Inhibition of HDAC6 activity through interaction with RanBPM and its associated CTLH complex

Ran binding protein M (RanBPM), also referred to as RanBP9, is a 90 kDa ubiquitous protein localized both in the nucleus and the cytoplasm which has been implicated in various cellular functions but has no intrinsic enzymatic activity. RanBPM contains a Spla kinase and ryanodine receptor (SPRY) domain, known to mediate protein-protein interactions. It also contains a lissencephaly type-1 like homology (LisH) domain, known to mediate protein dimerization and tetramer formation and a C-terminal to LisH (CTLH) domain, whose function remains unknown [1‐3]. The C terminal CT-11-RanBPM (CRA) domain is made up of six helices that resemble the death domain superfamily [4]. RanBPM has been shown to interact with numerous proteins, implicating RanBPM to function in a variety of cellular processes including cell adhesion, migration, microtubule dynamics, and gene transcription [5‐7]. It has been hypothesized that RanBPM functions as a scaffolding protein that is part of a large complex [5, 6, 8, 9]. RanBPM is well conserved in eukaryotes and a RanBPM counterpart, glucose-induced degradation-deficient 1 (Gid1), also called vacuolar import and degradation 30 (Vid30), has been identified in yeast [10, 11]. Gid1 also contains SPRY, LisH, CTLH and CRA domains and is a component of a large protein complex made up of several other Gid proteins [5, 10‐12]. The mammalian homologs of almost all of these proteins have also been found in a large complex, which has been called the CTLH complex [13]. The CTLH complex also includes the protein muskelin, which is not encoded in the yeast genome [13]. The yeast complex functions as an E3 ubiquitin ligase in the degradation of fructose-1,6-bisphosphatase (FBPase), a gluconeogenic enzyme required when yeast cells are grown in a carbon-poor medium [5, 10, 11]. The Really Interesting New Gene (RING) domains of Gid2 and Gid9 which confer the E3 ubiquitin ligase activity in the Gid complex are conserved in the human homologs required for meiotic nuclear division homolog A (Rmnd5A) and macrophage erythroblast attacher (MAEA), respectively [11, 13, 14]. This suggests that the human CTLH complex may also have E3 ubiquitin ligase activity, however this remains to be demonstrated.

Our previous studies have suggested that RanBPM has tumour-suppressive activities. Downregulation of RanBPM resulted in decreased apoptotic activation in response to ionizing radiation (IR) and caused increased expression of Bcl-2, an anti-apoptotic Bcl-2 family member [2]. Downregulation of RanBPM also resulted in loss of growth factor dependence and in increased cell motility, suggesting that RanBPM expression confers activities that restrict cell growth and cell migration [15].

Histone deacetylase 6 (HDAC6) is a class IIb HDAC and, unlike other HDAC enzymes, HDAC6 shows cytoplasmic localization. HDAC6 uniquely has duplicate deacetylase domains as well as a C-terminal binder of ubiquitin zinc finger (BUZ) domain, which is able to bind ubiquitin [16]. Although named HDAC6, it does not have detectable deacetylase activity toward histones in vivo [17, 18]. Its most characterized substrates include α-tubulin, heat shock protein 90 (Hsp90) and cortactin [17]. HDAC6 deacetylation activity is both negatively and positively regulated by post-translational modifications. Several kinases, such as protein kinase C (PKC) ζ, PKCα, G protein-coupled receptor kinase 2 (GRK2), glycogen synthase kinase 3 β (GSKβ), casein kinase 2 (CK2), ERK and Aurora A phosphorylate HDAC6 and promote HDAC6 α-tubulin deacetylase activity [19‐25]. Conversely, epidermal growth factor receptor (EGFR) phosphorylation of HDAC6 decreases tubulin deacetylase activity [26]. Acetylation of HDAC6 by p300 also inhibits HDAC6 α-tubulin deacetylase activity [27]. In addition to being regulated by post-translational modification, HDAC6 is also regulated by protein-protein interactions. HDAC6 association with dysferlin, p62, paxillin, tau and tubulin polymerization-promoting protein/p25 (TPPP/p25) result in decreased α-tubulin deacetylase activity [28‐32].

