Identification Of Small Molecule TRABID Deubiquitinase Inhibitors By Computation-Based Virtual Screen
- First Online:
- Cite this article as:
- Shi, T., Bao, J., Wang, N.X. et al. BMC Chem Biol (2012) 12: 4. doi:10.1186/1472-6769-12-4
Wnt/β-catenin-mediated gene transcription plays important roles in a wide range of biological and pathophysiological processes including tumorigenesis where β-catenin-mediated transcription activity frequently elevates. TRABID, a deubiquitinase, was shown to have a positive Wnt/β-catenin-mediated gene transcription and hence holds a promise as a putative anti-cancer target.
In this study, we used a combination of structure based virtual screening and an in vitro deubiquitinase (DUB) assay to identify several small molecules that inhibit TRABID DUB activity. However, these inhibitors failed to show inhibitory effects on β-catenin-mediated gene transcription. In addition, expression of TRABID shRNAs, wildtype TRABID, or the DUB activity-deficient mutant showed little effects on β-catenin-mediated gene transcription.
TRABID may not be a critical component in canonical Wnt/β-catenin signal transduction or that a minute amount of this protein is sufficient for its role in regulating Wnt activity.
The evolutionarily conserved Wnt signaling pathway regulates various developmental processes during embryogenesis and plays an important role for tissue homeostasis in adults. Wnt signaling pathway also plays an important role in tumorigenesis, particularly the formation of inherited and sporadic colorectal cancer as result of adenomatous polyposis coli (APC) mutation that leads to β-catenin accumulation in the nucleus [1–4]. Nuclear beta-catenin binds to and functions as a cofactor of lymphoid enhancer-binding factor (LEF-1)  and T cell factors (TCF)  to stimulate the transcription of Wnt target genes . Wnt-β-catenin signaling is essential in sustaining the cancer stem cell phenotype and is also involved in the transformation into malignant human squamous cell carcinomas . The loss of Wnt signaling pathway component APC in stem cells results in progressively growing neoplasia . Activation of Wnt/TCF pathway is also a determinant of lung adenocarcinoma metastasis to brain and bone. The phenotype of those metastatic derivatives of adenocarcinoma resembles the bronchioalveolar stem cells . Thus, it holds promises to prevent and/or treat cancers by targeting Wnt signaling pathway components.
TRAF-binding domain (TRABID), one of deubiquitination enzymes, was recently reported to specifically and positively regulate Wnt signaling pathway . TRABID is composed of a TRAF-binding domain in the C-terminus and three Zinc-finger (ZnF) motifs at the N-terminus . TRABID preferentially binds to lysine 63-linked polyubiquitin chains (K63 chains), but not K48-linked polyubiquitin chains (K48 chains), and specifically cleaves K63 chains . K63-linked ubiquitination was suggested to regulate substrate activity rather than protein stability [13–16]. The catalytic residues of TRABID reside in the OTU (ovarian tumor) domain within the TRAF-binding domain .
The OTU domain is conserved within the members of the OTU family of DUBs that include A20 and possess the cysteine protease activity . TRABID was also shown to bind to APC and may be responsible for its deubiquitination. However, the direct interaction between APC and TRABID was not detected . Knockdown of TRABID with RNAi resulted in downregulated expression of canonical Wnt target genes and decreased Wnt transcription activity, whereas its knockdown does not affect TNF-2 pathway . Epistasis analysis suggested that TRABID might act downstream of beta-catenin stabilization and affect the interaction of beta-catenin with LEF1 . Furthermore, TRABID heterozygosity suppressed the rough eye phenotype caused by ectopic Wingless expression in fly, but did not affect the phenotypes caused by the inhibition of Notch signaling or inhibition of EGF signaling. These results suggest that TRABID might be a potential drug target for controlling Wnt pathway activation .
In this study, we screened for small molecule inhibitors of TRABID by employing a combination of structure based virtual screening and an in vitro DUB assay. We searched for compounds in a chemical library from the National Cancer Institute that potentially bind to TRABID catalytic site based on the crystal structure of A20 catalytic domain . We identified several compounds that are able to inhibit the DUB activity of TRABID. However, these inhibitors failed to show inhibitory effects on Wnt activity. Furthermore, neither shRNAs that silenced TRABID efficiently nor overexpression of wild type (WT) TRABID or its DUB activity-deficient mutant showed inhibitory effects on canonical Wnt signaling activity.
