts overnight at 4°C while rotating. Beads were washed three times with immunoprecipitation (IP) wash buffer (25 mM HEPES pH 7.9, 60 mM KCl, 0.5 mM EDTA, 0.25% NP-40, 12% glycerol), resuspended in SDS loading buffer and boiled at 95°C prior to SDS-PAGE.

The washing and elution protocols were adapted from a previous study (Kaboord et al., 2015). After immunoprecipitation and bead washing with IP wash buffer, beads were washed three times in 50 mM ammonium bicarbonate (ABC), followed by five times in high-performance liquid chromatography (HPLC)-grade water. Proteins were eluted with two rounds of 5-min room temperature incubation in 0.5% formic acid (FA), 30% acetonitrile with 1250 rpm shaking on a Thermomixer C (Eppendorf, 2231000667). Pooled eluted samples were spun down briefly to collect any residual beads, transferred to a new tube and dried in a SpeedVac (Thermo Fisher Scientific). The dried protein was resuspended in 6 M urea and incubated in 10 mM dithiothreitol (DTT) for 30 min, followed by incubation in 10 mM iodoacetamide for 25 min in the dark, and the sample was then methanol precipitated as described previously (Kuljanin et al., 2017). The protein pellet was resuspended in 200 µl of ABC and subjected to a sequential digest first with 250 ng of LysC (125-05061, Wako Chemicals) for 4 h, then 500 ng of trypsin/LysC (V5071, Promega) for 16 h, followed by 500 ng of trypsin (V5111, Promega) for an additional 4 h. Digested samples were incubated at 37°C at 600 rpm with interval mixing (30 s mix, 2 min pause) on a Thermomixer C. After the last digestion, samples were acidified with 10% FA to pH 3–4 and centrifuged at 15,000 g to pellet insoluble material. The supernatant was filtered by passing through a >10 kDa cellulose membrane (UFC501096, Sigma-Aldrich) and the flowthrough was dried in a SpeedVac, then resuspended in 0.1% trifluoroacetic acid and desalted with C18 Ziptips (Z720070, Sigma-Aldrich).

The eluted peptides were dried in a SpeedVac and resuspended in 20 µl of 0.1% FA. 5 µl of the eluted peptides was injected into a Waters M-Class nanoAcquity HPLC system (Waters) coupled to an ESI Orbitrap mass spectrometer (Q Exactive plus) (Thermo Fisher Scientific) operating in positive mode. Buffer A consisted of mass spectrometry-grade water with 0.1% FA and buffer B consisted of acetonitrile with 0.1% FA (Thermo Fisher Scientific). All samples were trapped for 5 min at a flow rate of 5 ml min⁻¹ using 99% buffer A and 1% buffer B on a Symmetry BEH C18 Trapping Column (5 mm, 180 mm×20 mm, Waters). Peptides were separated using a Peptide BEH C18 Column (130 Å, 1.7 mm, 75 mm×250 mm) operating at a flow rate of 300 nl min⁻¹ at 35°C (Waters). Samples were separated using a non-linear gradient consisting of 1–7% buffer B over 1 min, 7–23% buffer B over 179 min and 23–35% buffer B over 60 min, before increasing to 98% buffer B and washing. The MS acquisition settings are provided in Table S1.

All MS raw files were searched in MaxQuant version 1.5.8.3 using the Human UniProt database (reviewed only; updated on May 2017 with 42,183 entries) (Bateman et al., 2015; Cox and Mann, 2008). The number of missed cleavages was set to three, cysteine carbamidomethylation was set as a fixed modification and oxidation (M), N-terminal acetylation (protein) and deamidation (NQ) were set as variable modifications (maximum number of modifications per peptide=5), peptide length ≥6. Protein and peptide FDR was set to 0.01 (1%) and the decoy database was set to revert. Match between runs was enabled and all other parameters left as default. The protein lists were loaded into Perseus and proteins identified by site, reverse and contaminants were removed. Proteins that were present in over 50% of negative controls in all HeLa cell experiments imported into the CRAPome database (updated June 30, 2020) were also removed (Mellacheruvu et al., 2013). The remaining proteins and corresponding label-free quantitation (LFQ) intensities were formatted and input into the MiST algorithm (Verschueren et al., 2015). Proteins with MiST scores ≥0.75 were considered HCIPs. GO analysis was performed using stringApp (Doncheva et al., 2019) on Cytoscape (Shannon et al., 2003), with genome used as the background and FDR ≤0.05. All interactomes were designed on Cytoscape. The BioGrid database was accessed on November 11, 2021 and used to isolate previously identified CTLH complex interactors (Oughtred et al., 2021).

Cells were seeded on coverslips 24 h before being fixed with 4% paraformaldehyde for 10 min at 4°C, permeabilized with 0.5% Triton X-100 for 10 min at room temperature, and blocked with 5% FBS for 1 h at room temperature. Duolink II (Sigma-Aldrich) in situ proximity ligation assay was then performed as previously described (Fell et al., 2016). Tankyrase-1/2 (1:1000; E-10, sc-365897, Santa Cruz Biotechnology), RanBPM (1:400; K-12, sc-46253, Santa Cruz Biotechnology) and macroH2A1 (1:2000; ab37264, Abcam) primary antibodies were diluted in 5% FBS in PBS. Duolink Orange reagent (DUO92007, Sigma-Aldrich) was used as the detection reagent mentioned in the protocol. Cells were mounted on glass slides with Prolong Gold antifade reagent with DAPI (Molecular Probes by Life Technologies). Images were captured using an Olympus BX51 microscope and analyzed using Image-Pro Plus v4.5 software (Media Cybernetics).

Differences between two groups were compared using unpaired two-tailed t-test on GraphPad PRISM 8.0 (GraphPad Software). Results were considered significant for P<0.05.

We would like to thank all current members of the Schild-Poulter and Lajoie labs for the helpful discussions and technical expertise. We would also like to acknowledge the Robarts Research Institute and the Department of Biochemistry for the communal reagents and equipment needed to complete this study.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259638.