HDAC6 associates with microtubules and has been shown to promote cell motility through deacetylation of α-tubulin and/or cortactin [18, 33, 34]. Overexpression of HDAC6 results in increased cell motility and knockout, downregulation or inhibition of HDAC6 by trichostatin A (TSA) or tubacin results in severely reduced cell migration [18, 33‐37]. This clearly indicates a role for HDAC6 catalytic activity in promoting cell motility.

HDAC6 has been implicated in cancer development and HDAC6-specific inhibitors have emerged as a chemotherapeutic agent to combat cancer. HDAC6 is required for in vitro oncogene-induced cell transformation and transforming growth factor (TGF) β1 induced epithelial-mesenchymal transition (EMT) [38, 39]. HDAC6 expression is also required to maintain anchorage-independent growth of established cancer cell lines [38]. HDAC6 has been shown to promote tumour formation in mouse models [38]. Upregulated HDAC6 levels have been observed in many cancer cell lines and in cohorts of oral squamous cell carcinoma (OSCC) and hepatocellular carcinoma (HCC) patients [40, 41]. HDAC6 has also been demonstrated to play a role in promoting angiogenesis [42, 43]. Several clinical trials are in progress using ACY-1215, an HDAC6 specific inhibitor alone or in combination with other agents. Preclinical studies for ACY-1215 have shown very promising results for the treatment of multiple myeloma, non-Hodgkin lymphoma and inflammatory breast cancer and the treatment was well tolerated in animals [44‐46].

We have previously demonstrated that RanBPM is able to form a complex with HDAC6 [47]. The LisH/CTLH domains of RanBPM were found necessary for association with HDAC6, as deletion of these domains resulted in loss of interaction [47]. We also reported that RanBPM is able to inhibit HDAC6 activity using an in vitro HDAC6 activity assay. Consistent with this, levels of acetylated α-tubulin, a specific HDAC6 substrate were found significantly reduced in cells stably expressing RanBPM shRNA compared to those expressing a control shRNA [47]. Interestingly, we also reported that RanBPM partially co-localizes with microtubules in both Hela and 3T3 mouse embryonic fibroblasts (MEFs) and that RanBPM associates with α-tubulin using co-immunoprecipitation [48].

In this study, we show that RanBPM-mediated inhibition of HDAC6 α-tubulin deacetylase activity is dependent on its association with HDAC6. The RanBPM-HDAC6 interaction requires the second catalytic domain of HDAC6 and the LisH domain of RanBPM. We show that HDAC6 does not require RanBPM to associate with microtubules, but that RanBPM colocalization to microtubules requires HDAC6. Furthermore, we demonstrate that components of the CTLH complex associate with both microtubules and HDAC6 and that muskelin, another component of the CTLH complex, associates with HDAC6 and mediates HDAC6 inhibition. Lastly, RanBPM was found to inhibit HDAC6 mediated cell migration. Our work suggests that RanBPM, together with the CTLH complex, associates with HDAC6 and inhibits HDAC6 activities.

Expanding on our previous findings that RanBPM is able to inhibit HDAC6 deacetylase activity, our study demonstrates a functional effect of RanBPM on HDAC6 activity in cell migration and reveals the involvement of the CTLH complex in HDAC6 inhibition. Investigation of the specific domains of RanBPM and HDAC6 that mediate their interaction revealed that, while their interaction does not appear to be direct, the LisH domain of RanBPM is required for association with HDAC6, while HDAC6 interaction with RanBPM requires its second catalytic domain. We show that RanBPM colocalization with α-tubulin is dependent on HDAC6 and that both the LisH and CTLH domains are required for this association. We demonstrate that members of the CTLH complex are able to colocalize with microtubules and inhibit HDAC6 activity and that muskelin is able to associate with HDAC6 independently of RanBPM, suggesting that muskelin may bridge RanBPM to HDAC6. Finally, we show that RanBPM’s association with HDAC6 inhibits HDAC6 activity and prevents HDAC6-mediated cell migration.