Results and discussion
Searching on the surface of the equilibrated structure of TRABID OTU domain modeled from the crystal structure of A20 OTU revealed that there was a potential pocket adjacent to the catalytic center of OTU (Figure 1E). This pocket is part of a distal ubiquitin binding site at the catalytic site in A20 . Because of its spatial proximity to the catalytic site and its role in ubiquitin binding, it is reasonable to postulate that targeting this site may block the entrance of the substrates and hence inhibit the OTU DUB activity. Of note, during the molecular dynamics refinement, considerable conformational changes occurred in this site, making it more expanded comparing with the pocket size in the A20 crystal structure (Figure 1F).
Wnt signaling plays a crucial role in cancer cell growth and cancer stem cell maintenance, which are believe to be responsible for cancer metastasis and poor prognosis of chemotherapy. Thereby, regulation of Wnt signaling holds promise in chemotherapy of cancer. TRABID, one of K63-specific DUB enzymes, became an attractive candidate to achieve this task since it has been reported to regulate Wnt signal positively , as inhibition of TRABID presumably decreases Wnt signaling activity. However, in this study, we failed to confirm the notion that TRABID positively regulates canonical Wnt signaling based on several lines of evidence. Firstly, the compounds, which blocked TRABID DUB activity in vitro, failed to inhibit Wnt transcription activity in two tumor cell lines whereβcatenin-mediated transcription are known to be elevated. Secondly, the TRABID shRNAs failed to downregulate Wnt/β-catenin-mediated transcription activity or endogenous Wnt target gene Axin2 mRNA levels in these tumor cells. Thirdly, overexpression of WT and C443S mutant TRABID failed to show any effect on Wnt target transcription activity. We did observe that the siRNA described in the previous report  was able to inhibit Wnt reporter gene activity (Figure 4C-E). One possibility is that the effect of the siRNA reported in  is off target. A less likely possibility is that a minute amount of TRABID is sufficient for its regulation of Wnt signaling in these cells, and our shRNAs or inhibitors were not efficient enough to alter Wnt activity. Nevertheless, the fly data  have already suggested that TRABID might not be a core component of Wnt signaling. The observation that the TRABID inhibitors were able to inhibit cell growth suggests that TRABID may have a role in cell growth regulation. However, the mechanisms for this role of TRABID need to be further investigated.
We have used computation-based virtual screen to identify chemical inhibitors for TRABID, but failed to confirm that TRABID has a significant role in Wnt signaling. Nevertheless, inhibition of TRABID may inhibit cell growth.
Construction of TRABID expression vectors and shRNA expression vectors Vectors expressing GST-fused full length (FL) and N-terminal fragment (NT) (amino acides: 1–354) of human TRABID were constructed using pET-GST vector (EMD Chemicals Inc., Gibbstown, NJ). Proteins were expressed in bacteria and purified using glutathione sepharose 4B (Roche Diagnostics, Indianapolis, IN). FLAG-expression vectors for expression of the full-length and DUB-deficient (C443S) TRABID in mammalian cells were gifts from Dr. Hoanh Tran. FLAG-tagged proteins were immunoprecipitated using the M2 agarose affinity gel (Sigma-Aldrich, St. Louis, MO). TRABID DNA sequences were confirmed by DNA sequencing. TRABID shRNA targeting sequences were selected using the RNAi Codex algorithm (http://katahdin.cshl.org:9331_portal_scripts_main2.pl)  and the shRNAs were expressed using Migr-CMV-YFP-miR30 vector as previously reported . The shRNA target sequences are: shTrabd1, TGTCTCAACAAGCAGCAAAGT; shTrabd2, AGGAGCTAGGTAATGAGGAAC; shTrabd3, TGTCAGAACGTGGAATTAAGT; shTrabd4, TGATCATCCCAGACCTAATAA; shTrabd5, CTGGCACATATTCTTAGACGA; shTrabd6, TTGGAAAAGTCCGATTGCTCT. β-catenin siRNA  and TRABID siRNA  target sequence were described previously.