A proteomic screen previously demonstrated that muskelin associates with HDAC6 [50]. We confirm here that muskelin does form a complex with HDAC6 through co-immunoprecipitation analysis and show that the HDAC6-muskelin association occurs independently of RanBPM, as muskelin and HDAC6 still associated in RanBPM shRNA cells. We previously showed that RanBPM associates with both HDAC6 and the microtubules, specifically α-tubulin [48], and RanBPM and muskelin have been shown to be associated and be a part of the CTLH complex [5, 13, 51]. Multiple domains of RanBPM are required for the interaction with muskelin, including the LisH domain [51]. In this study we demonstrated that other members of the CTLH complex, namely Twa1 and MAEA, are also localized at microtubules. This suggests that the entire CTLH complex is present at microtubules and associates with HDAC6 through muskelin. Interestingly, there is evidence of other LisH domains mediating interactions with HDAC proteins. Transducing β-like 1 (TBL1) and transducing β-like 1 receptor (TBLR1), LisH domain-containing proteins, were shown to be part of a large protein complex encompassing HDAC3 that functions as a transcriptional repressor. Similar to how the RanBPM association with HDAC6 was significantly reduced upon deletion of the LisH domain, both TBL1 and TBLR1 lose their ability to associate with HDAC3 when the LisH domain is removed [54]. Therefore, it is possible that the LisH domain of muskelin mediates its association with HDAC6, bringing the entire CTLH complex, including RanBPM, in proximity to HDAC6.

Our data demonstrate that RanBPM association with microtubules is mediated through its LisH and CTLH domains. LisH domains have been shown to be important for dimerization and also binding to microtubules [3], however the role of the CTLH domain in microtubule interaction is still unclear. Also, whether muskelin is required for RanBPM to associate with microtubules, via α-tubulin or if this association is dependent on another protein, which would bridge RanBPM and α-tubulin remains to be investigated. As all members of the CTLH complex, with the exception of Armc8, have LisH and CTLH domains [13], it is also possible that other CTLH complex members may function to recruit the CTLH complex to microtubules [48].

Our results suggest that the CTLH complex modulates HDAC6 activity, but how this is achieved remains to be investigated. The obvious conclusion would be that the CTLH complex, through its putative E3 ubiquitin ligase activity, inhibits HDAC6 by targeting it for degradation by the proteasome through the addition of ubiquitin. Recently, RMND5A from Xenopus laevis was demonstrated to have E3 ubiquitin ligase activity [55], but whether the mammalian CTLH complex retains E3 ubiquitin ligase activity still remains to be verified. We have previously demonstrated, however, that HDAC6 protein levels remain unchanged in control and RanBPM shRNA cells [47], indicating that RanBPM is not inhibiting HDAC6 activity through modulation of its protein levels. Thus, the CTLH complex activity could potentially affect an HDAC6 regulator, as HDAC6 activity is positively and negatively modulated through many post-translational modification events and interactions with several proteins. HDAC6 activity is positively regulated by phosphorylation by PKCζ, PKCα, GRK2, GSKβ, CK2, ERK and Aurora A and negatively regulated by phosphorylation by EGFR and acetylation by p300 [19‐27]. Therefore we speculate that RanBPM’s inhibitory effect on HDAC6 is mediated by the E3 ubiquitin ligase activity of the CTLH complex, which could target an activator of HDAC6 activity for proteasomal degradation.

HDAC6 activity is also modulated through its interaction with other proteins. Association with dysferlin, p62, paxillin, tau and TPPP/p25 have all been reported to result in decreased HDAC6 α-tubulin deacetylase activity [28‐32]. Similar to RanBPM, p62 associates with HDAC6 via the second catalytic domain and inhibits its tubulin deacetylase activity, but the precise mechanism of p62 inhibition of HDAC6 has not yet been elucidated [29]. The interaction of TPPP/p25 and HDAC6 was reduced by the presence of tubulin [32], suggesting that HDAC6 associates with TPPP/p25 and tubulin via the same domain. Similarly, dysferlin, a transmembrane protein, inhibits HDAC6 deacetylation of α-tubulin by associating with HDAC6, preventing it from interacting with α-tubulin [28]. However, it does not appear that this is the mechanism through which RanBPM inhibits HDAC6 α-tubulin deacetylase activity, as we observed no significant decrease in the colocalization of HDAC6 and α-tubulin in cells expressing RanBPM shRNA.

HDAC6 retains the ability to associate with α-tubulin even when its catalytic activity is inhibited, either by mutagenesis of critical residues within the catalytic domain or by treatment with an inhibitor [17]. RanBPM is still able to associate with HDAC6 even when HDAC6 is inactivated by chemical inhibition, tubacin treatment or by mutagenesis rendering it catalytically incompetent (Additional file 3: Fig. S3). This indicates that the catalytic activity of HDAC6 is not required for its association with RanBPM.