DUB enzymatic activity assay
For determining TRABID DUB enzymatic activity, hexa-K63 ubiquitin (0.01 ug/ul, Boston Biochem Inc., Cambridge, MA) and the compounds were incubated with TRABID immunoprecipitated from transfected HEK 293 T cells or prepared from bacteria for 3 hrs at 37°C in a DUB reaction buffer (50 mM HEPES at pH 7.4, 150 mM KCl, 10 mM DTT, 5% glycerol, 0.01% Triton X-100) . Samples were applied to SDS-PAGE under non-reducing conditions and then transferred onto members. Ubiquitin chains were detected using a rabbit anti-ubiquitin antibody (Bethyl Laboratories Inc., Montgomery, TX).For determining A20 DUB enzymatic activity, Penta-K48 ubiquitin (0.04 ug/ul, Boston Biochem Inc., Cambridge, MA), 1.5 uM of A20 catalytic domain (Boston Biochem Inc., Cambridge, MA), and various concentrations of compounds were incubated for 3 hrs at 37 C in an A20 DUB reaction buffer (50 mM HEPES at pH 8.0, 3 mM DTT, 0.01% Brij-35) .
Homology modeling and potential pocket identification There are two crystal structures of A20 OTU domains deposited in PDB with nearly identical sequences and structures. Because The OTU domain of A20, shared a high degree of amino acid homology with that of TRABID, its sturcutre is the ideal template for the homology modeling of TRABID OTU structure. We used one of the A20 OTU structures (from PDB ID 2VFJ ) as the template for homology modeling.. The homology model of TRABID was built by the Prime module of the Schrodinger molecular modeling package [22, 23]. A 3 ns molecular dynamics was performed by AMBER10  to equilibrate the initial model since there is considerable gap (15%) between the sequences of TRABID and A20 (Figure 1A). The final equilibrated structure is used for the further modeling. The ICM PocketFinder module based on an algorithm of utilizing a Van de Waals grid potential map and a carbon probe  was used to detect potential relevant pockets around the catalytic cysteine residue of the TRABID OTU domain.
Since there is no any knowledge about the characteristics and properties of the potential ligands, it is necessary to search all the compounds in the library to find potential hits. We thus carried out virtual screens of the National Cancer Institute (NCI) database (http://184.108.40.206/ncidb2) for chemical compounds that could potentially bind to the detected cavity. This database includes the coordinates of 250,251 small (M.W. < 1,000 Da) chemical compounds. In order to eliminate the large amount of irrelevant compounds we used Glide module [26, 27] in the Schrodinger package to perform hierarchical searches. The scoring function of High-throughput in Glide, which is less accurate but very fast, was used to perform the first round screening. Subsequently, the top 5000 compounds were selected and further tested with the normal scoring function in Glide. Finally, the most precise, but slower, scoring function in Glide  was used to perform the final screening with the top 500 compounds, and the top 200 compounds were requested from NCI and evaluated by the experimental assay.
Where ∆Gvw is the Van der Waals energy, ∆Ghb is the htdrogen bonding energy, ∆Gto is torsional energy, ∆Gel is electrostastic energy; and ∆sf is the surface energy.
TOPFLASH reporter assays
For testing the effect of compounds, cells were seeded in 48-well culture plates at a density of 0.2-0.3 million/ml, 250 ul/well and transfected with TOPFLASH or FOPFLASH and Renilla luciferase using the Lipofectamine Plus reagent (Invitrogen Corporation, Carlsbad, CA). The compounds were added 3 hrs later. For shRNA knockdown experiments, cells were transfected with an shRNA expressing plasmid and re-transfected with the Wnt reporter gene plasmids and the shRNA two days later. Wnt reporter gene activity was determined 24 hrs after the re-transfection and presented after normalization against the Renilla luciferase activity.
Cells were plated into 6 well plates at a density of 0.2 million/ml for 12–16 hrs. For compound inhibition experiments, cells were incubated with 20 μM of a compound for 24 hrs. For shRNA inhibition experiments, cells were transfected with the shRNAs and incubated for 72 hrs. Total RNAs were extracted using the RNAeasy mini kit (Qiagen, Valencia, CA) and transcripted into cDNAs. The levels of Wnt target gene mRNAs were determined using a MyiQ Single color real-time PCR detection system (Bio-Rad, Hercules, CA) with beta-actin as a control.
Cell growth assay
Cells were seeded at a density of 0.05 million/ml in 24-well culture plates and added with 20 μM compound. On indicated date, the numbers of living cells were counted using Guava Easycyte Mini Flow Cytometry System (Millipore Corporation, Billerica, MA).
We thank Xiaofeng Li and Yong Zhang for technical help. This work is supported by NIH grant to D.W. (CA132317).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.