HDAC6 has been shown to promote cell motility through deacetylation of α-tubulin and/or cortactin. Overexpression of HDAC6 results in increased cell motility, however, overexpression of a catalytically inactive HDAC6 mutant shows motility similar to that of control cells [33]. Consistently, downregulation, knockdown or chemical inhibition of HDAC6 results in reduced cell migration [18, 34‐37]. This evidence clearly indicates a role for HDAC6 catalytic activity in promoting cell motility. Interestingly, RanBPM has previously been shown to have a role in inhibiting cell migration. RanBPM was found to inhibit chemotactic migration by associating with leukotriene B4 receptor 2 (BLT2) [56]. Recently, a study performed in gastric cancer cells also reported that downregulation of RanBPM resulted in decreased cell adhesion and increased cell motility [57]. Our own studies also uncovered that cells with stable downregulation of RanBPM had increased migration compared to control cells in a wound healing assay, indicating that RanBPM functions to inhibit cell migration [15]. This led us to evaluate the effect of RanBPM inhibition of HDAC6 on cell migration. RanBPM shRNA cells exhibited increased cell migration when compared to control shRNA and expression of WT RanBPM but not a LisH/CTLH domain RanBPM deletion mutant, which no longer associates with HDAC6, was able to restore cell migration activity to that of control cells. We also demonstrated here that inhibition of HDAC6 with tubacin restored cell migration to that of control cells, indicating that the loss RanBPM modulation of HDAC6 activity is responsible for the increased migration observed in RanBPM shRNA cells. Interestingly, previous studies showed that siRNA-mediated downregulation of muskelin resulted in significantly increased cell migration compared to control siRNA [58], similar to what we observed with stable downregulation of RanBPM. This suggests that the CTLH complex functions to inhibit cell migration and disruption of the complex by downregulation of its components relieves the inhibition of cell migration.

As RanBPM is able to inhibit HDAC6-mediated cell migration, which promotes tumour invasiveness, this identifies a tumour suppressive function for RanBPM. RanBPM has been identified to have other tumour suppressor functions. We previously showed that RanBPM through its inhibition of ERK signaling promotes apoptosis by preventing expression of pro-apoptotic Bcl-2 family proteins at both the transcriptional and protein level [2, 15]. Similarly, RanBPM upregulates the mRNA levels of the pro-apoptotic transcription factor p73, contributing to the activation of apoptosis [59]. Furthermore, downregulation of RanBPM resulted in loss of growth factor dependence as RanBPM shRNA cells were shown to continue to survive and proliferate in the absence of growth serum [15].

Aside from promoting cell migration and consequently tumour invasiveness, HDAC6 contributes to other oncogenic activities. Increased expression of HDAC6 results in deacetylation of Hsp90, enhancing its chaperone activity. Hsp90 client proteins include breakpoint cluster/Abelson murine leukemia viral oncogene homolog 1 (Bcr/Abl), Raf, epidermal growth factor receptor 2 (ErbB2) among others. Many Hsp90 client proteins contribute to cell growth and survival pathways commonly exploited in cancer cells. In fact, Hsp90 inhibitors are also under investigation as chemotherapeutic agents [60]. HDAC6 has also been demonstrated to play a role in promoting angiogenesis, a hallmark of cancer [42, 43, 61]. It would be of interest to evaluate whether RanBPM and its associated CTLH complex are also able to restrict these other oncogenic activities of HDAC6.

In summary our results demonstrate a functional effect of RanBPM on HDAC6 activity in cell migration and reveal the involvement of the CTLH complex in HDAC6 inhibition. As HDAC6 has been demonstrated to have a causal role in the development of cancer, specific inhibitors have emerged as a promising target for cancer treatment and have been shown to result in decreased cell growth and decreased tumour formation in preclinical studies [44‐46]. Since RanBPM has been shown to have tumour suppressor functions and has been demonstrated to inhibit HDAC6 activity, this study provides mechanistic insights to understand the function of RanBPM and its associated CTLH complex in prevention of tumourigenesis and cellular transformation through regulation of HDAC6. Understanding the mechanism by which the CTLH complex is able to inhibit HDAC6 could provide insight on the development of small molecule HDAC6 inhibitors that could be used in chemotherapy treatment.